(Empirical Approaches To Language Typology 54) Silvia Luraghi., Tuomas Huuomo (Eds.) - Partitive Cases and Related Categories-De Gruyter Mouton (2014) PDF

Silvia Luraghi and Tuomas Huumo (Eds.
)
Partitive Cases and Related Categories
Empirical Approaches
to Language Typology
Editors
Georg Bossong
Bernard Comrie
Kristine Hildebrandt
Yaron Matras
Volume 54
Partitive Cases and
Related Categories
Edited by
Silvia Luraghi and Tuomas Huumo
ISBN 978-3-11-034404-2
e-ISBN (PDF) 978-3-11-034606-0
e-ISBN (EPUB) 978-3-11-039457-3
ISSN 0933-761X
Library of Congress Cataloging-in-Publication Data

A CIP catalog record for this book has been applied for at the Library of Congress.
Bibliographic information published by the Deutsche Nationalbibliothek

The Deutsche Nationalbibliothek lists this publication in the Deutsche Nationalbibliografie;
detailed bibliographic data are available on the Internet at http://dnb.dnb.de.
6 2014 Walter de Gruyter GmbH, Berlin/Boston

Typesetting: RoyalStandard, Hong Kong
Printing and binding: CPI books GmbH, Leck
♾ Printed on acid-free paper
Printed in Germany
www.degruyter.com
Table of contents
List of abbreviations vii
Introduction 1
I Typological aspects
Silvia Luraghi and Seppo Kittilä

1 Typology and diachrony of partitive case markers 17
Matti Miestamo
2 Partitives and negation: a cross-linguistic survey 63
II Uralic languages
Anne Tamm
3 The partitive concept versus linguistic partitives: from abstract concepts to
evidentiality in the Uralic languages 89
Tuomas Huumo and Liina Lindström

4 Partitives across constructions: on the range of uses of the Finnish and
Estonian “partitive subjects” 153
Helena Metslang
5 Partitive noun phrases in the Estonian core argument system 177
M. M. Jocelyne Fernandez-Vest
6 Finnish Partitive and resultativity in translation(s): a discourse-cognitive
approach 257
III Basque
Urtzi Etxeberria
7 The definite article and the partitive particle in Basque: dialectal
variation 291
Borja Ariztimuño López

8 The origin of the Basque partitive 323
vi Table of contents
IV Slavic languages
Michael Daniel
9 The second genitive in Russian 347
Katia Paykin
10 The Russian partitive and verbal aspect 379
Elżbieta Tabakowska
11 Double government in Polish: a case study 399
V Historical perspectives on Indo-European languages
Eystein Dahl
12 Partitive Subjects and Objects in Indo-Iranian and beyond 417
Luz Conti and Silvia Luraghi

13 The Ancient Greek partitive genitive in typological perspective 443
Anne Carlier and Béatrice Lamiroy

14 The grammaticalization of the prepositional partitive in Romance 477
VI Oceanic languages
Peter Budd
15 Partitives in Oceanic languages 523
Subject index 563

Author index 567
List of abbreviations
A the most actor-like argument of a transitive verb

ABE abessive
ABL ablative
ADD additive
ADE adessive
ADJ adjectivizing suffix
Adj adjective
ADV adverbial
AFF affirmative
AG pronominal clitic or affix referring to the agent of prototypical
action verbs; Ancient Greek
ALL allative
ANA anaphoric
AOR aorist
ART article
AUX auxiliary
BEN benefactive
BR basic root
CAUS causative
CEOC Central/Eastern Oceanic
CL clitic
CMPL completive
CMPR comparative
CNG connegative
COM comitative
COMM comment marker
COMN common noun
COMP complementizer
COND conditional
CONJ conjunction
CONN connector
CONST construct
COP copula
viii List of abbreviations
D dual
DAT dative
DCL drinkable classifier
DEF definite
DEIC deictic
DEM demonstrative
DET determiner
DELA delative
DIMIN diminutive suffix
DIP discourse particle
DIRECT. PASS direct passive
DOM differential object marking
DSM differential subject marking
e–NP the sole argument of the existential clause

EC existential clause
ECL edible classifier
ELA elative
EMPH emphatic
EOC Eastern Oceanic
ERG ergative
EX existential
EXC exclusive
EXTR extraction
F feminine
Fi Finnish
FIP Finnish partitive
FUT future
FV final vowel
GCL general possessor classifier

GEN genitive
GEN 2 Russian second genitive
GENP general possessor-indexing host
HAB habitual
IE Indo-European
ILL illative
IMM immediate future
List of abbreviations ix
IMP imperative
IMPF imperfect
INC inclusive
INDF indefinite
INE inessive
INF infinitive
INS instrumental
IPFV imperfective
IRR irrealis
IRS irresultative
K kind of
LK linker
LMT limiter
LOC locative
M masculine
MDA modal adverb
MID middle voice
MODIF modifier
M/P medio/passive
N neuter, noun
NAR narrative
NEG negation
NMLZ nominalization, nominalizer
NOM nominative
NP noun phrase
NPS non-past
NS non-singular
NSP non-specific
NT neutral modality
O object; the ‘not most actor–like argument’ of a transitive verb

OBJ object
OBL oblique
OPT optative
PAR partitive
PASS passive
PERS personal article
x List of abbreviations
PFV perfective
PIE Proto-Indo-European
PL plural
PMP Proto-Malayo Polynesian
PNCV Proto North and Central Vanuatu
PNLOC proper name location
POC Proto-Oceanic
POL politeness marker
POSS possessive
PP prepositional phrase
PPF pluperfect
PPN Proto-Polynesian
PREP prepositional case
PRES presentational
PREV preverb
PRF perfect
PRG progressive
PROX proximal/proximate
PRP preposition
PRS present
PRSP prospective
PSR possessor
PST past
PTC particle
PTCP participle
Q question particle
QLF qualifier
QUAN quantifier
R realis
RED reduplication
REF referential
REFL reflexive
REL relative
RES resultative
S single argument of canonical intransitive verb

Sa Sami
SAP speech act participant
SBJ subject
List of abbreviations xi
SBJV subjunctive
SG singular
STAT stative
SUP superlative
SUPERESS superessive
TA tense/aspect marker
TAM tense/aspect/mood marker
TELIC marker of telicity
TR transitiviser
TRAN translative
V verb
VOC vocative
VP verb phrase
WOC Western Oceanic
X verbal complement
1 first person
2 second person
3 third person
Silvia Luraghi and Tuomas Huumo
Introduction
This collection of papers is devoted to partitive case markers across languages.
In particular, the contributors describe various aspects of partitive cases in the
Finnic languages and Basque, of the partitive genitive in some Indo-European
languages, the development of the so-called partitive article in Romance lan-
guages, and a variety of morphemes usually referred to as partitives in Oceanic
languages. As we will discuss further on in this introduction, linguistic items in
this heterogeneous set all share the feature of expressing indefiniteness, at least
in certain contexts and to a certain extent. In spite of being morphologically
case markers (or prepositions), they do not behave as normal cases, in that
they are not connected with a specific grammatical relation. Rather, they partly
function as determiners or indefinite quantifiers. Diachronically, this peculiar
function seems to develop out of a more typical case marking function. The com-
plete change, including recategorization, from case marker to determiner is
borne out by the development of the Romance partitive article. Remarkably,
partitive markers treated in this volume should not be confused with partitives
occurring in so-called partitive constructions (e.g., A piece of that cake). We
elaborate on this point below.
A dedicated partitive case occurs in Finnic languages and in Basque, and is
usually said to indicate partial affectedness of patients (cf. Blake 2001: 151). This
characterization obviously concerns the object-marking partitive, but the overall
range of uses of these cases is wider and varies from one language to another.
Even in Finnic and Basque, this definition does not cover all uses of the partitive
case, as it does not reckon with the existence of partitive subjects. The name
partitive assigned to this case in grammatical analyses of these languages reflects
its core function as indicator of a ‘partial’ meaning. In some other languages a
similar a function is attributed to a number of other cases, as in the case of the
Hungarian partitive-ablative, and the partitive-genitive of various Indo-European
languages (a separate partitive, lexically restricted, also exists in Russian). In
grammatical analyses of those languages these cases are not called partitives,
though some of their functions resemble those of dedicated partitive cases.
Even though studies devoted to language specific partitive cases are com-
paratively numerous (see below), cross-linguistic research is virtually inexistent.
One reason for this is a certain confusion concerning the term ‘partitive’ (this
issue is taken up again in Tamm, this volume). Indeed, this term is most often
used to refer to partitive constructions, that is, part-whole constructions such as
the English a piece of that cake (see Koptjevskaja-Tamm 2006). Partitive cases
2 Silvia Luraghi and Tuomas Huumo
can also be found in such constructions, but not necessarily. Partitive construc-
tions indicate a part of a given whole, as in the English example above. Partitive
markers described in this book, instead, typically convey, at least in some
contexts, the meaning of indefiniteness, which is not characteristic of part-whole
relations. Compare the following examples:
Dutch (constructed example)

(1) Fred at van de aardbeien.
Fred eat:PST.3 SG of ART. PL strawberry:PL
‘Fred ate of the (previously identified, belonging to a given set) strawberries.’
Finnish (from Heine & Kuteva 2004: 32–33)

(2) Elmeri löys-i mansiko-i-ta.
Elmer find-3 SG . PST strawberry-PL-PAR
‘Elmer found some (i.e. and indefinite quantity of not previously identified)
strawberries.’
While the prepositional phrase van de aardbeien ‘of the strawberries’ in (1) is a
partitive construction and indicates a partition of a previously identified whole,
the partitive NP mansikoita ‘strawberries’ in (2) basically indicates indefinite-
ness, and does not refer to a part of a previously identified whole. Remarkably,
Finnish partitive constructions feature the elative case, as in (3):
Finnish (from Koptjevskaja-Tamm 2001)

(3) pala tä-stä hyvä-stä kaku-sta
bit:NOM this- ELA good- ELA cake- ELA
‘A bit of this good cake.’
The partitive case can also occurs in this construction, as shown in (4):
Finnish (personal knowledge)

(4) pala tä-tä hyvä-ä kakku-a
bit:NOM this-PAR good-PAR cake-PAR
‘A bit of this good cake.’
Note, however, that in this construction it is more typical for the partitive to have
the indefinite meaning, and that the part-whole reading in (4) is triggered by
the occurrence of the demonstrative ‘this’. Indeed, if no demonstrative occurs,
indefiniteness seems to be the basic function of the partitive in this construction,
as in (5):
Introduction 3

(5) pala hyvä-ä kakku-a
bit:NOM good-PAR cake-PAR
‘A bit of good cake.’ (an indefinite cake, not a quantity of a previously
identified whole).
In fact, even example (4), in spite of its demonstrative, allows the alternative
reading where the partitive phrase tätä hyvää kakkua refers to a particular type
of cake (‘this kind of good cake’), not necessarily a certain individual cake, as
the elative construction in (3). Thus (4) but not (3) can be used in the context of
a café where cut-off pieces of different kinds of cake are for sale on plates.
In the literature, partitive cases are often confused with partitive nominal
constructions. This confusion is clearly reflected in Heine & Kuteva (2004: 32–
33), who mix up the two categories, and quote the Finnish sentence in (2) along
with a German partitive construction in (6):
German (from Heine & Kuteva 2004: 33)

(6) Gib mir ein bisschen vom Käse!
give:IMP.2 SG 1 SG . DAT one bit from.the cheese
‘Give me a bit of the cheese!’
Similar to the Dutch example in (1), the prepositional phrase vom Käse in (6)
indicates a part of a certain, well identifiable piece of cheese, for example the
cheese which the speaker sees on the table. The authors must be aware of
some lack of homogeneity, as they remark: “Note, however, that “partitive”
does not appear to be a unified notion (Martin Haspelmath, personal communi-
cation).” However, they do not attempt to go deeper into this issue.
Another distinction must be made between partitive case markers studied
here and pseudo-partitives (see Selkirk 1977). According to Koptjevskaja-Tamm
(2006: 218), a pseudo-partitive “merely quantifies over the kind of entity . . .
indicated by the other nominal”, as the prepositional phrase ‘of coffee’ in a
cup of coffee. Indeed, partitive case markers can also occur in pseudo-partitive
constructions, as in (7):
Finnish (from Koptjevskaja-Tamm 2001: 523)

(7) säkki peruno-i-ta
sack:NOM potato- PL- PAR
‘A sack of potatoes.’
Crucially, pseudo-partitives are non-referential. As shown in (7), non-referentiality

is a possibility for the Finnish partitive case. Note however that the Finnish
partitive case does not always entail non-referentiality. Indeed, the Finnish
object-marking partitive that reflects unbounded aspect can be used for highly
topical, definite phrases and even pronominal phrases referring to the speech
act participants; cf. (8a) and its perfective counterpart (8b):

(8a) Houkuttel-i-n sinu-a ulos.
lure- PST.1 SG you- PAR out
‘I was luring [trying to lure] you out.’
(8b) Houkuttel-i-n sinu-t ulos.

lure- PST-1 SG you-ACC out
‘I lured you out’ [successfully].
In any case, non-referentiality is often connected with partitive cases, as

shown by possible co-occurrence of partitive cases with negation. Accordingly,
the use of partitive cases may be more or less restricted, depending on the
language. In Basque, for example, the partitive occurs in negative sentences
and it can mark either the object of transitive verbs or the subject on intransitive
verbs (in other words, it can substitute the absolutive case in negative sentences;
see Ariztimuño, this volume, and Etxeberria, this volume). A connection be-
tween negation and partitive case marking also occurs in the Finnic and Slavic
languages (in the latter case, genitive case marking). The alternation between
the partitive and other cases also has connections with aspect: this has been
argued for Finnic (e.g., Heinämäki 1984, Kiparsky 1998; Huumo 2010, and the
papers by Metslang, Tamm and Fernandez-Vest in this volume; for a generative
approach, see e.g., Vainikka & Maling 1996, Nelson 1998; Brattico 2012 and
the literature mentioned there), Slavic (see e.g. Fischer 2004; Tabakowska, this
volume), and possibly Sanskrit (Dahl 2009; this volume).
As noted above, partitive marking is not restricted to NPs indicating patients:
in Finnish existentials, for examples, even agentive intransitive verbs such as
juosta ‘run’, opiskella ‘study’, etc. (unergative), allow partitive subjects. In some
Indo-European languages, besides partitive objects and partitive subjects
(mostly with unaccusative verbs, cf. Conti & Luraghi, this volume), partitive
adverbials also exist, for example in time expressions (see Conti & Luraghi, this
volume, on Ancient Greek).
The indefiniteness meaning conveyed by partitive case markers seems to
be connected with their existential entailment. This is also a reason why the
Introduction 5
extension of partitive case markers to subjects seems to start with existential

clauses (see Carlier & Lamiroy this volume). In Finnish, the quantifying function
of the partitive object is directly related with indefiniteness: it may indicate an
indefinite quantity of a substance (with mass nouns) or of individuals (in the
plural). De Hoop (1998: 204) discusses examples (9) and (10):
Finnish (from De Hoop 1998: 204)

(9) Tunne-n ruotsalaise-t
know- PRS .1 SG Swede- NOM . PL
‘I know the Swedes.’
(10) Tunne-n ruotsalais-i-a

know- PRS .1 SG Swede- PL- PAR
‘I know some Swedes.’
According to De Hoop (1998), (9) has a strong (referential) interpretation, while

(10) has a weak (existential) interpretation. Note however that the connection
between indefiniteness and non-referentiality is not straightforward: depending
on different theoretical approaches, both definites and indefinites are conceived
as possibly referential or non-referential (see Lyons 1999: 165–166), with specificity
playing a role in the case of indefinites. Following Partee (2008), Seržant (ms.)
introduces the notion of ‘decreased referentiality’. Here, we are not going to
elaborate on that: rather, we limit our observations to noting that the difference
between (9) and (10) is better captured simply in terms of definiteness. While (9)
is about a particular group of Swedes or about Swedes in general (the English
translation conveys the same alternative meanings), example (10) is indefinite
and indicates an indefinite group of Swedes (a detailed Finnish-English contras-
tive study on the expression of indefiniteness is Chesterman 1991).
In one of the few existing cross-linguistic descriptions of partitive cases,
Moravcsik (1978: 272) summarizes their typical semantic correlates as follows:
a. the definiteness-indefiniteness of the noun phrase;
b. the extent to which the object is involved in the event;
c. the completedness versus non-completedness of the event;
d. whether the sentence is affirmative or negative.
To these semantic correlates we can, on the basis of the papers in the current
volume, add non-assertive modality, which motivates partitive marking in certain
contexts in Finnic and in Basque (see Huumo & Lindström, this volume, and
Aritzimuño, this volume).
Importantly, Moravcsik remarks that the marking difference brought about

by the partitive “does not correlate with any difference in semantic case func-
tion”. In sum, the use of the partitive case seems to be at odds with the basic
function of cases, that is “marking dependent nouns for the type of relationship
they bear to their heads” (Blake 2001: 1): rather than to indicate a specific gram-
matical or semantic relation that a NP bears to the verb, the partitive seems to
indicate indeterminacy (in various manners). This has been noted by several
authors. For example, Laka (1993: 158) suggests that “what is referred to as
‘partitive case’ in Basque is a polar determiner, much like English any”. In
Finnish and Estonian, the functions of the partitive are also related to indeter-
minacy, unboundedness and polarity (see the papers on Finnic in this volume),
and it is noteworthy that the partitive is not the sole marker of any grammatical
function but alternates with other cases in all of its main functions: as marker
of the object (PAR ~ ACC ), the existential subject (PAR ~ NOM ) and the predicate
nominal (PAR ~ NOM ), as well as marker of the complements of adpositions (PAR ~
GEN ). In addition to case affixes, functions reminiscent of the partitive are indi-
cated by the so-called partitive article of some Romance varieties, which derives
from the preposition that has substituted the Latin genitive (Latin de).
The brief survey above shows that there are striking similarities in partitives
across languages, which are not limited to the indication of partial affectedness.
However, research on partitives is mostly limited to individual languages. The
present volume sets out to fill this gap by introducing a cross-linguistic perspec-
tive to the study of partitives. Most of the papers are based on presentations
given at the workshop on partitives organized by Silvia Luraghi and Tuomas
Huumo at the 2010 Meeting of the Societas Linguistica Europaea in Vilnius. The
papers give detailed accounts on central uses of partitive cases and reminiscent
expressions in a number of different languages or language groups. The range of
languages covered in the volume include Finnic languages, Basque, Romance,
Ancient Greek, Indo-Iranian languages, Russian, including regional varieties,
Polish, and Oceanic languages. In addition, two papers have a cross-linguistic
focus, and discuss the typology and the diachrony of partitive markers, and the
use of partitives with negation.
As remarked above, what is common to all these case markers is that they
partly display features which are not typical of a case: this issue is specifically
addressed in the papers by Etxeberia on Basque, and Ancient Greek by Conti
and Luraghi. These papers point out that here is some similarity between parti-
tives and quantifiers, that the partitive marks indefiniteness, and that it does not
have the function typical of a case, i.e. that of indicating a grammatical relation.
This is especially clear for Indo-European languages and Finnic, where the
partitive can mark both (existential) subjects and objects. In Indo-European
Introduction 7
this is impossible for the nominative and the accusative, whereas in Finnic the
nominative can also mark the object (the third option is the accusative). Clearly
this evolution (case marker ! marker of indefiniteness not connected with a
specific grammatical relation) is what enabled the Romance former partitive to
become an indefinite article (see Carlier & Lamiroy, this volume).
The functions of partitive cases and similar elements are manifold, and
relate to many different levels and linguistic subsystems. As already noted, in
addition to different meanings of ‘partiality’ (which can be considered the core
meaning of partitives) and indefiniteness, these elements are often used to indi-
cate non-referentiality (especially under negation), unboundedness of quantity
and / or aspect, irrealis uses, incremental themes, low transitivity and low
affectedness. Partitives often participate in a semantically motivated alternation
with other cases marking the same syntactic argument. A good example is
Finnish, where the partitive can be used to mark the object, the existential sub-
ject and the predicate nominal (predicative), but where it is not the sole marker
for any of these functions, as it alternates with other cases (the accusative and
the nominative). Indeed, even low transitivity verbs that are normally said to
always take the partitive, such as rakastaa ‘love’, can take the accusative in
resultative constructions, as shown in (11):

(11) Rakast-i-n hänet hullu-ksi
love-PST-1 SG 3 SG . ACC crazy-TRA
‘I loved him/her crazy.’
Similarly, in several Indo-European languages the partitive genitive is in free

variation with the accusative, and in some of them, such as Ancient Greek and,
to a more limited extent, Indo-Iranian languages, with the nominative. However,
Indo-European languages also often feature obligatory partitive/genitive direct
objects with low transitivity verbs, as is the case in Ancient Greek (see Conti &
Luraghi, this volume).
There are also differences between NPs marked with the partitive vs. other
cases related to cross-referencing in the verb. In Finnish, for example, partitive
subjects do not trigger agreement with the verb, and this is also the case for
partitive genitive subjects in some, though not in all, Indo-European languages.
The functions of the partitive are thus manifold, intertwined with many linguistic
subsystems and functional levels.
The present volume consists of fifteen papers, which are divided into sub-
sections based on the languages or language groups they study.
The first section of the book contains two chapters that discuss typological
perspectives on partitives. The first chapter, The typology and diachrony of
partitives by Silvia Luraghi and Seppo Kittilä, discusses typological features of
partitive markers surveyed in this book, along with a description of attested
developments that led to their creation. In the first part of the chapter, the
authors analyze the various semantic correlates of partitive markers as indicated
in (a)–(d) above, and show how they are reflected in the languages described in
the rest of the book. In the second part, the origin of partitive markers is traced
back as far as is allowed by attested sources. It turns out that partitive case
markers often originate from ablatives or genitives, but that a different origin is
perhaps reconstructable for Oceanic languages, that relates partitive markers to
the accusative case. The second chapter, by Matti Miestamo, is titled Partitives and
negation: a cross-linguistic survey. It starts from the observation that in some Euro-
pean languages – Finnic, Baltic, Slavic and Basque – NPs in the scope of negation
are marked, either obligatorily or as a matter of preference, with a case that has
a partitive-marking function (partitive or genitive). The phenomenon is referred
to as the partitive of negation. Miestamo reports the results of a large-scale typo-
logical survey of the partitive of negation and related effects of negation on the
marking of grammatical categories in NPs. In a larger context, the effects on NPs
are one of the many ways in which negation can affect the structure of clauses,
or one of the many ways in which negatives can show structural asymmetry
vis-à-vis affirmatives. This larger context becomes relevant when we start look-
ing for explanations for the link between partitives and negation, which has
been attributed to semantic and pragmatic factors, such as quantification,
referentiality/specificity, and aspect.
The second section of the book contains four chapters devoted to Finnic
languages. In Chapter 3, Partitive semantics and semantic partitives in the Uralic
languages, Anne Tamm addresses some of the confusing issues regarding parti-
tives on the basis of the empirical material on various examples from the Uralic
languages. The main point of the chapter is that it proposes a distinction
between “Linguistic Partitives” and “Partitive Concepts” in describing the parti-
tive phenomena. In this framework, the Partitive Concept is an abstract concept
that serves for comparing the semantics of grammatical forms to the semantics
of “part-of-N”. A Linguistic Partitive is a grammatical form that is related to the
meaning of the Partitive Concept. In several Uralic languages, partitive cases
have developed their specific semantics and pragmatics. The author divides the
Linguistic Partitive into functional and structural categories, depending on the
semantics of the partitive in the structure of the language at hand, and follows
the works that divide the Partitive Concept in two metonymically related sub-
concepts: the partitive and the pseudopartitive. Chapter 4, by Tuomas Huumo
Introduction 9
and Liina Lindström, has the title The Partitives stretching borders: How well do
Finnish and Estonian partitive subjects serve as a criterion for the existential
clause? In this chapter, the authors examine the definitions given for the range
of uses of the partitive subject in both languages, and compare the range of
existentials, their definition and form. They argue that existentials form a radial
category, with a prototype and less canonical instances, where the prototype is
clearly definable but the actual borderline between existentials and other clause
types is fuzzy. Moreover, the authors discuss differences between Finnish and
Estonian as regards the range of uses of the partitive, showing that in Estonian
the (plural) nominative is often used in expressions where the partitive would be
the natural option in Finnish. On the other hand, the use of the partitive in
negated and interrogative existentials seems to be more widespread in Estonian
than in Finnish, and this use seems to be motivated by the irrealis meaning
expressed by these constructions. Chapter 5, Partitive noun phrases in the Estonian
core argument system by Helena Metslang, discusses the role of the Estonian par-
titive case in association with different argument types. The partitive-permitting
arguments – the object and the existential (presentational) clause NP – are
analyzed in contrast with each other and with the transitive and intransitive
clause subjects. The distributions of semantic, pragmatic and coding properties
are examined among these nominative, genitive and partitive arguments. The
chapter further analyzes how the typologically disputatious referential hierarchy
differentiates the arguments and case-uses under scrutiny. The author also com-
pares differential case-marking systems of Estonian objects and existential NPs
and the salience of each factor is measured in the empirical data. Chapter 6
by M. M. Jocelyne Fernandez-Vest has the title Finnish Partitive revisited: a
discourse-cognitive approach, in comparison with some other Finno-Ugric and
Indo-European languages. Fernandez-Vest compares the main criteria that moti-
vate the use of the Finnish partitive: context/cotext, situation, and the inherent
aspect of the verb. She uses a corpus of oral and written language where the
verbs have been classified into three categories – RES (resultative), IRRS (irresul-
tative) and RES-IRRS (resultative-irresultative), according to their most typical
aspectual meaning. This distinction aims to refine the traditional conception of
resultativity in Finnish grammars. It shows the great number of RES-IRRS verbs
(nearly 50%) and highlights the difference between activity verbs and mental
verbs (nearly 70% of the second take a partitive object). The discussion of the
respective choice criteria relies on a comparison between translations of Finnish
texts into another Uralic language (Northern Sami) that has no partitive case
and neighboring Indo-European (Scandinavian) languages, and vice-versa.
The third section of the book contains two chapters, which are devoted to
Basque. Chapter 7, The definite article and the partitive in Basque: dialectal varia-
tion by Urtzi Etxeberria, aims to an as thorough as possible description of both
the Basque so-called definite article and the partitive particle [-rik] across its
various dialects. Etxeberria starts from the observation that these two elements
are related to each other; the partitive [-rik] is taken to be the morpheme that is
used in polarity contexts instead of the existential interpretation of the definite
article [-a(k)]. Chapter 8, The origin of the Basque partitive by Borja Ariztimuño
López, describes the main contexts in which the Basque partitive suffix -(r)ik
can appear, and presents a proposal about the source and the development of
its different uses. The origin of the morpheme is then explained, drawing on
grammaticalization theory.
The fourth section is devoted to Slavic languages, and contains three chapters.
Chapter 9, by Michael Daniel, concerns The second genitive in Russian. The chapter
contains an overview of the so-called second genitive in Russian, a nominal
form available for a minority of Russian nouns but widely used in certain con-
texts. The second genitive is a secondary case, as it may always be substituted
with a regular genitive form, while the opposite is not true. Daniel argues that a
major subset of the contexts where the second genitive may be used fits into
what is known as a functional category of partitive: for this reason, this form is
sometimes called Russian partitive. To a certain extent, indeed, the second geni-
tive is the form with which the regular genitive may be substituted in partitive
contexts. However, the analysis of the distribution of the second genitive shows
that the partitive meaning is not the only function of this form, which also dis-
plays several idiomatic uses and uses in combinations with prepositions. Russian
is also the topic of chapter 10, Russian Partitive and Verbal Aspect, by Katia Paykin.
The goal of the chapter is to revisit the most recent analyses of this question,
providing several critiques and alternative solutions. Paykin starts by clarifying
what exactly the term “partitive” means when considered in the context of
verbal aspect. Having established that the reality covered by the label in this
particular context corresponds to the partitive genitive, she discusses its func-
tioning as opposed to that of the nominative and the accusative, with which it
can compete. Her main emphasis is on the relationship between the partitive
genitive and the aspect of the verb assigning it. Chapter 11, Double government
in Polish: a case study by Elżbieta Tabakowska, sets out to demonstrate that the
opposition between the accusative and the genitive marking the direct object
with (some) transitive verbs, which in traditional Polish linguistics has been
classified as free variation, is motivated by semantic and pragmatic factors. The
author substantiates the claim that the opposition between structures with tran-
sitive verbs followed by the direct object in either the genitive or the accusative
case reveals a significant difference in meaning, which results from an intricate
interplay of lexical semantics, aspectual meaning of verbs, the meaning of
verbal prefixes, pragmatic factors and discourse structure.
Introduction 11
The fifth section, on Indo-European languages, also contains three chapters.

Partitive subjects and objects in Indo-Iranian and Greek consitute the topic of
chapter 12, by Eystein Dahl. Dahl starts from the observation that the genitive
is quite frequently used as an object marker and somewhat less commonly
used as a subject marker with finite verbs in the Indo-Iranian languages as well
in Indo-European more generally. In his chapter, he explores the semantic
domain of these constructions with particular regard to the lexical semantic
properties of the predicates selecting genitive-marked core arguments, claiming
that, in general, partitive subjects and objects are restricted to predicates with
a relatively low degree of transitivity. Specifically, these constructions appear to
be restricted to cases where the argument is represented as being only partially
involved in the situation, either having a low degree of control over the outcome
of the situation or only being partially affected by the situation. Chapter 13, The
Ancient Greek partitive genitive in typological perspective by Luz Conti and Silvia
Luraghi, gives a description of the partitive genitive in Ancient Greek. The
authors show that the partitive genitive can occur in a wide variety of syntactic
constructions, partly parallel to the genitive in other Indo-European languages,
or to other partitive markers discussed in this book. Hence, the partitive genitive
is not connected with a specific grammatical relation: rather, it indicates partial
involvement of a referent in an event, which is usually reflected in a low degree
of transitivity. In addition, the partitive genitive may indicate indefiniteness.
From the point of view of discourse organization, partitive subjects, which often
occur in presentative clauses, convey new information and are never topical. These
various semantic and pragmatic features of the partitive genitive are captured by
the authors in terms of low participation of a participant in an event. Finally,
chapter 14 by Béatrice Lamiroy and Anne Carlier concerns The grammaticaliza-
tion of the prepositional partitive in Romance. Lamiroy and Carlier discuss the
adverbal use of the partitive in Romance. Romance languages developed a
partitive marker out of the Latin adposition de (+NP), whose original meaning
was ‘away from’. This grammaticalization process is represented in a five-step
model, defining the different stages of the shift of morpho-syntactic categoriza-
tion, from Late Latin onwards: primitively a preposition, de (+NP) turns into an
article. It is shown that, although the source expression of the partitive is the
same in French, Italian, and Spanish, the outcome of the evolution differs signif-
icantly according to the language. The authors argue that the different patterns
which characterize the emergence of the partitive in Romance can be linked to
global typological properties regarding word order and information structure.
The last section contains a single chapter, Partitives in Oceanic languages,
by Peter Budd. This chapter presents a survey of partitive items in the Oceanic
languages, the large subgroup of the vast Austronesian language family. The
Oceanic subgroup stretches from Papua New Guinea (PNG) and the Micronesian
islands in the west to New Zealand in the south, Hawai’i in the north, and Rapa
Nui (Easter Island) in the east. Morphological partitives can be identified in
languages throughout the Oceanic subgroup and there is considerable syntactic
and semantic diversity: the chapter aims to provide a synthesis of the findings.
Acknowledgements
We would like to thank EALT editor Kristine Hildebrandt for accepting our volume
for publication in the series. Our gratitude goes to all contributors for providing
high quality papers, as well as for their active involvement in the internal review
and in the editing process. Thanks are also due to a number of other reviewers,
who helped us with their comments, and to our assistants Tommaso Claudi, who
helped us prepare the final manuscript and compile the list of abbreviations
and the subject index, and Krista Teeri, who helped us read the proofs.
References
Blake, Barry. 2001. Case. Cambridge: Cambridge University Press.
Brattico, Pauli. 2012. Structural Case Assignment and Phi-Agreement in Finnish. SKY Journal of
Linguistics 25. 29–59.
Chesterman, Andrew. 1991. On definiteness. A study with special reference to English and Finnish.
Cambridge: Cambridge University Press.
Conti, Luz. 2008. Zum Genitiv bei impersonalen Konstruktionen im Altgriechischen. Paper read
at the XIII. Fachtagung der Indogermanischen Gesellschaft, Salzburg 22–27 September,
2008.
Dahl, Eystein. 2009. Some semantic and pragmatic aspects of object alternation in Early Vedic.
In Jóhanna Barðdal & Shobhana Chelliah (eds.), The Role of Semantics and Pragmatics in
the Development of Case, 23–55. Amsterdam: Benjamins.
Fischer, Susann. 2004. Partitive vs. Genitive in Russian and Polish: an empirical study on
case alternation in the object domain. In Susann Fischer, Ruben van de Vijver and Ralf
Vogel (eds.), Experimental Studies in Linguistics. I, LiP 21. 123–137.
Heinämäki, Orvokki. 1984. Aspect in Finnish. In Casper de Groot & Hannu Tommola (eds.),
Aspect Bound: A voyage into the realm of Germanic, Slavonic and Finno-Ugrian aspectol-
ogy, 153–178. Dordrecht: Foris Publications.
Heine, Bernd & Tania Kuteva. 2004. World Lexicon of Grammaticalization. Cambridge: Cam-
bridge University Press.
Helasvuo, Marja-Liisa. 2001. Syntax in the Making: The emergence of syntactic units in Finnish
conversation. Amsterdam: Benjamins.
Hoop, Helen de. 1998. Partitivity. Glot International 3. 3–10.
Introduction 13
Huumo, Tuomas. 2003. Incremental Existence: The World According to the Finnish Existential
Sentence. Linguistics 41(3). 461–493.
Huumo, Tuomas. 2009. Fictive dynamicity, nominal aspect, and the Finnish copulative construc-
tion. Cognitive Linguistics 20(1). 43–70.
Huumo, Tuomas. 2010. Nominal aspect, quantity, and time: The case of the Finnish object.
Journal of Linguistics 46. 1–43.
Kiparsky, Paul. 1998. Partitive Case and Aspect. In Miriam Butt and Wilhelm Geuder (eds.),
The Projection of Arguments: Lexical and Compositional Factors, 265–308. Stanford. CSLI
Publications.
Koptjevskaja-Tamm, Maria. 2001. “A piece of the cake” and “a cup of tea”: partitive and
pseudo-partitive nominal constructions in the Circum-Baltic languages. In Östen Dahl &
Maria Koptjevskaja-Tamm (eds.), The Circum-Baltic Languages: Typology and Contact,
vol. 2, 523–568. Amsterdam: Benjamins.
Kopjevskaja-Tamm Maria. 2006. Partitives. In Keith Brown (ed.), Encyclopedia of Languages
and Linguistics, vol. 9, 2nd edn., 218–221. Amsterdam: Elsevier.
Laka, Itziar. 1993. Unergatives that Assign Ergative, Unaccusatives that Assign Accusative.
MITWPL 18. 149–172.
Luraghi, Silvia. 2003. On the Meaning of Prepositions and Cases. A Study of the Expression of
Semantic Roles in Ancient Greek. Amsterdam: Benjamins.
Luraghi, Silvia. 2009. The internal structure of adpositional phrases. In Johannes Helmbrecht,
Yoko Nishina, Yong-Min Shin, Stavros Skopeteas, Elisabeth Verhoeven (eds.), Form and
Function in Language Research: Papers in honour of Christian Lehmann, 231–254. Berlin:
Mouton de Gruyter.
Lyons, Christopher. 1999. Definiteness. Cambridge: Cambridge University Press.
Moravcsik, Edith. 1978. On the case marking of objects. In Joseph Greenberg et al. (eds.),
Universals of Human Language, vol. IV. Syntax, 249–290. Stanford, Ca: Stanford University
Press.
Nelson, Diane. 1998. Grammatical case assignment in Finnish. New York: Garland Pub.
Partee, Barbara H. 2008. Negation, intensionality, and aspect: Interaction with NP semantics.
In Susan Rothstein (ed.), Theoretical and Crosslinguistic Approaches to the Semantics of
Aspect, 291–317. Amsterdam: Benjamins.
Sands, Kristina & Lyle Campbell. 2001. Non-canonical subjects and objects in Finnish. In
Alexandra Aikenvald, R. M. W. Dixon, and Masayuki Onishi (eds.), Non-canonical Marking
of Subjects and Objects, 251–305. Amsterdam: Benjamins.
Selkirk, Elisabeth O. 1977. Some remarks on noun phrase structure. In Adrian Akmajian, Peter
Culicover & Thomas Wasaw (eds.), MSSB-UC Irvine Conference on the Formal Syntax of
Natural Language, 225–245.
Vainikka, Anne & Joan Maling. 1996. Is partitive case inherent or structural? In Jacob Hoeksma
(ed.), Partitives: Studies on the Syntax and Semantics of Partitive and Related Constructions,
179–208. Berlin: Mouton de Gruyter.
I Typological aspects
Silvia Luraghi and Seppo Kittilä
1 Typology and diachrony of partitive
case markers
Partitive cases constitute a rather heterogeneous category. They may be defined

formally, when the notion is confined primarily to languages with a dedicated
partitive case, such as Finnish and Basque. From a functional perspective, in
turn, also other languages may have morphemes that express the same function,
e.g. other cases (such as the genitive in several Indo-European languages) or
other formal means (adpositions, verbal cross-reference etc.). Functionally, par-
titive case markers can be used to express an array of functions. Typically, they
are related to expressing partiality or indefiniteness, but often also to lower tran-
sitivity. What is most notable is that there is no link between partitive cases and
a specific grammatical relation: partitive cases can code subjects and direct
objects; in addition, they may code adverbials and appear with adpositions in
some languages. Formal and functional properties of partitive case markers are
discussed thoroughly in this paper from a cross-linguistic perspective. The paper
also includes a discussion of the diachrony of partitive case markers.
Keywords: Partitive case, partitive genitive, indefiniteness, low transitivity,

part-whole relation
1 Introduction
As noted in the introduction, partitive case markers are understood in this book
as markers that are related to the partial meaning, covering also functions such
as the expression of indefiniteness and partial affectedness. This definition is
also employed in this chapter. As will become clear, the notion is not associated
with a specific grammatical relation, e.g. partitive object: it is typical for the
cases reviewed in this book to occur both with subjects and with objects, and
even with certain adverbials. In this sense, partitives differ from other cases,
whose typical function is to signal grammatical relations (see Blake 2001: 1).
Partitive case markers display a wide range of variation cross-liguistically,
both in form and in functions. In the first place, they can be defined both for-
mally and functionally. Following formal definitions of partitive, we are dealing
with languages that have a dedicated partitive case that is used for coding some
18 Silvia Luraghi and Seppo Kittilä
specific meanings, such as partial affectedness, indefiniteness, non-telicity (see

below for a more detailed discussion). Examples include languages such as
Finnish and Basque that are well-known for their partitive case. Functional defi-
nitions apply to languages that use some other case form (frequently the geni-
tive) for expressing meanings which are similar to those expressed by partitive
cases in the first type of languages. This kind of partitive is found, for example,
in Ancient Greek and many Slavic languages. In the second place, this array of
formally and functionally defined cases can express a variety of functions across
languages, and display semantic differences. The functions expressed by parti-
tive cases include an obvious quantitative component, partly related to, or de-
riving from part-whole relationships, indefiniteness, low or non-referentiality.
In addition, in many languages partitive cases are somehow related to lower
transitivity and/or imperfective aspect. Of these, the expression of quantitativity
is among the most frequent functions of partitive cases across languages.
Partitive case markers are typically highly polysemous: this makes it not
an easy task to define their primary functions across languages. With some of
the functions, the partitiveness (i.e. the expression of some sort of ‘partiality’)
cannot be straightforwardly identified, although a functional link may be found.
As is often the case, diachrony sheds some light on the common features of par-
titive case markers. When considered diachronically, most partitive case markers
appear to share a common origin. Indeed, they most often originate in partitive
constructions, and, as we show in section 4, they share a common development,
leading them to express meanings that range from partial quantification to
indefiniteness. However, the Oceanic languages offer evidence for a different
origin and for different developments, as we also argue in section 4.
Let us start by considering the Finnish partitive case. Among its other numer-
ous functions, the partitive case can indicate an indefinite number of items (with
plural nouns), or an unbounded quantity of a certain referent (with singular
nouns or mass noun), as in:

(1) a. naise-t tul-i-vat koti-in
woman-PL come-PST-3 PL home-ILL
‘The women came home.’
b. nais-i-a tul-i koti-in

woman- PL- PAR come-PST.3 SG home-ILL
‘Some (of the) women came home.’
Typology and diachrony of partitive case markers 19
(2) a. Aino sö-i leivä-n

Aino eat-PST.3 SG bread-ACC
‘Aino ate the (whole) bread.’
b. Aino sö-i leipä-ä

Aino eat-PST.3 SG bread- PAR
‘Aino ate (some of the) bread.’
In (1b), the Finnish partitive expresses indefiniteness; the group of women in

question is not known in the context. In favorable conditions, the construction
may also be related to a part-whole relationship. This would be possible, for
example, in a context where someone is asking whether all the women (we had
been talking about) came, but the answer is that only some of them (i.e., a part)
arrived (see further section 3.2.). In (2), as in (1), accusative vs. partitive variation
is connected in the first place with definiteness. In addition, it can also indicate
whether the whole bread or only a part of it has been eaten, that is, it can indi-
cate a part-whole relation. As expected, the accusative is used when the whole
bread has been consumed, while the partitive indicates that only a part of the
bread has been affected by the given event. The whole-part-reading becomes
prominent if we emphasize the definiteness of the bread by adding a demonstra-
tive like tämän/tätä (this:ACC/this:PAR) to the clause; if this is not the case, the
indefinite reading is more prominent. Indeed, as noted in the Introduction, the
part-whole relation is unambiguously indicated by the elative case, rather than
by the partitive (see the discussion concerning examples (3)–(5) in the Introduc-
tion to this volume).1 Of the two meanings that the Finnish partitive has in (1b),
only the indefiniteness meaning is common to all languages treated in this
book. The part-whole meaning is also present in some other languages, though
not in all. In addition, partitives may have a variety of other functions, as will be
discussed below.
The goal of this chapter is to give a cross-linguistic overview of partitive
case markers from a formal, functional and diachronic perspective. The chapter
is largely based on the data retrieved from the other chapters in the volume. This
has the consequence that some languages are better represented than others.
The chapter does therefore not aim at presenting any statistical data on the
examined phenomena, but its goal is to give an overview of the partitive markers
in the world’s languages based on the languages we have data for. First, a typol-
ogy of the mechanisms employed for partitive coding will be proposed. This will
1 The examples in (2) may also have other readings, e.g. (2b) ‘Aino was eating the bread (when
something happened)’ which are not relevant here.
be followed by a thorough functional typology of partitive markers based on the

functions partitive markers express across languages. In section 4, the diachrony
of the partitive markers will be discussed.
2 Formal typology of partitive markers

In this section, we discuss the formal means that languages use for coding par-
titive meanings. Indeed, besides case endings partitive meanings can be coded
in several other ways, though marginally. In the first place, languages that have
no morphological case can use adpositions in their place. In addition, other lan-
guages have partitive clitics or partitive articles, and even partitive verbal affixes,
as we illustrate below.
2.1 Case markers

Languages in which partitive meanings are expressed by case can be divided
into two groups, based on whether they have a dedicated partitive case, or whether
another case form is used for expressing partitive meanings. Finnish, Estonian
and Basque are well-known for their partitive cases. Examples are found in
(3)–(4) (the relevant elements appear in bold face):
Finnish (personal knowledge):

(3) Aino ju-o mehu-a
Aino drink- PST.3 SG juice- PAR
‘Aino is drinking (some) juice.’
Basque (Etxeberria, this volume)

(4) Goxoki-rik nahi al duzu?
candy- PAR want Q AUX
‘Do you want any candy?’
Partitive markers typically contrast with nominative/accusative/absolutive

markers in languages that have a partitive case. Finnish and Basque have a
distinct case form labeled as partitive, whose primary function is to express
partitive meanings. Typically, partitive cases are not directly connected with the
encoding of a specific grammatical relation. This is most clear in Finnish, in
which the partitive case can occur with direct objects, as in (2), with subjects of
unaccusative verbs as in (1), and even, occasionally, with subjects of transitive

verbs, as in example (5):
Finnish (Tuomas Huumo p.c.)

(5) Use-i-ta ihmis-i-ä odott-i satee-ssa bussi-a.
Many-PAR- PL person-PAR- PL wait-PST.3 SG rain- INE bus- PAR
‘Many people were waiting for the bus in the rain.’
Notably, in this group of languages, the partitive marker, being a case ending,
shares the morphological distribution of case endings. This means that it cannot
co-occur with other case endings, even though it is not connected with a specific
grammatical relation (in some languages, it can even mark adverbials).
In the second type of languages, another case form (typically the genitive) is
associated with expressing partitive meanings in addition to its basic function.
This is the case in many ancient and modern Indo-European languages, in
which the genitive is used as a partitive. Example (6) is from Avestan, and con-
tains a partitive genitive subject; in example (7) from Ancient Greek we find a
partitive genitive object:
Avestan (in Yašt 7.4; Dahl, this volume)

(6) uruuaranąm zairi.gaonanąm zaramaēm paiti zəmāδa
plants:GEN yellow.coloured:GEN spring:ACC again earth:ABL
uzuxšiieiti
grow.forth:PRS .3. SG
‘Yellow-colored plants grow forth again across the earth in the spring.’
Ancient Greek (Arist. HA 612a24)

(7) he dè khelōnē hótan ékheōs phágçi epesthíei
ART. NOM PTC turtle when snake: GEN eat: SBJV. PRS .3 SG eat: PRS .3 SG
tēn oríganon
ART. ACC oregano:ACC
‘In case they eat snake, turtles take oregano (as an antidote).’
Modern Russian takes a position in between the two groups, as it features a

second genitive, which is lexically restricted, and mainly functions as a partitive.
Note that with nouns that only have one genitive, the latter can also function as
a partitive, as is shown in examples (8)–(9):
Russian (adapted from Daniel, this volume)

(8) Ja že xotel vypit’ soku
1SG PTC want:PST. IMPV. SG drink:INF. PFV juice:GEN 2
‘I wanted to drink some juice.’
Russian (personal knowledge)

(9) Ja vypil vody
1 SG drink:PST. PFV. M . SG water:GEN
‘I drank (some) water.’
The noun sok ‘juice’ in (8) has two genitive forms: soka, the normal genitive of
masculine nouns, and soku, the second genitive (see Daniel, this volume). The
word voda ‘water’ in (9) has only one genitive, vody. The meaning expressed by
the two genitives in the examples is the same, that is, they indicate an indefinite
quantity: ‘some juice’, ‘some water’. As in the other Indo-European languages,
the partitive genitive can occur as an object or as a subject. The extent to which
the partitive genitive can function as subject varies from language to language:
in Russian, Avestan, Vedic Sanskrit, Ancient Greek and Lithuanian it only occurs
with intransitive (possibly unaccusative) verbs. In addition, only in Ancient
Greek do partitive subjects trigger number agreement with the verb, while in
the other languages they do not. In spite of some restrictions, however, it
remains remarkable that the partitive genitive is not connected with a specific
grammatical relation (subject or object). Its semantic component, partly involv-
ing indefiniteness, overrides grammatical relations.
In the languages of the second group, and to a large extent also in Russian,
the genitive case can be contextually understood as functioning as a partitive.
Thus, it not only shares the distribution of other case markers, but also has a
two-fold function: as a true genitive it indicates nominal modification, and
semantic roles of nominal modifiers (typically possessor), while as a partitive it
indicates various notions connected with partitives cross-linguistically, which
we discuss further in section 3.
2.2 Adpositions
Closely related to genuine case markers are adpositions. No languages among
those treated in this book have adpositions that function in the same way as the
Finnish partitive case, and can also indicate indefiniteness. A possible example
of a preposition which is in the process of becoming an indefinite marker, and
derives from a partitive construction, is Dutch van in so-called ‘faded partitives’.
Consider example (10):
Dutch (De Hoop 1998: 194)

(10) Er lagen van die dikke boeken
there lay:PRS .3 PL of DEM . PL thick: PL book: PL
op de tafel
on ART table
‘Some thick books lay on the table.’
According to Zwart (1987), the effect of van in this construction is to cancel the
meaning of the determiner die ‘those’, so that the phrase van die boeken func-
tions as a bare plural. De Hoop (1998) discusses some differences between
what she calls ‘faded’ partitives, i.e. partitives whose original meaning is ‘fading
away’ toward the meaning of indefiniteness, and bare plurals. As her conclu-
sions are most interesting if set in the framework of the diachronic development
of indefiniteness markers from partitive constructions, we will leave the issue for
the time being and return to it in section 4. Here, it is important to remark that
faded partitives described in the literature always co-occur with demonstrative
determiners. Thus, even in this case, one cannot say that the partitive-indefinite
meaning is expressed by the preposition alone.
The Romance languages feature so-called ‘partitive articles’, a combination
of the preposition which also expresses genitive meanings (French de, Italian di)
and the definite article. This type of lexical item has undergone complete trans-
categorization and must be considered an indefinite article synchronically, with
little connection left to the partitive meaning. However, the same complex of
preposition plus article can still have prepositional usage: for this reason we
start by giving some examples in this section, but will discuss them in the next.
Let us consider examples (11)–(12):
Italian (personal knowledge)

(11) La casa dei ragazzi
ART. F. SG home of+ART. M . PL boys
‘The boys’ home.’
(12) Mangio dei panini

eat:PRS .1 SG of+ART. M . PL sandwiches
‘I eat some sandwiches.’
In (11), the form dei, which consists of the preposition di and the plural mascu-
line form of the definite article il has a genitive function, as it indicates nominal
dependency: the noun ragazzi modifies the head la casa. Semantically, dei
ragazzi indicates a possessor NP. In (12), on the other hand, dei is an indefinite
plural article. In spite of formal identity, the distribution of indefinite ‘partitive’
articles is not the same as the distribution of primary prepositions, as we show
in the next section.
2.3 Articles
According to Budd (this volume) “some Oceanic languages have separate partitive
markers for singular and plural reference, for example Central Pacific languages
often have rich paradigms of articles and the Samoan forms se, ni, and sina
shown above are an example of this.” An example is given in (13):
Samoan (from Mosel & Hovdhaugen 1992: 265, quoted in Budd, this volume)
(13) ‘aumai sina wai
bring ART. PAR . SG water
‘Bring a little water.’
As mentioned in section 2.2, some Romance varieties, notably French and

Italian, feature so-called partitive articles, which are formed with the genitive
preposition plus the definite article. Diachronically, such articles can be shown
to have originated within partitive constructions (see Carlier & Lamiroy, this
volume, and below, section 4). For this reason, the name ‘partitive article’ is still
used, even though these articles have little left to do with partitivity. In addition,
in spite of being analyzable as containing a primary preposition, partitive arti-
cles do not share the distribution of primary prepositions. Primary prepositions
normally do not co-occur with each other in French and Italian: one cannot
combine di, a, da, and con in Italian for example. However, primary preposi-
tions co-occur with the partitive article, as in (14):
Italian (personal knowledge)

(14) Ho condito l’ insalata con
have:PRS .1 SG dress:PTCP. PST. SG ART. F. SG salad with
dell’ olio di oliva
of:ART. M . SG oil of olive
‘I dressed the salad with olive oil.’
Thus, transcategorization is complete in the case of the complex di (or de in

French) plus definite article, even though the same complex still also functions
as true preposition as shown in (11).
In Classical Greek, the partitive genitive may or may not co-occur with de-
finite articles. Limited to direct objects of consumption verbs, there seems to be
a semantic difference connected with this type of variation: the definite article
occurs in partitive constructions, while the indefinite interpretation of the parti-
tive genitive is possible when no definite article occurs, as shown by examples
(15)–(16) (see Napoli 2010):
Ancient Greek (Hdt. 1.73.6; 2.37.18–19)

(15) Kuaxárēs kaì hoi pareóntes daitumónes
Cyaxares:NOM and ART. NOM . PL be.present: PTCP. PRS . NOM . PL guest: NOM . PL
tôn kreôn toútōn epásanto
ART. GEN . PL flesh: GEN . PL DEM .GEN . PL eat: AOR . MID.3 PL
‘Cyaxares and the guests who were with him ate of that flesh.’
(16) ikhthúōn dè oú sphi éxesti pásasthai

fish:GEN . PL PTC NEG 3 PL . DAT may:PRS .3 SG eat:INF. AOR
‘They may not eat fish.’
Note however that, in the case of partitive genitive subjects, things apparently
work the other way around, as the indefinite reading seems to be connected
with the occurrence of the definite article, as shown in (17):
Ancient Greek (Arist. HA 513a.9)

(17) eisì dè kaì tôn perì phúsin hoi toiaútēn
be:3 SG PTC and ART.GEN . PL about nature:ACC ART. NOM . PL such:ACC
mèn ouk epragmateúthēsan akribologían perì tâs phlébas
PTC NEG labor:AOR .3 PL precision:ACC about ART. ACC . PL vein: ACC . PL
‘There are also scientists who have not investigated the veins with so
much accuracy.’
2.4 Verbal morphology

In the preceding sections, we have examined languages in which partitive coding
is related to nouns. There are also languages in which partitive meanings are
associated with the verb. An example is provided in (18) by Paamese, where
the verbal affix –tei appears in the verb for expressing a partitive meaning:
Paamese (from Crowley 1982: 145, quoted in Budd, this volume)

(18) Ma-ani-tei raise
eat:1SG . IMM . PAR rice
‘I’d like to eat some rice’
The Estonian modus obliquus provides us with another example, where the
partitive marker as such is attached to the verb, in this case for expressing hear-
say evidentiality:
Estonian (Tamm, this volume)

(19) a. Mari tule-b.
Mary come-3SG
‘Mary will come.’
b. Mari tule-va-t.
Mary come-PRS . PTCP- PAR
‘Allegedly/reportedly, Mary will come.’
c. Mari armasta-b Jaanus-t.

Mary love:PRS -3 SG Janus-PAR
‘Mary loves John.’
The marker attached to the verb in (19b) is formally identical to the partitive
marker that appears on nouns, as shown in (19c). This kind of extension seems
to be very rare cross-linguistically (Estonian is the only example we have come
across). We return to the possible semantic motivation for this extension below,
section 3.5.
Finally, partitive subjects can bring about special patterns of indexation. In
Balto-Finnic languages, such as Finnish and Estonian, verbal agreement is
different with nominative and partitive subjects. The verb agrees in person and
number only with nominative subjects, as the examples in (20) show:

(20) a. poja-t tule-vat
boy-NOM . PL come-3 PL
‘The boys are coming.’
b. poik-i-a tule-e(*tulevat)
boy-PL- PAR come-3 SG
‘(Some) boys are coming.’
In (20a), the verb appears in the third person plural form, because the subject is
also plural (in colloquial speech, the verb may also appear in the third person
singular form in (20a)). In (20b), in turn, the verb is in the third person singular
form, even though the subject is plural. The third person singular form can be
seen as the least marked or default form, because the verb appears in this
form, for example, also with weather verbs and in existential constructions. It
is in order to note that the lack of verbal agreement is not the primary way
of expressing partitiveness in Finnish, as it is related to existential subjects in
general. In example (21), an existential subject in the nominative plural occurs
with a singular verb form:

(21) Pöydä-llä on shakkinappula-t
table-ADE be:PRS .3 SG chessmen-NOM . PL
‘There are chessmen [a set of them] on the table.’
However, lack of verbal agreement is obligatory with partitive subjects.

A connection between lack of agreement and existential subjects is common
across languages. Among languages surveyed in this book, for example, evidence
is provided by Ancient Greek. As shown in Conti & Luraghi (this volume), in
Classical Greek partitive genitive subjects usually trigger agreement with the
verb, but existential predicates in the singular may occasionally co-occur with
partitive genitive plural subjects. Note, however, that this is not connected with
partitiveness, as lack of agreement in similar contexts can also occur with nomina-
tive subjects, in much the same way as in Finnish:
Ancient Greek (Xen. Hell. 7.5.17, in Conti & Luraghi, this volume)
(22) tôn dè polemíōn ên hoùs
ART. GEN . PL PTC enemy:GEN . PL be:IMPF.3 SG REL . ACC . PL
hupospóndous apédosan
under.truce:ACC . PL return: AOR .3 PL
‘And there were some of the enemies, who they returned under a truce.’
Ancient Greek (Hdt. 7.34.4, in Conti & Luraghi, this volume)

(23) ésti dè heptà stádioi ex Abúdou
be:PRS .3 SG PTC seven stadia:NOM . PL from Abydos:GEN
es tḕn apantíon
to ART. ACC opposite.shore:ACC
‘There are seven stadia between Abydos and the opposite shore.’
In (22) the subject tôn polemíōn ‘the enemies’ is in the partitive genitive, whereas
the subject of (23) heptà stádioi ‘seven stadia’ appears in the nominative. Both
NPs are plural, but in both sentences the verb (existential ‘be’) is in the third
person singular, thus showing that the trigger of non-agreement is existentiality,
rather than case.
3 Functional typology of partitives

In the previous section, formal means used to express partitive meanings were
briefly discussed. In this section, we proceed to examining the functions partitive
markers code across the languages treated in this volume.
3.1 Indefiniteness and non-referentiality

Partitive markers reviewed in this book all express indefiniteness, at least to
some extent. When expressing indefiniteness, the partitive marking a direct
object typically contrasts with the accusative2 (or the absolutive depending on
the basic alignment of the language), which appears if the reference is to a
well-defined and definite set of entities, or to a complete entity. A typical example
is found in (24):

(24) a. opettaja näk-i lapse-t
teacher:NOM see-3 SG . PST child- PL . NOM
‘The teacher saw the children.’
b. opettaja näk-i laps-i-a
teacher:NOM see-3 SG . PST child- PL- PAR
‘The teacher saw some children.’
Example (24b) can simply indicate indefiniteness and unspecificity: the teacher
saw some not previously identified children. Instead, a possible reading of (24a)
is that all of the children in a specific group (that may, for example, have been
the topic of conversation previously) have been seen (see also Larjavaara 1991:
377 for a similar remark). This reading can be emphasized by adding the quan-
tifier kaikki (‘all’, i.e. kaikki lapset ‘all the children-PL . NOM ’) to the sentence.
Besides the indefinite unspecific reading, example (24b) can refer to a situation
2 Note that the accusative and the nominative do not have separate endings in the plural. For
this reason, conventionally, only the label ‘nominative’ is used for the nominative/accusaitive
plural form in Finnish linguistics. We conform to this convention here, but it must be kept in
mind that the case that we gloss ‘nominative’ in the plural also has the function of the accusa-
tive with personal pronouns, and functionally the -t form in (24) is an accusative case.
where a specific group of children is also thought of, but in which only a part of
the children in that group have been spotted. Thus, it can be indefinite specific.
This reading becomes more evident if we add a sentence like mutta ei kaikkia
heistä (‘but not all of them’) to (24b).
Let us now consider example (25):

(25) äiti hake-e poja-t /poik-i-a
mother pick.up-3 SG . PRS boy- NOM . PL /boy- PL- PAR
‘The mother will pick up the boys / some (of the) boys.’
With the accusative, (25) means that the mother will pick up a specific group of
boys that has been established earlier and that is activated in the discourse.
With the partitive, in turn, the identity of the boys to be picked up is not speci-
fied and the reference is thus indefinite. The function of the partitively marked
object is therefore more to specify the class of its referent leaving the exact iden-
tity out. Indefiniteness is directly related to other partiality meanings reviewed
in the next sections. When the whole group of potential entities is not referred
to, it is no longer clear which members of the group have been mentioned, and
the reference is thus indefinite.
As briefly noted above, in (25) the class the object represents is more impor-
tant than its exact identity. This is even more pronounced in the case of mass
nouns, as in (26):

(26) Aino juo limsa-a, muttei vissy-ä
Aino drink: PRS .3 SG soft.drink- PAR but.not mineral.water- PAR
‘Aino drinks soft drinks, but not mineral water’
In (26), the function of the object is only to specify the class of the beverage
Aino drinks, or does not drink, but the reference is not to a definite set of soft
drinks or mineral water. The partitively coded object is therefore best seen as
non-referential. The use of the accusative would render the expression referen-
tial and definite in (26). This is true of mass and abstract nouns: a count noun
object can be accusative and indefinite, as in Ostin auton ‘I bought a car.ACC ’.
Note however that in this case we have a specific (and thus referential) indefinite,
rather than a non-referential expression as in (26). A non-referential reading may
be coined for the accusative in a sentence like (26) only with adverbials that make
the construction in question generic in nature.3 It is worth noting in this context
3 Remarkably, this may be more true of singular objects than of plural ones, where the NOM
can also have a non-specific reading: Sihteeri lajittelee kirjeet ‘The secretary sorts [the] letters’
= any letters arriving at any time, as opposed to Sihteeri lajittelee kirjeitä [PAR] ’the secretary is
sorting [the/some] letters’, but the PAR version can also have a reading similar to (26): ‘the
secretary’s job is to sort letters’, as plurals behave like mass nouns in many ways.
that a similar (yet functionally reverse) case is attested in Russian, where the
genitive emphasizes quantity, while the accusative denotes a class (see Paykin
& Van Peteghem 2002 and below, ex. (28)).
Indefinites are often held to be referential if they are specific, and non-
referential if they are non-specific. However, the possible readings of the Finnish
partitive support a more complex, three-fold distinction, whereby non-specific
indefinites are further divided into two different categories. In the first one, we
have indefinites that do not refer to a previously identified set, but to a certain
quantity, while in the second we have expressions that indicate a class, and not
a quantity. Compare English ‘I drink wine’ (non-ref.) vs. ‘I drank some wine’
(indef., but not necessarily specific, i.e. not necessarily belonging to a previ-
ously identified quantity). Similar cases attested in other languages, such as
Russian, give further support to the three-fold distinction outlined above. In
Russian the partitive genitive typically gives rise to indefinite readings that can
either refer to a specific quantity (from a previously identified whole), or to a
non-specific quantity, as shown in (27a, b) which does not refer to a previously
identified entity/set of entitities, as shown in (27a).
Russian (courtesy K. Paykin)

(27) a. ja uže poel tvoego supa
1 SG . NOM already eat:PST POSS .2 SG .GEN soup:GEN
‘I already partook of your soup.’
b. muz kupil piroga s malinoj

husband:NOM buy:PST cake:GEN with raspberry:INS
‘My husband bought some cake with raspberry.’
If Russian needs to stress the class, it uses the accusative case, which also
gives an indefinite non-specific reading, but does not stress quantity:
Russian (courtesy K. Paykin)

(28) ja ljublju pirog s malinoj
1 SG . NOM like:PRS .1 SG cake:ACC with raspberry:INS
‘I like raspberry cake”
Thus, the partitive genitive in Russian only covers a part of the meanings covered
by the Finnish partitive. Summarizing, we have:
Table 1: Case and (in)definiteness in Finnish and Russian4
INDEFINITE INDEFINITE INDEFINITE DEFINITE

SPECIFIC NON -SPECIFIC NON -SPECIFIC
( QUANTIT Y READING ) ( CLASS READING )
F INNISH ACCUSATIVE +
PARTITIVE + + +
R USSIAN ACCUSATIVE + +
GENITIVE + +
Indefiniteness is one of the meanings associated with partitives also in exis-

tential constructions. Two examples are provided in (29) and (30):
Estonian (Tamm, this volume)

(29) Taigna sees on pipar-t
batter.GEN in be: PRS .3 SG pepper- PAR
‘There is (some) pepper in the batter.’

(30) Bada atzerritar-rik Donostian!
yes.is foreigner:PAR Donostia.in
‘There are foreigners in Donostia!’
The functional link between indefiniteness and existential constructions is

easily accounted for. The function of existential constructions is to refer to the
existence of an entity at a specified location. The entity referred to is either intro-
duced by an existential construction or the class of the entity, its exact identity
is less important. For example, in (29), the fact that there is some pepper in the
batter is more important than the exact identity of the pepper; it is not relevant
whether the pepper has been mentioned previously or not. If the identity of the
given entity is known, it is typically the topic of conversation and we then have
a locative construction, where the given entity functions as subject or topic.
This difference is manifest also in English in constructions such as there is a
book on the table (existential construction) vs. the book is on the table (locative
construction). Definiteness (or rather the lack of it) is especially true of predicate
4 With the term ‘quantity’ in this table we refer to unbounded quantity. Note further that the
position of the accusative as only definite refers to mass nouns and to plurals. In the case
of singular count nouns, the accusative can have an indefinite specific reading, or even non-
specific indefinite, as in Liisa haluaa tavata miljonäärin ‘Liisa wants to meet a millionaire-ACC ’
(implaying that any millionaire will do); see Vilkuna 1992.
adjectives, which in Finnish copular constructions are marked with the partitive
if the subject is a mass noun or a plural, as in (31):

(31) a. Tee on musta-a.
tea:NOM be:PRS .3 SG black-PAR
‘The tea is black.’
b. Pöytä on musta.
table.NOM be:PRS .3 SG black.NOM
‘The table is black.’
c. Pöydä-t o-vat must-i-a.

table-PL be-3 PL black-PL- PAR
‘The tables are black.’
In (31a), nominative would be rather infelicitous, if it is possible at all, while in

(31b), the partitive would yield a rather ungrammatical construction. In (31c), in
turn, both are possible, even though partitive coding of the predicate adjective
seems more natural. Note that here the use of the partitive on the predicate
noun reflects a property of the subject (denoting a mass noun or an unbounded
quantity of countable entities), and obviously not of the predicate noun itself,
which, being an adjective, predicates a quality of the subject and is per se non-
referential.
3.2 ‘Part of’-meanings in partitive constructions

In some languages, partitive case markers are also used in partitive construc-
tions, or when indicating a part-whole relation. As discussed in the Introduction
and in section 1 of this chapter, this is partly the case for the Finnish partitive; in
addition, the partitive genitive of several Indo-European languages also occurs
in partitive constructions. Indeed, in various languages the usage in partitive
constructions seems to be the starting point from which partitive case markers
originated. We discuss this issue thoroughly in section 4.
However, the connection between partitive markers, partitive constructions,
and part-whole relations is not universal. In other languages, such as Basque
and Oceanic, partitives occur in such constructions to different extents. In
Basque, the partitive case does not occur either in pseudopartitive or in partitive
constructions. Consider examples (32)–(34):
Basque (courtesy B. Aritzimuño)

(32) Tarta zati bat
cake piece one
‘A piece of (a) cake.’
(33) Tarta horren zati bat

cake that:GEN piece one
‘A piece of that cake.’
(34) Tarta horretatik zati bat

cake that:ABL piece one
‘A piece of that cake.’
In (32), the interpretation is unspecific, hence non-referential: this is a pseudo-

partitive construction, similar to part-whole constructions with unspecific mean-
ing; in (33) the reading is specific, and again the construction is similar to part-
whole constructions with specific meaning, as in ibai-ertz (river-edge) ‘river
bank’ or in ibaiaren ertz (river:DET.GEN edge:DET ) ‘the bank of the river’. The latter
NPs are partitive constructions but, while the former features the genitive, the
second is formed with the ablative in -tik (archaic -rik), which is etymologically
connected with the partitive ending -(r)ik. Thus, the existence of some fossilized
expressions and old attestations of a partitive-like use of -(r)ik with some quan-
tifiers, e.g. eskerrik asko [thank(s)-PAR many] ‘thank you (very much), many
thanks’, lit. ‘a lot of thanks’, might indicate that this suffix has been used in
partitive constructions at an unattested stage of the language (Borja Aritzimuño,
p.c.).
In Oceanic languages, part-whole relations are expressed through the geni-
tive case. Examples are:
Bierebo (courtesy P. Budd)

(35) Galgalnaviniada
Scale.na.fish
‘Fish scale(s)’
Lewo (from Early 1994: 225)

(36) lepas-ne-u na marua
side-POSS -1SG . P GEN right
‘My right side.’
This is quite distinct from the use of the so-called partitive particle ta, which
indicates an indefinite quantity of a referent, as in (37):
Bierebo (courtesy P. Budd)

(37) Ne-saniada ta
1SG . S -eat.fish PAR
‘I’ll eat some fish.’
Apparently, partitive constructions are not easily available in Oceanic. In Araki,

for example, the partitive marker is re, from an earlier *te (see section 4 for
possible etymologies). According to Alex François (p.c.), “there is no simple
way to say ‘give me some of that cake’, at least none using a genitive construc-
tion of any sort. One would need a periphrasis such as ‘Cut that cake and give
me some.’ The first object would not be marked with re since it is [+def][+spec];
the second object would have re since it is [-spec]. The same would apply to
something like ‘Buy these balloons and give me one.’: ‘one’ would be expressed
by re, except that this /re/ is under-specified for number, so the sentence would
be ambiguous between give me one and give me some. It can be disambiguated
with a numeral following re.”
3.3 Numerals and quantifiers

Partitives accompany numerals in languages such as Finnish and Estonian.
Moreover, in Basque, they occur with certain quantifiers, such as ‘many’, and
in Finnish certain postpositions require partitive. Consider (38) and (39):

(38) kolme poika-a / sata euro-a / puolitoista vuot-ta
three boy- PAR / hundred Euro- PAR / one.and.a.half year- PAR
‘Three boys/a hundred Euros/one and a half years.’

(39) Zeresan-ik asko / gizon-ik aski / lagun-ik franko
gossip-PAR many / man- PAR enough / friend- PAR many
‘many gossiping / enough men / many friends.’
Similarly, in Russian the genitive occurs in the same contexts, as shown in (40):

(40) Pjat’ / mnogo jablok / sto / mnogo rublej
five many apple:GEN . PL hundred many rubl:GEN . PL
‘Five/many apples / a hundred/many rubles.’
The use of the partitive, or of the partitive genitive, with numerals is rather
easily accounted for, because numerals favor partitive constructions, as they
choose a certain number of entities from a whole. For example, ‘three boys’
means ‘three entities from the whole (universal) group of boys’. Moreover, it
seems that the use of the partitive with numerals follows quite naturally the
grammaticalization path of partitive from the ablative, i.e. ‘two boys’ (partitive)
originates from ‘two of the boys’ (ablative). Quantifiers are functionally close to
numerals, and it comes thus as no surprise that certain quantifiers also govern
the partitive.
3.4 Negation and interrogatives

The use of partitives or partitive genitives with negation is also well-known and
also widely studied (see Miestamo, this volume for a more detailed discussion
and references). Examples are given in (41)–(43):
Estonian (courtesy T. Huumo)

(41) Ta ei tap-nud mesilas-t
s/he NEG kill- PTCP bee- PAR
‘S/he did not kill the bee.’

(42) Maia-k ez du ardo-rik edan.
Maia- ERG NEG AUX wine- PAR drink
‘Maia has not drunk wine.’

(43) Maria ne pila vina
Maria:NOM NEG drink:PST wine: GEN
‘Maria didn’t drink (any) wine.’
Examples in (41)–(43) are negated sentences and their object thus appears in the
partitive (or genitive) case. The close relation of partitives and negation is rather
easily accounted for. The function of negation is, naturally, to state that the
event/state referred to did not occur. This makes the patient of negated clauses
indefinite, because the reference is not to a specific entity, but rather to any
entity that corresponds semantically to the direct object referent. In other words,
the reference of the object in (41)–(43) is to any bee or any wine, not to a specific
bee or portion of wine. The function of the object argument is thus to specify the
class of the entity not targeted by the denoted event. Its function is thus very
similar to that of the partitively coded object in (26).
Note that In Finnish and Estonian, the quantity is under the scope of nega-
tion and equals to zero. The Finnic (bare) partitive of negation cannot express
indefinite quantities outside the scope of negation such as ‘Some people did
not come’, as Vilkuna (1992) points out. Consider example (44):

(44) Viera-i-ta ei tul-lut
guest-PL- PAR NEG .3 SG come-PTCP
‘No guests came.’
Vilkuna points out that if the partitive vieraita were not under the scope of nega-
tion, then the reading would be ‘some guests did not come’, but this is not the
case. Thus even clause-initial negated partitive subjects are under the scope of
negation and therefore non-referential.
Moreover, partitives are attested in interrogatives in certain languages:
Basque (Exteberria, this volume)

(45) Goxoki-rik nahi al duzu?
candy-PAR want Q AUX
Estonian (Huumo & Lindström, this volume)

(46) a. Kas su-l telefoni on?
Q 2 SG -ADE telephone: PAR be:PRS .3 SG
‘Do you have a telephone?’
b. Kas Renate-t on seal?

Q Renate- PAR be:PRS .3 SG there
‘Is Renate there?’
The semantic connection between interrogatives and indefiniteness is also

rather straightforward, which explains the occurrence of cases such as (45) and
(46). Indefiniteness is very evident in (45) and (46a), where the reference is to
any piece of candy or any telephone. The class of the entity referred to is even
here more important than its exact identity. Example in (46b) is more intriguing,
because proper names are inherently definite, but nevertheless partitive appears
in (46b) as well. In Finnish, partitive would normally not appear in (46b), but it
is not ruled out here either. An additional reading is that we are looking for
someone named Renate, and anyone with that name will do. Notably, the use
of the partitive in interrogatives is more productive in Estonian than in Finnish
(see Huumo & Lindström, this volume for more details). This is, however, not
the context in which (46b) is used, which means that indefiniteness is not the
triggering factor of partitive use in (46b), even though its contribution to the
use of partitive is very important in the other cases.
3.5 Non-assertive modality

Closely related to the use of partitive cases with negations or interrogatives is
its possible occurrence with moods other than the indicative. Indeed, negative,
interrogative and non-declarative sentences share the common feature that they
are not asserted. An example of partitive case with non-assertive modality is
represented by the usage of the partitive in the protasis of conditional sentences
in Basque, as shown in the following example:
Basque (Ariztimuño, this volume)

(47) Inon zorte txarr-ik izan baldin ba-du, Interlagos
anywhere luck bad-PAR have COND if-AUX .3 SG .3 SG Interlagos
zirkuitu-an izan du
track-LOC have AUX .3 SG .3 SG
‘If he has had bad luck anywhere, he has had it in the Interlagos track.’
In Estonian, as mentioned in section 2.4 above, the partitive suffix can be

added to verbs to give an evidential meaning. As already noted, in Estonian
the partitive case also appears in questions, where the source of information is
also less direct than in declaratives, which may explain the use of the partitive
for coding hearsay evidentials. In both cases, we are dealing with something less
directly observable than with the accusative, for example.
Even in Finnish, partitives may have an irrealis flavor in some cases. For
example, a partitive sounds a bit more natural in the conditional mood, as in
the examples below:

(48) oli-si-ko sinu-lla tä-tä kirja-a?
have- COND - Q 2 SG -ADESS this- PAR book- PAR
‘Would you/do you happen to have this book?’
In (48), the speaker is looking for a very specific book, which means that the use
of the partitive is not determined by indefiniteness. The nominative (which is the
case of the possessum in Finnish possessive constructions) is also possible in
(48), but it seems a bit less probable than the partitive. The use of the partitive
adds a nuance of doubt to the question, i.e. the speaker does not have any kind
of evidence for the fact that the addressee would actually have the book s/he is
asking for. With the nominative, a similar expectation may be present. In the
indicative, for its part, the nominative is more felicitous, even though partitive
cannot be ruled out. In other words, there is a tendency for the partitive to be
favored in the irrealis (conditional) mood, but this is not grammatically re-
quired in any way. It is also important to bear in mind that this kind of mood-
conditioned variation between accusative and partitive is limited to interroga-
tives, it is not possible in declarative clauses.
3.6 Imperfective aspect

One of the most important functions of partitives related directly to verbal mean-
ings is illustrated by aspect, which is associated with partitives in Finnish and
Estonian. Also in this case, the partitive case is in variation with the accusative
case; the partitive expresses imperfective aspect (and/or unbounded activity), while
the accusative is associated with perfective aspect (and/or bounded activities).
Consider (49):

(49) a. Assistentti kirjoitt-i väitöskirja-n
Assistant:NOM write-PST.3 SG dissertation-ACC
‘The assistant wrote a dissertation.’
b. Assistentti kirjoitt-i väitöskirja-a

Assistant:NOM write- PST.3 SG dissertation- PAR
‘The assistant was writing a dissertation.’
Finnish lacks aspect as a verbal category, but with some transitive verbs allow-
ing an accusative object aspect can be expressed by case.5 The accusative case is
5 It is interesting to note that in Russian, the variation in the object coding is not sensitive to
aspect, as shown by Paykin (this volume). Aspect is marked primarily on the verb in Russian,
and nominal marking has no special relevance.
associated with perfective aspect and completed events; the reading of (49a)
is that the dissertation in question was finished. In (49b), for its part, aspect is
imperfective, which means that the activity in question was not finished.
A somewhat different example of an aspectual variation is illustrated in (50):
Estonian (Metslang, this volume)

(50) a. Ost-si-me oma tehnika / masina-d Austria-st.
buy- PST-1 PL our technology: GEN / machine- NOM . PL Austria- ELA
‘We bought our technology/machines from from Austria.’
b. Osta-me Austria-st tehnika-t / masina-id

buy- PRS .1 PL Austria-ELA technology- PAR . SG / machine- PAR . PL
‘We are buying technology/machines from Austria.’
In (50), both aspect and quantity are relevant to object coding. In (50a), aspect
and quantity are bounded (according to the terminology Metslang is using),
because of which the object appears in the genitive or in the nominative (the
latter is possible and also obligatory in the plural). In (50b), in turn, the aspect
and quantity are unbounded, and the object thus appears in the partitive. Differ-
ently from (49) the event in (50b) is not necessarily uncompleted, but it rather
refers to a habitual event (see also (51)). This habitual reading is possible also
in Finnish (see Huumo 2010 for a more detailed discussion); with the accusative
the sentence would get a ‘continuously all’ (Huumo 2010: 99) reading meaning
that we always buy our technology from Austria. An example of the partitive in
a habitual construction is found in (51):

(51) a. Matti maala-a talo-n
Matti paint- PRS .3 SG house-ACC
‘Matti is painting a/the house.’
b. Matti maala-a talo-ja

Matti paint-PRS .3 SG house-PAR . PL
‘Matti paints houses (e.g. as his profession).’
In (51a), Matti has finished painting a specific (definite or indefinite) house that
is known in the context, which makes the NP in question referential and moti-
vates the use of the accusative. Example (51b), in the reading given in the trans-
lation, does not refer to a concrete event, but to a habitually occurring event
that we are not witnessing as we speak. Notably, (51b) is also lower in transi-
tivity, as the direct object is non-referential (see Gerstner-Link 1998 for a more
detailed discussion of the lower transitivity of habituals). Non-referentiaity is

also relevant to the use of the partitive in this case.
3.7 Low transitivity

Aspect, discussed in the previous section, is also directly related to transitivity,
as has been shown, for example, in Hopper & Thompson (1980: 252). Perfec-
tive aspect (and thus successfully completed events) is related to high transitivity
and imperfective aspect (on-going events, or events that were not completed
successfully) to lower transitivity. Partitive coding is also in many other ways
associated with lower degrees of transitivity. Partitives, for example, appear with
many low transitivity-predicates in languages such as Finnish and Estonian, in
which case there is no aspectual alternation with the accusative, while the parti-
tive genitive occurs with certain low transitivity verbs in Ancient Greek and
Indo-Aryan languages. In this section, the use of partitives as markers of lower
transitivity is discussed in more detail. The discussed cases include partial/lower
affectedness, the already noted use of partitives with many low transitivity-
predicates, and lower degrees of agency.
3.7.1 Partial/lower affectedness
Partial affectedness means in the present context that only a part of an entity is
affected, as opposed to the whole entity. The affected part can be thoroughly
affected, but affectedness concerns only this part of the entity in question. Ex-
amples are (8) and (9) from Russian, (15), (52b) from Vedic, (53) from Finnish
and (54a) from Ancient Greek:
Vedic (RV I 104.9, in Dahl, this volume)

(52) a. ápaḥ sómam. ástam indra prá yāhi
drink: AOR .2 SG soma:ACC home:ACC Indra:VOC go:IMP.2 SG
‘You have drunk the soma. Go home, Indra!’ (RV III 53.6)
b. arvāṅ éhi sómakamaṃ tvā āhur

hither come:IMP.2 SG soma.lover:ACC 2 SG . ACC say:PRF.3 PL
ayáṃ sutás tásya pibā mádāya
this:NOM juice:NOM it:GEN drink:IMP.2 SG inebriation:DAT
‘Come hither! They say you are a lover of soma. This is the juice.
Drink of it to inebriation.’

(53) a. Lapsi sö-i kaku-n / kakku-a
child: NOM eat-PST.3 SG cake-ACC / cake- PAR
‘The child ate the cake/ some cake.’
b. Opettaja maala-si talo-n / talo-a

teacher:NOM paint- PST.3 SG house-ACC / house- PAR
‘The teacher painted the house/a part of the house.’
Ancient Greek (Thuc. 1.30.2; 5.31.3, in Conti & Luraghi, this volume)
(54) a. pleúsantes es Leukáda tḕn Korinthíōn
sail:PTCP. PRS . NOM . PL to Leucas: ACC ART. ACC Corinthian: GEN . PL
apoikían tês gês étemon
colony: ACC ART.GEN land:GEN ravage:AOR .3 PL
‘Sailing to Leucas, the colony of the Corinthians, they ravaged part
of the country.’;
b. hoi Ēleîoi . . . Lepreatōn tḕn gên étemon

ART. NOM . PL Elean: NOM . PL Lepreate: GEN . PL ART. ACC land: ACC ravage: AOR .3 PL
‘The Eleans ravaged the land of the Lepreates.’
The examples (52) and in (53a) cointain verbs of consumption. In (53a), the use
of the accusative indicates that the whole cake was eaten, while the partitive is
used when only a piece was consumed. In (52a), reference is made to a bounded
quantitiy of soma, while in (52b) it is said that an unspecified quantity of soma
must be drunk up to the achievement of a certain state (a verb of consumption
also occurs in (15)). In (53b), the accusative is used when the whole house has
been painted, while the partitive appears when only a part of the house has
been affected by the painting event. In both cases, the affected part has been
thoroughly affected (i.e. the given piece of cake has been fully consumed and
the part of the house in question has been fully painted), but the difference
to total affectedness follows from the fact that the whole entity has not been
targeted and the rest remains unaffected. Similarly, in (54a) it is said that only
a part of the country has been ravaged, while the accusative in (54b) indicates
that the whole referent has been affected.
Partial affectedness is especially connected with direct objects that are incre-
mental themes. An example is provided in (55):

(55) Vahtimestari lämmitt-i luentosali-n / luentosali-a
Janitor:NOM heat- PST.3 SG lecture.hall-ACC / lecture.hall- PAR
‘The janitor heated the lecture room completely/lecture room somewhat.’
Example in (55) illustrates a rather clear instance where an entity as a whole is

affected, but the overall degree of affectedness varies according to which of the
two cases is used. The accusative occurs if the lecture hall has been completely
heated and the temperature has reached the desired level, while the use of the
partitive means that the lecture hall has become somewhat warmer than it was,
but the temperature aimed at has not yet been reached. In other words, the
lecture hall undergoes a more radical change-of-state in the former case. The
reading that only a part of the lecture hall has been heated is perhaps possible,
but it is far less likely than in the version on (55) that contains the partitive.
The partitive may also have the reading that the agent failed to complete the
event s/he attempted. Examples from Ancient Greek and from Finnish are given in
(56)–(58) (see also Dahl, this volume, for similar cases in Indo-Aryan languages).
Ancient Greek (Hom. Il. 23.805; Il. 6.466, in Conti & Luraghi, this volume)
(56) hoppóterós ke phthêisin orexámenos khróa kalón
INDF. NOM . SG PTC overtake: SBJV. AOR .3 SG reach: PTCP. AOR . NOM flesh: ACC fair: ACC
‘Which of the two will first reach the other’s fair flesh.’
(57) hṑs eipṑn hou paidòs oréxato phaídimos Héktōr

so say:AOR .3 SG 3 SG .GEN child:GEN reach: AOR .3 SG glorious: NOM Hector: NOM
‘So saying, glorious Hector tried to reach his boy.’

(58) Mestästäjä ampu-i linnu-n / lintu-a
hunter: NOM shoot-3 SG . PST bird-ACC / bird-PAR
‘The hunter shot a bird/at a bird.’
In (56) and (57) the verb orégein ‘reach’ occurs once with the accusative and
once with the genitive. While in (56) the accusative indicates that the fighter
will try to actually reach each other with their weapons, in (57) the genitive
indicates that the father only stretched out his arms to reach his boy, but was
not able to actually reach him (the following context says that the boy moved
back scared by the father’s weapons). In (58), the accusative codes, expectedly,
an event where the hunter successfully shoots (and hits) the bird thus causing
its death. With the partitive, the construction has two (likely) readings. The first
of these is a lower affectedness reading very similar to that in (58); the hunter
shot a bird, but only wounded it, i.e. the degree of affectedness is lower than if
the bird had been killed. The other possible reading, relevant in this context, is
that the hunter tried to shoot the bird, but missed. In this case, the hunter fails
to complete the event s/he was aiming at, because the bird has not been hit and
killed. The use of the accusative cannot yield this reading. Similarly, in Indo-
Aryan languages, the use of the genitive as an object marker with a partitive
meaning is related to underspecification of the change of state feature, and the
use of the partitive with subject may yield a conative-like reading meaning that
the subject referent is attempting to perform an action (Dahl, this volume).
3.7.2 Use with low transitivity predicates
Thus far, cases have been examined in which the partitive has been related (pri-
marily) to a low degree of affectedness. In such cases, the partitive varies with
the accusative and its function is thus to underline a decrease in affectedness
that would otherwise not be inferable. In this section, we examine cases in
which the partitive appears with predicates ranking inherently lower for transi-
tivity. What is important in the present context is that all these verbs are low
transitivity predicates, which do not indicate a change of state.
In Finnish, for example, typical examples of verbs taking partitive objects
include mental verbs or verbs of cognition and experience, as in (59):

(59) a. Lapsi rakasta-a äiti-ä-än
child: NOM love- PRS .3 SG mother- PAR- POSS .3 SG
‘The child loves his/her mother.’
b. Henkilö ajattele-e kesä-ä

person: NOM think- PRS .3 SG summer- PAR
‘A person is thinking about the summer.’
Verbs rakastaa (‘love’) and ajatella (‘think’) normally govern the partitive in
Finnish, the only exception being resultative constructions (see the discussion
of ex. (11) in the Introduction). In this respect, they differ from many of the other
verbs examined thus far which allow variation between partitive and accusative.
The obligatory partitive coding may be claimed to follow from the inherently low
(semantic) transitivity associated with the verbs, for example in the spirit of
Hopper & Thompson (1980). Verbs governing the partitive only denote events
that do not involve a canonical agent or a saliently affected patient, which

explains their infelicity with the accusative that can be seen as a marker of
high transitivity. Similar examples are attested also in Estonian (see Metslang,
this volume) and Indo-Iranian (see Dahl, this volume).
Verbs of emotion and perception frequently take the partitive genitive in
Ancient Greek as well. The verb ‘love’ érasthai, for example, always takes a
genitive stimulus, as shown in (60):
Ancient Greek (Hom.Il. 16.182)

(60) tês dè kratùs argeïphóntēs ērásat(o)
DEM .GEN . SG . F PTC mighty:NOM slayer.of.Argus:NOM love:AOR .3 SG
‘Mighty Hermes loved her.’
With some other low transitivity verbs, the genitive can alternate with the accu-
sative in Ancient Greek, with little semantic difference. This is the case with
akoúein ‘hear’ in (61) (see further Conti & Luraghi, this volume):
Ancient Greek (Hom. Od. 12.265–266)

(61) mukēthmoû t’ ēkousa boôn aulizomenáōn
lowing:GEN PTC hear:AOR .1 SG cow:GEN . PL lodge:PTCP. PRS .GEN . PL
oiôn te blēkhēn
sheep:GEN . PL PTC bleating:ACC
‘I heard the lowing of the cattle lying (in the courtyard) and the bleating
of the sheep.’
3.7.3 Lower degree of agency
In 3.7.1 and 3.7.2, cases were discussed in which the use of partitives is somehow
related to a lower degree of affectedness. The mirror image of this is attested in
languages in which partitive cases may also mark a decrease in agency. An
example is (62b):

(62) a. Aino laula-a
Aino sing- PRS .3 SG
‘Aino is singing.’
b. Aino-a laula-tta-a
Aino- PAR sing- CAUS - PRS .3 SG
‘Aino feels like singing.’
Example (62a) is a normal intransitive construction, where the subject referent

performs an act of singing willfully and on purpose. In (62b), in turn the only
participant available feels an urge to sing, but is not necessarily singing, when
(62b) is uttered. The degree of agency associated with the agent participant is
thus lower than in (62a).
A lower degree of agency is also reflected in what Kempf (2007: 98ff) has
labeled “the partitivus of a weakened activity”. Tabakowska (this volume) dis-
cusses the alternation between the accusative and the partitive genitive in Polish
in cases in which no partitivity seems to be involeved. Compare the following
examples:
Polish (Tabakowska, this volume)

(63) a. Panie, wysłuchaj nasze modlitwy
Lord:VOC hear:IMP.2 SG our:ACC . PL prayers:ACC . PL
b. Panie, wysłuchaj naszych modlitw

Lord:VOC hear:IMP.2 SG our:GEN . PL prayers:GEN . PL
‘Lord, hear our prayers.’
The meaning of the two sentences is the same, but the degree of agency differs.
As Tabakowska remarks, “while the wysłuchać + Nominal:GEN construction im-
plies an interpretation in terms of low involvement of the agent in an event, the
wysłuchać + N: ACC evokes the “holistic” interpretation, involving full engage-
ment in an activity or/and its all-embracing character.”
Partitive or genitive NPs can also function as non-canonical subjects of
so-called impersonal constructions in some languages. In Finnish, the partitive
NP in such constructions indicates the experiencer, as in (64):

(64) a. minu-a hävettä-ä
1SG - PAR feel.ashamed- PRS .3 SG
‘I feel ashamed.’
b. minu-a janotta-a
1SG - PAR be.thirsty- PRS .3 SG
‘I am thirsty.’
Examples in (64) are rather similar to those in (62). However, the relevant differ-
ence between (62) and (64) is that examples in (64) lack an agentive counter-
part, and their predicates refer to inherently uncontrollable events (or rather
states). Examples in (64) can thus be seen as instances of impersonal construc-

tions, and their only participant appears obligatorily in the partitive.
In Ancient Greek, some impersonal constructions feature dative experiencers
and nominative or genitive stimuli. Genitive stimuli share some features of sub-
jects, and can qualify as non-canonical subjects. An example is given in (65):
Ancient Greek (Hom.Od. 9.19–20, in Conti & Luraghi, this volume)

(65) eím’ Oduseùs Laertiádēs, hòs . . .
be:PRS .1 SG Odysseus:NOM . SG son.of.Laertes:NOM . SG REL . NOM . SG
ánthrṓpoisi mélō . . .
man:DAT. PL care:PRS .1 SG
‘I am Odysseus, son of Laertes, of great interest to men.’
3.8 Other uses

So far, the most typical functions that partitive cases have across languages have
been illustrated and discussed. We end the functional typology by discussing
some other, sporadic uses that are harder to classify under one heading than
the functions examined thus far. The examined functions include the use of par-
titives with certain postpositions, and the use of partitives in adverbial func-
tions.
In Finnish, the partitive and the genitive are the cases typically governed
by adpositions. The use of the partitive becomes very understandable especially
if we consider the variation between genitive and partitive with so-called bi-
positions (illustrated in (66b)) that can function both as prepositions and post-
positions. Aspect (among other things) is also relevant to this variation. In (66b),
the partitive appears in the prepositional construction, whose meaning is ‘all over
the city/around the city’. The combination genitive + ympäri, in turn, appears in
sentences/constructions coding boundedness, such as ‘he ran around the city’
(he circled the city). In other words, the partitive is also associated with un-
boundedness here.

(66) a. minu-a vastaan / Turku-a kohti
1 SG - PAR against / Turku- PAR towards
‘Against me/towards Turku.’
b. ympäri kaupunki-a / kaupungi-n ympäri

around city- PAR / city- GEN around
‘Around in the city, all over the city/around the city.’
In Homeric Greek, the partitive genitive also occurs with prepositions. It

indicates a single path, rather than multiple path, as indicated by the accusa-
tive. Examples are:
Ancient Greek (Hom. Il. 13.202–204; Il. 17.283, in Conti & Luraghi, this volume)
(67) kephalēn d’ hapalês apò deirês kópsen
head:ACC PTC tender:GEN from neck:GEN cut:AOR .3 SG
Oïliádēs . . . hêke dé min sphairēdòn
son.of.O.:NOM throw:AOR .3 SGPTC 3 SG . ACC like.a.ball
helixámenos di’ homílou
roll:PTCP. AOR . MID. NOM through crowd:GEN
‘The son of Oïleus cut the head from the tender neck, and with a swing
he sent it rolling through the throng like a ball.’
(68) helixámenos dià bēssas

turn:PTCP. AOR . MID. NOM through glen:ACC . PL
‘(A wild boar) turning around through the glens.’
As noted in Conti & Luraghi (this volume) “in (67) and (68) the same verb form,
helixámenos, indicates two different types of motion. The head of the champion
in (67), cut off from his neck, rolls on itself along a straight trajectory inside an
area defined by the crowd: here, the genitive landmark is a surface which can be
divided into parts. Hence, the trajectory can be traced down. The wild boar in
(68), instead, runs around in different directions among the glens. The accusa-
tive landmark does not allow for precise tracking of the trajector, and movement
is performed at random.”
Finally, partitive cases have a number of adverbial functions across languages
and they are used for coding, for example, causal and spatial meanings. Exam-
ples from Finnish are given in (69) (see also Ariztimuño, this volume, for similar
cases in Basque and Conti & Luraghi, this volume, for Ancient Greek):

(69) a. hän tek-i se-n ilkeyt-tä-än
s/he do- PST.3 SG it-ACC nastiness- PAR- POSS .3 SG
‘S/he did it out of nastiness.’
b. hän on suku-a minu-lle

s/he be: PRS .3 SG family- PAR 1 SG -ALL
‘S/he is of my family.’
c. Ville tule-e koto-a

Ville come- PRS .3 SG home- PAR
‘Ville comes from home.’
In (69), the partitive case expresses different kinds of adverbial functions, such
as causal (a) and spatial meanings (c), and they may also appear in so-called
inclusion clauses functionally resembling predicate nominals. All of the exam-
ples in (69) manifest the spatial origins (a separative case) of the Finnish parti-
tive rather well. The spatial origin is most evident in (69c), where the partitive is
still used in a separative function; kotoa is one of the lexicalized uses of the old
partitive meaning still present in the modern language. In (69a), the original
semantics is also rather manifest, because the denoted event may be said to
originate from nastiness, and the ablative (or a similar case/adposition) could
be used to express this function in languages without partitive. Finally, in
(69b), the referent of the subject comes from my family, which also underlines
the spatial origin of the partitive.
In Basque, the adverbial usage of partitives possibly goes back to its origin
as an ablative. It is still preserved in a small set of adverbs, listed in (70):
Basque (Aritzimuño, this volume)

(70) alde-rik alde: ‘from one side to the other, right through’
esku-rik esku: ‘from hand to hand, hand in hand’
herri-rik herri: ‘from town to town’
egun-ik egun: ‘day by day, from day to day’
In Ancient Greek, the partitive genitive occurs in some time adverbials, such as
nuktós (night:GEN ) ‘at night’, and, limited to Homer, in space adverbial as well,
in which it indicates locative (see further Conti & Luraghi, this volume).
3.9 The partitive case as a radial category

In this section, we aim at describing the meaning of partitive morphemes as a
radial category (see Lakoff 1987, Nikiforidou 1991), and map more and less cen-
tral meanings onto semantic space, in order to come up with a representation
that shows the relations among the different functions of partitives and their
position relative to each other. The likelihood of such representation will be
tested with the diachronic data discussed in part (b). A central area in the
semantics of partitives is shared by partitive constructions in the sense of
Koptjevskja-Tamm (2006; see also Tamm, this volume); the representation pro-
vided for the description of the functions of partitives will be shown to account
for this polysemy.
Figure 1: The meanings of partitive cases
4 The origin of partitives and their possible

evolution
So far, we have discussed partitives across languages from a rather purely syn-
chronic perspective. In this section, we will take a look at origin of partitives and
their possible grammaticalization paths. In sections 4.1 and 4.2, we will focus on
the formal development of partitives, while in 4.3 the focus will be the func-
tional changes necessary for the emergence of partitives. Formal and functional
developments cannot be completely separated from each other, because formal
changes (apart from pure sound changes) are typically functionally and/or
semantically motivated. However, the two aspects are examined separately in
this section for making a more detailed discussion possible.
4.1 From ablative or genitive to partitive

Partitives are often seen as originating from ablatives (or separative cases in
more general terms) or genitives (see Heine & Kuteva 2004: 32–33). This path is
rather manifest, for example, for Finnic languages and Basque. The Finnic parti-
tive originates from the Finnic-Mordvinian separative (‘away from’) case, which
was used as a rudimentary partial object, but also as a kind of partitive attribute.
(The separative case as such may be older, and ultimately go back to Proto-Uralic,
but its grammaticalization can be seen in the Finnic-Saami-Mordvin branches.)
The separative (and in more general terms, spatial) origins of the partitive are
very visible in many of its uses in the modern language, as the examples dis-
cussed in this paper have shown. The most evident of these is the ‘part-of’-
meaning that is a clear case of separation. The other uses of the partitive are
also motivated by this function of the partitive (Larjavaara 1991: 401–2).
Similar to Finnic, the partitive case ending also originated from a former
ablative in Basque. As shown in Aritzimuño (this volume), the partitive is an
allomorph of the present ablative: the two case forms became differentiated at
a pre-literary stage when, according to Aritzimuño, the features of number and
definiteness in spatial cases had not yet emerged. At a later stage, the partitive
remained indefinite, while the ablative acquired definiteness and singular number.
This development is partly different from Finnic: in Finnic, the older ablative
case lost its local function, while in Basque two distinct cases were created.
Notably, in neither language we find simply the extension to a new meaning
(partitive) that accompanies the old one (ablative): no ablative-partitive poly-
semy arises either in Finnic or in Basque. In the former case, an ablative case
marker loses the local function and becomes a partitive, while in the latter
allomorphs are exploited to convey the two different meanings, thus becoming
different case markers.
Another interesting development, partly similar to Basque, is found in Russian.
In this language, a number of second declension nouns feature the so-called
second genitive, which is partly used in partitive contexts (see Daniel, this
volume). The morpheme involved was in origin the genitive ending of the -u
declesion, which later merged with consonant stems (second declension). Some
of the former -u stems preserved the older genitve, while also acquiring the new
one; in addition, the second genitive ending spread to some other nouns that
were not older -u stems. Remarkably, the second genitive can only partly be viewed
as a dedicated partitive: as shown in Daniel (this volume), it also appears out-
side partitive contexts. Moreover, nouns that do not have the second genitive
feature the regular genitive in partitive contexts. Comparative data further show
that the genitive already functioned as a partitive before the second genitive was
created (indeed, this function of the genitive was inherited from Proto-indo-
European). Thus, as we discuss in more detail below, the development in Russian
and other Indo-European languages is different from the developments illustrated
above for Finnic and Basque, in that partitive is one of the meanings of the
genitive case, and genitive cases generally speaking allow synchronic polysemy,
contrary to ablative cases, as we have shown above.
From the point of view of the morphology, the Russian development can be
seen as an instance of exaptation (see Lass 1990). The -u stem genitive, which
had lost its function following the disappearance of the inflectional class to
which it belonged, was so to speak re-cycled, and acquired a new function, for
which there was previously no dedicated morpheme.
The extension of genitive cases to partitive meanings is typical of the Indo-
European languages. In some of them, one might think of a development parallel
to Basque and Finnic, and based on an earlier ablative, as ablatives often also
constitute the source for genitives. However, this does not seem to be the case:
indeed, in Indo-European languages in which the genitive and the ablative have
different endings (e.g. Latin and Indo-Aryan), the partitive function is typical of
the genitive, rather than of the ablative.
Remarkably, there is a fundamental difference between Balto-Finnic and
Basque on the one hand, and Indo-European languages (including Romance)
on the other in the development of the partitive. Balto-Finnic languages and
Basque have an independent partitive, which historically derives from an abla-
tive, but no longer functions as an ablative (apart from some lexicalized uses,
such as (49c), and a genitive, which is formally and functionally distinct from
the partitive. Indo-European languages, for their part, have a genitive which
also functions as a partitive, and moreover some of these languages have an
independent ablative distinct from the partitive genitive. This is illustrated below:
Type a. Balto-Finnic, Basque: partitive ≠ genitive ≠ ablative

Type b. Indo-Aryan, Latin, Italian: partitive = genitive ≠ ablative
Type c. Ancient Greek, French: partitive = genitive = ablative (the original
situation was as type b)).
The above list seems to indicate that, even though partitives can historically
originate from ablatives, synchronically they tend to be distinct, unless both the
ablative and the partitive merge with the genitive. Indeed, the type ablative =
partitive ≠ genitive is not attested in any of the languages we have data for.
This could also mean that there are two separate sources for partitives, namely
ablatives and genitives (see also Moravcsik 1978, Heine & Kuteva 2004). In the
first case, the two meanings appear to be incompatible, if they do not also
include genitive. In the second case, the two meanings tend to co-exist.
In this respect, the development from Latin to Romance is of particular
interest. Among the Romance languages, French and Italian feature so-called
partitive articles, which originated from Late Latin partitive construction (see
section 4.3.4). The development shows interesting differences in the two lan-
guages. Latin had a genitive case that expressed both partitive and genitive
meanings. It also had a separate ablative case, which indicated source especially
with prepositions. The most used source preposition with the ablative case in
Classical Latin was ab. In Late Latin, de took over as the marker of source with
the ablative case. Later still, cases were lost, and de developed into a marker of
genitive as well, while retaining the function of marking ablative. In Proto-
Romance, a prepositional phrase with de expressed the genitive and the abla-
tive; it also occurred in patitive constructions (see the discussion of examples
(73)–(74) below). French has retained the Proto-Romance system, while Italian
features two different prepositions: di (from de) is used for the genitive and
also occurs in the partitive article, while da (from de ab) is used for the ablative.
What seems to have happened when Late Latin started losing its case system
and the genitive was replaced by a prepositional phrase is an extension of the
ablative preposition de to all functions of the genitive, not only a separate
change from ablative to partitive. Then, when the two functions, genitive and
ablative, became distinct again in Italian, the partitive meaning remained with
the genitive, as it was in origin in Classical Latin (see Carlier & Lamiroy, this
volume).
4.2 Oceanic developments

The diachronic paths for the creation of partitive morphemes in Oceanic is not
as easy to indicate as those described in section 4.1. The most likely reconstruc-
tion connects the partitive marker with the ending that marked the undergoer of
intransitive verbs in Proto-Malayo-Polynesian. According to Malcom Ross (p.c.),
all these languages showed ergative alignment. The S and the O appeared in the
absolutive, and the A in the genitive. The O of a regular ergative clause could
only be definite. However, antipassive intransitive verbs indicating activity (as
in ‘I am eating’) could also take an undergoer argument. The latter was in-
definite (non-referential), and was marked as an oblique. In various Oceanic lan-
guages, the partitive morpheme apparently descended from the Proto-Oceanic
*ta, which comes from the Proto-Malayo-Polynesian *ta, used for marking such
indefinite undergoers. This ending has often been labelled accusative. As these
languages have a separate genitive ending (the same ending of the ergative
case, in those that have preserved ergative alignment), and because the genitive
ending occurs in part-whole constructions (see section 3.2),one can have the
puzzling impression that the opposition between the genitive and the accusative
works contrarywise in Oceanic. However, once the origin of the morpheme *ta is
taken into account, the picture becomes clearer: this was not the marker of the
definite, referential direct object, as the accusative is (at least in one of its possi-
ble functions) in the languages we discussed in the previous section. Concerning
the more remote origin of *ta, it apparently goes back to some local oblique, but
not necessarily an ablative. Indeed, some Polynesian languages provide evidence
for the locative preposition i in partitive constructions, as shown in (71)–(72):
Tongan (from Clark 1973: 600)

(71) Na’e kai ’a e ika ’e he tamasi’i.
PST eat ABS REF fish ERG ref boy
‘The boy ate (up) a fish.’
(72) Na’e kai ’a e tamasi’i ’i he ika.

PST eat ABS REF boy in REF fish
‘The boy ate some of a fish.’
Another, completely different grammaticalization path attested in some

Polynesian languages is the numeral ‘one’ (Proto-Malayo-Polynesian *esa) that
may develop into an indefinite marker and further into a partitive morpheme.
The relation between the numeral ‘one’ and indefiniteness is also known from
other languages, so this grammaticalization path is also rather understandable.
In addition, even motion verbs in serial verb languages can develop into
partitive markers: due to the documented connection between partitives and
ablatives, this development is not surprising. Indeed, the Basque partitive might
ultimately go back to the verb *din (Modern Basque jin), ‘come from’, as arguend
in Aritzimuño (this volume). Note that following this reconstruction, the connec-
tion between the verb and the partitive case is not direct, as the verb root gave
origin to the ablative case marker. Only at a later time, an allomorph of this case
ending was reanalyzed as the partitive case ending.
A direct connection between a verb and a partitive marker might be attested
in Oceanic. According to Budd (this volume) a possible explanation for the pres-
ence of post-verbal partitive forms in some Vanuatu languages connects partitive
markers with verbs: “Bierebo ja seems likely to be an accreted form of gicha, a
variant form which occurs in some more conservative dialects of the language.
In turn, gicha can be diachronically analysed as gi-cha, where gi is a serial verb
meaning ‘take time’ which is still productive in the contemporary language and
cha is perhaps a phonologically-conditioned variant of ta. What seems likely
here then, is that a modified serial verb with a literal meaning of ‘take a little
time’ has developed into a more general VP partitive meaning ‘do a little’. It
remains to be seen whether this analysis would be valid for other Vanuatu lan-
guages with VP partitives.”
4.3 Semantic/functional development
4.3.1 Cases as prototypical categories
In section 4.1 we have highlighted the affinity of partitive cases with genitives
and ablatives. Diachronically, the development from a genitive or an ablative to
a partitive case starts within partitive constructions, as we will show in section
4.4. In this setion, we briefly discuss the semantic motivation that links geni-
tives, ablatives and partitives to one another.
As is well known, case markers are best described as polysemous cate-
gories, which express meanings that are interconnected in a complex manner.
A very suitable way to concieve of case meaning is in terms of protypical cate-
gories, in which each meaning is directly connected only with a small set of
other meanings. The meaning of the genitive has been described in such a
manner by Nikiforidou (1991), who refers to it as an instance of structured poly-
semy. As basic for the category, Nikiforidou indicates the possessor meaning.
The partitive meaning is readily derived from it by the metaphor ‘Wholes are
possessors of their parts’, which operates pervasively in languages. Nikiforidou
further explains the cross-linguistically comparatively frequent polysemy of
genitive and ablative through a metaphor by which ‘Wholes are origins.’ This
metaphor connects the source (ablative) meaning directly with the part-whole
meaning. Thus, semantic extension precedes as follows: possessor/possession
! whole/parts ! originating element/origin. Note however that this is not the
direction of semantic extension in diachrony. In the diachronic survey in 4.1, we
found extensive evidence for the extension of ablatives to partitives, without
an intervening genitive (i.e. possessive) meaning. The diachronic development
actually attested procedes in the contrary direction: ablative (source/origin) !
partitive (whole/part relation). In addition, in the case that a genitive derives
from an ablative, the partitive meaning may constitute the bridging link between
the two.
In section 4.3 we also described a less frequent spatial source for partitives,
locatives, found in some Polynesian languages. The semantic motivation for this
extension can be conceived as based on a variant of the container metaphor,
such as ‘Wholes are containers, and parts are entities contained.’ Based on this
metaphor, a locative indicates that only a part of what is in a container is affected
by the relevant event.
Summarizing, the follwing metaphors can be singled out as responsible for
the extension of other case markers to partitive:
a. Genitive to partitive ! Wholes are possessors, Parts are entities possessed

b. Ablative to partitive ! Wholes are origins, Parts originate from wholes
c. Locative to partitive ! Wholes are containers, Parts are entities contained
The part-whole relation appears to be relevant to all the above metaphors. In the
next section, we describe an attested case of the historical change that led from
the expression of a part-whole relation to a partitive case marker and later to an
indefinite article.
4.4.2 Late Latin partitive constructions
The construction out of which partitives originate occurs in Late Latin, as shown
in the following examples:6
Latin (New Testament)

(73) et misit ad agricolas in tempore servum ut
and send:PRF.3 SG to peasant: ACC . PL in time: ABL servant: ACC for
ab agricolis acciperet de fructu vineae
from peasant: ABL . PL collect:SBJV. IMPF.3 SG from fruit: ABL vineyard:GEN
‘At harvest time he sent a servant to the tenants to collect from them
some of the fruit of the vineyard.’ (Mark 12,2)
(74) dicit eis Iesus adferte de

say:PRS .3 SG 3 PL . DAT. Jesus: NOM bring: IMP. PRS .3 PL from
piscibus quos prendidistis nunc
fish: ABL . PL REL . ACC . PL catch:PRF.2 PL now
‘Jesus said to them, “Bring some of the fish you have just caught!”’
(John 21,10)
(75) et ipse in nobis quoniam de Spiritu suo

and 3 SG . NOM in 1 PL . ABL because from spirit: ABL POSS .3 SG . ABL
dedit nobis
give: PRF.3 SG 1 PL . DAT
‘[We know that we live in him] and he in us, because he has given us of
his Spirit.’ (John 4,13)
6 The Greek Bible, which has been the main source for the Latin translation, also contains a few
genitive subjects; they are sometimes regarded as due to Hebrew influence, even though Classi-
cal Greek did have genitive (partitive) subjects, as we will see below (see also Conti 2010 fn. 3
and Luraghi 2013 for discussion). They are variously translated into Latin with the nominative or
with a prepositional phrase with ex. We are leaving aside the issue of translation here.
(76) probet autem se ipsum homo et sic de

examine:SBJV.3 SG indeed REFL . ACC self: ACC man: NOM and so from
pane illo edat et de calice bibat
bread: ABL DEM . ABL eat:SBJV. PRS .3 SG and from cup: ABL drink: SBJV. PRS .3 SG
‘A man ought to examine himself before he eats of the bread and drinks
of/from the cup.’ (Corinthians 11.28).
The above examples contain a variety of different types of noun that function as
source in the partitive construction: a collective noun in (73), a plural count in
(74), an abstract noun in (75) and a mass noun in (76). The last example also
indicates that the source meaning was clearly expressed by this construction:
indeed, the second de phrase, de calice is ambiguous between a partitive (‘of
that cup’, with cup metonymically understood as its content), and a source read-
ing (‘from that cup’). The Latin examples all contain partitive constructions, im-
plying that the de phrase indicates a specific referent, a part of which is affected
by the state of affairs, and do not have a possible indefinite interpretation, as
grammaticalized partitives can have to varying extents. This is shown, among
other things, by the fact that they are all accompanied by some type of deter-
miner or modifier that identifies the specific referent. Thus, de piscibus in (74)
means ‘some of those specific fish that you have caught’, and could not possibly
mean ‘some (indefinite) fish’, as Italian dei pesci or French des poissons normally
mean.
As the examples show, this construction typically occurs in the place of a
direct object: as already remarked, it indicates that a part of a referent under-
goes the process, possibly a change of state, indicated by the verb. As often
noted, the partitive construction indicates partial affectedness: crucially, how-
ever, at this stage partial affectedness does not exactly coincide with a low
degree of affectedness. To the contrary, the part of the referent which undergoes
the effects of the state of affairs may be affected to any degree, including high,
as it can undergo a change of state: typically, partitive expressions occur with
verbs of ingestion, which imply that the referent of the direct object is consumed.
4.3.3 From partitive construction to indefiniteness and low transitivity
Different inferences may arise from the occurrence of a partitive construction in

the place of a direct object, which can be represented as in Figure 2.
Figure 2: Inferences arising from the partitive construction
In one case, the fact that only a part of a referent is affected gives rise to the
implicature that affectedness is partial (i.e. the feature of partiality is profiled):
this leads to the use of the partitive in low transitivity contexts. Low transitivity
may be understood in various manners, including as implying imperfectivity, as
in Finnish, or non-assertivity, as in Basque.
The other implicature, represented on the right branch of Figure 1, arises
from the same notion, but leads to an indefinite interpretation, according to
which “a non-specified quantity is necessarily non-uniquely identifiable to the
hearer.” (Carlier 2007: 27). In this latter case, it is the feature of non-specificity
which is profiled and eventually gains relevance. A further step in grammatical-
ization and semantic change is achieved when such pragmatic implicatures
become more prominent than the original notion out of which they originated
(see Traugott & Dasher 2002 and Traugott 2003: 635). These two clines of gram-
maticalization coexist to varying extents in the languages with grammaticalized
partitives that are treated in this book. It needs to be stressed that affectedness is
not a property connected with a certain grammatical relation (e.g. direct object).
Rather, it is a semantic property: accordingly, even though partitives seem to
originate from partitive constructions that occur in the place of direct objects,
they may extend to other syntactic positions, as they in fact do.
Indefiniteness is also compatible with the semantics of ‘part of’.7 As noted
above, indefiniteness becomes one of the natural readings for a group of entities
taken from a (specified) whole. The removed part is less definite than the speci-
fied whole, which explains the use of (original) separative cases for expressing
7 The earlier Finnish tradition tried to separate strictly quantity and definiteness, thus creating
confusion. The analysis given here is in line with more recent accounts such as Vilkuna (1992).
indefiniteness. The indefiniteness is especially evident when the whole is a uni-

versal set of all possible referents of the noun in question. Next, partial affected-
ness is also rather directly explainable by separation. Similarly to taking a part
of a whole, a part of an affected entity is chosen, whose affectedness is specified.
Lower degree of affectedness is a logical next step following from this, because
in this case the whole entity has been only partially affected, which is intimately
related to imperfective aspect (a ‘part of the verb’ is affected). These uses pave
the way for the use of partitives with low transitivity-predicates; not only lower
degree of affectedness is relevant, but rather the lower transitivity in general.
Finally, the use with negation follows naturally from the low transitivity uses,
because in negation, affected entities are completely missing. In brief, partitives
undergo a grammaticalization process that leads them to lose their primary case
marking function and take over other functions: indicate indefiniteness, degrees
of affectedness, imperfective aspect. All of these semantic shifts are, however,
understandable in light of their original semantics.
4.3.4 Structural development
As shown in section 4.3.3, partitive constructions start developing the meanings

connected to partitive cases when they function as direct objects. However,
virtually all languages described in this book show the possible usage of partitive
cases for subjects too, although to a more limited extent. In general, partitive
subjects occur with unaccusative verbs, and they are frequently found in exis-
tential clauses. It is a well-known fact that subjects of unaccusative verbs share
direct object properties: hence, the extension of partitive cases to such argu-
ments is consistent with other of their chracteristic features. Subjects of existen-
tials also share some typical features of direct objects: they are not topical, as
they introduce new participants into the discourse.
In Finnish, as we have shown in section 3.1 above, partitive subjects, which
express indiefiniteness productively, are largely limited to existential clauses or
to unaccusative. In addition, they are used with some transitive verbs that do
not indicate change of state, especially if there is an existential implication.
Recently, however, an ongoing change has been reported, whereby partitive sub-
jects are being extended to change-of-state verbs, with no existential implica-
tion, and tend to convey simple indefiniteness (Vilkuna 1989: 260). Apparently,
the Finnish partitive is undergoing a further step in grammaticalization, similar
to the Romance partitive when it turned into an article. Such extension is having
the effect that partitive subjects are extending to all types of verb, in very much
the same way as nominative subjects. Still, the fact that partitive subjects are
‘real’ subjects has been questioned (see Sand & Campbell 2001: 266–269 for
discussion and references therein; cf. also Helasvuo & Huumo 2010). Notably,
partitive subjects, contrary to nominative, do not trigger verb agreement, but are
always accompanied by third person singular verbs. However, syntactic tests
seem to indicate subjecthood also in the case of partitive arguments, as shown
in example (77):
Finnish (from Sands & Campbell 2001: 167)

(77) Tuli miehiä ja naisia ja tekivät
come:PST.3 SG man: PAR . PL and woman: PAR . PL and make:PST.3 PL
tupasiaan kaikille rinteille
cabin. PAR . PL . POSS 3 PL all. ALL . PL slope. ALL . PL
‘There came men and women and they made their cabins on all the
hillsides.’
Example (77) contains two coordinated sentences; the first one features a parti-
tive subject (miehiä ja naisia ‘men and women’) and a verb in the third person
singular (tuli ‘(there) came’); the second has no overt subject: the elliptical sub-
ject refers to the partitive NP of the preceding sentence, and the verb is in the
third person plural. This indicates that the partitive NP in the first sentence
does indeed have some subject properties.8
The fact that agreement with the verb has not (yet?) developed may be taken
as a consequence of the behavior-before-coding principle (Haspelmath 2010):
in very much the same way as in Romance, the partitive started out marking
objects, then extended to subject starting from existential clauses and unaccusa-
tive verbs, and is presently gaining further ground in its extension to all types of
verb and its function of indicating indefiniteness. Unlike the Romance partitive,
the Finnish partitive has acquired behavioral properties of subjects, such as
null-anaphora control (although to a limited extent), but not coding properties,
among which verb agreement. Similar to the Finnish (and Estonian) partitive,
partitive genitive subjects do not trigger verb agreement in most Indo-European
languages in which they occur, such as Russian (see Paykin, this volume) and
Indo-Iranic (see Dahl, this volume). An exception is Ancient Greek, in which
genitive plural partitive subjects trigger agreement with the verb (see Conti &
Luraghi, this volume).
At leat in Finnic languages and in Early Romance, partitive NPs can also
function as predicate nouns. The relative chronology of the extension for direct
object to subject and predicate noun is not clear, as argued for Old French in
8 It must be noted that, even though Campbell and Sands have found this example in their
corpus, example (77) is marginal for some speakers of Finnish.
Carlier & Lamiroy (this volume). In any case, the extension to predicate noun may
have started in contexts such as Finnish (69b) that we repeat for convenience:
(69) b. hän on suku-a minu-lle

s/he be: PRS .3 SG family- PAR 1 SG -ALL
‘S/he is of my family.’
In this example, belonging (to a whole) and inclusion are indicated. As remarked,
this is presumably the historically oldest usage of the partitive in predicate
nominals, where the ‘part of’ meaning is evident.
A different formal development of partitive case markers is their extension
to verbs, attested in Estonian. As we discussed in sections 2.4 and 3.5, verbs
can take the partitive ending in this language, and acquire an evidential mean-
ing. Formally and semantically, the development can be summarized as follows
(see also Metslang, this volume):
a. “part of N”
b. ! “partof V” (N-obj has the morphological partitive marking)
c. ! (N-object is a non-finite, deverbal nominalization and partitive marked)
d. ! “indirect evidence” (V-nonfin (main predicate) has the morphological
partitive formative)
e. ! “partof/incomplete evidence” (V-nonfin (main or embedded predicate) has
the morphological partitive formative)
f. ! “partof/incomplete evidence for the completion/completability of the event”
(partitive object case)
5 Conclusion
In this chapter, form, functions and development of partitive case markers have
been discussed. Dedicated partitive cases (such as those in Finnic languages
and Basque) seem to be rather rare cross-linguistically; even the usage of a
different case, such as the genitive, is not especially frequent, as it seems to be
confined to Indo-European languages.
An important feature of partitive cases, or other cases used in the function
of partitives, is their possible occurrence with NPs that bear different gram-
matical relations. In this sense, partitive cases are at odds with the definition of
case in Blake (2001: 1), that is ‘marking dependent nouns for the type of relation-
ship they bear to their heads’ (see further the discussion in Luraghi 2009: 243–
249). The reason for this is that partitive cases indicate other properties of the
NPs they occur with, notably indefiniteness, and low or partial affectedness,
and that the indication of such properties overrides the expression of grammati-
cal relations. Often, indefiniteness is accompanied by non-referentiality, as

noted in sections 3.1 and 3.6.
As we have shown in this chapter, the above features of partitive case markers
explain their frequent occurrence in various contexts in which indefiniteness or
non-referentiality are relevant, such as under negation, with quantifiers, with
non-assertive modality, in existential clauses, and so on. In addition, partial
affectedness may have a bearing on the interpretation of verbal aspect, and
convey an imperfective meaning, as in the Finnic languages.
Diachronically, partitive case markers originate either from separative cases
or from genitives; to a more limited extent, they can also apparently originate
from locatives, as in some Oceanic languages. Attested developments (mostly
from the Indo-European languages) indicate that partitive case markers may orig-
inate inside partitive nominal constructions, in the sense of Koptjevsaja-Tamm
(2006). Partitive constructions appear as direct objects for example in Late Latin;
they refer to a part of a specific, already identified whole. From such construc-
tions, two inferences can originate, one leading to the indefiniteness meaning,
and one to the implication of low affectedness. Such acquired meanings of
partitive cases make them suitable for the various contexts in which they are
found cross-linguistically. Oceanic languages also point to another possible origin
of partitive markers, that is, from markers of indefiniteness, thus highlighting the
importance between the concept of partitivity and the concept of indefiniteness.
References
Besnier, Niko. 2000. Tuvaluan: A Polynesian language of the Central Pacific. London: Routledge.
Carlier, Anne. 2007. From preposition to article: The Grammaticalization of the French partitive.
Studies in Language 31. 1–49.
Clark, Ross. 1973. Transitivity and case in eastern Oceanic languages: Oceanic Linguistics 12.
559–606.
Conti, Luz. 2010. Weiteres zum genitiv als semisubjekt im Altgriechischen: Analyse des Kasus
bei impersonalen Konstruktionen. Historische Sprachwissenschaft 122. 182–207.
Crowley, Terry. 1982. The Paamese language of Vanuatu. Canberra: Pacific Linguistics.
Hoop, Helen de. 1998. Partitivity. Glot International 3. 3–10.
Gerstner-Link, Claudia. 1998. How transitive are habituals? Sprachtypologie und Universalien-
forschung 51(4). 327–354.
Haspelmath, Martin. 2010. The behaviour-before-coding principle in syntactic change. In Franck
Floricic (ed.), Mélanges Denis Creissels, 541–554. Paris: Presses de L’École Normale
Supérieure.
Heine, Bernd & Tania Kuteva. 2004. World lexicon of grammaticalization. Cambridge: Cambridge
University Press.
Helasvuo, Marja-Liisa & Tuomas Huumo. 2010. Mikä subjekti on? [On the subject of subject in
Finnish]. Virittäjä 114(3). 165–194.
Hoeksema, Jacob (ed.). 1996. Partitives. Studies on the syntax and semantics of partitive and
related constructions. Berlin: Mouton de Gruyter.
Hopper, Paul J. & Sandra A. Thompson. 1980. Transitivity in grammar and discourse. Language
56. 251–299.
Journal of Linguistics 46. 83–125.
Kempf, Zdzisław. 2007. Próba teorii przypadków: Część II. Opole: Wydawnictwo Uniwersytetu
Opolskiego.
Koptjevskaja-Tamm, Maria 2001. “A piece of the cake” and “a cup of tea”: partitive and pseudo-
partitive nominal constructions in the Circum-Baltic languages. In Östen Dahl & Maria
Koptjevskaja-Tamm (eds.), The Circum-Baltic Languages: Typology and Contact, vol. 2,
523–568. Amsterdam: Benjamins.
Koptjevskaja-Tamm, Maria. 2006. Partitives. In Keith Brown (ed.), Encyclopedia of Languages
and Linguistics, 2nd edn., 218–221. Oxford: Elsevier.
Lakoff, George. 1987. Women, fire, and dangerous things. Chicago: University of Chicago Press.
Larjavaara, Matti. 1991. Aspektuaalisen objektin synty [The emergence of the aspectual object].
Virittäjä 95. 372–408.
Lass, Roger. 1990. How to do things with junk: Exaptation in language evolution. Journal of
Mouton de Gruyter.
Luraghi, Silvia. 2013. Partitivi nel latino bibblico. Studi Italiani di Linguistica Teorica e Applicata
42(1). 41–60.
Moravcsik, Edith. 1978. On the case marking of objects. In Joseph Greenberg et al. (eds.)
Universals of Human Language, vol IV. Syntax, 249–290. Stanford University Press.
Mosel, Ulrike & Even Hovdhaugen. 1992. Samoan reference grammar. Oslo: Scandinavian Uni-
versity Press.
Napoli, Maria. 2010. The case for the partitive case: The contribution of Ancient Greek.
Transactions of the Philological Society 108(1). 15–40.
Nikiforidou, Kiki. 1991. The meanings of the genitive: A case study in semantic structure and
semantic change. Cognitive Linguistics 2(2). 149–206.
Paykin, Katia & Marleen Van Peteghem. 2002. Definiteness in a language without articles: A
case-study of Russian. Recherches linguistiques de Vincennes 31. 97–112.
Sands, Kristina & Lyle Campbell. 2001. Non-canonical subjects and objects in Finnish. In
Alexandra Aikhenvald, R.M.W. Dixon & Masayuki Onishi (eds.), Non-canonical marking of
subjects and objects, 251–305. Amsterdam: Benjamins.
Traugott, Elizabeth C. 2003. Constructions in grammaticalization. In Brian J. Joseph & Richard J.
Janda (eds.), Handbook of Historical Linguistics 624–647. Oxford: Blackwell.
Traugott, Elizabeth C. & Richard B. Dasher. 2002. Regularity in semantic change. Cambridge:
Cambridge University Press.
Vilkuna, Maria. 1989. Free word order in Finnish: Its syntax and discourse functions. Helsinki:
Suomalaisen Kirjallisuuden Seura.
Vilkuna, Maria. 1992. Referenssi ja määräisyys suomenkielisten tekstien tulkinnassa [Reference
and definiteness in the interpretation of Finnish texts]. Helsinki: SKS.
Zwarts, Frans. 1987. Paradigmaloos van. TABU 17. 184–192.
Matti Miestamo
2 Partitives and negation: A cross-linguistic
survey
The partitive of negation, i.e. partitive marking of NPs under the scope of nega-
tion, is known to be found in some European languages, namely Finnic, Baltic,
Slavic and Basque. Based on an extensive and representative language sample,
this typological study surveys the cross-linguistic distribution of the partitive of
negation and other asymmetries between affirmatives and negatives in the mark-
ing of noun phrases. Instances of the partitive of negation realized as part of a
case marking system are not found outside the European languages mentioned.
Nonetheless, negation is found to affect the use of articles and other determiners,
e.g., in Polynesian languages and in French, as well as the use of class markers,
e.g., in some Bantu languages, in which the class markers in question actually
function as determiners. Effects on focus marking and alignment are also ob-
served in some languages. There is a pragmatically motivated tendency for an
indefinite noun phrase to have a non-referential reading under the scope of
negation. The grammatical effects of negation on the use of articles and deter-
miners, including class markers, result in marking the noun phrases as non-
referential and are thus motivated by the connection between negation and
non-referentiality. It is further argued that partitives, too, contribute to marking
noun phrases as non-referential, and the partitive of negation can thus be seen
as arising from similar motivations.
Keywords: negation, partitive, case marking, determiners, referentiality
1 Introduction
In some European languages – Finnic, Baltic, Slavic and Basque – noun phrases
in the scope of negation are marked, either obligatorily or as a matter of prefer-
ence, with a case that has a partitive-marking function (partitive or genitive). In
this paper, the phenomenon will be referred to as the partitive of negation.
Although the link between partitives and negation is relatively well-studied in
these European languages, it has not been systematically addressed in typolog-
ical research. Related phenomena have been reported in some language groups
outside Europe, e.g., in some Oceanic languages, but their cross-linguistic distri-
bution is not known. The present paper aims to fill this gap. It will report the
64 Matti Miestamo
results of a large-scale typological survey of the partitive of negation and related

effects of negation on the marking of grammatical categories in noun phrases. In
a larger context, the effects of negation on noun phrases are one of the many
ways in which negation can affect the structure of clauses, or in Miestamo’s
(2005) terms, one of the many ways in which negatives can show structural
asymmetry vis-à-vis affirmatives. This larger context becomes relevant when we
start looking for explanations for the link between partitives and negation. In
the literature, it has been attributed to semantic and pragmatic factors, such as
quantification, referentiality/specificity and aspect (cf., e.g., Krasovitsky et al.
2011).
The remainder of this introduction gives a short presentation of the phe-
nomenon as we know it from familiar European languages. Section 2 discusses
the methodology followed in the typological survey, while Section 3 presents the
results. Section 4 discusses the cross-linguistic findings in a functional perspec-
tive and concludes the paper.
The following examples (1) give a preliminary illustration of the case alter-
nations in Finnish.
(1) Finnish (constructed examples)1

a. söin banaani-n
eat.PST.1 SG banana-GEN
‘I ate {a/the} banana.’
b. söin banaani-a
eat.PST.1 SG banana-PAR
‘I {ate some / was eating {a/the}} banana.’
c. en syönyt banaani-a
NEG .1 SG eat.PST. PTCP banana-PAR
‘I {didn’t eat / wasn’t eating} {a/the} banana.’
In the affirmative, a distinction between total and partial objects2 can be made
using different case forms – the total object is marked by genitive3 case as in (1a)
1 The abbreviations used in the glosses are listed in the beginning of the collective volume.
2 The terms total and partial object are used here, but as already pointed out above, quantifica-
tion is only one factor determining their use, others being, e.g., aspectuality and referentiality.
3 Traditionally this form in this function has been called accusative despite the fact that it is
formally identical to the genitive; only personal pronouns and the pronoun ‘who’ have separate
accusative forms. In this article I will adopt the usage in the most up-to-date and comprehen-
sive grammar of Finnish (Hakulinen et al. 2004: 108) and restrict the term accusative to the
separate accusative forms of pronouns.
Partitives and negation: A cross-linguistic survey 65
and the partial object by partitive case as in (1b). In the negative (1c) only the
partitive can be used and the distinction between total and partial objects is
lost.
In a similar way, in a number of European languages, noun phrases in the
scope of negation are marked by a case that has partitive semantics. These lan-
guages include Finnish and Estonian (Finnic); Lithuanian, and to some extent
also Latvian (Baltic); Russian, Ukrainian, Polish (Slavic); and Basque. In the
Baltic and Slavic languages mentioned, the case with partitive functions involved
in the alternation is the genitive. With the exception of Basque, these languages
belong to the Circum-Baltic languages as defined by Dahl and Koptjevskaja-
Tamm (2001: xviii–xix). According to Koptjevskaja-Tamm & Wälchli (2001: 663),
case alternations between total and partial objects (usually involving the parti-
tive of negation) are also found in older stages of Indo-European languages,
e.g., Classical Greek, Sanskrit, Gothic, Old High German and Middle Low
German, total objects being marked by the accusative and partial objects by
the genitive.
The partitive of negation has been studied quite extensively in many of the
European languages in which it occurs. For Russian, for example, there is a
separate bibliography devoted to the partitive (=genitive) of negation (Corbett
1986). Koptjevskaja-Tamm & Wälchli (2001: 729) characterize the case alterna-
tions between total and partial objects (usually involving the partitive of nega-
tion in one way or another) as typologically “probably unusual but not unique”
to Circum-Baltic languages; their observation is not based on a cross-linguistic
survey going beyond their Circum-Baltic areal focus. The partitive of negation has
also been noted in general typological literature on negation, e.g., in Payne (1985),
Forest (1993), Honda (1996), and Miestamo (2005), as well as in Moravcsik’s (1978)
article on the typology of object marking. Payne lists it as one of the “secondary
modifications” that may be found in negatives in addition to negative marking
itself. In Miestamo (2005), I noted that it can be seen as one of the many ways
in which negatives show structural asymmetry vis-à-vis affirmatives, but did not
include it in the scope of my typological survey. In the literature the examples
are taken from the familiar European languages mentioned above, and no typo-
logical information is available on the cross-linguistic frequency or areal distri-
bution of the phenomenon. The same lack of typological information is true of
other effects of negation on the marking of noun phrases. To fill this lacuna in
the typological literature, this chapter presents a typological survey of the parti-
tive of negation and related effects of negation on the marking of grammatical
categories in noun phrases. It should perhaps be noted that negative polarity
items, although closely related to the issue of the marking of grammatical cate-
gories in noun phrases under the scope of negation, are beyond the scope of the
present survey. I will now turn to the material and method of the survey.
66 Matti Miestamo
2 Material and method

This section will explicate the methodological choices adopted in the cross-
linguistic survey of the nature and spread of the partitive of negation and
related effects on the marking of noun phrases under negation. The partitive of
negation is here defined as the obligatory or preferred use of partitive marking,
or marking that has partitivity as one of its functions, on noun phrases under
the scope of negation, in contrast to the corresponding noun phrases in affir-
matives showing less or no partitive marking. Partitive marking is here defined
simply as marking that denotes a part of a whole, or more generally, an in-
definite quantity (cf. the function of partitive case in the Finnish example in 1b
above).
In the typological survey, I have paid attention to all realizations of the
partitive of negation found in the languages surveyed. The partitive of negation
is part of the broader question of how the marking of noun phrases is affected
under negation. Other effects of negation on the marking of grammatical cate-
gories within noun phrases are also paid attention to in the survey, with a focus
on effects that are connected to the domains of quantification, referentiality/
specificity and aspect, i.e. the domains that have been suggested to be relevant
in finding functional motivations for the partitive of negation. Referentiality and
specificity are used roughly synonymously in this paper; in referential/specific
use of noun phrases the identity of the referent is established, i.e. the speaker
has in mind a specific entity or entities to which the noun phrase refers.
This study is primarily based on a sample of 240 languages.4 The sample
languages come from different genera (in the sense of Dryer 1989; see also
Haspelmath et al. 2005; Dryer & Haspelmath 2011), i.e. no two languages come
from the same genus. The same language sample has been used in my earlier
work on negation (see Miestamo 2005: 27–39, 241–254). In that study I focused
on structural differences between negatives and affirmatives manifested on the
verbal and clausal levels, but did not pay attention to the marking of noun
phrases. When examining the sources of the sample languages, I did, however,
try to make notes of everything the sources said on negation. The present paper
is based on a re-examination of those notes and going back to the original sources
in case the notes indicated a given language might show some effects of
4 To the extent possible, all language names in this paper are given in the form in which
they appear in The world atlas of language structures (Haspelmath et al. 2005; Dryer &
Haspelmath 2011).
negation on the marking of noun phrases.5 The survey of the sample languages
was supplemented by a query on the Lingtyp mailing list, asking for pointers to
languages that have the partitive of negation or any other changes in the mark-
ing of noun phrases induced by negation.6
In the survey I have taken into account all languages in which I have found
some effects of negation on the marking of noun phrases, regardless of whether
they belong to the 240-language sample, or whether I have become aware of the
data through the Lingtyp query or in other ways. I have then classified the ob-
served effects into types according to their structural and functional properties.
When making observations about the cross-linguistic frequency of the different
types, it is important to base these observations on an areally and geneal-
ogically balanced language sample. The 240-language sample provides such a
basis, and in principle it would be possible to balance it even further in areal-
genealogical terms following the principles introduced in (Miestamo 2005).
However, as the following section shows, the types are all quite rare and areally
and genealogically constrained so that quantitative analysis would not make
much sense in this study. The cross-linguistic observations and generalizations
are presented in Section 3, and their possible functional motivations are dis-
cussed in Section 4.
3 Results
The broad cross-linguistic survey conducted here confirms Koptjevskaja-Tamm
& Wälchli’s (2001: 729) estimation that the partitive of negation is typologi-
cally unusual. In fact, clear instances of the partitive of negation realized as
part of a case marking system were not found outside the European languages
already known to exhibit the phenomenon. In this section, I will start with a
closer look at the effects on case marking, and then move on to other types of
elements affected by negation, such as articles and other determiners, and
5 In (Miestamo 2005) the sample size was actually 297 languages, but only 240 languages
belonged to the core sample in which every language comes from a different genus. For the 57
extra languages notes were not made on negation-related phenomena that were outside the
scope of the study, and thus no systematic notes were available on effects of negation on the
noun phrase level for these languages.
6 The original query as well as a summary of the replies is available in the archives of the mail-
ing list at <http://listserv.linguistlist.org/cgi-bin/wa?A0=LINGTYP> (see October 2009, Week 5;
November 2009, Week 1).
68 Matti Miestamo
class-markers, and finally say a few words on the effects of negation on focus
marking and alignment.
As mentioned in Section 1, a number of European languages (Finnic, Baltic,
Slavic and Basque) use a case with a partitive function on noun phrases in the
scope of negation, either obligatorily or as a matter of preference. The main
characteristics of these systems will now be discussed (for a more detailed pre-
sentation, see Koptjevskaja-Tamm & Wälchli 2001: 650–671).
Finnish has a system in which noun phrases in the scope of negation are
marked with the partitive case. Both objects of transitive sentences and subjects
of existential sentences are affected. Let us first look at the objects of transitives,
see the examples in (2).
(2) Finnish (constructed examples)

a. söin banaani-n
eat.PST.1 SG banana-GEN
‘I ate {a/the} banana.’
b. söin banaani-t
eat.PST.1 SG banana-PL . NOM
‘I ate the bananas.’
c. söin banaani-a
eat.PST.1 SG banana-PAR
‘I {ate some / was eating {a/the}} banana.’
d. söin banaane-j-a
eat.PST.1 SG banana-PL- PAR
‘I {ate (some) / was eating {(some)/the}}bananas.’
e. en syönyt banaani-a
NEG .1 SG eat.PST. PTCP banana-PAR
‘I {didn’t eat / wasn’t eating} {a/the} banana.’
f. en syönyt banaane-j-a
NEG .1 SG eat.PST. PTCP banana-PL- PAR
‘I {didn’t eat / wasn’t eating} (the) bananas.’
In the affirmative, total and partial objects can be distinguished. I will not go
into details about their semantics, but as a general rule, it can be said that total
objects are interpreted as totally affected and the sentence gets a perfective
aspectual reading whereas partial objects are partially affected and usually give
an imperfective reading to the sentence. Total objects are marked with the
genitive or nominative case, depending on the number of the object and its
morphosyntactic environment, e.g., clause type; the examples in (2a–b) are
simple affirmative declaratives. Partial objects are marked with the partitive
(2c–d). In the negative, only the partitive is possible (2e–f), and the distinction
between total and partial objects cannot be made. The requirement of the parti-
tive on objects of negated clauses is highly grammaticalized in Finnish. Only
marginally, under specific semantic-pragmatic conditions, is it possible to have
non-partitive objects in negatives (see Almqvist 1987).7 Note that there are verbs
that require their objects to be in the partitive in the affirmative as well, but I
will not go into these lexical issues here.
Case marking differs between affirmatives and negatives in a related way in
some existential sentences as well, see the examples in (3).8
(3) Finnish (constructed examples)

a. pöydällä on omena
table.ADE be.3SG apple.NOM
‘There is an apple on the table.’
b. pöydällä on omena-a
table.ADE be.3SG apple-PAR
‘There is some apple on the table.’
c. pöydällä ei ole omena-a

table.ADE NEG .3 SG be.CNG apple-PAR
‘There is {not an apple / no apple} on the table.’
d. pöydällä on omeno-i-ta
table.ADE be.3SG apple-PL- PAR
‘There are apples on the table.’
e. pöydällä ei ole omeno-i-ta

table.ADE NEG .3 SG be.CNG apple-PL- PAR
‘There are no apples on the table.’
7 Essentially, these are cases in which there is a positive implication despite the overtly nega-
tive form of the sentence. An example would be Ei liene järin ylivoimaista toteuttaa tuo pyyntö.
(NEG .3 SG be.POT. CNG very insurmountable fullfill.INF that.NOM request.NOM ) ‘It shouldn’t be too
hard to fullfill that request’ (Almqvist 1987: 163). In most cases, like here, the total object under
negation is the object of an infinitive itself under the scope of a negated finite verb.
8 These are existential predications. Locative predications would have the subject in nomina-
tive case and exhibit some other word order than LOCATION + COPULA + SUBJECT, most typically
SUBJECT + COPULA + LOCATIVE .
70 Matti Miestamo
In affirmatives, singular subjects can be in the nominative or in the partitive

with a quantificational difference in semantics (3a,b), and plural subjects are in
the partitive (3d) (the nominative could replace the partitive in 3d but it would
produce a definite reading and require a special contrastive context in this clause
type). In negatives, the subject of the existential is in the partitive in both singular
and plural (3c,e). The details of the Finnish case alternations are complicated
and have generated a lot of literature (e.g., Almqvist 1987), but in this typologi-
cal survey I will not delve deeper into those details. Note finally that at an
earlier (reconstructed) stage of the language, the function of the partitive case
was separative (movement from), see for example (Hakulinen 1961: 93).
The system is essentially similar in most other Finnic languages, e.g., Esto-
nian (Erelt 2003: 95–97; Metslang, this volume) and Votic (Ariste 1968: 21). There
are some differences, of course, e.g., according to Metslang (2001), the role of
case in the marking of aspect is not as important in Estonian as it is in Finnish,
since the verbal particle ära is used together with total objects to mark perfec-
tivity. In Liv, as noted by Koptjevskaja-Tamm & Wälchli (2001: 652), a develop-
ment has been observed whereby the partitive would be spreading as a general
object case, making thus the alternation obsolete. As to existential clause sub-
ject marking in other Finnic languages, the system is, in the main, similar to
Finnish (see Huumo & Lindström, this volume, for more details on the uses of
the partitive in Estonian and Finnish existentials).
In Lithuanian (Dambriunas 1972: 27, 39, 123, 139–141, 166; Ambrazas et al.
1997: 500–506, 667–668), the direct object is in the genitive in negatives. In
non-negatives, most verbs take accusative objects, but the genitive may be used
to refer to an indefinite amount or quantity where the nominative or accusative
would otherwise be used. Subjects of negative existentials are in the genitive if
they are in the scope of negation; in the affirmative, genitive subjects may be
used with selected verbs to denote indefinite quantity. In Latvian, subjects of
negated existentials are in the genitive, but objects of transitive sentences do
not show the partitive of negation, see (Lazdiņa 1966: 24, 28; Holst 2001: 207–
210; Fennel & Gelsen 1980: 22–23, 26).
In Polish (Bielec 1998: 69–70, 103, 117–118; Swan 2002: 333–335), the direct
object in negatives is in the genitive instead of the accusative used in affirma-
tives (4a,b). The subjects of negative existentials are also in the genitive instead
of the nominative used in affirmatives (4c,d). The genitive has partitive uses in
non-negatives (4e).
(4) Polish (Swan 2002: 333; Bielec 1998: 70, 121)

a. Oglądam telewizję. b. Nie oglądam telewizji.
watch.1SG television.ACC NEG watch.1SG television.GEN
‘I watch televison.’ ‘I don’t watch television.’
c. W parku jest fontanna d. W parku nie ma fontanny

in park.LOC is fountain.NOM in park.LOC NEG have fountain.GEN
‘There is a fountain in the park.’ ‘There is no fountain in the park.’
e. Kupię mleka i sera.

buy.1SG milk.GEN and cheese.GEN
‘I’ll buy some milk and cheese.’
In Russian (Wade 2000: 111–115; Krasovitsky et al. 2011), the use of the geni-
tive is not obligatory in direct objects under negation and depends on various
factors, including definiteness and specificity, definite and specific objects being
more readily accusative-marked; verbal aspect and the lexical semantics (e.g.,
abstractness) of the noun also play a role. Subjects of negative existentials
are in the genitive instead of the nominative used in affirmative existentials. In
Ukrainian the system is largely similar to Russian (Pugh & Press 1999: 97–99).
In Czech (Naughton 2005: 196–198), the genitive is used in these functions only
occasionally, in contemporary language largely restricted to fixed phrases. In
other Slavic languages the partitive of negation is even more restricted or non-
existant.
The only language outside Finnic, Baltic and Slavic exhibiting the partitive
of negation expressed within a morphological case marking system is Basque.9
In Basque (Hualde & Urbina 2003: 124–126, 549–554), the partitive case marked
with -(r)ik regularly occurs on transitive objects and intransitive subjects in
negative sentences. It may also occur in other contexts, e.g., polar interrogatives
and conditionals. It is interpreted as non-specific. Its diachronic origin is in the
ablative suffix (Hualde & Urbina 2003: 551). If negatives have an object with a
specific/definite reading, the partitive is not used (cf. Etxeberria, this volume;
Koptjevskaja-Tamm & Wälchli 2001: 666).10
I will now leave case marking and turn to the effects of negation on other
types of marking in noun phrases. In French, as example (5) shows, the deter-
miner de (glossed here simply as DET ) replaces the indefinite article under the
scope of negation both in regular transitives and in existentials.
9 Note that its status as a case is questioned by Hualde & Urbina (2003: 124). In any case, it is
a partitive marker and thus relevant in the present context.
10 In Evenki, noun phrases in privative constructions with the negative noun a:chin take the
indefinite accusative (partitive) case, but this case does not seem to be used in existential
predications with a:chin (see Nedyalkov 1994: 4, 27–29; Pakendorf 2007: 162–164).
72 Matti Miestamo
(5) French (constructed examples)

a. je vois un chien
1 SG . NOM see.1 SG INDF. M dog
‘I see a dog.’
b. je ne vois pas de chien

1 SG . NOM NEG see.1 SG NEG DET dog
‘I do not see a dog.’
c. il y a un livre sur la table

EX INDF. Mbook on DEF. F table
‘There is a book on the table.’
d. il n’y a pas de livre sur la table

EX . NEG DET book on DEF. F table
‘There is no book on the table.’
e. je bois du lait
1 SG . NOM drink.1SG PAR . M milk
‘I’m drinking (some) milk.’
f. je ne bois pas de lait

1 SG . NOM NEG drink.1 SG NEG DET milk
‘I’m not drinking milk.’
g. je vois des chiens

1 SG . NOM see.1 SG PAR . PL dog
‘I see (some) dogs.’
h. je ne vois pas de chiens

1 SG . NOM NEG see.1 SG NEG DET dog.PL
‘I do not see dogs.’
In most contexts, the determiner de occurs in front of the noun phrase in the
scope of negation (5b,d) where the corresponding affirmatives have indefinite
articles (5a,c). With non-count and plural count nouns, the determiner de replaces
the so-called partitive article, formed by the combination of the separative prepo-
sition de and the definite article (see 5e–h).
It may now be asked, how the alternation between indefinite/partitive articles
used in affirmatives vs. the determiner de used in negatives is related to the
partitive of negation – after all, the marker appearing under negation is not a
partitive marker per se and there seems to be a partitive marker used in affirma-
tives but not in negatives. First of all, although the partitive article appearing
in affirmatives derives historically from a partitive construction expressing a
part of a whole, in contemporary French it has lost this function and functions
primarily as an indefinite article with non-count and plural count nouns (see
Carlier 2007). Thus, we are actually dealing with an alternation between indefinite
articles and the determiner de in these cases, too. Secondly, the determiner de
can be seen as related to the partitive in the following way. As mentioned above,
the partitive article, formed by the combination of the separative preposition de
and the definite article, had a truly partitive meaning in older stages of French.
In the combination of de and a definite article, it was the separative preposition
de that contributed the meaning of extracting a part of the whole and was thus
responsible for the partitive meaning of the combination. The determiner de is
etymologically the same element as the separative preposition (cf. the diachrony
of the Finnish and Basque partitive case markers above). Its connection to parti-
tives is further demonstrated by the fact that it is required to appear before
nouns with most quantifiers, e.g., peu de lait ‘little milk’, beaucoup de lait ‘a
lot of milk’. The French alternation thus shows clear similarities to the partitive
of negation as defined in this paper.
Taking a closer look at the function of the alternation in contemporary
French, we may note that the so-called partitive article may actually appear
under negation, but it then gets a referential reading as in (6a).
(6) French (constructed examples)

a. je ne bois pas du lait qu’ il m’ offre
1 SG . NOM NEG drink.1 SG NEG PAR . M milk REL 3 SG . NOM 1SG . DAT offer
‘I’m not drinking (any of) the milk he’s offering me.’
This shows us that the use of the determiner de under negation is in fact con-
nected to the referentiality of the noun phrase. I will not enter into a more
detailed discussion of the French facts here, but I will come back to the issue
of referentiality at many points later in this paper.
Article usage is affected by negation in Albanian as well: according to
(Newmark et al. 1982: 152), negative generic expressions (with meaning ‘(there
is) not a / no’) take indefinite nouns without indefinite article. Givón (1978: 74,
citing Robert Hetzron, p.c.) notes that in Hungarian the indefinite article convey-
ing a referential indefinite reading cannot be used with the object of a negated
sentence. Another case of the use of articles affected under negation is found in
the Brazilian language Nambikuára (7).
74 Matti Miestamo
(7) Nambikuára (Kroeker 2001: 34)

a. hu3kx-a2 yũ3n-a1-wa2
bow-DEF own-1 SG - IPFV
‘I have a bow.’
b. hu3ki3-la2 yũn2-nxa3-wa2
bow-NEG own-1 SG . NEG - IPFV
‘I don’t have a bow.’
In transitive and nonverbal negative sentences, the negative clitic -la3 occurs on
the object and replaces the definite or indefinite article suffix, see (Kroeker 2001:
34, 43, 46, 76). The effects on article usage found in Albanian, Hungarian and
Nambikuára are not instances of the partitive of negation in the sense defined
above, and the same is, in the main, true for the effects of negation on the mark-
ing of noun phrases discussed in what follows. However, as will be shown, most
of the phenomena to be discussed are relevant for understanding the motivation
of the partitive of negation.
According to Creissels (2009: 90, 165), Kita Maninka, spoken in Mali, has a
marker of definiteness on the noun. In affirmatives, this marker, despite its
name, is used in both definite and indefinite contexts (8a–b; note that this
marker is purely tonal: yirı̍ vs. yiri), and definiteness may be specified by deter-
miners as in (8c–d). In negatives, the presence of the definiteness marker on the
noun is not obligatory and a distinction between indefinite (non-referential) vs.
definite may be made by the form of the noun alone (8e–f).
(8) Kita Maninka (Creissels 2009: 90–91)

a. n dí yirı̍ tège
1 CMPL . AFF tree.DEF cut
‘I cut a/the tree.’
b. *n dí yiri tège
1 CMPL . AFF tree cut
c. n dí yirì dò tège
1 CMPL . AFF tree.DEF EXTR cut
‘I cut a tree.’
d. n dí yirı̍ ’n tège
1 CMPL . AFF tree.DEF DEM cut
‘I cut the tree.’
e. n mán yiri tège

1 CMPL . NEG tree cut
‘I didn’t cut a/any tree.’
f. n mán yirı̍ tège

1 CMPL . NEG tree.DEF cut
‘I didn’t cut the tree.’
The definiteness marker could be characterized as a default determiner whose

absence is licensed by certain contexts only. Negative is one of these contexts,
and this choice is also available in interrogatives.
In a number of Bantu languages, negation has an effect on the form of the
noun class prefixes, see the Xhosa examples in (9).
(9) Xhosa (Taraldsen 2010: 1526–1527)

a. ndi-bon-a a-ba-fundi
1 SG -see-FV DET-CL-student
‘I see the/some students’
b. a-ndi-bon-i ba-fundi
NEG -1 SG -see-FV CL -student
‘I don’t see any students’
c. a-ndi-ba-bon-i a-ba-fundi
NEG -1 SG - CL -see-FV DET- CL-student
‘I don’t see the students’
‘There are some students I don’t see.’
In the affirmative (9a) the noun class prefix -ba- is preceded by a pre-prefix, a
kind of default determiner that can be absent in certain contexts only (very
much like the definiteness marker in Kita Maninka above, Denis Creissels, p.c.).
Negation is one of the contexts in which the default determiner can be omitted
(9b). The determiner can be found in negatives, as well, if the object prefix also
appears on the verb (9c), i.e. when the object is definite (or, more rarely, specific
indefinite). The same phenomenon is found in closely related languages such as
Zulu (see Doke 1961: 300–301), as well as in some more distantly related Bantu
languages such as Bemba, Bobangi, Kinyarwanda and Luganda (see Givón 1978:
74–75). Doke’s examples show that interrogatives form another context in which
the absence of the pre-prefix is possible in Zulu (cf. Kita Maninka above).
In the Australian language Nunggubuyu, according to Heath (1984: 526–
531), negation has a number of effects on the structure of the clause. One of
76 Matti Miestamo
these effects is that nominals in the scope of negation obligatorily have a noun
class prefix and furthermore non-humans must have the continuous rather than
the punctual aspect noun class prefix. In non-negative contexts, the choice of
overt prefix and continuous prefix for nonhumans depends on a multitude of
factors including case, givenness/definiteness (for the functions of noun class
prefixes, see Heath 1984: 163–173). Heath (1984: 169) notes that in the nomina-
tive (the case used for subjects and objects) where the opposition is the most
significant, the presence of the (continuous) prefix correlates with definiteness
and givenness and its absence with focus and foregrounding.
Some Oceanic languages, e.g., Araki spoken in Vanuatu, show an interest-
ing interaction between negation and determiners that are often termed partitive
markers. Consider the Araki examples in (10) (Alexandre François, p.c.; see also
François 2002: 54–68).
(10) Araki (Alexandre François, p.c.)

a. nam les-i-a jau lo lep̈ a
1SG . R see-OBJ. REF-3 SG coconut.crab LOC ground
‘I’ve seen a/the coconut crab on the ground.’
b. nam les-i-a jau mo-hese lo lep̈ a

1SG . R see-OBJ. REF-3 SG coconut.crab 3.R-one LOC ground
‘I’ve seen a coconut crab on the ground.’
c. nam je les re jau lo lep̈ a

1SG . R NEG see PAR coconut.crab LOC ground
‘I haven’t seen a/any coconut crab on the ground.’
d. nam je les-i-a jau lo lep̈ a

1SG . R NEG see-OBJ. REF-3 SG coconut.crab LOC ground
‘I haven’t seen the coconut crab on the ground.’
[but not *‘I haven’t seen a coconut crab on the ground.’]
In realis affirmatives, as in (10a), objects are bare noun phrases and the verb
bears a marker of referential object and person-number cross-reference. The
object may be further specified as indefinite by the specific indefinite marker
mo-hese (10b). In the negative, there is no cross-reference on the verb and the
object is marked by the partitive marker re (10c). Referential marking and
cross-reference on the verb is possible in negatives, but then the reading is
definite (10d); in this case re does not occur.
The specific indefinite marker mo-hese is impossible in negatives (11a) and the
partitive re is impossible in realis affirmatives (11b). As can be seen in (11c–d),
both specific and non-specific are possible in irrealis affirmatives. In the negative,
realis and irrealis behave in the same way.
(11) Araki (Alexandre François, p.c.)

a. *nam je les-i-a jau mo-hese lo lep̈ a
1SG . R NEG see-OBJ. REF-3 SG coconut.crab 3.R-one LOC ground
*‘I haven’t seen a coconut crab on the ground.’
b. *nam les re jau lo lep̈ a

1SG . R see PAR coconut.crab LOC ground
*‘I have seen any coconut crab on the ground.’
c. na pa han re jau
1 SG . IRR FUT eat PAR coconut.crab
‘I will eat a/some coconut crab.’
d. na pa han-i-a jau
1 SG . IRR FUT eat-OBJ. REF-3 SG coconut.crab
‘I will eat the coconut crab.’
Alexandre François (p.c.) summarizes the situation as in Table 1. The marker

re is also found in negative existentials (see François 2002: 65–66), but there it
seems to have grammaticalized as part of the negative existential predicate.
Affirmative existentials do not use re.
Table 1: Verbal cross-reference and marking of referentiality in noun phrases in Araki (Alexandre
François, p.c.)
AFF REALIS AFF IRREALIS NEG REALIS NEG IRREALIS
OBJECT [+def] V-i-a+N V-i-a+N V-i-a+N V-i-a+N

OBJECT [–def, +ref] V-i-a+N (mo-hese) * * *
OBJECT [–def, –ref] * V+re N V+re N V+re N
The marker re is referred to as the partitive-indefinite marker by François (2002:

59), because it has among its prototypical uses the partitive function (‘some
[water]’), but on the same page, the author describes it as a marker of non-
specific indefinite reference, and this seems to be the core meaning of the
marker. Given that re has a partitive function, too, the Araki case can be seen
as an instance of the partitive of negation.
A situation closely similar to what was just described for Araki obtains
in Mav̈ea, another Oceanic language of Vanuatu (Guérin 2007), a marker of
non-referentiality appearing in negative contexts. Samoan is another Oceanic
78 Matti Miestamo
language with a similar alternation between markers of specificity and non-

specificity. In this language (see Mosel & Hovdhaugen 1992: 263–264, 480), a
distinction between specific and non-specific reference can be made using articles.
In the present context, it is worth noting that the non-specific article is obliga-
tory with the absolutive noun phrase of the negative existential verb leai. This is
a highly grammaticalized restriction, since it also applies to proper names. The
absolutive noun phrase arguments of lē/leʻi maua ‘do not get’ and lē/leʻi lagona
‘do not feel’ behave similarly. Furthermore, whereas in positive generic verbal
clauses the arguments are determined by the singular specific article, in negatives
the non-specific article is used. In negative (and polar interrogative) equational
clauses the predicate noun phrase is obligatorily non-specific if it expresses a
quality, but predicates of identification are specific. According to Claire Moyse-
Faurie (p.c.), a pattern similar to what is found in Samoan is common in Polyne-
sian languages (see also Budd, this volume).
Another case of negation affecting the marking of referentiality is reported
from the Chadic language Hdi (Frajzyngier & Shay 2002: 333–334; Wolff 2009:
49). In this language, the suffix -ta on the verb marks the referentiality of the
event; one characteristic of a referential event is the referentiality of the object.
The referential suffix does not occur in negative clauses. This is not an effect of
negation on the marking of noun phrases, but it is clearly functionally related to
the effects discussed above.
Finally, it is worth mentioning in this discussion of the relationship between
partitives and negation, that some Oceanic languages show an interaction
between negation and partitive markers appearing on the verb. In Paamese
(Crowley 1982), yet another language of Vanuatu, as can be seen in the examples
in (12), the verb receives a partitive marker when negated.
(12) Paamese (Crowley 1982: 144, 145)

a. long-e b. ro-longe-tei
3SG . R .hear-3SG .OBJ NEG -3 SG . R . hear-PAR
‘He heard him.’ ‘He did not hear him.’
c. longe-nV ree-ku
3 SG . R . hear-COMM .OBJ voice-1 SG
‘He heard my voice.’
d. ro-longe-tei ree-ku
NEG -3 SG . R .hear-PAR
voice-1 SG
‘He didn’t hear my voice.’
e. ma-ani-tei raise
1SG . IM -eat-PAR rice
‘I would like to eat some rice.’
The partitive suffix appears on the verb in negated intransitives and transitives
with non-generic objects. It can also be used in affirmatives to convey partitive
meaning (12e). In intransitive affirmatives, the function of the partitive is to
express “that the action or the state depicted by the verb is attained only a little
and is not a major performance of the action or a complete achievement of
the state”, and in transitive affirmatives “that the referent of the object is an
indefinite subset of the total possible class of objects” (Crowley 1982: 144). The
function of the partitive is thus similar to the functions of the partitive markers
appearing in noun phrases in other languages seen above, but in Paamese the
partitive marker is a verbal suffix. What we find in Paamese is closely related to
the partitive of negation. It is interesting to note that in the related language
Lewo, the partitive marker re modifying the verb has grammaticalized as a
negative marker; the partitive marker still exists in the same phonological form
but cannot cooccur with the negator re, and the diachronic development, moti-
vated by the functional connection between negation and partitive, has thus led
to a situation in which negation and partitive are mutually exclusive (see Early
1994 for details).
We have seen that Araki (and some other Polynesian languages) show an
explicit connection between negation and the marking of referentiality. It will
be argued in Section 4 that the partitive of negation and most of the other effects
of negation on the marking of noun phrases taken up so far can also be seen as
functionally motivated by the effects of negation on the referentiality of argu-
ments. The above discussion of these cases is therefore interesting in view of
placing the partitive of negation in a typological-functional context. Before con-
cluding this section, I will briefly mention some other effects that negation may
have on the marking of noun phrases, but that cannot be directly linked with
referentiality, and are therefore not central to the aims of this paper.
In the Bantu language Aghem (Hyman 2010; Larry Hyman, p.c.), negation
is treated as inherently focused, which has the effect that objects are treated as
obliques in negatives. A connection between focus and negation is found in
many languages of Africa, resulting in different structural asymmetries between
affirmation and negation. An incompatibility between negation and focus in
Kanuri is noted by Cyffer (2009: 87, 89–90; see also some other papers in that
volume for the relationship between negation and focus in African languages).
In Lavukaleve (Terrill 2003), too, negation has some effects on the marking of
focus (see also Miestamo 2005: 137). The relationship between negation and
80 Matti Miestamo
focus has not received a lot of attention in typological studies of negation, and
remains a fruitful topic for future work.
Finally, it may be noted that in some languages, the alignment system is
affected by negation. According to Eduardo Ribeiro (p.c.), in Northern Ge lan-
guages (Central Brazil), negation triggers an ergative alignment pattern, and
the verb is nominalized (see Silva 2001 for Kayapó, Alves 2004 for Canela, and
Oliveira 2005 for Apinayé). In fact, the ergative pattern appears with nominaliza-
tion in other contexts as well, not only negation, and it is thus the nominaliza-
tion that is responsible for the ergative pattern, and ergativity is triggered by
negation only indirectly. In Yimas, one of the effects of negation is that the
alignment pattern of the clause changes with respect to the corresponding affir-
mative; in Yimas, too, this is connected with nominalization (cf. Foley 1991, see
Miestamo 2005: 146–149 for discussion). It is not rare in the world’s languages
that negation requires a nominalized or otherwise non-finite verb (see Miestamo
2005: 73–96, 172–174).
Table 2 recapitulates the effects of negation observed in this section. Since
these phenomena are cross-linguistically rather uncommon, it is not possible to
make quantitative analyses of their frequency and distribution. All relevant
cases found in the survey are mentioned here. I will now move on to discuss
their possible functional motivations.
Table 2: Summary of the effects of negation discussed
Partitive/genitive case used Finnic: Finnish, Estonian, Votic, Liv, etc.;

Baltic: Lithuanian, Latvian; Slavic: Russian,
Ukrainian, Polish; Basque
Non-specific (partitive) determiner used Araki, Mav̈ea, Samoan and various other
Polynesian languages; French
Omission / restrictions on use of article Albanian; Hungarian; Nambikuára
Absence of default determiner possible Kita Maninka; Bantu: Xhosa, Zulu, Bemba,
Bobangi, Kinyarwanda, Luganda
Class marker obligatory Nunggubuyu
Absence of referentiality marker on verb Hdi
Partitive marker on verb Paamese
Effects on focus marking Aghem; Kanuri; Lavukaleve
Effects on alignment Northern Ge: Kayapó, Canela, Apinayé; Yimas
4 Discussion and conclusion

In Section 1, it was briefly mentioned that the partitive of negation and the other
effects of negation on the marking of noun phrases discussed in Section 3 can be
seen as instances of structural asymmetry between affirmation and negation. In

Miestamo (2005), I made a distinction between symmetric and asymmetric nega-
tion according to whether there are structural differences between negatives and
their affirmative counterparts in addition to the presence of negative markers.
Different types of asymmetry between affirmation and negation were discussed,
but the focus was on markings appearing on the verbal and clausal levels, not
within noun phrases. In this paper I have looked at asymmetry phenomena on
the noun phrase level.
Functional motivations behind symmetric and asymmetric negatives can be
understood in terms of the concepts of language-internal and language-external
analogy (cf. Itkonen 2005). Symmetric negation, like any regularity in linguistic
structure, is motivated by language-internal analogy: negatives simply copy the
structure of the affirmative and are thus language-internally analogous to the
structure of the affirmative, ultimately motivated by structural cohesion that
helps processing and storage. Asymmetric negatives copy different aspects of
the functional (semantic and pragmatic) properties of negation that differ from
the functional properties of affirmation, and are thus language-externally analo-
gous to these functional-level asymmetries between affirmation and negation.
In Miestamo (2005: 195–235) I discussed the different functional properties of
negation motivating the different types of asymmetry between affirmatives and
negatives established in that study. In the following, I will address the func-
tional motivations of the noun phrase-level asymmetries observed in Section 3.
Givón (1978) discusses the relationship between negation and referentiality.
Consider the examples in (13).
(13) English (Givón 1978: 72)

a. John met a girl yesterday b. John didn’t meet a girl yesterday
. . . and Fred met one too . . . and Fred didn’t meet one either
. . . and Fred met her too *. . . and Fred didn’t meet her either
The object noun phrase in the affirmative sentence (13a) can get either a nonre-
ferential or a referential reading as shown by the possible ways of continuing
the sentence. In the negative (13b), only a non-referential reading of the object
noun phrase is felicitous, and the continuation compatible with a referential
reading is odd. There is a general tendency for indefinites in the scope of nega-
tion to be non-referential. This tendency is motivated by the discourse context of
negation. Negative sentences are used in contexts in which the corresponding
affirmative is present in one way or another; typically, the speaker assumes that
the hearer believes the corresponding affirmative to be the case. Negatives are
therefore not used to introduce new referents to the discourse. Referential objects
82 Matti Miestamo
are first introduced in affirmatives and appear as definite in negatives. The con-
nection between negation and non-referentiality is also observed, e.g., by
Hopper & Thompson (1980). In their transitivity criteria, negation and non-
referentiality of the object (O non-individuated) are among the correlates of
low transitivity and affirmation and referentiality of the object (O highly individ-
uated) are among the correlates of high transitivity.
Many of the effects of negation observed in Section 3 can be linked to the
connection between negation and non-referentiality. In the case of the Oceanic
determiners this is clear: determiners with the expression of non-referentiality as
their primary function are used with indefinite noun phrases under negation. In
French, too, the determiner de appears with non-referential nouns. In Hdi, a
marker of referentiality is omitted under negation. The absence of the default
determiners in Kita Maninka and the Bantu languages mentioned above are
also connected to referentiality; the default determiner is absent when the noun
phrase is non-referential, and the determiner is used with referential noun
phrases (which are overwhelmingly definite under negation). As to the effects
on article usage in Hungarian, it is the indefinite article conveying a referential
reading that is not used under negation. Referentiality effects are possibly also
behind the asymmetry found in Nunggubuyu, since definiteness and givenness
are mentioned as correlates of the obligatory use of class markers.
As to the motivations of the partitive of negation in Finnic, Baltic, Slavic and
Basque, I want to suggest that the functional connection between negation and
non-referentiality (and ultimately the discourse context of negation) plays an im-
portant role in motivating these cases as well. Partitives refer to a non-individuated
mass, rather than a clearly delimited entity and they thereby provide a useful
form for expressing indefinite non-referential meanings. A connection between
partitive (genitive) case and non-referentiality has been observed in the litera-
ture, e.g., for Russian (see Krasovitsky et al. 2011: Section 2) and Basque (see
above); see also Luraghi & Kittilä (this volume). As to Finnish, the distinction
between total and partial objects in affirmatives is linked with referentiality,
total objects being correlated with referentiality and partial objects with non-
referentiality, and the referentiality of the object is one of the factors that in-
creases the (very low) probability of a total object appearing under negation
(see Almqvist 1987: 26, 156). In addition to referentiality, the use of the partitive
in negatives is also motivated by its quantificational function. In negative sentences
the action is not carried out completely, or not at all, and therefore objects are not
affected by the action completely, or not at all. These motivations taken together
may lead to the grammaticalization of the partitive of negation in some languages.
Another factor connected to the partitive is aspectuality. In Finnish, partial

objects correlate with imperfective aspect and total objects with perfective
aspect. It has sometimes been claimed (e.g., Schmid 1980) that perfective aspect
and negation are incompatible in the world’s languages so that imperfective
aspect would be more likely to appear under negation. In (Miestamo 2005:
180–181) I showed, however, that this does not hold true in a wider typological
perspective, perfective- and imperfective-type aspects being equally likely to be
excluded in negatives (see also Miestamo & van der Auwera 2011). No wider
connection between aspect and negation thus seems to exist in a broad typolog-
ical perspective. Consequently, although clear connections between aspect and
case marking are found in some languages, e.g., in Finnish, the cross-linguistic
facts do not provide support for aspectuality as a motivation behind the partitive
of negation.
This paper examined the interaction between negation and partitive mark-
ing in noun phrases in a typological perspective. The requirement that a case
with a partitive function be used on noun phrases under the scope of negation
is not found outside the familiar European languages. Other effects of negation
on the marking of noun phrases were also observed, with a special focus on
cases that have a connection with referentiality. There is a tendency of indefinite
noun phrases in the scope of negation to be non-referential. The partitive of
negation was also claimed to be motivated by this connection between negation
and non-referentiality. This paper focused on partitives and other effects that
bear a functional similarity to them. In future research, a more comprehensive
view of the effects of negation on the marking of noun phrases is needed.
Acknowledgements
I am grateful to the audience at the partitives workshop at SLE in Vilnius,
September 2010, especially Denis Creissels, Juha Janhunen, and Brigitte Paken-
dorf. I wish to thank the following people who replied to my query on the Lin-
guist List (see Section 2): Bernard Comrie, Greville Corbett, Alexandre François,
Paul Hopper, Larry Hyman, Larisa Leisiö, Claire Moyse-Faurie, Eduardo Ribeiro,
Wolfgang Schultze, John Stewart, Bernhard Wälchli, and Ljuba Veselinova. Spe-
cial thanks to Ljuba Veselinova, Silvia Luraghi and an anonymous referee for
comments on the manuscript. Thanks are also due to the Helsinki Collegium for
Advanced Studies, my working place at the time of conducting the research
behind this paper, and Stockholm University where the paper was revised.
84 Matti Miestamo
References
Almqvist, Ingrid. 1987. Om objektsmarkering vid negation i finskan [On object marking with
negation in Finnish] (Acta Universitatis Stockholmiensis, Studia Fennica Stockholmiensia
1). Stockholm: Almqvist & Wiksell International.
Alves, Flávia de Castro. 2004. O timbira falado pelos Canela Apaniekrá: uma contribuição aos
estudos da morfossintaxe de uma língua Jê [The Timbira spoken by the Apaniekrá Canela:
A contribution to the study of the morphosyntax of a Ge language]. Campinas: University
of Campinas dissertation.
Ambrazas, Vytautas (ed.). 1997. Lithuanian grammar. Vilnius: Baltos Lankos.
Ariste, Paul. 1968. A grammar of the Votic language (Indiana University Publications, Uralic and
Altaic Series 68). Bloomington: Indiana University / The Hague: Mouton.
Bielec, Dana. 1998. Polish: An essential grammar. London: Routledge.
Carlier, Anne. 2007. From preposition to article: The grammaticalization of the French partitive.
Studies in Language 31(1). 1–49.
Corbett, Greville. 1986. The use of the genitive or accusative for the direct object of negated
verbs in Russian: A bibliography. In Richard D. Brecht & James S. Levine (eds.), Case in
Slavic, 361–372. Columbus: Slavica.
Creissels, Denis. 2009. Le malinké de Kita: Un parler mandingue de l’ouest du Mali (Langues et
Linguistique Mandé 9). Cologne: Rüdiger Köppe.
Crowley, Terry. 1982. The Paamese language of Vanuatu (Pacific Linguistics B 87). Canberra:
Australian National University.
Cyffer, Norbert. 2009. Negation patterns in Kanuri. In Norbert Cyffer, Erwin Ebermann, and
Georg Ziegelmeyer (eds.), Negation patterns in West African languages and beyond,
(Typological Studies in Language 87), 71–92. Amsterdam: Benjamins.
Dahl, Östen & Maria Koptjevskaja-Tamm. 2001. The Circum-Baltic languages: Introduction to
the volume. In Östen Dahl & Maria Koptjevskaja-Tamm (eds.), Circum-Baltic Languages,
volume 2: Grammar and typology (Studies in Language Companion Series 55), xv–xx.
Amsterdam: Benjamins.
Dambriunas, Leonardas, Antanas Klimas, and William R. Schmalstieg. 1972. Introduction to
Modern Lithuanian, 2nd revised edn. New York: Darbininkas.
Doke, Clement M. 1961. Textbook of Zulu grammar, 6th edn. Cape Town: Maskew Miller
Longman.
Dryer, Matthew S. 1989. Large linguistic areas and language sampling. Studies in Language 13.
257–292.
Dryer, Matthew S. & Martin Haspelmath (eds.). 2011. The world atlas of language structures
online. Munich: Max Planck Digital Library. http://wals.info/
Early, Robert. 1994. Lewo. In Peter Kahrel and René van den Berg (eds.), Typological studies in
negation (Typological Studies in Language 29), 65–92. Amsterdam: Benjamins.
Erelt, Mati. 2003. Syntax. In Mati Erelt (ed.), Estonian language (Linguistica Uralica Supplemen-
tary Series 1), 93–129. Tallinn: Estonian Academy Publishers.
Fennel, Trevor G. & Henry Gelsen. 1980. A grammar of Modern Latvian, vol. I–III (Slavistic
Printings and Reprintings 304). The Hague: Mouton.
Foley, William A. 1991. The Yimas language of New Guinea. Stanford: Stanford University Press.
Forest, Robert. 1993. Négations: Essai de syntaxe et de typologie linguistique (Collection
Linguistique LXXVII). Paris: Klincksieck.
Frajzyngier, Zygmunt, with Erin Shay. 2002. A grammar of Hdi (Mouton Grammar Library 21).
Berlin/New York: Mouton de Gruyter.
François, Alexandre. 2002. Araki. A disappearing language of Vanuatu (Pacific Linguistics 522).
Canberra: Australian National University.
Givón, Talmy. 1978. Negation in language: Pragmatics, function, ontology. In Peter Cole (ed.),
Syntax and semantics, vol. 9, Pragmatics, 69–112. New York: Academic Press.
Guérin, Valérie. 2007. Definiteness and specificity in Mav̈ea. Oceanic Linguistics 46(2). 538–
553.
Hakulinen, Auli, Maria Vilkuna, Riitta Korhonen, Vesa Koivisto, Tarja Riitta Heinonen & Irja Alho.
2004. Iso suomen kielioppi [The big grammar of Finnish]. Helsinki: Suomalaisen Kirjalli-
suuden Seura.
Hakulinen, Lauri. 1961. Suomen kielen rakenne ja kehitys [The structure and development of
Finnish], 2nd, revised edn. Helsinki: Otava.
Haspelmath, Martin, Matthew S. Dryer, David Gil & Bernard Comrie (eds.). 2005. The world
atlas of language structures. Oxford: Oxford University Press.
Heath, Jeffrey. 1984. Functional grammar of Nunggubuyu. Canberra: Australian Institute of
Aboriginal Studies.
Holst, Jan Henrik. 2001. Lettische Grammatik. Hamburg: Helmut Buske Verlag.
Honda, Isao. 1996. Negation: A cross-linguistic study. Buffalo: State University of New York at
Buffalo dissertation.
56. 251–299.
Hualde, José Ignacio & Jon Ortiz de Urbina. 2003. A grammar of Basque (Mouton Grammar
Library 26). Berlin/New York: Mouton de Gruyter.
Hyman, Larry M. 2010. Focus marking in Aghem: Syntax or semantics? In Ines Fiedler and Anne
Schwarz (eds.), The expression of information structure: A documentation of its diversity
across Africa (Typological Studies in Language 91), 95–116. Amsterdam: Benjamins.
Itkonen, Esa. 2005. Analogy as structure and process: Approaches in linguistics, cognitive
psychology and philosophy of science (Human Cognitive Processing 14). Amsterdam:
Benjamins.
Koptjevskaja-Tamm, Maria & Bernhard Wälchli. 2001. The Circum-Baltic languages: An areal-
typological approach. In Östen Dahl & Maria Koptjevskaja-Tamm (eds.), Circum-Baltic lan-
guages, volume 2: Grammar and typology (Studies in Language Companion Series 55),
615–750. Amsterdam: Benjamins.
Krasovitsky, Alexander, Matthew Baerman, Dunstan Brown & Greville G. Corbett. 2011. Chang-
ing semantic factors in case selection: Russian evidence from the last two centuries.
Morphology 21(3). 573–592.
Kroeker, Menno. 2001. A descriptive grammar of Nambikuara. International Journal of American
Lazdiņa, Terẽza B. 1966. Latvian (Teach Yourself Books). London: The English Universities
Press.
Metslang, Helle. 2001. On the developments of the Estonian aspect: The verbal particle ära.
In Östen Dahl & Maria Koptjevskaja-Tamm (eds.), Circum-Baltic languages, volume 2,
Grammar and typology (Studies in Language Companion Series 55), 443–479. Amsterdam:
Benjamins.
86 Matti Miestamo
Miestamo, Matti. 2005. Standard negation: The negation of declarative verbal main clauses
in a typological perspective (Empirical Approaches to Language Typology 31). Berlin/New
York: Mouton de Gruyter.
Miestamo, Matti & Johan van der Auwera. 2011. Negation and perfective vs. imperfective
aspect. In Jesse Mortelmans, Tanja Mortelmans & Walter De Mulder (eds.), From now to
eternity (Cahiers Chronos 22), 65–84. Amsterdam/New York: Rodopi.
Moravcsik, Edith A. 1978. Case marking of objects. In Joseph Greenberg (ed.), Universals of
human language, volume 4: Syntax, 250–289. Stanford: Stanford University Press.
Mosel, Ulrike & Even Hovdhaugen. 1992. Samoan reference grammar. Oslo: Scandinavian
University Press.
Naughton, James. 2005. Czech: An essential grammar. London: Routledge.
Nedyalkov, Igor. 1994. Evenki. In Peter Kahrel & René van den Berg (eds.), Typological studies
in negation (Typological Studies in Language), 1–34. Amsterdam: Benjamins.
Newmark, Leonard, Philip Hubbard & Peter Prifti. 1982. Standard Albanian: A reference gram-
mar for students. Stanford: Stanford University Press.
Oliveira, Christiane Cunha de. 2005. The language of the Apinajé people of Central Brazil.
Portland: University of Oregon dissertation.
Pakendorf, Brigitte. 2007. Contact in the prehistory of the Sakha (Yakuts): Linguistic and
genetic perspectives (LOT Dissertation Series 170). Utrecht: LOT.
Payne, John. R. 1985. Negation. In Timothy Shopen (ed.), Language typology and syntactic
description, vol. I: Clause structure, 197–242. Cambridge: Cambridge University Press.
Pugh, Stefan M. & Ian Press. 1999. Ukrainian: A comprehensive grammar. London: Routledge.
Schmid, Maureen Alicia. 1980. Co-occurrence restrictions in negative, interrogative, and condi-
tional clauses: A cross-linguistic study. Buffalo: State University of New York at Buffalo
dissertation.
Silva, Maria Amélia Reis. 2001. Pronomes, ordem e ergatividade em Mebengokrê (Kayapó).
Campinas: University of Campinas MA Thesis.
Swan, Oscar E. 2002. A grammar of contemporary Polish. Bloomington: Slavica.
Taraldsen, Knut Tarald. 2010. The nanosyntax of Nguni noun class prefixes and concords.
Lingua 120. 1522–1548.
Terrill, Angela. 2003. A grammar of Lavukaleve (Mouton Grammar Library 30). Berlin/New York:
Mouton de Gruyter.
Wade, Terence. 2000. A comprehensive Russian grammar, 2nd revised and expanded edn.
(Blackwell Reference Grammars). Oxford: Blackwell.
Wolff, H. Ekkehard. 2009. The impact of clause types and focus control, aspect, modality, and
referentiality on negation in Lamang and Hdi (Central Chadic). In Norbert Cyffer, Erwin
Ebermann & Georg Ziegelmeyer (eds.), Negation patterns in West African languages and
beyond (Typological Studies in Language 87), 21–56. Amsterdam: Benjamins.
II Uralic languages
Anne Tamm
3 The Partitive Concept versus Linguistic
Partitives: From Abstract Concepts to
Evidentiality in the Uralic Languages1
Finnic and Sámi (Uralic, Finno-Ugric) languages have a morphological case that
is referred to as the partitive. The meaning of the dedicated partitive case, how-
ever, diverges from the generally assumed partitive concept of part-whole rela-
tionships. Instead, the meaning of the partitives is either bleached, or it has
developed to express other categories, such as aspect. Since the partitive also
combines with non-finites, further developments have resulted in grammatical
markers that are based on the partitive but belong to epistemic modality and
evidentiality. As for the part-whole concepts, the Uralic languages tend to express
them rather by juxtaposed bare nouns, elatives, or ablatives than by morpholog-
ical partitives. The article places this mismatch between form and meaning in a
wider context of using abstract concepts for comparing grammatical categories
across languages. Examining the grammaticalization of TAM categories in Esto-
nian (Finnic) and the partitives among the rich system of Uralic separatives, the
analysis employs two terms: the Partitive Concept – a heuristic tool and a basic
concept used for comparison – and Linguistic Partitives, dedicated morphologi-
cal partitive cases and their language-specific further developments. Linguistic
Partitives are described via two (or more) concepts. Firstly, they are described
via the concepts that they express, such as aspect or evidentiality. Secondly,
they are also described in terms of the Partitive Concept. This multiple linking
to concepts, the Partitive Concept and for instance TAM concepts, is intended
to guarantee that the description of language-specific categories, such as Esto-
nian aspect or evidentiality, reflects the aspectual and evidential nature of these
categories as well as the overarching system of partitivity in the Finnic grammat-
ical system. In addition, the Partitive Concept is constructed to serve as a suit-
able basis for further psycholinguistic testing with the goal of finding out if the
abstract concept corresponds to a cognitively motivated linguistic category.
1 Many thanks to Tuomas Huumo and Silvia Luraghi for creating a lively forum for discussing
partitives across languages and theories. I am indebted to an anonymous reviewer and Silvia
Luraghi for insightful discussions and to Leyla Caglar and Tommaso Claudi for their work with
the manuscript.
90 Anne Tamm
Keywords: source, separative, concepts, grammatical categories, cross-linguistic

comparison, Uralic.
1 Introduction
Partitive is among the most theory-dependent terms for a case in modern lin-
guistics. This paper targets some of the confusing issues concerning the parti-
tives from the Uralic languages. The main aim is to reduce the uninformed cross-
talk between linguists of different traditions and help them see where their
coverage of the term “partitive” converges or diverges. What is the uniform con-
ceptual toolkit to tackle the examples discussed under the term “partitive” in
(1)?
(1) a. Russian
čaška čaj-u
cup[ F. NOM ] tea- GEN 2
‘a cup of tea’
b. Hungarian
gyerek-e-i-m-ből a leg-fiatal-a-bb
child- LK- PL- POSS .1 SG - ELA DEF SUP-young- LK-CMPR
‘the youngest of my children’
c. Finnish
ilman rahaa
without money.PAR
‘without money’
d. Mari sõ-i õuna.

M[NOM ] eat-PST 3 S apple.PAR
‘Mary was eating an apple.’
e. Mari tule-va-t koju.

M[NOM ] come-PERS . PRS . PTCP- PAR home.ILL
‘Allegedly/reportedly, Mary will come home.’
Scholars referring to the category expressed by the Russian -u on čaj-u in

čaška čaj-u ‘a cup of tea’ in (1a) may understand why the category expressed by
the Hungarian elative -ből in the example gyerekeim-ből a legfiatalabb ‘the
youngest of my children’ (1b) should be referred to as “partitive”, but they will
The Partitive Concept versus Linguistic Partitives 91
find it problematic to extend the category to the Finnish partitive -a in the com-
plement of ilman rahaa ‘without money’ (1c). Scholars concentrating on the
Finnic partitive forms, for instance, the partitives occurring as part of the object
(1d) and predicate (1e) in the Estonian, ‘Mary was eating an apple’ and ‘Allegedly/
reportedly, Mary will come home’ find it unintuitive to refer to genitives and ela-
tives as “the partitive”. The main puzzle is thus how to compare these and many
other partitives in relation to each other. The main solutions proposed in this
paper follow in (i).
(i) A distinction between “Linguistic Partitives” and “Partitive Concepts” in de-
scribing the partitive phenomena in Uralic is useful for better understanding
of the partitive phenomena. Distinguishing the two helps to compare the
mismatches between the partitive form and the part-whole meaning across
the individual languages.
– The Partitive Concept is an abstract concept that serves for comparing
the semantics of grammatical forms to the “part-of-N” (1b) and “amount-
of-N” (1a) concepts.
– The Partitive Concept comprises two metonymically related subconcepts:
the partitive (N-of-the N, 1b) and the pseudopartitive (N-of-N, 1a).
– A Linguistic Partitive is a grammatical form that is conceptually related
to the meaning of the Partitive Concept. The partitive cases have developed
their specific semantics and pragmatics in each Uralic language where
the case appears.
– The Linguistic Partitive is divided into functional (e.g., aspectual, 1d,
evidential, 1e) and structural categories (e.g., complement case, default
case 1c), depending on the semantics of the partitive in the structure of
the language at hand.
The term “partitive” is most frequently applied to a type of Indo-European

genitive (Koptjevskaja-Tamm 2001: 525). The term typically covers phenomena
that are not morphological partitives, but still similar to (1a), the Russian -u on
čaj-u in čaška čaj-u ‘a cup of tea’ or to the Hungarian elative -ből in gyerekeim-
ből a legfiatalabb ‘the youngest of my children’. Other aspects of the forms
referred to as the partitive have enjoyed considerable attention in theoretical
linguistic literature due to the special thematic relationship that the partitive
encodes between the predicate and the object as in (1d) (Krifka 1992). Hoeksema
(1996) contains several formal and generative papers on the partitive. De Hoop
(1996) gives an influential account in terms of quantifiers, and de Hoop (1998)
offers an extensive overview of the theoretical approaches to the phenomenon
and a thorough bibliography. At the interface between cognition and language,
92 Anne Tamm
Jackendoff (1991) relates cognition and theoretical linguistics with the concept of
partitivity and parts, and interpretational peculiarities of some partitive phenom-
ena via discourse and pragmatics. In psycholinguistics, Reed (1991) discusses
several constraints of the partitive constructions, arguing that these constraints
stem from discourse requirements.
Some of these influential typological, theoretical and cognitive accounts
mention a lesser known, but nevertheless theoretically and typologically intrigu-
ing phenomenon: a dedicated partitive case. The Uralic languages, more specifi-
cally, the Finnic and Sámi languages are a whole group of related languages
where there is a morphological partitive. This case marks objects, subjects, pre-
dicatives, other complements, and even measure adjuncts. It combines with
various non-finites (nominalizations) that vary in their degree of finiteness. The
combinations of nonfinites and the partitive case formant are either transparent
or opaque, and are combined productively or have completely grammaticalized
as intersubjective markers. Especially several recently analyzed Estonian parti-
tive case phenomena are an interesting object of study between form, meaning,
cognition, and communication in general. Modern Estonian has a wide range of
furher developments of the partitive, covering distinct functional categories such
as aspect or evidentiality. The partitive evidential is an intersubjective marker
the understanding of which requires a Theory of Mind. The partitive has also
developed a wide variety of desemanticized uses and is a structural case
because it behaves as a formal, semantically opaque complement case.
There are many questions this contribution wishes to address. What is the
relationship between the idea we have about partitivity and the various occur-
rences of a dedicated partitive form? How are these occurrences related? How
does the partitive distinguish itself from other similar cases, such as the accusa-
tive, or the source cases? Is there a uniform partitive concept that can be taken
as a reference point for comparisons? How should we map the partitive mean-
ings to the partitive forms across languages? What is the function of the partitive
in communication? A wider perspective on partitive and partitive-like concepts
in the Uralic languages provides valuable material for further studies into the
relationship between language, communication, and the cognition of concepts.
Up until now, the description has been fragmented for the specific purposes
of particular studies. This article wishes to give a representative overview of
the properties of the morphological partitives and related cases in the Uralic
languages.
The partitive is by now a well-studied grammatical phenomenon in Uralic
linguistics. It has been discussed in grammar books of Finnic languages and
general collections on Uralic languages such as Abondolo (1998), or particular
descriptive or diachronic studies on the partitive case in Finnic, such as Tveite
(2004) on Livonian, Ritter (1989) on Veps, or Denison (1957) on Finnish. There

are informative works on the semantic or syntactic conditions of the partitive
case in Finnic languages, such as aspect, indefiniteness, or grammatical rela-
tions (e.g., Heinämäki 1984, Rajandi and Metslang 1979, Metslang 2001, 1994,
Hiietam 2003). In-depth studies on the partitive across the Finnic languages
include parallel text-based quantified comparisons (Lees 2005) and diachronic
investigations (Larsson 1983). Some questionnaire-based targeted typological
research on Uralic languages includes the topic of the partitive, as the edited
volume on negation (Miestamo et al. forthc.). More theory-oriented and less data
oriented approaches are almost exclusively on Finnish: cognitive (e.g. Huumo
2009), formal (Kiparsky 1998), generative (Brattico 2011, Vainikka and Maling
1996), or Optimality Theoretic (Kiparsky 2001, Anttila and Fong 2000). Several
phenomena where the Estonian partitive is central are discussed from typologi-
cal, cognitive and comparative generative and lexical viewpoint in the research
of Tamm (e.g., Tamm 1999, 2012c). Master’s theses or articles and monographs
based on doctoral dissertations discuss the partitive in Estonian (e.g., Tauli
1968, 1980, Nemvalts 2000, Vaiss 2004, Metslang 2007). This volume also con-
tains analyses of the Estonian partitive phenomenon, mainly, the subjects, in a
typological framework (Metslang, this volume, Huumo and Lindström, this
volume).
The current understanding in typology, most recently argued for in Haspel-
math (2010) is that one cannot do comparative and language-specific analyses
simultaneously by using cross-linguistic categories. Instead, comparative linguists
should try to reach an agreement on concepts that can serve as comparison.
As opposed to previous approaches, my article wishes to improve the current
understanding of this issue by comparing the partitive concepts and various
uses of separative (source) cases with a dedicated partitive form. The expression
of the partitive concept in Uralic separatives or source cases will be compared
to the concepts and functions expressed by the Finnic partitives. Most of the
examples are taken from a language that has the highest type and token fre-
quency of morphologically encoded partitive phenomena in spoken as well as
written registers – Estonian. Any theoretical bias is deliberately avoided in order
to be “legible” for linguists across frameworks. A classical generative and typo-
logical mapping or linking problem is central in this paper. Methods from several
frameworks are applied in order to reach a better understanding of the partitive.
Hopefully, this wall-to-wall linguistic approach contributes more to the inter-
disciplinary study of the phenomenon than a strictly specialized analysis, carried
out within the confines of a single theory and methodology.
More specifically, this paper first applies the method that is typical for typolo-
gists, establishing the correpondences from a concept to forms. Then it addresses
94 Anne Tamm
the data with descriptive linguists’ methods, linking forms to meaning and use
and, finally, makes use of pragmatics and cognitive semantics to explain the
current polyfunctionality, cross-linguistic variation, and diachronic aspects of
the partitive phenomena. Section 2 introduces the conceptual coverage of the
Partitive Concept and points at data that are not covered by the Partitive Concept.
Section 3 classifies the partitive forms that do not always match the Partitive
Concept under Linguistic Partitives and introduces the term Linguistic Partitive
in more detail. Section 4 shows that the Uralic languages are a suitable test-bed
for studying various kinds of the partitive, with the Finnic, Skolt and Inari Sámi
languages, which all have a dedicated partitive case, and that have both func-
tional and structural instances of the partitive. The section also describes the
empirical data and discusses some of the cases that express the Partitive Con-
cept in the Uralic languages. Section 5 concentrates on the structure and the
conceptual content of one instance of the linguistic partitive: the Estonian
partitive evidential. The conclusion can be found in Section 6.
2 The conceptual coverage of the Partitive

Concept
The discussion of this section is about a general conceptual notion of the parti-
tive. It also makes reference to language-specific conceptualizations and the
desemanticized instances of the partitive to emphasize the distinction.
2.1 Comparative concepts and psychologically real concepts

The labeling of the categories and the problem of language-specific versus more
universal heuristic concepts has been subject of debates and research in typology,
recently taken up in a number of articles by Haspelmath (2007, 2010) and works
cited in these two sources. The discussion of the Uralic partitives will relate to
the current debate on comparative concepts versus descriptive categories, as
there is a discrepancy between the forms and meanings across the various
Uralic languages. This paper envisages the Partitive Concept as a sort of com-
parative concept, distinguished from other comparative concepts in that a cogni-
tive basis is also intended to be part of the concept (following Jackendoff 1991).
The Linguistic Partitives of languages and language groups are diachronic descrip-
tive categories, in the sense that they form a diachronically motivated partitive
category in each of the languages. Synchronically, the Linguistic Partitives

express a number of other concepts such as complementhood, subjecthood,
objecthood, aspect, epistemic modality, indefiniteness, or evidentiality.
I aim at constructing a cognitively plausible concept for comparison, whereas
typologists face the problem of the comparability of languages via artificial or
language-specific concepts. Haspelmath (2010) argues that descriptive categories
and comparative concepts should be distinguished. More specifically, descriptive
categories are the categories of particular languages, and comparative concepts
are constructed for cross-linguistic comparison. The latter are created for the pur-
pose of comparison, since the criteria for category-assignment are different from
language to language, and so descriptive categories cannot be equated across
languages. Linguists compare languages and need a toolkit for describing the
correspondences, which is why they identify some cross-linguistic categories.
The status of the latter is also a matter of debate. Many typologists do not support
the idea of cross-linguistic formal categories. Newmeyer (2007: 133) criticizes
typology, pointing out the contradiction between rejecting the idea of cross-
linguistic formal grammatical categories, and the policy of utilizing cross-
linguistic formal categories, which is opted for in many entries of World Atlas
of Language Structures (WALS, Haspelmath et al. 2005), despite the position
of two of the editors that such categories do not exist (Newmeyer 2007: 138).
Haspelmath (2007, 2010) argues that the categories used in the WALS are rather
comparative concepts, not cross-linguistic formal concepts. Haspelmath suggests
that since semantic analysis provides a method for determining meaning, it can
be used for making categories of languages comparable, even on a large scale.
What are comparative concepts in typology according to Haspelmath? Com-
parative concepts are concepts created by comparative linguists, for the specific
purpose of cross-linguistic comparison. Comparative concepts are designed to
be universally applicable, and they are defined on the basis of other universally
applicable concepts: universal meanings and universal formal notions, or on the
basis of other comparative concepts. Comparative concepts are more or less
suited for the task of permitting cross-linguistic comparison; each comparative
linguist can make their own comparative concepts. Comparative concepts are
heuristics that help to identify comparable phenomena across languages and to
formulate cross-linguistic generalizations. Comparative concepts have to be uni-
versally applicable in typology. Language typology is made possible because
of comparative concepts that serve as tertia comparationis, as argued for by
Haspelmath (2010).
The Partitive Concept is also a concept created for comparison, and it is
linked to several language-specific categories. The different levels of semantics
96 Anne Tamm
are seen as follows: the Partitive Concept is a comparative concept, an abstrac-

tion and a standard of comparison, and thus a heuristic tool. Language-specific
categories expressed by morphological partitives are called Linguistic Partitives.
Linguistic Partitives are described along two (or more) comparative concepts.
Firstly, Linguistic Partitives are described via other concepts, such as subject,
object, complement, evidentiality, aspect, genericity, or definiteness. Secondly,
Linguistic Partitives are also described in terms of the Partitive Concept. This
multiple linking to concepts, the Partitive Concept and other concepts, is intended
to guarantee that the description of the particularly Finnic categories such as
Finnish objects, Votic aspect, or Estonian evidentiality is linked to objecthood,
aspect, and evidentiality across languages and also to the overarching system
of partitivity in the Finnic grammatical system. The semantic features of the
Finnic objects, aspect, and evidentiality are related to the comparative concept
either directly or indirectly, by multiple metonymical or metaphorical extensions.
In addition, there are instances where the conceptual content has disappeared
altogether (it has bleached), and the partitive encodes a syntactic function of a
complement or functions as a default.
However, it should be emphasized that there is a difference between the
Partitive Concept and a comparative concept. The partitive concept that I con-
struct for comparative purposes is constructed in a way that it could serve as
a hypothesis for cognitive testing. Comparative concepts are like measuring
units – arbitrary, but indispensable units for comparing objects of study. As
opposed to comparative concepts that are created by comparative linguists for
the purpose of formulating readily testable cross-linguistic generalizations, the
Partitive Concept is created in order to be able to test at a later stage whether
the concept is also psychologically existent. While comparative conceps are like
liters, meters or kilograms, the Partitive Concept could be understood more like
items in the analog magnitude system (Carey 2009). This means that independ-
ently of language, there is a partitive relation that is perceived differently from
other relations, for instance, a possessive relation. Linguists can construct
nonlinguistic stimuli (pictures, videos, artificial social situations) and observe
speakers’ reaction in a systematic way. The present paper has the aim of creating
a suitable concept for further investigations into the cognitive properties of
quantificational and spatial-separative linguistic concepts such as the partitive.
One of the most intriguing questions in modern cognitive science concerns
how the human prelinguistic perceptual concepts are related to linguistic and
language-specific, more refined concepts (Carey 2009). A basic concept that has
multiple uses with highly varying domains of application in natural language –
from spatial to intersubjective – is a promising area for discovery in cognitive
science as well.
2.1 The labeling problem of some instances of the partitive in

synchronic linguistics
It would be perfectly justified to refer to the case that is referred to as “the par-
titive” in the Estonian grammar rather as “the accusative”. After all, it is the case
that marks most of the objects in Estonian transitive clauses, as in (2).
(2) Estonian
Mari armasta-b Jaanus-t.
M[NOM ] love-3SG J-PAR
Only because of the awareness of the historical origin of this case morpheme,
which typically but not always ends in -t, one of the object cases, “the accusa-
tives”, is referred to as “the partitive” and not as “the accusative number 1”. The
Estonian partitive is thus a kind of accusative in its function of being the pre-
dominant object case. The peculiarity of the Estonian object case system is that
there are two object cases, the total and the partitive. The object case alternation
has been linked to various semantic and pragmatic distinctions, predominantly to
the aspectual ones in the linguistic literature since the link between the Finnish
partitive and the Russian imperfective was established (Dahl and Karlsson 1975).
The total and the partitive are semantic as well as structural-grammatical cases
(cf. Kipasky 1998). Since historical records are available, the more frequent
object case is not referred to as the accusative but as the partitive.
Linguists can agree or disagree about labeling one of the many structural
or semantic relationships in language structure as “the partitive”. The term
Partitive Concept is no less a matter of convention. The example in (2) has little
to do with anything related to parts. The object John is marked with the partitive
case, but the sentence does not express that Mary loves only some concrete or
abstract part of John, e.g. his eyes or his good manners, as opposed to someone
else, whom she would love in total, regardless of his parts. In order to keep
apart the diachronic motivation behind the changes that have led to the current
polyfunctionality in the synchronic situation, it is useful to make a distinction
between the assumed original semantics of the formative that is now referred
to as the partitive and its later developments. The example in (2) is thus not an
instance of the core meaning, the Partitive Concept, but an instance of a typical
Linguistic Partitive in the Finnic languages.
(ii) A typical Linguistic Partitive (the morphological partitive case) in the Finnic
languages covers a wider spectrum of meanings and functions than the
Partitive Concept.
98 Anne Tamm
What is the core partitive meaning, then, and is it possible to establish it on the
basis of the variety that is covered by the morpheme called “the partitive”? Do
we need one more grammatical term, a new label; why cannot we just have the
labels Accusative 1 and Accusative 2 for Finnic object cases, as in some freshly
described languages with no written historical records and grammaticographic
traditions? The rationale behind finding a suitable label for a phenomenon is
the existence of a grammatically encoded distinction or an expression in a lan-
guage, preferably in many languages. Ideally, the distinction – the concept or
relationship – would make an impression of being basic, real, and clear-cut.
The Partitive Concept can be understood as the term that stands for the
most primitive or elementary concept of the partitive. The core meaning is rela-
tional. More specifically, it belongs to spatial relationships and identity conditions.
(iii) In terms of spatial relationships, the Partitive Concept instantiates a sepa-

rative relationship of an individual or matter to another individual or
matter.
(iv) In terms of identity, the partitive instantiates the same kind identity (not
difference or similarity).
The Partitive Concept stands for separation from identical matter. The fol-
lowing passages will go through some instances that are typical or possible
with the European genitive-based partitives, but not typical or possible with the
Uralic ones.
There is an illustrative example from Estonian culture that suits the expla-
nation of the partitive concept. The figure indicated on a website under example
(3) is the most famous and highly ambiguous political cartoon in the Estonian
art history. It nicely illustrates the essence of the partitive relationship. It has a
cognitive linguistic pointe based on the visual image and the partitive concept in
the nominal phrase Sitta kah. One reading of Sitta kah ‘Some manure, too’ (3a)
corresponds to a literal Partitive Concept, and the other reading is idiomatic, a
colloquial pragmatic phrase ‘I don’t give a shit/damn. / Who cares’ (3b).2
2 There are several other political interpretations that I do not discusss here, related to the
protest movement referred to as the Phosphorite War in spring 1987. Also, the cartoon marks
a tipping point in history, see the details in Lõhmus (2004). The more prevailing English title
of the cartoon deviates from the translation in order to convey the spirit of the anti-Soviet
sentiments in Estonia more adequately to the outside world: “Just shit”. Estonians wanted the
outsiders to know that they were treated by the Soviets as “just shit”.
(3) Sitta kah. . .

manure. PAR too
a. ‘(Let us take/throw some) manure (to the field), too.’
b. ‘I don’t give a shit/damn. / Who cares.’
c. ‘Just shit.’ (The English title for outsiders, which can be googled.)
The Partitive Concept, as in: ‘throw (some of the) manure from the cart onto
the field’ (Pärn 1987: 16) can be viewed on the Wikipedia page at: http://en.
wikipedia.org/wiki/File:Priit_P%C3%A4rn_Sitta_kah.jpg (9 November 2012).
I provide now some background knowledge to explain why this typical instance
of Soviet “writing between the lines” has found its way to a book on the parti-
tives. This cartoon plays with the perception of the viewer in categorizing the
parts of a whole as identical or different, and the interpretational ambiguity in
the partitive noun phrase. First of all, please look at the lumps that are taken
from the heap of manure on the cart and thrown onto the field. If you see
one bigger lump of manure and two smaller lumps, then you get the explicit
message exactly as any innocent Soviet citizen at the end of the eighties would
have. The political cartoon was published in May 1987 in the time of deep stag-
nation and much before the Berlin wall came down in 1989, and even before
Reagan’s speech in Berlin (“Mr Gorbachov, tear down this wall”) took place in
June 1987. So it was crucial to create an image that, after falling on the retina of
the viewer, patterned differently for the two opposite ideological camps because
of the related visual associations. For those whose memory was imprinted with
images of art depicting humans and animals at work, the image could have been
interpreted as applauding a weary but heroic collective farm worker of the
Peasant Class, in his efforts to fertilize the field in the face of adversity. Thus, it
is an impeccable picture for the Party leadership and its censorship (if you write
sitta kah in Google Translate, it would give you an innocent ‘of shit, too’). This
is what the censor or a random Soviet citizen would get: the picture of an un-
intellectual peasant working hard, suitable to boost the waning morale of the
starving people.
However what is written and drawn between the lines? On the basis of the
nominal partitive case marking of the text (“some of the manure, too”), the
separated mass is of the same kind with the rest of the manure mass on the
cart.3 The shape of the manure that is thrown out corresponds to the contours of
Estonia. The message between the lines is readable only by those who recognize
3 For some, the horse lacks one eye and, therefore, pulls the rickety carriage of the party to a
wrong direction, and for others, the peasant looks obviously dumb, uninspired and unelegant
to assume responsibility for any progress.
100 Anne Tamm
the difference in the identity between the mass on the cart and the separated
object that is thrown on the field. The separated object is clearly Estonia. In the
eyes of the more informed beholder there is the violation of the identity condi-
tion of the partitive, which causes the exclusion of the reading (3a) ‘(Let us take)
some manure (to the field), too.’ For the separation of objects belonging to dif-
ferent kinds, a native speaker of Estonian would use the elative instead of the
partitive. Instead of “take some of the manure” – something of the same kind,
partitive case – the action is interpreted as “take something from the manure” –
something of a different kind. Therefore, the literal meaning is not the default
one for those who recognize the shape. Only the idiomatic reading of the collo-
quial Sitta kah “I don’t give a shit / Who cares” and not “some manure, too” is
activated for those who see the contours of Estonia instead of the three blots of
manure. In the context of several countries seeking independence of the Soviet
Union, the disgraceful heap of manure transforms into the Soviet Union. The
half-blind horse pulling a half-wheeled cart turns into the senseless state
machinery, and the unelegant and uninspired peasant becomes the mindless
party leadership. The field is the free world. The message between the lines
is the wishful thinking of those who would soon be tearing down the wall
in Berlin, coaxing the mindless leaders: “just throw some of this manure out,
anyway, this cart is too heavy, take this lump from the manure for instance,
who cares (sitta kah), throw it well, far onto the field, please!”
The moral of the story is that the elative separative is less specific about the
identity of the object or matter that is separated, since the identity of it can be
different; the partitive separative requires identity between the parts and wholes.
Note that the picture emphasizes motion or dynamicity. An entity that is per-
ceived as having a clear identity (manure) of its own is divided in two in the
course of movement. It is the question what is the kind of the two newly created
objects is, that is, how the two parts are categorized: manure + manure or
manure + something of a different kind. The two separated parts are transformed
into new objects. The two entities differ in shape and size from each other and
from the original entity.4 In many other respects, however, the two new entities
belong to the same kind. The parts share all other properties, such as color,
texture, smell, and most plausibly, function related to human use (something
to drink, something to eat, something to feed animals with, something to make
clothes of, something to fertilize the fields with) and so forth.
In static relationships, parts of matter with identical properties cannot be
perceived easily; in order to be perceivable, the part should be somewhat differ-
ent from the whole. In order to be perceived as a part, the entity must be focused
4 In some cases, they may differ in number (as in an instance of strawberries in strawberry-
picking).
on or highly salient in its essential and perceivable properties. Focusing is con-

troversial, because the focused part would be categorized away as something of
a different kind located inside or on top, or in the close vicinity. Therefore, if the
identity of the parts and the whole is the same, then the salience of the parts has
to be created in one or another way, since the image of the part as such has to
be perceived as a part belonging to the same kind. Note again how the picture
emphasizes the dynamicity of the separation of the part. It is problematic to
envisage proper partitive relationships without motion. In the course of actions
of removing, some of the matter of the same kind becomes a part without other
salient features interfering with the categorization of the part.
Possibly, separation of some matter from matter of the same kind is a con-
cept that humans (and possibly, animals) can form a category of. Moreover, it
is likely that it can be expressed by all human languages. These plausible
hypotheses remain to be tested. The core semantics of the Partitive Concept
is “part of N that can be referred to as N”. As a matter of regular polysemy –
metonymy – the referent of a part can become analyzed as a whole, and the
partitive semantics corresponds to “amount-of-N”, referring to a part or quantity
out of a group or amount of substance.
(v) a. “separable part of N that belongs to the same kind with N” (4a) !
b. ! “amount of N” (metonymic extension of v-a) (4c)
Whereas the core partitive meaning, as in (v-a), “part-of-N” is relational, the

basic extension of this meaning, (v-b), “amount-of-N” is not. In the Hungarian
sentence (4-a), the youngest of my children is in a relation to other individuals
referred to as my children, while in the Hungarian sentence (4b) (“a glass of
wine”), the wine does not necessarily belong to any larger amount of wine, it is
just an amount. I set the partitive forms of the forms carrying the partitive mean-
ing boldface in this paper. Sentence (4a) illustrates (v-a), part-of-N, and sentence
(4c) illustrates the meaning (v-b).5
5 The reading in (v-a) corresponds to real partitives and the reading (v-b) to pseudopartitives in
the typological literature, as in Koptjevskaja-Tamm (2001). Pseudopartitives are generally taken
to refer to an amount or quantity of some (indefinite) substance (e.g., a cup of tea, a package of
butter, a box of chocolates). Real partitives refer to a part/subset of a (definite) superset (e.g.,
a hot cup of this green tea, a large package of this Danish butter, a small box of these
chocolates). Languages differ in terms of how they encode real partitives and pseudopartitives.
Koptjevskaja Tamm (2001) establishes that there is considerable variation between languages
in the grammatical marking of the substance-denoting expression in partitive and pseudo-
partitive constructions, ranging from case inflections to prepositions to zero marking, denoting
typically conventionalized measures (a litre of x), fractions (a slice of x), quanta (a lump of x),
collections (a group of x), or forms (a pile of x).
102 Anne Tamm
(4) Hungarian
a. gyerek-e-i-m-ből a leg-fiatal-a-bb
child- LK- PL- POSS .1 SG - ELA DEF SUP-young- LK- CMPR
b. az egyik belől-ük
DEF one[ NOM ] ELA-3 PL
‘one of them’
c. egy pohár bor

INDF glass[ NOM ] wine[NOM ]
‘a glass of wine’
Examples (4a) and (4b) illustrate the core meaning of the Hungarian Partitive
Concept: the youngest of my children, one of them. The construction with the
elative represents the core meaning, because it refers to an individual, “one
child of mine”, namely, the youngest, or “one of them”, that has the same iden-
tity with other members of the set where it belongs to (the set “my children”). It
is not an “amount-of partitive”: I cannot refer to my child as the youngest of my
children or as one of my children if I do not have more children than one or
two.6 All of my children form a whole, and the youngest one is separated from
them in conversation as part of this whole.
The relationships between parts and wholes are different with the wine in
the glass and wine. The wine in the glass can exist without any more wine in a
bottle, a cask or a barrel. The wine as expressed in the bare nominal construc-
tion is not necessarily a proper part of another amount of wine, that is, anything
else of the same identity. It is an “amount-of partitive”, because the wine in the
glass can exist without being in relation to any other amount of wine, that is,
without the existence of anything else of the same kind. The partitive noun
rather refers to a kind.
3 The conceptual coverage of a Linguistic

Partitive
If the Partitive Concept is a heuristic tool to get a grip on the ways a specific
meaning content can be expressed in a wide variety of languages, then what is
the term Linguistic Partitive good for? In the Estonian example (3), Sitta kah
6 Perhaps the infelicity is evoked by a scalar implicature. If you have one child, it is the youngest
and the oldest child, but it is not felicitous to evoke a comparison when there is none, and the
superlative is infelicitous if there are no more than two children.
‘some manure, too,’ there is a straightforward link between the morphological

partitive and the Partitive Concept. The morphological partitive matches the
definition in (v-a), encoding the concept of “separable part of N that belongs to
the same kind with N”. The morphological partitive in (1d), Mari sõ-i õuna M
[NOM ] eat-PST 3SG apple.PAR ‘Mary was eating an apple’ matches the definition in
(v-b), encoding the concept “amount of N”, which is a metonymic extension of
(v-a). These examples are not sufficient to argue for the necessity of extra items
in the conceptual toolkit of describing and comparing partitives. Theoretically, a
new term is justified only if (a) there are instances that clearly show a mismatch
between partitive forms and the Partitive Concept or (b) that demonstrate the
synchronically arbitrary relationship between morphological partitives and the
Partitive Concept. A working definition about the Linguistic Partitive is worded
in (vi) to proceed with clear examples as well as borderline cases.
(vi) A Linguistic Partitive is a form that is derived from a special subtype of

separative (source argument) form.
On the basis of numeral and adpositional complement systems I will argue that
there are instances where the link between the partitive form and concept is not
consistently realized in the grammatical system, and that there are instances
that clearly display a mismatch between partitive forms and the concept.
The link between the partitive form and concept is not consistently realized
in the grammatical system of Inari Sámi. The Inari Sámi example of the partitive
on ‘river’ in ‘eight rivers’ appears with numerals higher than 7, as in (5).
(5) Inari Sámi

Mun uáinám käävci juuhâd.
1SG . NOM see.1 SG eight river.PAR
(Toivonen 2003: 66)
In example (5), it is not clear if the link to the Conceptual Partitive is direct or
the partitive is lexical-constructional. Synchronically, the system case-marking
in numeral complements is opaque. The Partitive Concepts in the numeral phrases
represent the amount partitive concept: numerals and their complements are
partitive concepts in any language. However, only in numbers higher than seven,
the Partitive Concept is expressed by a Linguistic Partitive in this language, which
indicates the lack of consistent cognitive link between the amount-partitive con-
cept and number complement marking. More specifically, lower numbers do not
require the morphological partitive, and the non-partitive cases do not instan-
tiate a link between Conceptual and Linguistic Partitives. The partitive in higher
numbers displays a match between Conceptual and Linguistic Partitives, eight
104 Anne Tamm
of the rivers (the core meaning), or an amount of rivers, eight of the kind ‘river’.
However, the partitive in this construction is not a matter of regular meaning
extension; the distinction between the lower and higher numbers is not cogni-
tively motivated. Instead, it is a matter of convention.7 In ‘six rivers’ there is no
partitive on ‘river’, although there is no known perceptionally, conceptually,
or psycholinguistically significant distinction between more or less than seven
rivers that would motivate the difference in case encoding. A difference in
encoding would be plausibly motivated if it occurred between one and more,
two or more, or three or more. In sum, although all complements of the Inari
Sámi numerals are instances of Partitive Concepts, because they match the
amount partitive concept, only the 7+ ones are also Linguistic Partitives, because
they are instantiated by a formative that corresponds to the historically moti-
vated Partitive Concept. The fact that the divergence in the system of numeral
complements is cognitively not motivated shows that the partitive assignment
is arbitrary, that is, lexically and not cognitively determined.
The correspondence may be conceptually motivated, meaning that the use
can be conceptually linked to the Partitive Concept, but the relationship with
the Partitive Concept can be opaque as well. The morphological partitive seman-
tics is opaque in (5), since the Partitive Concept does not unambiguously deter-
mine the morphological partitive case encoding in the construction. The partitive
marking in (5) is restricted more by convention than by semantics, even if
the semantics of the Partitive Concept is present. Therefore, the term Linguistic
Partitive is useful to distinguish between Partitive Concepts that are only partly
realized by morphological partitives.
Are there any Linguistic Partitives that do not match Partitive Concepts at
all? The example above is both Partitive Concept and Linguistic Partitive. Now I
turn to instances that clearly display a mismatch between partitive forms and
the concept and therefore justify the inclusion of a separate Linguistic Partitive
heuristic in the conceptual toolkit of talking about partitives. Many Finnic parti-
tives that mark structural relationships such as complementhood are an instance
of Linguistic Partitives that are not Partitive Concepts and, synchronically, they
can be considered instances of the structural or default partitive. The noun jõgi
‘river’ is marked with the same morphological partitive (the partitive form is
jõge) if it is the complement of the numeral ‘two’ and the verb ‘look at’ in examples
(6a) and (6b), as well as if it is the complement of a preposition or a postposition
in examples (6c) and (6d).
7 See also Nelson (2003) for a comparison between the partitive use of Inari Sámi and Finnish.
This convention of numbers does not have a conceptual but rather arbitrary basis, perhaps
best to be compared to the arbitrary relation between the number concept of forty and the
Russian word for 40 (sorok). This simplex lexeme for the number originates from fur trade –
forty furs were a unit necessary for sewing a coat (Shanskiy and Bobrova 1994).
(6) Estonian
a. kaks jõge
two[ NOM ] river.PAR
‘two rivers’
b. vaata-n jõge
look-1SG river.PAR
‘I am looking at the river.’
c. mööda jõge
along river.PAR
‘along the river (prepositional phrase)’
d. jõge mööda
river.PAR along
‘along the river (postpositional phrase)’
The complement of the numeral ‘two’ is marked with the partitive in example
(6a) and it is a Conceptual and a Linguistic Partitive. In the complement of the
verb ‘look at’ in example (6b), the link with the Partitive Concept is semantically
motivated via the amount-of partitive but not as obviously as in the case of the
complement of the numeral quantifier. The amount pertains to the unbounded
looking-at event and not to parts of the river, as on a map. As the complement
of a preposition or a postposition in examples (6c) and (6d), the partitive noun is
not an instance of a Partitive Concept, but an instance of a Linguistic Partitive
that has evolved into a general grammatical complement marker.8
8 Previous formal linguistics literature contains many good examples of morphological

partitives without clear semantic content or function. Vainikka and Maling (1995) analyze the
partitive as a structural relationship, the default complement case of complements in Finnish.
Kratzer (2004) also ‘de-semanticizes’ the Finnish partitive in the sense of regarding it as a
default. The use of the partitive in these examples bears striking resemblance to the bare
nominal use of Hungarian nouns; namely, bare uninflected nouns are used in Hungarian in the
same environment. Hungarian has otherwise cases such as the accusative and the dative
that could hypothetically be “recruited” as default complement cases. The dependents of the
numeral heads and postpositions are bare uninflected nouns in Hungarian. In addition, the
predicative use of Hungarian bare and plural nouns and the predicative use of the Estonian
partitive (Tamm 2008a), as well and the pseudo-semantically incorporated Hungarian bare
nouns and a subset of Estonian abstract or deadjectival partitive objects and subjects (Tamm
2014) are also an interesting parallel to explore in the light of an indefiniteness account, such
as Luraghi and Kittilä (this volume).
106 Anne Tamm
(vii) Linguistic Partitives are forms that have evolved in the course of language
change from a morpheme or a construction in a language that stands for a
Partitive Concept (as defined in (v)). They may but do not have to corre-
spond to the Partitive Concept in one of their synchronic uses. Their rela-
tion to the Conceptual Partitive can be transparent, partly or wholly opa-
que in the grammatical system of a language.
The Linguistic Partitives may have bleached meaning; they have diverse seman-
tics and syntax but are “kept together” as a category by their morphological
form. An illustration of a highly diverse semantics couched under an identical
formative is exemplified by an instance of a partitive expressing part-whole
quantity, definiteness, boundedness, aspectual, epistemic modal, irrealis, and
evidential meanings.
In the following, I illustrate briefly two paths of grammaticalization of the
Estonian Linguistic Partitive as a functional and structural category in (vii) and
(viii), respectively.
(vii) The emergence of functional partitives, the TAM categories

a. “part of N” !
b. ! “part of V” (N-obj has the morphological partitive marking)
c. ! (N-object is a non-finite, deverbal nominalization and partitive
marked)
d. ! “indirect evidence” (V-nonfin (main predicate) has the
morphological partitive formative)
e. ! “part of/incomplete evidence” (V-nonfin (main or embedded
predicate) has the morphological partitive formative)
f. ! “part of/incomplete evidence for the completion/completability
of the event” (partitive object case)
(viii) The emergence of default (structural) partitives, complement marking

a. “part of N” (the morphological partitive marking has disappeared) !
b. ! “amount of N” (has the morphological partitive marking)
c. ! “amount of V” (N has the morphological partitive marking)
d. ! “N-obj” (N has the morphological partitive marking)
e. ! Adpostion “N-obj” (N has the morphological partitive marking)
The path in (vii) sketches the emergence of epistemic modals, evidentiality, and
epistemic modal object case alternation. More discussion of the examples will
follow in Sections 4 and 5.
(vii)0 The emergence of functional partitives, the TAM categories
(vii) a.0 “part of N, incomplete N” !
The original part-of-N meaning (vii-a) is expressed by the elative in Estonian (7),
and it can be a marginal meaning of the partitive in consumption verbs. I signal
this fact by means of a question mark in front of the partitive form, and a sepa-
rative without any indication of ill-formedness in grammaticality judgments.
This basic meaning gives rise to part-of-V meaning in (vii-b), which is the main
interpretation of the partitive in example (7).
(7) Estonian
Mari sõ-i ?õuna / õuna-st.
M[NOM ] eat-PST. 3SG apple.PAR apple-ELA
‘Mary ate some quantity of the apple.’ (bounded event, (non)quantized apple)
(vii) b.0 ! “part of V, incomplete V” (N-obj has the morphological partitive

marking).
The original part-of-N meaning (vii-a) gives rise to part-of-V meaning in (vii-b)
via metonymy. More specifically, a nominal quantization meaning, as in Mary
ate a part of apple becomes an event quantization meaning, as in Mary is
halfway eating the apple. This is the extension of the nominal meaning to the
aspectual meaning of the partitive (7), (8a), and (8b).9
The partitive of negation falls under this development, since the incomplete-
ness of the event coincides conceptually with the negated event (8c). Incom-
pleteness or insufficiency compared to a predicate-related norm is a possible
explanation for the theme-subject partitives as well, in positive or negative, as
in (8d). These examples overlap with the partitive types in (viii-b), (viii-c), and
(viii-d) and are presented here without repeating them under (viii).
(8) Estonian
a. Mari sõ-i õuna.
M[NOM ] eat-PST. 3SG apple.PAR
‘Mary was eating an apple.’ (unbounded event, quantized or
nonquantized apple)
9 Historical overview can be found in two sources concentrating on Finnic, Larjavaara (1991) on
the development of an aspectual object and Campbell (1991) and Ikola (1953) on the develop-
ment of the partitive evidential.
108 Anne Tamm
b. Mari kuul-is lindu.

M[NOM ] hear-PST.3 SG bird.PAR
‘Mary heard a bird.’ (unbounded event, quantized bird)
c. Mari ei söö-nud õuna ära.

M[NOM ] NEG eat-CNG . PST apple.PAR TELIC
‘Mary was not eating an apple.’
‘Mary did not eat an apple (up).’
(unbounded, telic or atelic, quantized or nonquantized apple)
d. Mari-l on / ei ole õunu.

M[NOM ] be.3SG NEG eat-CNG apple.PAR . PL
‘Mary has/ does not have apples.’
In instances that underly the diachronic development to evidential and epis-

temic modal partitives, the aspectual partitive object of verbs of perception and
mental verbs has a deverbal modifier (as in I heard a sing+ing bird), which agrees
in case with the head noun of the object (9). Thus it has a partitive encoding.
(9) Mari kuul-is lindu laul-va-t.

M[NOM ] hear-PST.3 SG bird.PAR sing-PERS . PRS . PTCP- PAR
‘Mary heard a singing bird.’
(vii) c.0 ! N-object is a non-finite, deverbal nominalization and partitive

marked
When a deverbal modifier of the object was reanalyzed as a predicate on its

own, the partitive did not disappear but started to be interpreted as a modal
marker, a kind of subjunctive or conjunctive in instances like (10).
(10) Mari kuul-is, et lind laul-va-t.

M[NOM ] hear-PST.3 SG that bird[NOM ] sing-PERS . PRS . PTCP- PAR
‘Mary heard that the bird was singing.’
(vii) d.0 ! “indirect evidence”

(V-nonfin (main predicate) has the morphological partitive formative)
As the complement of saying verbs, the meaning of the subjunctive-conjunctive

marking is associated with the matrix saying verb, as in (11).
(11) Mari ütle-s, et lind laul-va-t.

M[NOM ] say-PST.3 SG that bird[NOM ] sing-PERS . PRS . PTCP- PAR
‘Mary said that the bird was singing.’
As the complement of saying verbs, the matrix saying verb was dropped in the
construction, giving rise to the indirect evidential, which still preserved the
formative of the partitive, as in (12).
(12) (Mari ütle-s, et) lind laul-va-t.
M[NOM ] say-PST.3 SG that bird[NOM ] sing-PERS . PRS . PTCP- PAR
‘(Mary said that) the bird was singing.’
(vii) e.0 ! “part of/incomplete evidence”

(V-nonfin (main or embedded predicate) has the morphological
partitive formative)
However, in communication the hearer infers that indirect evidence is incomplete
evidence, because it is not taken for granted by the speaker. Therefore, the parti-
tive evidential has a double function: to encode the multiple speaker-hearer rela-
tionships and epistemic modality, indirect and incomplete evidence in (13).
(13) Lind laul-va-t.

bird[NOM ] sing-PERS . PRS . PTCP- PAR
‘Allegedly, the bird is singing.’
(vii) f.0 ! “part of/incomplete evidence for the completion/completability

of the event” (partitive object case)
The epistemic modal meaning of the partitive objects is also reinforced. The
strengthened meaning is understood as incomplete evidence about the comple-
tion or completability of an event instead of incomplete event in a number of
achievement verbs, as in (14).
(14) Silvi üllata-s Toomas-t.
S[NOM ] surprise-PST.3 SG T-PAR
‘Silvia surprised Thomas.’
The examples illustrating the sketch can be found in the following subsec-
tions in the context of other examples, phenomena, and languages.
The path in (viii) sketches the emergence of the “default partitive” with
completely bleached semantic content.
(viii)0 The emergence of default (structural) partitives, complement marking
(viii) a.0 “part of N” (the morphological partitive marking has disappeared) !
First of all, it should be pointed out that the semantic part-of meaning has lost
the link with morphological partitive marking; it is not known when it dis-
110 Anne Tamm
appeared in contexts such as (15). More discussion of this example can be found
in (vii-a), concerning example (7). Example (15b) is an instance of a typical
partitive construction, which is realized by the elative. Erzya (Mordvinian) is
discussed as an example of thwarted development of an aspectual partitive.
The ablative complement of verbs of consumption, as in (15c) is referred to as
an object (Collinder 1960: 124). Therefore, Erzya seems to have an object case
alternation comparable to instances of the aspectual object case alternation in
Finnic. However, although the ablative is used in more abstract contexts, as in
kortams mezedejak ‘speak about something’ (Niina Aasmäe, p.c.), the case allows
an interpretation where the identity of the separated matter is different. The
example can be understood as part-of and amount-of partitive.
(15) Estonian
a. Mari sõ-i ?õuna / õuna-st.
M[NOM ] eat-PST. 3SG apple.PAR apple-ELA
‘Mary ate some quantity of the apple.’ (bounded event, (non)quantized
apple)
b. noorim mu laste-st
young.SUP 1 SG .GEN child.PL- ELA
Mordvin
c. kšede jarcy, vinado simi
bread.ABL eats wine.ABL drinks
‘he eats bread, drinks wine’ (Collinder 1965: 125)
(viii) b.0 ! “amount of N” (has the morphological partitive marking)
Only the “amount of N” has the morphological partitive marking (viii-b). Here
the partitive nouns have kind reference. See the examples and discussion in
(vii-b) as well.
(16) Estonian
a. klaas veini
glass[ NOM ] wine.PAR
b. kaks jõge
two[ NOM ] river.PAR
‘two rivers’
c. viis kraadi sooja
five[ NOM ] degree. PAR warm. PAR
‘plus 5 degrees’
(viii) c.0 ! “amount of V” (N has the morphological partitive marking)
This meaning is metonymically carried over to the predicate level, as in (17),

to mean “amount of an event”, but it is not the verb that is marked but the
object noun that retains the partitive marking. See the examples and discussion
in (vii-b) as well.
(17) Estonian
Mari sõ-i õuna.
M[NOM ] eat-PST.3 S apple.PAR
‘Mary was eating an apple.’ (unbounded event, quantized or
nonquantized apple)
(viii) d.0 ! “N-obj” (N has the morphological partitive marking)
In the next step, the partitive nouns become the general object markers without
any partial meaning element related to the object noun (18). The partitive is used
for conveying aspectual unboundedness.10 Since the non-finite temporally un-
linked telic verbs without any finite environment display partitive objects, the
partitive is a default and not determined by lexical aspect (telicity, 18d) or indef-
initeness (see 18a, 19d). See the examples and discussion in (vii-b) as well.
(18) Estonian
a. Mari armasta-b Jaanus-t.
M[NOM ] love-3SG J-PAR
b. Mari vaata-b jõge.

M[NOM ] look-1SG river.PAR
‘Mary is looking at the river.’
c. Mari kuul-is lindu.

M[NOM ] hear-PST.3 SG bird.PAR
‘Mary heard a bird.’
d. seda jõge ära reosta-ma

this.PAR river.PAR TELIC pollute- M _ ILL
‘to completely pollute this river’
(viii) e. ! Adpostion “N-obj” (N has the morphological partitive marking)
10 Only a few constructions have retained the accusative-total as a default object option (more
analysis can be found in Tamm 2008a, 2008b).
112 Anne Tamm
The object is a complement of a verb, but the partitive is generalized into a more
general complement case in a language, perhaps also on the analogy of the par-
titives found in numeral and other measure phrases (19). Partitive tends towards
becoming a general complement case in Estonian.
(19) Estonian
a. mööda jõge
along river.PAR
‘along the river (prepositional phrase)’
b. jõge mööda
river.PAR along
‘along the river (postpositional phrase)’
I put aside constructions where the partitive has little or no semantic content.
Generally, semantic content is missing in combinations with adpositions. There-
fore, the partitive could be viewed as developing into a general complement
case occurring with certain prepositions, postpositions, numeral phrases, and
verbs in the Finnic languages.
4 The Partitive Concept in the Uralic languages

This section takes a look at the empirical data and discusses some of the cases
that express the Partitive Concept in the Uralic languages. Before the discussion
I emphasize that the case appears only in argument or predicative functions and
should be understood as such. I illustrate the meaning extension of the partitive
concept, the pseudopartitive, without embedding the form into a sentence.
4.1 The empirical data

What is special about the Uralic partitives? Embeddedness in rich case systems:
the multitude of source (separative) cases and a mismatch between the Partitive
Concept and the semantics of the various linguistic partitives. Thus, firstly, the
Uralic languages have many cases that express Partitive Concepts as well as
morphological Linguistic Partitives. The Linguistic Partitives either express the
Partitive Concept or related concepts, some of which have developed far from
the original concept. There is a wide spectrum of “what the partitive is used
for” in languages with the morphological Linguistic Partitive – the partitive
interacts with various aspects of the TAM system. The interaction between TAM
and the partitive is exceptionally clear in these languages, which display the
aspectual DOM, DSM, and DAM, definiteness effects, telicity, and partitive argu-
ments. These languages have a range of other cases than the morphological
Linguistic Partitive, covering the Partitive Concept. Uralic languages are special
about their cross-categorial case – case on non-finites and verb stems. In several
Uralic languages, the Linguistic Partitives and cases that express the Partitive
Concept are an integral part of these unusually rich cross-categorial case systems
and constitute, therefore, a unique area of partitive studies.
Uralic languages are typically characterized by rich case systems with
approximately ten members, and many have case systems of approximately fifteen
or twenty cases. In the selection of languages in the WALS, on the map by Iggesen
(2008), there are 24 languages recorded with black dots, which stand for systems
with more than 10 cases.11 Five of those listed are Uralic (Erzya Mordvin, Estonian,
Finnish, Hungarian, and Udmurt). The North-Eastern European and North-Western
Asian area would be studded with black dots on the WALS map if all Finno-
Ugric languages were represented. Table 1 summarizes the number of cases in
some Uralic languages.12
11 The following languages have “black dots” in WALS: Awa Pit, Basque, Brahui, Chukchi, Epena
Pedee, Estonian, Evenki, Finnish, Gooniyandi, Hamtai, Hungarian, Hunzib, Ingush, Kayardild,
Ket, Lak, Lezgian, Martuthunira, Mordvin (Erzya), Nez Perce, Nunggubuyu, Pitjantjatjara, Toda,
Udmurt.
12 Erzya Mordvin has twelve cases: nominative, genitive/accusative, dative/allative, interior
illative, inessive, elative, exterior ablative, lative, prolative triplets, translative, abessive,
comparative, and Moksha Mordvin thirteen cases (Zaicz 1998: 192–194), with the additional
causative. Eastern Mari has eight productive and three nonproductive cases (Kangasmaa-Minn
1998: 226). Udmurt sixteen cases (Riese 1998: 268), nominative, accusative, genitive, dative,
approximative, genitive/ablative, inessive, elative, ablative, terminative, instrumental, egres-
sive, caritive, adverbial, prolative 1 and 2. Komi has eighteen cases (Riese 1998: 268), nomina-
tive, accusative, genitive, dative, approximative, genitive/ablative, inessive, elative, ablative,
terminative, instrumental, egressive, caritive, adverbial, prolative 1 and 2, consecutive,
comitative. Komi Permyak has seventeen cases (Lytkin et al. 1962: 184). Tundra Nenets seven
(Salminen 1998: 537), nominative, accusative, genitive, dative, locative, ablative, prosecutive
(this is the suggested Proto-Samoyedic inventory, Janhunen 1998: 469). Kamas has seven
cases (Szimoncsics 1998: 585–586), nominative, accusative, genitive, lative, locative, ablative,
instrumental. Selkup has thirteen (Helimski 1998b: 560–561), nominative, accusative, genitive,
instrumental, co-ordinative, caritive, translative, dative/allative, illative, locative, elative, prola-
tive, vocative, Nganasan eight to eleven (Helimski 1998a: 496), nominative (=absolute form),
accusative, genitive, lative (=dative, or dative-lative), locative (=locative/instructive), elative
(=ablative), prolative (=prosecutive). The Sámi languages are described having systems with
six to nine cases. Inari, Pite, Skolt Sámi nine or eight, Southern Sámi eight or seven, Lule
114 Anne Tamm
Table 1: The number of cases in some Uralic languages
Language Number of cases

Erzya Mordvin 12
Moksha Mordvin 13
Eastern Mari 8+3
Udmurt 16
Komi 18
Komi Permyak 17
Votic 14
Võro 13 (+3 less productive cases)
Northern Sámi 7 or 6
Lule Sámi 7
Southern Sámi 8 or 7
Inari, Pite, Skolt Sámi 9 or 8
Khanty 3–11
Mansi 6–7
Nganasan 8–11
Selkup 13
Kamas 7
Tundra Nenets 7
Estonian 14
Finnish 15
Meänkieli 13
Ingrian 10
Karelian 12–16
Veps 22–23
In sum, the Uralic languages are a suitable testbed for studying several kinds of
partitive. On the one hand, there are several cases that denote separation and
source. There is a whole sub-branch of languages, the Finnic ones, that all have
a dedicated partitive case.
Sámi seven, Northern Sámi seven or six (Wikipedia). Khanty (three to eleven, including the fact
that the alignment system has variants, e.g. the Khanty Vakh dialect may have an ergative-
accusative alignment), Mansi (six to seven) (Honti 1998: 343). Hungarian eighteen cases (but
there are heavy debates whether what has been referred to as case is in fact case, or nominal
marking of different nature). Veps has 22–23 cases (Viitso 1998), Karelian twelve to sixteen
(Markianova 2002), Ingrian more than ten (Viitso 1998). Meänkieli (Finnish in Sweden) has
two cases less than Standard Finnish, which has fifteen cases. Võro is described as having
thirteen productive and three nonproductive cases (Iva 2007: 41). The Votic dialect reported
by Tsvetkov (2008: 27) has an inventory identical to that of Estonian, consisting of fourteen
cases. The additional unproductive excessive and instructive, and the accusative object case
are recorded in the dialect studied by Ariste (Ariste 1968: 17).
Table 2: Languages with a linguistic partitive and languages with a morphological separative
case
Languages with a linguistic partitive Languages with a morphological separative case

Finnic, some Sámi languages The rest of the Uralic languages, with the exception
of some dialects
Source cases in poor paradigms. The Samoyedic, Sámi, and the Siberian Ob-
Ugric languages have somewhat poorer case systems where the inventory of
source cases is also poorer than in the rest of the Uralic languages. This tendency
seems to be more a matter of a South-North opposition. Source cases in the poorer
paradigms express the part-of partitive relationships, and the amount-of partitive
is typically expresed by juxtaposition.13
4.2 Partitive cases that denote separation from source

In most of the Uralic languages, there are several cases that denote separation
from source. The specialization of source cases is typical but not unique for
Finnic languages. What makes the Finnic system special is the specialization of
source cases so that one of them denotes the separation or division into two
entities of the same kind.
Source cases in the Uralic languages are the ablative, elative, delative,
egressive, genitive-ablative and exessive. The ablative (Erzya, Estonian, Finnish,
Hungarian, Mansi, Vepsian, Votic, etc) denotes movement away from something
(e.g., away from the house). The elative (Erzya, Estonian, Finnish, Hungarian,
Lule Sámi, Pite Sámi, Votic, etc) denotes “out of something” (e.g., out of the
house). The delative (Hungarian) indicates movement from the surface, e.g.,
from (the top of) the house. The egressive (Veps, Udmurt) marks the beginning
of a movement or time (e.g., beginning from the house). The excessive (Karelian,
13 The following lists present the case systems of poorer Uralic languages. and the source
cases that can express the conceptual partitive are set in the boldface. The Tundra Nenets
system of seven cases (Salminen 1998: 537, the nominative, accusative, genitive, dative,
locative, prosecutive) has the ablative. The Kamas system of seven cases (Szimoncsics 1998:
585–586) comprises nominative, accusative, genitive, lative, locative, ablative, instrumental.
The Selkup system of thirteen cases (Helimski 1998: 560–561), comprises the nominative,
accusative, genitive, instrumental, co-ordinative, caritive, translative, dative/allative, illative,
locative, prolative, vocative has the elative. The Nganasan system of eight to eleven cases
(Helimski 1998: 496) contains the nominative (=absolute form), accusative, genitive, lative
(=dative, or dative-lative), locative (=locative/instructive), prolative (=prosecutive) and it has an
elative (=ablative). The poorest system can be encountered in some dialects of Khanty.
116 Anne Tamm
Ingrian, Livonian, Votic, Estonian, etc) denotes a transition away from some-
thing (from a house). The genitive-ablative of Komi stands for a source of infor-
mation, or for a resource; this case is interesting in combining the genitive and
separative meaning in a discourse setting.
In the Uralic languages, the Partitive Concept seems to be generally expressed
by the elative case. If there is no dedicated elative case, then the Partitive Concept
is expressed by a case called the ablative and understood as the most general
separative case. The genitive case is not present in all Uralic languages. Most
notably, the largest Uralic language in terms of number of speakers, Hungarian,
lacks it altogether. It seems that the Uralic partitives are not marginal exten-
sions of genitives, but they are specific, dedicated same-kind separatives. At
this point of research, a hypothesis can be worded as in (ix).
(ix) Hypothesis: the Uralic partitives are specialized separatives. More specifi-
cally, the Uralic partitives are same-kind separatives expressed by a case.
It is worth investigating how the Partitive Concepts are expressed in languages

that do have a genitive alongside with an elaborate system of separative cases.
The Indo-European languages do not have any elaborate system of source cases
that would be comparable to the Finnic ones, but they have genitives, and their
partitives are frequently based on the genitive semantics. In the Finnic lan-
guages, the typical partitive functions of the Indo-European genitives, such as
in Russian in the example in (1a) (čaška čaj-u ‘a cup of tea’), are never covered
by genitives but by partitives. If there are no partitives, then these functions are
expressed by elatives or other general separatives, or by no case marking. It is a
good question if some Partitive Concept meanings are covered by genitives in
languages combining a rich separative system with a genitive in the case para-
digm. The following sections present the types of Uralic languages according
to the place of the partitive within the overall system of separatives. Type 1 is
illustrated by the Partitive as one of the source cases. Type 2 has no linguistic
partitive but several source cases. Type 3 exemplifies a system with a linguistic
partitive but no grammatically encoded source cases. Type 4 illustrates a system
with no morphological partitive or separative cases.
All of these languages have also postpositions that can expresse source, but
the present article confines itself to one illustrative example. The system of
parallel ways of expressions in a more elaborate system of postpositions is
from Estonian separative postpositions. Relationships that are similar to case
relationships are instantiated by a rich system of postpositions. In Estonian, a
separated position or movement away from an entity can be expressed by post-
positions that stand (in terms of language history) in the elative or the ablative
case, as in (20). They typically end in -st or -lt, thus the elative or ablative end-
ings. The examples containing postpositions that are semantically roughly inter-
changeable with source cases are set bold in the example.
(20) Estonian
a. paadi see-st
boat.GEN inside-ELA
‘from inside a boat’
b. paadi juure-st
boat.GEN close-ELA
‘from the vicinity of a boat’
c. paadi ääre-st
boat.GEN near-ELA
‘from the (outer, longish) side of a boat’
d. paadi külje-st
boat.GEN side-ELA
‘from the (inner) side of a boat, from being attached to a boat’
e. paadi otsa-st
boat.GEN top-ELA
‘from the topmost, outmost, close-fitting, or sharp top or end of a boat’
f. paati-de sea-st
boat-GEN . PL among-ELA
‘from among the boats’
g. paati-de hulga-st
boat-GEN . PL amount-ELA
‘from among the boats’
h. paadi pea-lt
boat.GEN top-ABL
‘from top of a boat’
i. paadi koha-lt
boat.GEN over-ABL
‘from above a boat’
j. paadi kõrva-lt
boat.GEN side-ABL
‘from the outer side of a boat’
118 Anne Tamm
k. paadi ümbert
boat.GEN around.ABL
‘from around a boat’
l. paadi alt
boat.GEN under.ABL
‘from under a boat’
The composite forms that are based on postpositions that have a recognizable
elative component are the following: ‘from inside a boat’ (20a), ‘from the vicinity
of a boat’ (20b), ‘from the (outer, longish) side of a boat’ (20c), ‘from the (inner)
side of a boat, from being attached to a boat’ (20d), ‘from the topmost, outmost,
close-fitting, or sharp top or end of a boat’ (20e), ‘from among the boats’ (20f)
and (20g). Other composite forms are based on postpositions that have a recog-
nizable ablative component: ‘from top of a boat’ (20h), ‘from above a boat’ (20i),
‘from the (outer) side of a boat’ (20j). Some forms have just a recognizable -t in
the postposition, which we see in the elative and ablative formants as well:
‘from around a boat’ (20k) and ‘from under a boat’ (20l). The present study is
carried out proceeding from the assumption about the relevance of the difference
between cases and adpositions in carving up the conceptual space, as in (x).
(x) The distinction between cases and adpositions is grammatically relevant in

languages that have cases as well as adpositions expressing many separa-
tive concepts. More particularly, if the semantic field of separative motion is
diverse, then the linguistic partitive meanings are specialized so that they
express separation of matter of the same kind.
The digression into the postpositions was necessary to illustrate the position of
partitives and separatives encoded by cases in a larger system of numerous and
complex conventionalized means of expressing separation. In several instances,
it can be observed that the multitude of forms are composite in the sense that
the elative or ablative cases can be reconstructed as parts of the composite
forms. There are also forms that are formally similar, containing an ending with
a -t, but not reconstructable with the elative or ablative cases (e.g. ümbert ‘from
around’). This paper will only be addressing cases.
Note that the works explicitly contrasting partitive meanings are missing in
languages that grammaticalize partitives as case and as adposition simultane-
ously. This is a curious fact, since there are several studies that demonstrate
a contrast between a spatial case with a spatial adposition, or that contrast
adpositional and non-adpositional forms of genitives. This brief digression into
separatives has hopefully demonstrated that there are still vast gaps in the
research on the topic.
4.3 Source (separative) cases and the partitive

This subsection examines which languages have source cases and the partitive,
and what their patters are.
4.3.1 The partitive as one of the source cases
The system incorporating a partitive and several source cases is illustrated by

Estonian in Table 3. The source cases are set boldface.
Table 3: The Estonian case system
case translation N example N

Nominative book raamat
Genitive of a book raamatu
Partitive (of ) a book raamatu-t
Illative into the book raamatu-sse
Inessive in a book raamatu-s
Elative from (inside) a book raamatu-st
Allative onto a book raamatu-le
Adessive on a book raamatu-l
Ablative from the book raamatu-lt
Translative in(to), as a book raamatu-ks
Terminative until a book raamatu-ni
Essive as a book raamatu-na
Abessive without a book raamatu-ta
Comitative with a book raamatu-ga
The development from source to origin or cause is a possible but rather rare
meaning extension of the partitive in a system that is rich in source cases and
has a partitive. The following example illustrates the cause partitive on the
so-called “infinitives”, that is, non-finites in Karelian (21).
(21) Karelian
Suurdu keittä-miä pada musten-i.
big.PAR cook-M . NMLZ . PAR pot[NOM ] blacken-PST.3 SG
‘Intensive cooking made the pot turn black.’
120 Anne Tamm
Since the partitive marked adjective suurdu ‘big, intensive’ as in (21) can modify
the partitive form in question, the latter cannot be an infinitive, but another type
of nominalization with more nominal properties than infinitives would have. The
meaning of a cause event emerges with the Karelian event predicates and parti-
tive marking. This rare instance of linguistic partitive gives evidence of the
meaning element of causation and event structural properties of the predicates
involved.
4.3.2 No partitive, several source cases
A system without the partitive but with several source cases can be found in
Udmurt, illustrated by Table 4. The source cases – ablative, elative, and egressive
– are set boldface.
Table 4: Cases in the Udmurt noun paradigms
Case noun + case

Nominative s’ik
Genitive s’ik-len
Accusative s’ik/s’ik-ez
Ablative s’ik-les’
Dative s’ik-ly
Adessive s’ik-len
Instrumental s’ik-en
Abessive s’ik-tek
Inessive s’ik-yn
Illative s’ik-e
Elative s’ik-ys’(t)
Terminative s’ik-oz’
Egressive s’ik-ys’en
Prolative s’ik-eti
Approximative s’ik-lan’
Hungarian lacks a morphological partitive, but its inventory of three separative

cases allows interpreting these cases, especially the elative, as parts of con-
structions expressing Partitive Concepts. The elative, illustrated in (22a), denotes
separation from a container, the ablative denotes separation from the vicinity
of something (22b), and the delative denotes the separation from the object by
movement from a surface (22c).
(22) Hungarian
a. ház-ból
house- ELA
‘from (inside) a house’
b. ház-tól
house-ABL
‘from (the vicinity of) a house’
c. ház-ról
house- DELA
‘from (the top of) a house’
Also, several other Uralic languages have separative cases that are not referred
to as partitive, but their semantics is that of a prototypical partitive. As in
Hungarian, the typical Partitive Concepts are realized by the elative (or the abla-
tive, if there is no elative). Example (23) illustrates the Estonian Partitive Concept
realized by the elative (the youngest of my children).
(23) Estonian
noorim mu laste-st
young.SUP SG 1.GEN child.PL- ELA
4.3.3 Partitive and no source case: Sámi
Inari Sámi, which I use for illustration, has nominative, genitive, accusative,
illative, locative, comitative, abessive, essive and the partitive – the last two
cases only exist in their singular form, as described in Toivonen (2003: 36) in
Table 5. The partitive is set bold.
Table 5: The Inari Sámi case paradigm for kietâ ‘hand’ (Toivonen 2003: 36)
singular plural
nominative kietâ kieδah
genitive kieδâ kieδâi
accusative kieδâ kieδâid
illative kietân kieδáid
locative kieδâst kieδâin
comitative kieδáin kie'δâigui'm
abessive kie'δâttáá kie'δâittáá
essive kiettân
partitive kiettâd
122 Anne Tamm
Sámi comparative constructions are a possible extension of the partitive in a

system with no source or separative cases but a partitive that appears in singular
only, as in (24a). In Inari and Skolt Sámi, the partitive case cannot be regarded
as a general complement case like the partitives of the Finnic languages, since
its use is restricted to specific constructions only (e.g. restricted postpositional,
number phrases, etc.).
At this point it is worthwhile to make a digression into an interesting varia-
tion that regularly involves the partitive or a source case in Uralic languages.
There are several comparative constuction types in the Uralic languages. Among
them, there is one where the standard of comparison is marked with a source
case, which is frequently the elative or the partitive. In Inari Sámi, the standard
of comparison is marked with the partitive, as in (24a). Marking the meaning of
‘N has the property Adj more than someone/something’ is not rendered (only)
by ‘N is Adj-er than someone/something’, but (also) by ‘N is of/from someone/
something Y(-er)’. In source case systems with source cases and the partitive,
the marking of the standard of comparison varies. In Estonian, the standard of
comparison is marked with the elative (24d), but Finnish has variation between
the partitive (24b) and the elative (24c).
(24) a. Inari Sámi

Muorâ lii táállud ucceeb.
tree[NOM ] be.3 SG house.PAR smaller
‘The tree is smaller than the house.’
(Toivonen 2003: 65)
b. Finnish
Miehe-ni on 3 kk minu-a vanh-empi.
man[NOM ]-POSS .1 SG be.3SG 3 months 1 SG - PAR old-CMPR
‘My husband is three months older than I.’
c. Finnish
. . .nainen ol-i minu-sta vanh-empi14
woman[NOM ] be-3 SG . PST 1 SG - ELA old-CMPR
‘The woman was older than I.’
d. Estonian
Toomas on Peetri-st van-em.
T[NOM ] be.3 SG P-ELA old- CMPR
‘Tom is older than Peter.’
14 http://keskustelu.suomi24.fi/node/10906453
The example sentences in (24) are instances of Linguistic Partitive where

the link to the Conceptual Partitive is considerably weaker than, for instance,
example (23).
4.3.4 No partitive or separative
Almost all Uralic languages have a source case that is used for marking the
source argument. The poorest case system can be encountered in some dialects
of Khanty, where there is a distinction between four cases only. In the Kazim
dialect, the separative cases are missing altogether, in others, there is still an
ablative.
Table 6: Tha case system of some Khanty dialects (on the basis of Ruttkay 2003: 20)
case Kazym dialect Vakh dialect Vasyugan dialect

nominative Ø Ø Ø
accusative -t -t -t
lative-dative -a -a -a
approximative -pa/-pä
translative -γǝ/-γǝ̆ , -γ -γǝ/-γǝ̆ , -γ
locative -ǝn -nǝ/-nǝ̆ -nǝ/-nǝ̆
instrumental-final -tǝ/-tǝ̆ , -ǝ/-ǝ̆ -tǝ/-tǝ̆ , -ǝ/-ǝ̆
instrumental-comitative -nat/-nät -nat/-nät
distributive -tǝ̆ lta/-tǝltä -tǝ̆ l/-tǝl
ablative missing -oγ/-öγ -ow/-öw, -oγ/-öγ
abessive – -lǝ̆ γ/-lǝγ -lǝ̆ γ/-lǝγ
comparative – -ni ̮ηi̮ ̮t -niηǝ̮
4.4 Pseudo-partitives
This subsection deals with the reading (xi-b), the amount-partitive or the pseu-
dopartitive.
(xi) a. “separable part of N that belongs to the same kind with N” !

b. ! “amount of N” (metonymic extension of xi-a)
In this subsection, I collect the evidence for the following hypotheses.

124 Anne Tamm
(xii) The Linguistic Partitive is more characteristic of pseudopartitive

constructions in the Uralic languages that have a Linguistic Partitive.
(xiii) Pseudopartitive constructions are expressed predominantly via

juxtaposition in Uralic languages that do not have a Linguistic Partitive.
First of all, it is not completely clear at this stage if all Uralic languages have
any partitive constructions with the structure N-measure + N-substance (e.g., a
glass of wine). Many Uralic languages express pseudo-partitives with juxta-
position, as in N and W Sámi, Hungarian, Mari, Mordvinian, Komi, and Udmurt,
according to Koptjevskaja-Tamm (2001: 555). Juxtaposition (pseudo-partitive) is
illustrated by a Hungarian example in (25).
(25) Hungarian
egy pohár bor
INDF glass[NOM ] wine[NOM ]
Estonian has a morphological partitive. Note that the proper partitive relation-
ship is realized by the elative, but the pseudo-partitive (a glass of wine) is realized
with the partitive case-marking, as in (26).
(26) Estonian
klaas veini
glass[ NOM ] wine.PAR
There are languages where the morphosyntactic encoding of the two types of
partitive semantics clearly differs as shown by Koptjevskaja-Tamm (2001). This
seems to be the case in Estonian as well. Koptjevskaja-Tamm (2001) contains a
detailed typological study on Finnish and Estonian pseudopartitives. The illus-
tration contrasting the two concepts is taken from Finnish. Example (27a) is a
partitive nominal construction (PC) and (27b) is a pseudo-partitive nominal con-
struction (PPC) in Koptjevskaja-Tamm (2001). See also the discussion in Luraghi
and Kittilä (this volume).
(27) Finnish
a. pala tä-stä hyvä-stä kaku-sta
bit[NOM ] this-ELA good-ELA cake-ELA
‘a bit of this good cake’
b. säkki peruno-ita
sack[NOM ] potato-PAR . PL
‘a sack of potatoes’
Not all Uralic languages use case in these expressions, and it is not clear if the
pseudopartitves are clearly different from the real partitives. Anttila and Fong
(2000) discuss Finnish examples where the distinction does not emerge clearly,
and Tamm (2011b) demonstrates examples where even Anttila’s improved account
would not work either for the data set of abstract nouns in Estonian. The state of
the art in the data research of conventionalized measure structures in tempera-
ture in rarer Uralic languages is as follows. Khanty (28), Udmurt (29), and Komi-
Permyak (30) have a juxtaposition of nominative nouns in these constructions.
(28) Khanty, Eszter Ruttkay p.c. email, July 1st, 2010

a. kamәn ńăljaŋ grad iśki
outside 40 degree[NOM ] cold[NOM ]
‘The outside temperature is 40 below zero.’
b. ńăljaŋ grad ťemperatura tăjl

40 degree[NOM ] temperature[NOM ] be.3 SG
‘He has a fever of 40 degrees.’
(29) Udmurt, Svetlana Edygarova, email, June 28th, 2010

a. temperatura Celsi-ja 40 gradus
temperature[NOM ] c-ADV 40 degree[NOM ]
‘40 degrees Celsius’
b. n'yl-don gradus kez'yt/ shunyt

40 degree[NOM ] cold[NOM ]/ warm[NOM ]
‘40 degrees below/above zero’
(30) Komi-Permyak, Larisa Ponomareva, email, July 2nd, 2010

a. t'emperatura 40 gradus celsija
temperature[NOM ] 40 degree[NOM ] c-ADV
‘40 degrees Celsius’
b. n'ol'das gradus ködzyt/ shonyt

40 degree[NOM ] cold[NOM ]/ warm[NOM ]
‘40 degrees below/above zero’
Hungarian does not allow partitive-like constructions at all, as seen in (31), and
Estonian realizes a linguistic partitive, as in (32). The linguistic partitive is possi-
ble if the semantics of the noun that is measured is linguistically scalar, since it
126 Anne Tamm
is possible with derivations of warm and cold, but not with the underived noun
temperature in Estonian.
(31) Hungarian
a. kint negyven fokos meleg/ hideg van
outside 40 degree-ADJ warm[NOM ] cold[NOM ] be.3 SG
‘The outside temperature is 40 above/below zero.’
b. negyven fok-o-s láza van

40 degree-LK-ADJ fever[NOM ].POSS .3 SG be.3 S
‘He has a fever of 40 degrees.’
(32) Estonian
a. viis kraadi sooja
five[ NOM ] degree. PAR warm. PAR
‘plus 5 degrees’
b. viis kraadi külma

five[ NOM ] degree. PAR cold. PAR
‘minus 5 degrees’
c. #viis kraadi temperatuuri

five[ NOM ] degree. PAR temperature. PAR
‘5 degrees’
Note that the Hungarian lacks the construction and there is an adjectival con-
struction expressing the content in measure phrases, whereas complements of
the complements are marked partitive in the Estonian measure phrases in (32).
4.5 The partitive concept and aspect

The Hungarian elative is a form in the case inventory of Hungarian that corre-
sponds closely to the Partitive Concept, as illustrated in (xi-a). Bare nouns com-
pare to other Uralic zero marked bare nouns. The following examples illustrate
Udmurt (33) and Komi (34) part-of-partitives, the (xi-a) type, in their embedded
environment. Aspect and definiteness related to explicit accusative marking
seem to coincide in these languages.
(33) Udmurt: accusative, Partitive Concept, unmarked/accusative opposition

a. n'an' s'i-i (odig judes)
bread[ACC ] eat-INF (one[ACC ] piece[ACC ])
‘to eat (a piece of) bread.’
b. n'an'-ez judes s'i-i

bread-ACC piece[ACC ] eat-INF
‘to eat a piece of this bread.’
c. n'an'-ez s'i-i
bread-ACC eat-INF
‘to eat (a piece of) this bread up.’
(Svetlana Edygarova, p.c.)
(34) Komi, elative, Partitive Concept, unmarked/accusative alternation

a. kurčč-i n'an'-s'ys tor.
bite-PST.1 SG bread-ELA piece[ACC ]
‘I have bitten some bread.’
b. n'an' söj-i.
bread[ACC ] eat-PST.1 SG
‘I was eating bread, I ate some bread.’
c. söj-i n'an'-sö.
eat-PST.1 SG bread-ACC . DEF
‘I ate the bread (some of the bread).’
(Nikolay Kuznetsov, p.c.)
Interaction with the aspectual properties of the clause vary with different parti-
tive uses involving the verbs that trigger partitive meanings. Example (35) illus-
trates Hungarian (see the discussion on the partitive function of the elative and
examples in Moravcsik 1978: 261). The partitive-elative argument and the telic
particle do not constitute a well-formed sentence (35c).
(35) Hungarian
a. Evett a pizzá-ból.
eat.PST 3 S DEF pizza-ELA
‘She ate some of the pizza.’
b. Meg-ette a pizzá-t.
TELIC -eat.PST 3 S DEF pizza-ACC
‘She ate up the pizza.’
c. *Meg-ette a pizzá-ból.
TELIC -eat.PST 3 S DEF pizza-ELA
(‘She ate up of the pizza.’)
d. Pizzá-t evett.
pizza-ACC eat.PST 3 S
‘She was eating pizza.’ (unbounded)
128 Anne Tamm
The partitive yields an unbounded reading for Estonian, as in (36).
(36) Estonian
a. Mari sõ-i (neid) pitsa-sid.
M[NOM ] eat-PST.3 SG this.PAR . PL pizza-PAR . PL
‘Mary was eating (these) pizzas.’ (unbounded, nonquantized)
b. Mari sõ-i pitsa-d / %pitsa-sid ära.

M[NOM ] eat-PST.3 SG pizza-NOM . PL pizza-PAR . PL up
‘Mary ate the pizzas (up).’ (bounded, quantized)
Table 7 summarizes the situation with affectedness and the object cases in the
studied languages.
Table 7: Affectedness of the incremental theme and the object case
Incremental theme argument totally affected Incremental theme argument partially affected
Accusative ELATIVE (Hu)
PARTITIVE (Est)
Unmarked (Hu,Kh,U)
The question however is, whether the data in (36) is an instance of the Partitive
Concept or the Linguistic Partitive? It is both. The object has partitive marking,
so it is a Linguistic Partitive. It is a Partitive Concept, since the meaning exten-
sion has come into being by means of metonymy: the disappearance of the
pizzas is temporally related to the event where the pizzas are disappearing.
Therefore, the pizzas in the event and the temporal evolving of the event itself
are related by spatiotemporal contiguity. The boundedness of the event is not
determined by the boundedness of the object matter any more in the partitive
languages such as Modern Estonian. The loss of the relationship is demon-
strated by the combination of the quantized “these pizzas”, which is bounded,
and the partitive case on the noun denoting pizzas.
In presentational or existential sentences and certain transitive sentences
with achievement verbs, the mass or count properties of the argument, as in
(37a), are observed to influence the possibility of partitive case encoding.15 On
the basis of typical examples illustrating this regularity, one could argue that in
presentational or existential sentences, partitive is possible with mass (abstract,
unbounded, or bare plural) nouns. The examples that those scholars would
15 See also Erelt et al. 1997, available at http://www.eki.ee/books/ekk09/index.php?p=5&p1=

2&id=387 (accessed on 17 November 2012).
employ would be as presented in the Estonian intransitive sentences in (37a)

and as in the transitive sentences in (37b). I have added the semantically un-
acceptable options with a hash mark.
(37) a. Taigna sees on pipar-t /#sõrmus-t.

batter.GEN in be-3SG pepper- PAR /ring- PAR
‘There is (some) pepper in the batter.’
Not possible: ‘There is (some/a) ring in the batter.’
b. Mari sokuta-s taigna sisse pipar-t /#sõrmus-t.

M[ NOM ] manage- PST.3 SG batter.GEN into pepper- PAR /ring- PAR
‘Mary managed to add (secretly) pepper to the batter.’
Not possible: ‘Mary managed to add a ring to the batter.’
Further research has shown that there should be a another subdivision among
mass nouns, and that concrete mass nouns belong to just one of these types.
Abstract nouns that are also mass nouns can be divided according to their
appearance with the partitive marking in the same environment with other, con-
crete mass nouns, or not. Some abstract mass nouns cannot be marked with the
partitive if all other conditions hold equally. The contrast is demonstrated by
the nouns valgus ‘light’ versus pimedus ‘darkness’. In its case-marking behavior,
the mass noun valgus ‘light’ patterns with other mass nouns, such as pipar
‘pepper’, but the mass noun pimedus ‘darkness’ does not, pattening with the
count noun sõrmus ‘ring’ instead. The contrast is illustrated in transitive and
intransitive sentences (38).
(38) Estonian
a. Saali tekki-s valgus-t /#pimedus-t.
hall. ILL appear- PST.3 SG light- PAR darkness- PAR
‘Light/darkness emerged in the hall.’
b. Mari tekita-s saali-s valgus-t /#pimedus-t.

M[ NOM ] create- PST.3 SG hall- INE light- PAR /darkness- PAR
‘Mary created light/darkness in the hall.’
The possibility of the partitive is determined by factors that are similar to exam-
ple (32). In example (32), the linguistic partitive emerges if the noun measured
is semantically scalar. The contrast is observable in the difference between the
derivations of soe ‘warm’ and külm ‘cold’, which can be partitive, and the non-
derived noun temperatuur ‘temperature’, which cannot. In example (38), the dif-
ference between the nouns ‘light’ and ‘darkness’ consists in the scalar properties
of the base adjective, and also the existence of pragmatic standards or norms
130 Anne Tamm
(see the details in Tamm 2014). As opposed to the Finnish Partitive Concepts that
are expressed by a linguistic partitive, there are examples from Uralic where
partitive-like phenomena are actually neither Partitive Concepts nor linguistic
partitives. Consider the so-called Nenets partitive objects as in jī-kʔ tādaʔ ‘give
[me some] water’ (Hajdú 1968/1982: 69). Is the Tundra Nenets dative plural a
partitive (-kʔ above, the more general dative plural suffix form is xVʔ Hajdú
1982: 38)? A Linguistic Partitive is ruled out, since the dative plural has no links
to nominal morphology denoting source. Hajdú (1982: 69) interprets the dative
plural NPs as “partial objects”. As the result of grammaticalization processes
that significantly diverge from those leading to the Finnic case alternation, the
Nenets NP that stands for the event participant with the semantic role of a
Theme can be marked with either the accusative or dative plural ( jī-kʔ water-
dat.pl). Tereščenko (1973), however, refrains from classifying the nominal marker
as case, and the NP as an object, and since the dative plural NP does not trigger
objective conjugation in Nenets, there are reasons to prefer Tereščenko’s analysis.
Do the specialized uses of the Tundra Nenets dative plural correspond to the
Partitive Concept? No part of a quantized whole is implied. Some examples in
Tereščenko (1973) are close to the Partitive Concept, namely, close to the exten-
sion of the amount-of partitive in that the nouns in the examples seem to refer to
a kind rather than concrete – moreover, plural – referents. These uses seem to
coincide with the “Mordvinian Partitive”, namely, the Erzya ablative in example
(15c), but Tereščenko (1973: 185–187) provides a range of many other uses.
The noun ‘meat’ does not refer to a whole in the accusative example either.
Tereščenko finds parallels with the Finnic linguistic partitive-accusative object
case alternation only in the semantic totality and partiality and indefiniteness,
which later in the explanation boils down to a broader notion of nominal divi-
sivity (Tereščenko 1973: 187). This suggests that there is no evidence of any part-
whole opposition in the Nenets grammar system. In sum, the dative plural-
accusative alternation has led researches acquainted with the Finnic gramma-
ticographic tradition to consider the parallel with the Finnic Linguistic Partitive
as part of total-partial opposition. We have rather the case where the language-
specific categories of Finnic grammaticography have influenced the description
of the categories in another Uralic language.
4.6 Cross-categorial cases and separative relationships in

source cases
4.6.1 Partitives among the source cases as cross-categorial case
This subsection takes a closer look on how separative cases and partitives
appear as cross-categorial cases, because having extensive cross-categorial case
systems in various levels of development is a particular feature of the Uralic

languages. Cross-categorial case refers to the appearance of nominal markers,
in particular, case formants, as markers that pertain semantically to verbal cate-
gories, sometimes forming case paradigms in the verbal marking system, such
as the TAM categories or negation. The goal of this subsection is to present
several points.
– Firstly, Uralic languages are particularly rich in cross-categorial source cases
and the partitive. There are several strategies, two of which are disussed in
connection with the source cases in the cross-categorial case systems.
– Secondly, the partitive developments in the Finnic languages must be seen
as an exceptional development among the source cases. The source cases
diverge from each other in languages with several source cases in terms of
cross-categorial case. Source spatial cases are also different from non-source
spatial cases. Especially in the Hungarian aspectual preverb system, the
contrast between source and non-source cases is observable.
– Last but not least, the discussion of the non-finites of this subsection is
necessary to understand the role of the partitive case in the TAM categories.
More specifically, it is a necessary introduction in order to move on to the
topics of how the TAM categories have developed their current spectrum of
meanings in the instance of the Estonian linguistic partitive. The Estonian
partitive evidential, discussed in Section 5, is an instance of cross-categorial
case.
Unusual TAM marking by nominal case is attested in many languages

(Nordlinger and Sadler 2004). Recent research has drawn attention to “verbal”
or “versatile” case that appears in the verbal paradigm (Aikhenvald 2008, Butt
2006, Spencer 2009). Blake (2001) and Butt (2006) belong to the classic literature
addressing case that contains instances of cross-categorial case, and Aikhenvald
(2008) has published a journal article dedicated to case that appears on nouns
as well as verbs in language typology. Spencer (2009) discusses instances of the
phenomenon as “case marking on verbs” (39).
(39) Quechua
Rima-y-ta xalayu-ru-n.
speak- INF-ACC begin-PRF-3 SG
‘He began to speak.’
(Adelaar and Muysken [2004: 226] in Spencer [2009: 189])
According to the author’s description, the non-finite verb form is marked with
the accusative and it functions as the object of the verb ‘begin’ in Quechua.
132 Anne Tamm
This example adds another dimension of cross-categorial case ridden with

puzzles – non-finites combining with cases.
The semantic regularities of cases appearing cross-categorially have been
discussed in a more detailed account by Aikhenvald (2008) in recent literature.
Her term for the phenomenon is versatile case (Aikhenvald 2008: 565). It can
express temporal, causal and other relationships between clauses, or aspectual
and modal meanings within a clause. Versatile case comprises case on various
verb forms and falls in three types on the basis of its distributional characteris-
tics: appearing on verb roots, on fully or partially inflected verbs, or on non-
finite verbs. Aikhenvald describes versatile case as “chameleon morphemes”;
these morphemes can mark different categories and have related but also differ-
ent meanings. As one instance from her rich typological sample, Aikhenvald
provides examples of case on nouns and verbs in Manambu, where the objective-
locative case marks a core or oblique argument, as demonstrated in (40). The
locative case appears on the verb as well, as on wukemar ‘forget’, adding com-
pletivity to the event structure; locative case on a verb triggers the aspectual
completive interpretation of ‘completely forget’.
(40) Manambu
Wun [de-ke-m] wukemar-e-m
I he- LK- OBJ / LOC forget-LK- OBJ / LOC
‘I completely forgot him.’
(Aikhenvald 2008: 587)
The Manambu case expresses TAM categories; in the Quechua example it is not
clear if the accusative marks aspectual inchoativity on the verb or simply the
object. Aikhenvald generalizes that core cases tend to express aspectual and
modal meanings, while oblique cases tend to be used as clause-linkers.
In addition to attaching to nouns (and in languages with adjective-noun
agreement, to adjectives), case in Uralic also attaches to verbs (verb stems) and
to verbs with a nominalizing suffix, forming infinitives and in-between forms.
These are the forms that are also referred to as cross-categorical cases. Some
examples of the types of cross-categorial case are listed in (xiv) in order to intro-
duce in a nutshell the basic types of examples discussed below in more detail.
(xiv) Cross-categorical case types
1. Cross categorical case-like form attaches to verb stems
– Hungarian case form in preverbs+verb (be-megy ‘in-go, enter’), or
– Udmurt V+abessive, or
– to verb stems forming a non-finite, such as the Selkup infinitive
marker: V+translative.
2. Cross-categorical case attaches to nominalizations (Udmurt cases V+m

+case, V+n+case),
3. Cross-categorical case appears as a part of non-finites (Finnic, the case
formants are part of a morpheme of a non-finite verb)
4. Cross-categorical case attaches to an argument or adjunct but carries
TAM meanings (Finnish aspectual accusative-partitive case alternation,
accusative case on temporal adverbials that bound the situation).
In a cross-categorical case paradigm, not all nominal cases are necessarily

cross-categorial; in addition, cases are cross-categorial in various ways as speci-
fied in (xiv). However, the Uralic languages rarely combine all types of cross-
categorical case and have their particular language-specific strategies in using
case formatives for expressing the TAM meanings. It is also typical that there is
more than one strategy within one language, but across the Uralic languages,
the pattern of the strategies does not overlap.
In order to understand the extended semantics of linguistic partitives, the
difference between simple nouns and nominalizations should be considered. A
simple noun has relatively concrete properties, and a case relates it to a verb in
a rather unambiguous way. The result of relating verbs with nominalizations is
more complicated. A nominalization has its own predicate-argument structure
and a case relates events or abstract entities. In this linguistic environment,
which is semantically and pragmatically similar to the combinations of adposi-
tions and nominalizations, several TAM meanings may develop; Bybee et al.
(1994) discuss for instance how the progressive develops from locative expressions
(Tamm 2011a or Ross et al. 2010 can be consulted for the details of the develop-
ment of the progressive semantics and pragmatics in the Finnic languages).
4.6.2 Rich source case system with asymmetries in combining with the
nominalization paradigm, and case on bare verb: Udmurt
The goal of this subsection is to illustrate a stage in the development of semantics

that goes beyond encoding only argument relationships in a predicate. The
example is from Udmurt, where cases that mark nouns can mark bare verb
stems and two types of non-finites or nominalizations. The data presented in
Table 8 illustrates the case system in Udmurt.16 It highlights the source cases and
16 The Udmurt data are provided by Svetlana Edygarova, p.c, to whom I am grateful for her
help.
134 Anne Tamm
the abessive as they are found on non-derived nouns, and in the forms between
verbs and nouns containing -n- or -m- as bound nominalizing morphemes.
Table 8: Cases in the Udmurt noun (house) and non-finite case paradigms (the verb go)
Case noun + case verb + n + case verb + m + case verb + case

Nominative s’ik myn-on myn-em
Genitive s’ik-len myn-on-len myn-em-len
Accusative s’ik/s’ik-ez myn-on-ez myn-em-ez
Ablative s’ik-les’ myn-on-les’ myn-em-les’
Dative s’ik-ly myn-on-ly myn-em-ly
Adessive s’ik-len
Instrumental s’ik-en myn-on-en myn-em-en
Abessive s’ik-tek missing missing myny-tek
Inessive s’ik-yn myn-on-yn myn-em-yn
Illative s’ik-e myn-on-e myn-em-e
Elative s’ik-ys’(t) missing myn-em-ys’
Terminative s’ik-oz’ myn-on-oz’ myn-em-oz’
Egressive s’ik-ys’en missing missing
Prolative s’ik-eti
Approximative s’ik-lan’
The source case-marking of nominalizations is relatively uninteresting from the

point of view of syntax. Once a verb is nominalized, it can be case-marked
as any other argument in the source role. Therefore, one would predict an even
distribution of forms all over the nominalizations and non-finites. However, we
do not see an even distribution.
In Table 8 illustrating the case system in Udmurt, also the abessive on bare
stems is included for illustration of the nature of the phenomenon, in addition
to the case system on n-nominalizations and on m-nominalizations. Although
the system is regular, not all cases appear on the two nominalizations, and
only one morpheme is cross-categorial in the sense of combining with a verb
stem (the abessive, see Tamm forthc., 2011c). This demonstrates a system with
transparent cross-categorial case, that developes some additional meanings to
the argument relationships. Table 1 on Estonian records a more restricted and
opaque cross-categorial case system, with more TAM meanings.
In sum, Udmurt presents a rich source case system with asymmetries in
combining with verbs and the nominalization paradigm. The relationships are
transparent, but the paradigms are deficient. The elative source case does not
occur with the n-nominalizations, but it is there with the m-nominalizations.
The ablative source case occurs with the n-nominalizations as well as with the
m-nominalizations. The abessive, which combines directly with the bare stem,
does not combine with the two non-finites. It is also clear that the elative and
the illative appear asymmetrically, the elative is missing where the illative is
present, showing the constraints on combining with more abstract meanings.
The source-goal asymmetry is the topic of the following subsection as well.
4.6.3 Source-goal asymmetry in cross-categorial case
One of the best examples of cross-categorial case in the aspectual domain is the
Hungarian verbal particles (preverbs). I illustrate the Hungarian aspectual teli-
cizing particles that have the same origin with the pronominal forms of goal
cases INTO (-ba/-be), ONTO (-ra/-re), and TO (-hez/-hoz/-höz) in (41).17
(41) Hungarian
a. Feri be-ment az épület-be.
F[NOM ] into-go. PST.3 SG DEF building-INTO
‘Ferenc entered the building.’ (into-went into the building)
b. Gábor rá-lépett a sajt-ra.

G[NOM ] onto-step. PST.3 SG DEF cheese-ONTO
‘Gábor stepped on cheese.’ (on-stepped on the cheese)
c. Gregor Bernadett hozzá-ment egy sámán-hoz.18

GB[NOM ] to.him-go. PST.3 SG INDF shaman-TO
‘Gregor Bernadett married a shaman.’ (to-went to a shaman)
The Hungarian stative particles have the form of the adverb or pronominal case
forms denoting the location IN, ON, or IN THE VICINITY OF with the verb marad
‘stay’. The point is that it is not only interesting what there is in the grammatical
inventory of a language, but what there isn’t. In the triad of stative location pre-
verbs, IN and ON appear with an obligatory argument in the case that corre-
sponds to the preverb, as in (42a), (42b). The form nála ‘near’ that stands for IN
THE VICINITY location may appear with the verb marad ‘stay’ (42c), but there is
no obligatory argument marked with -nál/-nél.
17 Surányi (2009) treats several of these examples as incorporated locative adverbials in

Hungarian. See É. Kiss (2006) and Kiefer (2006) for the event structural properties of preverbs
in Hungarian.
18 http://velvet.hu/celeb/gregor0610/ 05/10/2011 16:12
136 Anne Tamm
(42) Hungarian
a. Nikolas megsirat-t-a, hogy benn-marad-t a verseny-ben.19
N[NOM ] lament-PST.3 SG - DEF that in-remain-PST.3 SG DEF competition-INE
‘Nikolas was sad that he remained in the competition.’
(Literally: in-stayed in the competition)
b. Rajt-a marad-t Thaci-n a szervkereskedelem

on-3 SG remain-PST.3 SG T-SUPERESS DEF organ.trafficking[NOM ]
vád-ja20
accusation[NOM ]-POSS .3 SG
‘The accusations of organ trafficking remained on Thaci.’
(Literally: on-stayed on Thaci.)
c. *nála marad-t a barát-jái-nál

near remain-PST.3 SG DEF friend-POSS .3 PL-ADE
‘He remained at his friends’ place.’
(Literally: at-stayed at his friends’ place)
In (43), the separative direction is not realized in the aspectual preverb system.
Separative preverbs are something that is paradoxically and unexpectedly miss-
ing in the highly developed Hungarian aspectual preverb system.
(43) Hungarian
a. *ról-jött a tető-ről
from.top.of-come.PST.3 SG DEF roof-DELA
The following content cannot be expressed with these grammatical
means: ‘He came down from top of the roof.’
b. *ből-jött a ház-ból
from.inside-come.PST.3 SG DEF from-ELA
means: ‘He came out of the house.’
c. *től-jött től-e
from.the.vicinity-come.PST.3 SG ABL-him/her
means: ‘He came from his place.’
19 www.borsonline.hu/news.php?hid=36212 02/10/2011 13:45:28

20 http://www.mr1-kossuth.hu/hirek/kulhon/rajta-maradt-thacin-a-szervkereskedelem-vadja.
html 02/10/2011 13:45:28
4.6.4 Source-goal cross-categorial case and the special status of the

Linguistic Partitive
Partitives and source cases appear on non-finites in Estonian. Non-finite forms

in Estonian frequently originate from case-marked non-finite verb forms, which
originally were complements but developed into base predicates of larger predi-
cate complexes. This process resembles the Udmurt paradigm, which is more
regular but less grammaticalized. These cross-categorial case complexes have
developed case-related semantics and TAM meanings in Estonian.
Table 9 The Estonian case system and case in the non-finite system
case translation N example N example V translation V

Nominative book raamat
Genitive of a book raamatu
Partitive (of) a book raamatu-t tule-va-t allegedly, coming
Illative into the book raamatu-sse tule-ma to come
Inessive in a book raamatu-s tule-ma-s, tulle-s coming
Elative from (inside) a book raamatu-st tule-ma-st from coming
Allative onto a book raamatu-le
Adessive on a book raamatu-l
Ablative from the book raamatu-lt
Translative in(to), as a book raamatu-ks tule-ma-ks in order to come
Terminative until a book raamatu-ni
Essive as a book raamatu-na
Abessive without a book raamatu-ta tule-ma-ta not having come
Comitative with a book raamatu-ga
The embeddeness of case in the system of non-finites helps understand the role
of source cases in the TAM categories and, more specifically, how the TAM cate-
gories have developed their current spectrum of meanings. The partitive devel-
opments in the Finnic languages must be seen as an exceptional development
among the source cases, because of the asymmetry between the source and
goal cases in the cross-categorical case systems. The exceptionally developed
source case cannot be explained with frequency, because source cases are less
frequent as parts of non-finites. This is shown in well-known case studies on
the development of non-finites (infinitives), which are usually based on goal
and not source cases. For instance, Haspelmath (1989) discusses the develop-
ment of infinitives on the basis of allatives (e.g. I go to eat and not I come from
eat). He also discusses the Hungarian infinitive as a form that has developed
from a nominalizer and a lative case form in the Hungarian language history.
Moreover, Pajusalu and Orav (2008) have shown that in Estonian the elative
138 Anne Tamm
source case is statistically far more rare as part of the non-finite with the m-
formative compared to the other m-formative spatial cases in non-finites. The
Estonian partitive evidential is therefore fairly special, and this is discussed in
Section 5.
4.7 Summary
After the discussion of this section, the quantificational relationships between
the nominal parts of partitives can be presented in a table. The relationships
between the two nouns in the two Partitive Concepts are represented in Table 10.
Table 10: Partitive Concepts and the quantification of the nouns in the constructions
Part-of-N Amount-of-N
N quantified Yes/no no
Construction quantified Quantified nonspecifically Either specific quantity (if
amount is specified) or non-
specific quantity (if amount is
not specified or non-quantized)
Combined with verbs Quantizes quantifyable activity De-quantizes a quantifiable
activity
The different morphological and syntactic realizations with extended or core

meanings in the Linguistic Partitives and Partitive Concepts are in Table 11. The
equation mark stands for identical kinds of matter, whereas the hash mark
stands for different kinds of matter. N1 stands for measure unit and N2 for the
substance.
Table 11: Linguistic Partitives and Partitive Concepts and the relationship between the two
nouns
Morphological Separatives in lan- Separatives in

partitive N1-of-N2 guages with morpho- languages without a
logical partitives dedicated partitive
Identity of the N1 = N2 The relationship is The relationship is
relationship between underspecified, but underspecified, but
the partitives typically, N1# N2 typically, N1# N2
The form Case Elative, ablative, Another case, one
delative, several function of which is
source/separative the partitive, or an
adpositions adposition, juxta-
position
5 The semantics and pragmatics of the Estonian

partitive evidential
5.1 An instance of a Linguistic Partitive
This section presents some instances of the partitive that have already less
relation with the Partitive Concept (Section 5.1) because they have entered the
epistemic modal and evidential domain. Then the section concentrates on the
conceptual content of the Estonian partitive evidential in the light of cross-
categorial cases and partitive relationships (Section 5.2).
In most Uralic languages that have a partitive case, the latter is semantically
motivated, especially in Finnic. Instead of the “part-of-N” semantics, the semantics
of the Finnic partitives is related to other semantic notions, typically, event
structural properties or aspect, as in (44) (Ackerman and Moore 2001, Kont
1963, Tveite 2004). Indefiniteness and irrealis have also been mentioned in
some sources. Sentences with the partitive object are referred to as non-bounded,
irresultative, imperfective, and atelic. Sentences with an accusative (total) object
are referred to as bounded, resultative, perfective, and telic. Presently, the
Estonian partitive object case appears in sentences that have the semantics of
incomplete event realization, unboundedness, atelicity, or imperfective aspect
(Kiparsky 1998, Erelt et al. 1993, Metslang 1994, Metslang 2001, Larsson 1983,
Lees 2005, Sulkala 1996), irrespective of part-whole relationships or partial affect-
edness (Tamm 2012c, 2007, 2004). The partitive marking of incremental themes
denoting the part-of the object relationship with the verb, (44a) versus (44b),
have given rise to a general aspectual marker of aspectual unboundedness,
unrelated to affectedness. Hearing does not affect, let alone, piecewise, or in
any possible way, the one who is being heard in the event, as in (44c). The
noun phrase ‘Thomas’ is not an affected object, since it refers to a quantized
and specific referent. Therefore, the case marking on the basis of affectedness
of the incremental theme has given rise to the meanings of atelicity/telicity and
aspect in general.
(44) a. Silvi sõ-i pitsa-sid.
S[NOM ] eat-PST.3 SG pizza-PAR . PL
‘Silvia was eating the pizzas.’
b. Silvi sõ-i pitsa-d ära.
S[NOM ] eat-PST.3 SG pizza-NOM . PL up/TELIC
‘Silvia ate the pizzas up.’
c. Silvi kuul-is Toomas-t.
S[NOM ] hear-PST.3 SG T-PAR
‘Silvia heard Thomas.’
140 Anne Tamm
Another analysis of these examples can be given in terms of epistemic mo-

dality and is connected in an intricate way to evidentiality. This analysis integra-
tes the discourse and perceptional properties of the partitive objects (Tamm
2012b). The multiple previous attempts at defining the Finnic partitive can be
replaced by an overarching epistemic modal approach. Instead of viewing the
Finnic partitive as a matter of encoding definiteness, boundedness, quantization,
perfectivity, or telicity oppositions, it is possible to view it as encoding an opposi-
tion in terms of having sufficient evidence for definiteness, boundedness, quan-
tization, perfectivity, or telicity. This approach explains many unexplained facts
about verb classes as well. The epistemic modal feature completely overrules the
aspectual feature – previously thought to determine object case encoding – in a
whole class of psychological predicates (üllatama ‘surprise’, ehmatama ‘frighten’,
solvama ‘offend’).
(45) a. Silvi üllata-s Toomas-t.

S[NOM ] surprise-PST.3 SG T-PAR
‘Silvia surprised Thomas.’
b. Silvi ehmata-s Toomas-t.

S[NOM ] frighten-PST.3 SG T-PAR
‘Silvia frightened Thomas.’
c. Silvi solva-s Toomas-t.

S[NOM ] offend-PST.3 SG T-PAR
‘Silvia offended Thomas.’
The psych-verbs denote events with clear temporal endpoints and are,
thus, aspectually telic in (45). Therefore, they are predicted to have total (non-
partitive, accusative, nominative) objects according to the hypothesis of linking
partitive to aspect. However, these verbs have partitive objects. An utterance that
originally encodes the speaker’s aspectual meaning and conveys the endpoint
also encodes a lower degree of evidence, thus, epistemic modality. The crucial
link between the aspect and evidential categories is instantiated when the
hearer overgeneralizes some examples where there is clearly no cognitively reli-
able immediate evidence of the endpoint (psych-verbs) to more fuzzy cases.
The deviant behavior of the psych-verbs can be explained in terms of epistemic
modality: it is difficult to have evidence about the endpoint of offending, for
instance. In an event of surprising or frightening as well, it is not easy to have
evidence when an event reaches its inherent endpoint and how effectively the
endpoint is reached. Therefore, the evidence for events that are encoded by
psych-verbs is incomplete. The incompleteness of the evidence and not just

aspect determines the partitive encoding of the object’s case with these verbs.
This analysis is based on the speaker’s cognition and inferences combined
with the hearer’s inferences. The pragmatic and semantic account concentrates
on the property of partitive to encode a speaker’s evaluation of evidence and
certainty. It is not surprising to find that the partitive evidential expresses a
speaker’s evaluation about the evidence for the truth of a proposition and certainty
about the proposition. However, unexpectedly, this pragmatics can be found in
the Estonian core case marking system. More specifically, the data show that the
inference about the total objects is that they express total evidence and certainty
and that the partitive objects express incomplete evidence and uncertainty.
This explanation can capture some previously unexplained Finnish data as
well. Finnish mental epistemic verbs with translative secondary predicates dis-
play an object case alternation that cannot be fit into an explanation based on
the telicity-atelicity opposition either. The two sentences illustrated below, with
tietää ‘know’ in (46a) and luulla ‘think, believe’ in (46b), are atelic and should,
therefore, have partitive object case marking if the telicity hypothesis is correct.
However, the use of tietää ‘know’ versus luulla ‘think, believe’ expresses the
beliefs of the speaker about Mary’s knowledge about the smartness of George.
The degree of the speaker’s evidence for George being smart is lower if luulla
‘think, believe’ is used, and the degree of evidence is complete if tietää ‘know’
is used. The difference in evidence is reflected in the speaker’s choice of the
accusative case with tietää ‘know’ and confirmed by the choice of the matrix
verb.
(46) Finnish
a. Mari tietä-ä Jyri-n viisaa-ksi.
M[ NOM ] know- PST.3 S J-ACC smart-TRAN
‘Mary knows that George is smart.’
(Tuomas Huumo, p.c.)
b. Mari luule-e Jyri-ä viisaa-ksi.

M[ NOM ] believe-PST.3 S J-PAR smart-TRAN
‘Mary believes that George is smart.’
(Tuomas Huumo, p.c.)
Mary (the subject) may believe that George is smart in (46a) and (46b), but the
belief of the subject does not matter for the object case encoding. What matters
are the speaker and her beliefs. The speaker believes that Mary is right in (46a)
and George is smart, but that she (the speaker) has insufficient evidence in
(46b). In (46b), the evidence may be insufficient, for instance, because (a) the
142 Anne Tamm
speaker does not trust Mary when Mary says George is smart, or I know that
George is smart or, (b) the speaker trusts Mary, but Mary utters I think that
George is smart (but I do not know for sure).
The Finnish examples with the mental epistemic verbs (tietää ‘know’ versus
luulla ‘think, believe’) seem to have lexicalized the epistemic modal distinction
in the verbal features. The epistemic modal feature completely overrules the
aspectual feature in Finnish object marking in this minimal pair. Several Finnish
perception verbs and mental epistemic verbs have accusative objects in Finnish
as in (47a), unlike Estonian, which has partitive objects with this group of verbs
(47b) (tuntea/tundma ‘feel’ and nähdä/nägema ‘see’). It is difficult to say if epis-
temic modality or the categorization of events is involved in this difference. The
same Finnish verb may also appear with a partitive object because of aspectual
reasons, which means perception as a lexicalized change of state from not feel-
ing to feeling (corresponding to accusative marking, 47a) versus a continuous
state of feeling (corresponding to partitive marking, as in 47c).
(47) a. Finnish
Tun-si-n sen melkoisen selvästi.21
feel- PST-1 SG this.ACC pretty clearly
‘I could feel it pretty clearly.’
b. Estonian
Tund-si-n seda üsna selgesti.
feel- PST-1 SG this.PAR pretty clearly
‘I could feel it pretty clearly.’
c. Finnish
Osa minu-sta tun-si sitä tuska-a,
part[ NOM ] 1 SG - ELA feel- PST.3 S this.PAR anguish-PAR
mitä Jacob tun-si.22
that.PAR J[NOM ] feel- PST.3 S
‘Part of me felt the kind of pain that Jacob felt.’
In any case the lexicalization pattern involves a whole distinct cognitive area of
perception and reasoning in a language, and this fact is too conspicuous to
ignore. Table 12 summarizes the correspondence between the speaker’s evidence
about the endpoint and the object case marking in Estonian.
21 http://keskustelu.suomi24.fi/node/9776187 (accessed 18 September 2011).

22 http://www.vampirelove.net/fanfiction/onnieiseuraa.php (accessed 18 September 2011).
Table 12: The speaker’s evidence about the endpoint and the object case
The speaker’s evidence, certainty Complete Incomplete

Object case total (accusative) partitive
5.2 Cross-categorial cases and partitive relationships

Another peculiarity of the Uralic languages-case on non-finite verb forms (cf.
Aikhenvald 2008) – has lead to the situation where the partitive semantics has
spread to the domain of epistemic modality and evidentiality (the vat-form, as in
Erelt et al. 2006, Kehayov 2008). As the result of the diachronic development
of the partitive case, the Estonian evidential, epistemic modal, aspectual and
NP categories share similarities. The overlap between the epistemic modal and
evidential categories (Van der Auwera & Plungian 1998) is due to these develop-
ments. Aspects of the diachronic development have been dealt with in Larjavaara
(1991), Campbell (1991), or Aikhenvald (2004), but a coherent motivated link
between the sematics of the various Uralic partitives will be given in this presen-
tation is a new research question that is addressed in this article only. Partitive
has the following stages of development in Estonian:
1. an NP-stage (Krifka 1992), that is, the stage where the meaning of the
partitive pertains to parts of a whole
2. an aspectual stage (Larjavaara 1991, Laanest 1975),
3. epistemic modal and evidential phase (Campbell 1991).
The aspectual partitive marks objects in sentences describing incomplete events,

and the partitive evidential appears in sentences that encode incomplete evidence
compared to the expectation of complete evidence (Tamm 2009).
The partitive case provides the example of cross-categorial case, which in
present-day Estonian preserves the diachronic evolution path from a spatial
case to an aspectual case and further, to a marker of epistemic modality and
evidentiality. The categories of aspect and evidentiality preserve the basic
semantics of the spatial partitive; the example provides an illustration of the
shared structure of these categories.
Once the participle was used with an object, the partitive case marking was
applied to participles as well if they modified the object of an atelic verb. As a
consequence, epistemic modal meanings emerged in the embedded clauses
(Wälchli 2000, Tamm 2008b, 2009). In modern Estonian, there is still con-
siderable variation, but clear tendencies can be noticed as well. With auditory
evidence, the partitive form is used (47a), since evidence from hearing is not
as reliable as evidence from seeing. Visual evidence is not partial; another non-
finite tends to be used with visual evidence (47b). In independent clauses, the
144 Anne Tamm
partitive-marked participle began to be used as an indirect evidential (47c). If

evidence is not in question, no partitive evidential is used (47d).
(47) a. Mari kuul-is teda koju tule-va-t.

M[NOM ] hear-PST.3 S s/he.PAR home. ILL come-PERS . PRS . PTCP- PAR
‘Mary heard him/her come home.’
b. Mari näg-i Jüri-t koju tule-mas.

M[NOM ] see-PST.3 SG J-PAR home. ILL come-M _ INE
‘Mary saw George coming home.’
c. Mari tule-va-t.
M[NOM ] come-PERS . PRS . PTCP- PAR
‘Allegedly/reportedly, supposedly Mary will come.’
d. Mari tule-b.
M[NOM ] come-3SG
‘Mary will come.’
Table 13 summarizes the correspondences between the completeness of the

speaker’s evidence or certainty and the presence or absence of the partitive evi-
dential. Complete evidence and certainty is not specifically marked, but incom-
plete evidence or certainty of the speaker is marked by the partitive evidential.
Table 13: The completeness of the speaker’s evidence or certainty and the partitive evidential
speaker’s evidence, certainty complete incomplete

mood/modality no partitive evidential partitive evidential
6 Discussion and conclusions

The partitive has stood in the center of descriptive studies for long decades in
the Finnic linguistic tradition, and in the last two decades the phenomena of
partitivity have thrilled formal linguists as well. A recent debate on the com-
parability of grammatical concepts across languages has involved cognitive,
generative, and typological linguistic frameworks. The present paper has united
the agendas of these pursuits with the some older and some more recent studies
and analyses of the partitives and partitive like concepts across the Uralic lan-
guages, with the aim of creating a suitable basis for further investigations into
the cognitive properties of linguistic concepts such as the partitive. One of the
intriguing questions concerns how the human prelinguistic perceptual concepts

are related to linguistic and language-specific, more refined concepts. The diverse
patterns of the Uralic separative and partitive phenomena have been studied
in more detail in the present article in order to establish the correspondence
between variations of a central concept.
In the hope that further empirical, experimental or language historical
methods will confirm the existence of the partitive as a naturally occurring –
that is, a perceived and communicated – category, this article has proposed a
hypothesis about the cognitive content of the Partitive Concept. It has provided
examples from the Uralic languages, and it has discussed some case studies of
particular instances of the Linguistic Partitives.
Several Uralic languages have cases that are referred to as the partitive;
however, the semantics of these cases diverge from the generally assumed
notion of partitive. All Uralic languages can, however, express the concept of
part-whole relationships by means of a restricted set of constructions that typi-
cally contain juxtaposed bare nouns, elatives, or ablatives. Therefore, on the
basis of examples of one language family, this paper made a distinction between
Partitive Concepts that can be expressed by all Uralic languages and the mor-
phological, linguistic partitives.
A characteristic of the Uralic languages is that there are many source (sepa-
rative) cases, and this article concentrated on the place of the partitive within
the system of source cases. The interaction between TAM, definiteness, and the
partitive can be observed in many areas: the aspectual DOM, definiteness
effects, telicity, and case on non-finites. The Uralic languages are particularly
rich in cross-categorial case, that is, the use of case formants as markers of
verbal categories such as TAM categories. There are several processes that lead
to case formants developing into TAM categories, two of which have been dis-
cussed in connection with the source cases in the cross-categorial case systems.
The source cases are different from each other in languages with many source
cases in terms of cross-categorial case, and also different from non-source spa-
tial cases. Especially, in the Hungarian aspectual preverb system, the contrast
between source and non-source cases has been made visible. Therefore, the
partitive developments in the Finnic languages must be seen as an exceptional
development among the source cases.
This article has illustrated the current spectrum of TAM meanings in the
instance of the Estonian linguistic partitive, in particular, the Estonian partitive
evidential. The Estonian partitive case provides an example of cross-categorial
case, which in present-day Estonian preserves the diachronic evolutionary path
in present-day uses. The present diversity in the meanings and functions allows
us to assume a development of a concept from a spatial relation, to an aspectual
146 Anne Tamm
relation and further, to epistemic modal and evidential interpretations. The

categories of aspect and evidentiality preserve the basic semantics of the spatial
partitive; the example provides an illustration of the shared structure of these
categories.
The Partitive Concept is a heuristic, a comparative concept to enable us to
compare variations of it across languages. Additionally the Partitive Concept
can be hypothesized to correspond to a prelinguistic perceptual concept. In
the descriptions given in the present article, it has fixed “standard” semantic
properties, and the Linguistic Partitive cases have developed their specific
semantics and pragmatics, or have bleached meanings in each Uralic language
where the partitive case appears. The message for linguistic typology is that
typology needs cross-linguistic cognitive conceptual categories, which can be
established in pre-linguistic humans by experimental testing.
References
Abondolo, Daniel (ed.). 1998. The Uralic languages (Routledge Language Family Descriptions
Series). London & New York: Routledge.
Ackerman, Farrell & John Moore. 2001. Proto-Properties and Grammatical Encoding: A
Correspondence Theory of Argument Selection. Stanford, CA: CSLI.
Adelaar, Willem F. H. & Pieter C. Muysken. 2004. The languages of the Andes (Cambridge
language surveys). Cambridge: Cambridge University Press.
Aikhenvald, Alexandra Y. 2008. Versatile cases. Journal of Linguistics 44(3). 565–603.
Aikhenvald, Alexandra. 2004. Evidentiality (Oxford Linguistics). Oxford: Oxford University Press.
Anttila, Arto & Vivienne Fong. 2000. The Partitive Constraint in Optimality Theory. Journal of
Semantics 17. 281–314.
Ariste, Paul. 1968. A grammar of the Votic language. Bloomington: Indiana University.
Auwera, Johan Van der & Vladimir Plungian. 1998. Modality’s semantic map. Linguistic
Typology 2. 79–124.
Brattico, Pauli. 2011. Case assignment, case concord, and the quantificational case construc-
tion. Lingua 121. 1042–1066.
Butt, Miriam. 2006. Theories of case. Cambridge: Cambridge University Press.
Bybee, Joan L., Revere Perkins & William Pagliuca. 1994. The evolution of grammar: Tense,
aspect, and modality in the languages of the world. Chicago & London: University of
Chicago Press.
Campbell, Lyle. 1991. Some grammaticalization changes in Estonian and their implications. In
Elizabeth C. Traugott & Bernd Heine (eds.), Approaches to Grammaticalization (1), 285–
299. Amsterdam: Benjamins.
Carey, Susan. 2009. The Origin of Concepts. Oxford: Oxford University Press.
Collinder, Bjőrn. 1960. Comparative Grammar of the Uralic Languages. Stockholm: Almqvist &
Wiksell.
Dahl, Östen & Fred Karlsson. 1975. Verbal aspects and objects marking: a comparison between
Finnish and Russian. Logical grammar reports, Göteborg: Department of Linguistics,
University of Göteborg.
Denison, Norman. 1957. The Partitive in Finnish. Annales Academiae scientiarum Fennicae.
Series B. 108. Helsinki: Suomalaisen tiedeakatemia.
Erelt, Mati, Tiiu Erelt & Kristiina Ross. 1997. Eesti keele käsiraamat [The handbook of Estonian].
Tallinn: Eesti Keele Sihtasutus.
Erelt, Mati, Reet Kasik, Helle Metslang, Henno Rajandi, Kristiina Ross, Henn Saari, Kaja Tael
& Silvi Vare. 1993. Eesti Keele Grammatika II. Süntaks. Lisa: Kiri. [The Grammar of the
Estonian Language II. Syntax]. Tallinn: Eesti Teaduste Akadeemia Keele ja Kirjanduse
Instituut.
Erelt, Mati, Helle Metslang & Karl Pajusalu. 2006. Tense and evidentiality in Estonian. In Bert
Cornillie & Nicole Delbecque (eds.), Topics in Subjectification and Modalization (Belgian
Journal of Linguistics 20), 125–136. Amsterdam: Benjamins.
Hajdú, Péter. 1968/1982. Chrestomathia Samoiedica. Budapest: Tankönyvkiadó.
Haspelmath, Martin. 1989. From purposive to infinitive – a universal path of grammaticization.
Folia Linguistica Historica 10(1–2). 287–310.
Haspelmath, Martin. 2007. Pre-established categories don't exist: Consequences for language
description and typology. Linguistic Typology 11. 119–132.
Haspelmath, Martin. 2010. Comparative concepts and descriptive categories in cross-linguistic
studies. Language 86(3). 663–687.
Haspelmath, Martin, Matthew S. Dryer, David Gil & Bernard Comrie (eds.). 2005. The World
Atlas of Language Structures. Oxford: Oxford University Press.
Heinämäki, Orvokki. 1984. Aspect in Finnish. In Casper de Groot & Hannu Tommola (eds.), Aspect
bound: A voyage into the realm of Germanic, Slavonic and Finno-Ugrian aspectology, 153–
177. Dordrecht & Cinnaminson: Foris.
Helimski, Eugen. 1998a. Nganasan. In Daniel Abondolo (ed.), The Uralic languages (Routledge
Language Family Descriptions Series), 480–515. London & New York: Routledge.
Helimski, Eugen. 1998b. Selkup. In Daniel Abondolo (ed.), The Uralic languages (Routledge
Language Family Descriptions Series). 548–579. London & New York: Routledge.
Hiietam, Katrin. 2003. Definiteness and grammatical relations in Estonian. Manchester: Univer-
sity of Manchester PhD dissertation.
Hoeksema, Jack (ed.). 1996. Partitives. Studies on the Distribution and Meaning of Partitive
Expressions. Berlin: Mouton de Gruyter.
Honti, László. 1998. Ob-Ugrian. In Daniel Abondolo (ed.), The Uralic languages (Routledge
Hoop, Helen de. 1996. Case configuration and NP interpretation. New York: Garland.
Hoop, Helen de. 1998. Partitivity. Glot International 3(2). 3–10.
tion. Cognitive Linguistics 20. 43–70.
Iggesen, Oliver A. 2008. Number of cases. In Martin Haspelmath, Matthew S. Dryer, David Gil
& Bernard Comrie (eds.), The World Atlas of Language Structures Online (Chapter 49).
Munich: Max Planck Digital Library. http://wals.info/feature/49 (accessed 17 January 2010).
Ikola, Osmo. 1953. Viron ja liivin modus obliquuksen historiaa. Suomi 106(4). Helsinki: SKS.
Iva, Sulev. 2007. Võru kirjakeele sõnamuutmissüsteem [Inflectional morphology in the Võro
Literary Language] (Dissertationes Philologiae Estonicae Universitas Tartuensis 20). Tartu:
Tartu Ülikooli Kirjastus. http://dspace.utlib.ee/dspace/bitstream/10062/4860/1/iva_
sulev.pdf (accessed 1 January 2010).
148 Anne Tamm
Jackendoff, Ray. 1991. Parts and boundaries. Cognition 41. 9–45.

Janhunen, Juha. 1998. Samoyedic. In Daniel Abondolo (ed.), The Uralic languages (Routledge
Kangasmaa-Minn, Eeva. 1998. Mari. In Daniel Abondolo (ed.), The Uralic languages (Routledge
Kehayov, Petar. 2008. An Areal-Typological Perspective to Evidentiality: the Cases of the Balkan
and Baltic Linguistic Areas. Dissertationes Linguisticae Universitatis Tartuensis 10.
Kiefer, Ferenc. 2006. Aspektus és akcióminőség, különös tekintettel a magyar nyelvre [Aspect
and Aktionsart in Hungarian]. Akadémiai Kiadó. Budapest.
Kiparsky, Paul. 1998. Partitive Case and Aspect. In Miriam Butt & Willem Geuder (eds.), The
Projection of Arguments, 265–307. Stanford: CSLI Publications.
Kiparsky, Paul. 2001. Structural case in Finnish. Lingua 111. 315–376.
É. Kiss, Katalin. 2006. The function and the syntax of the verbal particle. In: É. Kiss K. (ed.)
Event structure and the Left Periphery. Studies in Natural Language and Linguistic Theory.
Springer, Dordrecht.
Klaas, Birute. 1999. Dependence of the object case on the semantics of the verb in Estonian,
Finnish and Lithuanian. In Mati Erelt (ed.), Estonian: Typological studies III (Publications
of the department of Estonian of the University of Tartu 11), 47–83. Tartu: Tartu Ülikool.
Kont, Karl. 1963. Käändsõnaline objekt läänemeresoome keeltes. [The declined object in Finnic
languages]. ENSV Teaduste Akadeemia Keele ja Kirjanduse Instituudi uurimused IX,
Tallinn: ENSV Teaduste Akadeemia Keel ja Kirjanduse Instituut.
Koptjevskaja-Tamm, Maria. 2001. ‘A piece of the cake’ and ‘a cup of tea’: Partitive and pseudo-
partitive nominal constructions in the Circum-Baltic languages. In Östen Dahl & Maria
Koptjevskaja-Tamm (eds.), Circum-Baltic Languages Volume 2. Grammar and Typology
(Studies in Language Companion Series: 55), 523–568. Amsterdam: Benjamins.
Kratzer, Angelika. 2004. Telicity and the meaning of objective case. In Jaqueline Gueron &
Jacqueline Lecarme (eds.), The Syntax of Time, 389–424. Cambridge, MA: The MIT Press.
Krifka, Manfred. 1992. Thematic Relations as Links between Nominal Reference and Temporal
Constitution. In Ivan Sag & Anna Szabolcsi (eds.), Lexical Matters, 29–53. Stanford: CSLI
Publications.
Laanest, A. 1975. Sissejuhatus läänemeresoome keeltesse. Tallinn: Eesti NSV Teaduste Akadeemia
Keele ja Kirjanduse Instituut.
Larjavaara, Matti. 1991. Aspektuaalisen objektin synty [The birth of the aspectual object].
Virittäjä. 372–408.
Larsson, Lars-Gunnar. 1983. Studien zum Partitivgebrauch in den ostseefinnischen Sprachen.
Acta Universitatis Upsaliensis, Studia Uralica et Altaica Upsaliensia 15. Uppsala: Univer-
sity of Uppsala.
Lees, Aet. 2005. The case of the object in early Estonian and Finnish texts. In Chris Muskovsky
(ed.), Proceedings of the 2004 Conference of the Australian Linguistic Society.
Lõhmus, Maarja. 2004. An effect of meaning-breaker: Analysis of the cartoon ‘Just shit’.
Semiotica 150 (1–4).
Lytkin, Vasilij I. (ed.), 1962. Komi-Permjatskij jazyk [Komi-Permyak]. Kudymkar: Komi-Permjatskoe
knizhnoe izdatel’stvo.
Markianova, Ludmila. 2002. Karjalan kielioppih [Grammar of Karelian]. Periodika.
Metslang, Helena. 2007. Two types of aspectual opposition in Estonian duration adverbials? In
Danielle Monticelli & Anu Treikelder (eds.), Aspect in languages and theories: Similarities
and differences, in Tartu, 2006 (Studia Romanica Tartuensia 6), 77–90. Tartu: University of
Tartu Press.
Metslang, Helle. 2001. On the developments of the Estonian aspect: The verbal particle ära. In
Östen Dahl & Maria Koptjevskaja-Tamm (eds.), The circum-Baltic languages: typology and
contact: grammar and typology, 443–479. Amsterdam: Benjamins.
Metslang, Helle. 1994. Temporal Relations in the Predicate and the Grammatical System of
Estonian and Finnish. Oulun Yliopiston Suomen ja saamen kielen laitoksen tutkimusra-
portteja 39. Oulu: Oulun Yliopisto.
Miestamo, Matti, Anne Tamm & Beáta Wagner-Nagy (eds.). forthc. Negation in Uralic Languages
(Typological Studies in Language XX). Amsterdam: Benjamins.
Moravcsik, Edith A. 1978. On the Case Marking of Objects, in Joseph H. Greenberg. (ed.), Uni-
versals of Human Language, vol. 4, Syntax, 249–289. Stanford/CA: Stanford University
Press.
Nelson, Diane. 2003. Case and adverbials in Inari Saami and Finnish. Nordlyd 31(4). 708–722.
Nemvalts, Peep. 2000. Aluse sisu ja vorm. Alusfraasi käändevaheldus tänapäeva eesti
kirjakeeles [The content and form of the subject. The case alternation of the subject
phrase in Modern Standard Estonian]. Tallinn: Eesti Keele sihtasutus.
Newmeyer, Frederick J. 2007. Linguistic typology requires crosslinguistic formal categories.
Linguistic Typology 11. 133–157.
Nordlinger, Rachel & Louisa Sadler. 2004. Tense beyond the verb: Encoding clausal tense/aspect/
mood on nominal dependents. Natural Language and Linguistic Theory 22. 597–641.
Pajusalu, Renate, Heili Orav. 2008. Supiinid koha väljendajana: Liikumissündmuste keelendamise
asümmeetriast [Supine constructions encoding spatial entities: asymmetry in expressing
motion event]. In Erelt, Mati (ed.), Emakeele Seltsi Aastaraamat, 104–121. Tallinn: Teaduste
Akadeemia Kirjastus.
Pärn, Priit. 1987. Cartoon ‘Just shit’. In Sirp ja Vasar (Sickle and Hammer), 8/5/1987, p. 16.
Rajandi, Henno & Helle Metslang. 1979. Määratud ja määramata objekt [Defined and undefined
object]. ENSV TA KKI. Tallinn: Valgus.
Reed, Ann M. 1991. On interpreting partitives. In Donna Jo Napoli & Judy Anne Kegl (eds.),
Bridges between psychology and linguistics: A Swarthmore Festschrift for Lila Gleitman,
207–223. Hillsdale, NJ: Erlbaum.
Riese, Timothy. 1998. Permian. In Daniel Abondolo (ed.), The Uralic languages (Routledge
Ritter, Ralf-Peter. 1989. Untersuchungen zum Partitiv im Vepsischen (Veröffentlichungen der
Societas Uralo-Altaica 26). Wiesbaden: Otto Harrassowitz.
Ross, Kristiina, Heete Sahkai & Anne Tamm. 2010. Finiteness and Non-Finiteness in Finno-Ugric
Languages. Synchronic, Diachronic, and Cross-Modular Issues. Linguistica Uralica 46(3).
229–232.
Ruttkay-Miklián, Eszter. 2003. Az alany és a tárgy jelölése a hanti nyelvben (különös tekintettel
az ergatív szerkezetre) [The marking of subjects and objects in the Khanty language: focus
on ergative structures]. In Beatrix Oszkó & Maria Sipos (eds.), Uráli TÁRGYaló: 2002.
január 10-12. (Budapesti Uráli Műhely; 3.), 114–137. Budapest: MTA Nyelvtudományi
Intézet.
Salminen, Tapani. 1998. Nenets. In Daniel Abondolo (ed.), The Uralic languages (Routledge
Shanskiy, Nikolaj M. & Tat’jana A. Bobrova. 1994. Etimologicheskij Slovar’ Russkogo Jazyka
[Ethymological Dictionary of the Russian Language], Moscow: Prozerpina.
Spencer, Andrew. 2009. Case as a morphological phenomenon. In Andrej Malchukov & Andrew
Spencer (eds.), The Oxford Handbook of Case, 185–199. Oxford: Oxford University Press.
150 Anne Tamm
Sulkala, Helena. 1996. Expression of aspectual meanings in Finnish and Estonian. In Mati Erelt
(ed.), Estonian: Typological studies 1, 165–217. Tartu: Publications of the department of
Estonian of the university of Tartu.
Surányi, Balázs. 2009. Incorporated locative adverbials in Hungarian. In Katalin É. Kiss (ed.),
Adverbs and adverbial adjuncts at the interfaces, 37–72. Berlin: Mouton de Gruyter.
Szimoncsics, Péter. 1998. Kamassian. In Daniel Abondolo (ed.), The Uralic languages (Rout-
ledge Language Family Descriptions Series), 580–601. London & New York: Routledge.
Talmy, Leonard. 1996. Fictive Motion in Language and “Ception”. In Paul Bloom, Mary Peder-
son, Lynn Nadel & Merrill F. Garrett (eds.), Language and Space, 211–276. Cambridge,
MA & London: MIT Press.
Tamm, Anne. 1999. Specificity, aspect and the two partitives of Estonian. DOXIMP 3. Graduate
Students’ Third Linguistics Symposium. June 24th, 1999, Selected Papers (221–235).
Research Institute for Linguistics, Hungarian Academy of Sciences.
Tamm, Anne. 2004. Relations between Estonian aspect, verbs, and case. PhD dissertation, ELTE,
Budapest. <http://www.dipfilmod-suf.unifi.it/upload/sub/phd_tamm/Tamm_Ch123.PDF>,
<http://www.dipfilmod-suf.unifi.it/upload/sub/phd_tamm2/TAMM_Ch4567bibl.PDF>
(accessed 6 September 2011).
Tamm, Anne. 2007. Perfectivity, telicity and Estonian verbs. Nordic Journal of Linguistics 30(2).
229–255.
Tamm, Anne. 2008a. Aspect and the Estonian partitive objects: a review of arguments for
analysing partitive NPs as instances of incorporation. In Daniele Monticelli & Anu Treikelder
(eds.), Aspect in Languages and Theories: Similarities and Differences in Tartu, 2006 (Studia
Romanica Tartuensia 6), 205–226. Tartu: Tartu University Press.
Tamm, Anne. 2008b. Partitive Morphosemantics across Estonian Grammatical Categories,
and Case Variation with Equi and Raising. In Miriam Butt & Tracy Holloway King (eds.),
Proceedings of the LFG08 Conference. Stanford, CA: CSLI online Publications. 473–493.
<http://csli-publications.stanford.edu/LFG/13/papers/lfg08tamm.pdf> (accessed 6 September
2011).
Tamm, Anne. 2009. The Estonian partitive evidential: Some notes on the semantic parallels
between the aspect and evidential categories. In Lotte Hogeweg, Helen de Hoop & Andrej
Malchukov (eds.), Papers from TAM TAM: Cross-linguistic semantics of Tense, Aspect, and
Modality, 365–401. Amsterdam: Benjamins.
Tamm, Anne. 2011a. Cross-categorial spatial case in the Finnic non-finite system: focus on the
absentive TAM semantics and pragmatics of the Estonian inessive m-formative non-finites.
Linguistics: An Interdisciplinary Journal of the Language Sciences 49(4). 835–944.
Tamm, Anne. 2011b. Scalarity and dimensionality across categories: Estonian pseudopartitive
constructions. Linguistica Uralica 47(1). 22–40.
Tamm, Anne. 2011c. The Case of Nincstelenek (‘the destitute’): Nounless and Verbless in the
Uralic Grammar. In Sándor Csúcs, Nóra Falk, Viktória Tóth & Gábor Zaicz (eds.), Pro-
ceedings of Congressus XI Internationalis Fenno-Ugristarum: Piliscsaba, Hungary, August
9–14, 2010 (Dissertationes symposiorum ad linguisticam), 270–276. Piliscsaba: Reguly
Társaság.
Tamm, Anne. 2012a. Let me introduce the Estonian analytical causatives: the permissive and
factitive laskma ‘let, make, have, allow, permit’. Jaakko Leino & Ruprecht von Waldenfelds
(eds.), Analytical causatives: from ‘give’ and ‘come’ to ‘let’ and ‘make’. (LINCOM Studies in
Language Typology 24), 247–289. Munich: Lincom Europa.
Tamm, Anne. 2012b. Partitive objects and the partitive evidential marker -vat in Estonian
express incomplete evidence. Finnisch-ugrische Mitteilungen 35. 97–140.
Tamm, Anne. 2012c. Scalar Verb Classes. Scalarity, thematic roles, and arguments in the
Estonian aspectual lexicon. (Biblioteca di Studi di Filologia Moderna 14). Florence: Firenze
University Press. <http://www.fupress.com/Archivio/pdf\5326.pdf> (accessed 16 November
2012)
Tamm, Anne. 2013. Social cognitive and developmental psychology explaining linguistic
patterns: Theory of Mind, epistemic vigilance, natural pedagogy, and the typology of Uralic
generics, evidentials, and possessives. In Márta Csepregi, Kata Kubínyi & Jari Sivonen
(ed.), Grammar and Context – New Approaches to the Uralic Languages III. Grammatika
és Kontextus – Új szempontok az uráli nyelvek kutatásában III. (Urálisztikai Tanulmányok
20.) 3. 260–285. Budapest: ELTE BTK.
Tamm, Anne. 2014. Cross-categorial scalar properties explaining Differential Object Marking.
Linguistics: An Interdisciplinary Journal of the Language Sciences. 52(2). 469–511.
Tamm, Anne. forthc. Negation in Estonian. In Matti Miestamo, Anne Tamm & Beáta Wagner-
Nagy (eds.), Negation in Uralic Languages (Typological Studies in Language XX), xxx–xxx.
Tamm, Anne & Ülle Viks. 2009. Morphology based glossing of Estonian and the Leipzig glossing
rules: problems and solutions for Uralic languages. Poster at the Uralic Typology Days,
Tallinn, November 26–27, 2009. Part 2 available online at <http://glossing.pbworks.com/f/
TabelViksTamm2009Tallinn2.doc> (accessed 6 September 2011).
Tauli, Valter. 1968. Totaalobjekt eesti kirjakeeles [Total object in Estonian]. (Suomalais-
ugrilaisen seuran toimituksia 145), 216–224. Helsinki: Suomalais-ugrilainen Seura.
Tauli, Valter. 1980. Eesti grammatika II. Lauseõpetus [Estonian grammar II. Syntax.]. Uppsala:
Finsk-ugriska institutionen.
Tereščenko, Natalija Mitrofanovna. 1973. Sintaksis samodijskih jazykov [The Syntax of Samoyedic
Languages]. Leningrad: Nauka.
Toivonen, Ida. 2003. Inari Sámi. Paper presented at LAGB, University of Canterbury, 09 March
2003. http://http-server.carleton.ca/~toivonen/pdf/lagb.pdf (accessed on 22 October 2011).
Tsvetkov, Dmitri. 2008. Vadja keele grammatika [Grammar of Votic]. Tallinn: Eesti Keele
Sihtasutus.
Tveite, Tor. 2004. The case of the object in Livonian: A corpus based study (Castrenianumin
toimitteita 62). Helsinki: Finno-Ugrian Society.
Vainikka, Anne & Maling, Joan. 1996. Is partitive case inherent or structural? In Jacob Hoeksema
(ed.), Partitives. Studies on the Distribution and Meaning of Partitive Expressions, 179–
208. Berlin: Mouton de Gruyter.
Vaiss, Natalja. 2004. Eesti keele aspekti väljendusvõimalusi vene keele taustal. [Some ways of
expressing the Estonian aspect against the background of Russian]. Tallinn: University of
Tallinn MA thesis.
Viitso, Tiit-Rein. 1998. Fennic. In Daniel Abondolo (ed.), The Uralic languages (Routledge
Wälchli, Bernhard. 2000. Infinite predication as a marker of evidentiality and modality in the
languages of the Baltic region. Sprachtypologie and Universalienforschung 53(2). 186–
210.
Zaicz, Gábor. 1998. Mordva. In Daniel Abondolo (ed.), The Uralic languages (Routledge
Tuomas Huumo and Liina Lindström
4 Partitives across constructions:
on the range of uses of the Finnish and
Estonian “partitive subjects”
In classical accounts of Finnish and Estonian grammar, the possibility of using

the so-called partitive subject has been one definitional criterion for the category
of existential clauses. However, in both languages there are other syntactic
clause types (or constructions) with an NP that resembles the partitive subject to
a lesser or a greater extent. The partitive subject is also used in many instances
of so-called pragmatic clause types such as negated and interrogative clauses,
whose affirmative counterparts would not count as existentials. In this paper
we examine the definitions given for the range of use of the partitive subject in
both languages, and compare the range of existentials, their definition and
form. We argue that existentials form a radial category, with a prototype and
less canonical instances, where the prototype is clearly definable but the actual
borderline between existentials and other clause types is fuzzy. We also discuss
differences between Finnish and Estonian as regards the range of uses of the
partitive, showing that in Estonian the (plural) nominative is often used in ex-
pressions where the partitive would be the natural option in Finnish. On the
other hand, the use of the partitive in negated and interrogative existentials
seems to be more widespread in Estonian than in Finnish, and this use seems
to be motivated by the irrealis meaning expressed by these constructions.
Keywords: Finnish, Estonian, constructions, existential clause, subject
1 Introduction: The existential partitive subject

Estonian and Finnish are two closely related Baltic Finnic (Uralic) languages,
with many similar constructions that in grammars are described by proposing
similar rules and restrictions. One construction that has received significant
attention in both traditions is the existential construction, the typical structure
of which is NPLOC + V3 S G + NPNOM/PAR . The existential construction has the over-
all function of introducing a discourse-new referent into a location indicated by
the clause-initial locative. It can also be argued that the function of the con-
struction is to characterize the location by giving a predication of its content.
154 Tuomas Huumo and Liina Lindström
The construction thus appears to have a double function: on the one hand, it
introduces new (indefinite) referents into the discourse, and on the other hand,
it characterizes the location indicated by the typically clause-initial locative
expression by predicating what the location contains, starts to contain or ceases
to contain (cf. 1 and 2).
(1) Lasi-ssa on mehu-a. (Finnish, FI)

glass-INE be.PRS .3 SG juice-PAR
‘There is juice in the glass.’
(2) Klaasi-s on mahla. (Estonian, ES)

glass-INE be.PRS .3 SG juice.PAR
‘There is juice in the glass.’
The (typically) clause-final NP that introduces a new referent has classically been
analyzed as the subject of the construction, though it has been pointed out in
many studies that this element differs from canonical subjects in many ways.
This NP is often though not always marked with the partitive case – hence
the widely-used term partitive subject. However, as many scholars especially in
Finland have pointed out (a recent overview is Huumo and Helasvuo, forth-
coming), the subject status of this NP is questionable for a number of reasons
that are related to its grammatical, semantic, and discourse features: for
instance, it fails to trigger agreement in the verb, it does not occupy a topical
position in the clause (i.e. it is not a clausal topic), its referent is typically not
an agentive animate participant but a substance or multiplicity that occupies a
location, and it is not tracked in the discourse (see also Helasvuo 2001).
The case marking of this so-called existential subject varies between the
nominative and the partitive, depending on whether it is headed by a count
noun (typically nominative) or a mass noun (typically partitive), with plurals
often though not always behaving as mass nouns and marked with the partitive.
Under negation, all existential subjects, including those headed by a count
noun, take the partitive case. Huumo and Helasvuo (forthcoming) propose that the
NP called the existential subject does not meet most of the classical criteria for sub-
jects, and should therefore be called, in lack of a better term, the E(xistential)-
NP. In Estonian, however, the existential subject sometimes triggers agreement
in the verb, which is a feature that makes it more like a canonical subject than
its Finnish counterpart – therefore there seem to be reasons to treat the Estonian
existential subject as one (though not prototypical) kind of subject. For unity, we
use the terms existential subject and e-subject (when we refer to the function of
the element in the overall construction) and partitive subject (when we refer to
the case marking) throughout this paper.
Partitives across constructions 155
Especially in the Finnish tradition it has been typical to base the definition
of the existential construction on the possibility of using the partitive subject:
existentials are defined as clauses that either have a partitive subject or have a
nominative (count noun) subject that corresponds to the partitive subject, i.e.,
turns into the partitive under negation or if pluralized. Estonian grammars are
more cautious: they argue that the partitive subject can only occur in existential
clauses, but also admit that its range of use does not extend to all (affirmative)
existentials. Other grammatical criteria for the clause type include the inverse
word order (XVS in both languages), and (especially as far as Finnish is con-
cerned), lack of agreement between the verb and the subject – however, as
pointed out above, this criterion does not work for Estonian, which has some
existential constructions where agreement occurs. Canonical existentials of the
two languages are illustrated by examples (1) and (2) above. These examples
correspond to the prototype of existentials, and meet all criteria typically pre-
sented for the clause type: there is a topical, clause-initial locative, the pale exis-
tential verb meaning ‘be’ and a postverbal, discourse-new subject that refers to
an indefinite quantity of a substance.
In both languages, there are expression types that fail to meet some of the
criteria of existentials but that have nevertheless been classified as instances of
the existential clause. For instance in Estonian, a number of clauses traditionally
classified as existentials show subject–verb agreement (3). Note that since the
canonical form of the verb in both Finnish and Estonian existentials is a crystal-
lized 3rd person singular form (irrespective of the number and person of the
existential subject), this agreement can only be observed in instances where
the existential subject is not a 3rd person singular form. For instance, example
(3) has a nominative plural e-subject, and the verb shows agreement with it.
Unlike the nominative subject, the partitive never triggers person or number
agreement in the verb, as illustrated by the Finnish example (4) with a plural
partitive subject and a singular 3rd person verb form; the situation is the same
in Estonian. Under negation, however, even singular count noun subjects take
the partitive both in Finnish (5) and Estonian (6).
(3) Saali taga.osa-s istu-vad kooli.lapse-d. (ES)

room.GEN back.part-INE sit-PRS .3 PL school.child-PL . NOM
‘In the back of the room sit [some] schoolchildren; There are schoolchildren
sitting in the back of the room.’
(4) Sali-n taka.osa-ssa istu-u koululais-i-a. (FI)

room-GEN back.part-INE sit-PRS .3 SG schoolchild-PL- PAR
‘In the back of the room sit [some] schoolchildren; There are schoolchildren
sitting in the back of the room.’
(5) Koulu-ssa ei ole rehtori-a. (FI)

school-INE NEG .3 SG be.CNG headmaster-PAR
‘There is no headmaster in the school / The school has no headmaster.’
(6) Kooli-s ei ole rektori-t. (ES)

school-INE NEG be.CNG headmaster-PAR
‘There is no headmaster in the school / The school has no headmaster.’
A comparison of the definitions given for existentials in Finnish and Estonian

grammars reveals that the definitions given to the Finnish existential clause
have been more directly based on formal features than those given to Estonian
existentials, which rely more on semantic and discourse (e.g. given vs. new
information) criteria. It has also been pointed out that in both languages there
are borderline cases which are difficult to classify as existential or non-existen-
tial clause types. In his seminal paper on Finnish existentials (which he contrasts
with non-existential normal clauses), Hakanen (1972) gives three form-based
criteria that define existentials: 1) The possibility of using the partitive subject
under the condition that the subject is a mass noun or a plural form (not a
singular count noun), and when the clause is negated, in which case all existen-
tial subjects take the partitive; 2) the inverse (VS) word order, and 3) lack of
subject–verb agreement. In normal clauses, which comprise frequent schematic
clause types such as transitive, intransitive and copular clauses, the subject is
always in the nominative, triggers person and number agreement in the verb,
and typically precedes the verb (in standard written Finnish). Hakanen also
points out that the three criteria for existentials differ in strength: in particular,
word order is a weaker criterion than the partitive case and the lack of subject–
verb agreement, since the existential subject quite often precedes the verb in
actual usage. In such instances, it is the partitive marking of the subject that
marks the expression as an existential; consider example (7) where the partitive
subject precedes the verb (note that the information structure is not neutral but
emphatic).
(7) Mehu-a on lasi-ssa. (FI)

juice-PAR be.PRS .3 SG glass-INE
‘There is juice in the glass / As for juice, there’s some in the glass.’
Example (7) shows that the partitive case keeps the indefinite interpretation of
the subject in spite of its clause-initial position. Since Finnish lacks articles, it is
usually the case that the clausal position of NPs determines, or at least suggests,
whether they are to be interpreted as definite or indefinite (see Chesterman 1991
for a detailed discussion on definiteness and the Finnish partitive). In example

(7), however, the partitive indicates indefiniteness and quantitative unbounded-
ness (‘some’) of the referent and therefore corroborates the existential interpre-
tation of the example. The same word order with a nominative subject results in
a reading where the subject is definite; cf. (8).
(8) Mehu on lasi-ssa. (FI)

juice.NOM be.PRS .3 SG glass-INE
‘The juice is in the glass.’ 1
Because the subject in (8) is in the nominative singular, the verb is now under-
stood as agreeing with it. This shows that there is a discrepancy between exis-
tentials with a partitive subject and those with a nominative subject: in the latter
group, it is only the XVS word order that marks existentiality; cf. the contrast
between examples (9) and (10) below. In contrast, the partitive subject can also
be positioned before the verb without losing the existential reading.
(9) Lattia-lla on sukka. (FI)

floor-ADE be.PRS .3 SG sock.NOM
‘There is a sock on the floor.’
(10) Sukka on lattia-lla. (FI)

sock.NOM be.PRS .3 SG floor-ADE
‘The sock is on the floor.’
Thus the differences observed between (7) and (8) only distinguish existentials
with a mass noun or a plural subject from corresponding non-existential intran-
sitive clauses (normal clauses in the terminology of Hakanen 1972), but fail to
do so if the subject is a singular count noun, because such subjects are in the
nominative even in existentials.
As already pointed out, in the Estonian tradition existentials are defined
more vaguely than in the Finnish one, and the definition is based more on
meaning than on form, mainly because of the lack of strict formal criteria that
all existentials would meet. The main function of the existential clause in Estonian
has been argued to be that of asserting the existence of the subject referent in
1 In principle, a nominative mass noun can also follow the verb, as in Lasi-ssa on mehu [glass-
INE is juice.NOM], but such a construction fails to fulfil the function of introducing a discourse-
new referent. The reading can be contrastive (‘It is the juice that is in the glass’), and pre-
supposes the existence of the subject referent, as the traditional terminology goes (‘There is
the juice (or: a portion of juice) in the glass’).
general or with respect to a certain time or place. This is also how the existential
clause is generally defined (Rannut 1964; Mihkla et al. 1974: 63; EKG II: 42).
Nemvalts (1978; 2000) has defined the existential clause in Estonian only by its
semantic structure [LOC] – [EX] – [REF] where [LOC] is a locative or temporal
component, and [EX] is an existential component marking the existence of the
referent [REF] (Nemvalts 2000: 44–47). The Estonian existential clause is some-
times also called a presentative clause as its main function is to introduce new
referents into a certain time or place (EKG II: 15). Some authors have argued that
the possibility of using the partitive (e.g. in negation) is the main formal criterion
to distinguish existentials in Estonian (Erelt 1978), while others have emphasized
factors related to information structure (the occurence of the subject in a rhematic
position, resulting in XVS word order; Sang 1983, Nemvalts 1978; 2000).
Hence, according to the last-mentioned view, there are no strict formal criteria
that would distinguish Estonian existentials from the other clause types but only
the semantic criterion of introducing or asserting the existence of the subject
referent. Other typical features of the Estonian existential clause cannot be
handled as criteria for their definition, since they do not extend to all existential
clauses. Like in Finnish, these criteria include the XVS word order, the possibility
of using a partitive subject, and (regarding expressions with the partitive subject)
lack of agreement between the subject and the verb. In contrast to Finnish, how-
ever, existentials with a nominative subject show subject–verb agreement (cf.
example 3 above). It is noteworthy that Finnish also allows constructions such
as (3), with a postverbal plural nominative subject triggering verb agreement,
though in Finnish such expressions do not count as existentials but rather
as discourse-pragmatically motivated, non-neutral variants of normal (=non-
existential) clauses:
(11) Sali-n taka-osa-ssa istu-vat lapse-t. (FI)

room-GEN back-part-INE sit-PRS .3 PL child-PL . NOM
‘It is [the] children who sit at the back of the room / At the back of the
room sit the children.’
Thus (11) is formally similar to the Estonian existential construction (3), but their
discourse function is different: (11) is not a presentational construction but
causes a contrastive reading on the verb-final NP, which is also understood as
definite. (For Finnish word order in general, see Vilkuna 1989.)
There has been a lot of discussion concerning the circumstances under
which the partitive subject can be used in Estonian, but generally it has been
concluded that the partitive subject can only be used in existentials, with a
few exceptions (Rannut 1964, Mihkla et al. 1974: 63). The Estonian academic
grammar EKG II distinguishes three basic clause types: existential clauses,

possessor-experiencer clauses and normal clauses (a term that has its origin in
Hakanen’s 1972; 1978; 1980 works on Finnish). Normal clauses always have a
nominative subject that triggers agreement in the verb, and typically the canonical
SVX word order, while the other (exitential and possessor-experiencer) clause
types show XVS word order. Among these three clause types, the partitive sub-
ject may occur in existential and possessor-experiencer clauses, which are re-
miniscent of existentials in many respects.
2 From existentials to neighboring constructions

Estonian
In more recent literature, the classification of Estonian basic clause types has
been modified in such a way that the possibility of using a partitive subject has
been analyzed as a feature typical of so-called marked basic clauses which are
defined mainly by their inverted (XVS) basic word order – these include existen-
tials, possessive clauses (12 and 13), source-marking resultative clauses (14,
below) and experiential clauses (15, below) (Erelt & Metslang 2006). As examples
(12) and (13) show, Estonian possessives are based on a locative-existential meta-
phor where the possessor is indicated by a clause-initial NP in the adessive case
and the possessee by a rhematic NP that resembles the existential subject.
(12) Mu-l on hea-d veini. (ES)

1 SG -ADE be.PRS .3 SG good-PAR wine.PAR
‘I have some good wine.’
(13) Mu-l on punane auto. (ES)

1 SG -ADE be.PRS .3 SG red.NOM car.NOM
‘I have a red car.’
Source-marking resultatives (14) indicate a change of state where the undergoer

of the change is indicated by a clause-initial elative-case (the basic spatial
meaning of which is ‘from’) NP and the acquired state or role by a rhematic NP
that resembles the existential subject.
(14) Tema-st ei saa-nud hea-d õpetaja-t. (ES)

3 SG - ELA NEG become-PTCP good-PAR teacher-PAR
‘He/she did not become a good teacher.’
Experiential clauses (15) resemble possessives in that they have an initial adessive-
marked animate participant to which they attribute an internal (typically psycho-
physiological) state by a rhematic NP that again resembles the existential subject,
except that it does not turn into the partitive under negation and it is not refer-
ential, as it can also be an adjective as in (15).
(15) Urma-l on igav. (ES)

name-ADE be.PRS .3 SG boring.NOM
‘Urmas is bored.’
Thus all these constructions are reminiscent of the existential clause at a more
schematic level. However, the use of the partitive subject is not equal in all
marked basic clauses. In existential clauses it is far from obligatory, whereas in
possessive clauses it is obligatory in the case of a non-count (i.e. mass or plural)
subject (Erelt & Metslang 2006). In other types of marked basic clauses, the
partitive subject occurs less frequently and mainly in negative clauses, cf. (14)
for source-marking resultatives.
The use of the partitive subject as a central criterion for clause types is most
problematic in the category of experiential clauses (as classically defined), as
there are more than one construction type to express an experience. Never-
theless, for semantic reasons, grammars classify these as instances of the same
category. The partitive subject can only be found in those experiential clauses
which formally resemble existential or possessive clauses, i.e. ones that start
with a typically animate possessor-experiencer in the adessive case and where
the stimulus is indicated by a rhematic subject, e.g. (16), (17). Example (16) is
an exceptional instance compared with more typical possessive or existential
clauses, as it has a definite partitive subject sind ‘you’ (indicating the stimulus
of the experience). An alternative, and in our view more natural, analysis of
(16) is one where the verb ‘be’ and the adverb vaja together behave as a transi-
tive predicate and thus the phrase sind ‘you’ is an object. In examples like (17),
the subject can be in the partitive in negative clauses, but partitive case-marking
is not obligatory.
(16) Mu-l on sin-d vaja. (ES)

1 SG -ADE be.PRS .3 SG you-PAR necessary
‘I need you.’
(17) Mu-l ei ole täna halba enesetunne-t. (ES)

1 SG -ADE NEG be.CNG today bad.PAR self.feeling-PAR
‘I don’t have this ill-being feeling today.’
In spite of certain differences between these construction types, one can say that
at a more schematic level they all (except for the other construction types classi-
fied under the category of the experiential clauses) roughly correspond to the
semantic definition of existentials and could be analyzed as subtypes of these.
This is most obvious for possessive and experiential clauses like the one in (17),
as both of these constructions assert the existence of the subject referent. In the
academic grammar of Estonian (EKG II), these and also the source-marking
result clause are classified as subtypes of existentials.
In this paper, we use the term existentials later mainly for “pure” existen-
tials, for possessive clauses and experiential clauses resembling possessives as
these types meet the criteria of existentials in the most systematic way (they
are used mainly to present discourse-new referents in the clause-final position).
This brings Estonian data closer to the Finnish tradition where possessives are
typically classified as a sub-category of existentials (e.g. Hakulinen & Karlsson
1979).
Finnish
In Finnish, there are similar crystallized clause types that are not canonical ex-
istentials but display some features typical of these (see also Huumo & Helas-
vuo, forthcoming). In some classifications, e.g. the comprehensive grammar by
Hakulinen et al. (2004), such expression types are analyzed as independent con-
structions, but it is also pointed out that at a more schematic level they can be
analyzed as a subtype of existentials. As examples of such constructions, we
next discuss shortly what are called state clauses and result clauses by Hakulinen
et al. (2004). The category of result clauses corresponds to its Estonian counter-
part just discussed above (example 14).
The term state clause (SC), as it is used by Hakulinen et al. (2004), is
actually a cover term for a heterogeneous group of constructions that resemble
each other semantically and grammatically to a lesser or a greater extent. Some
subtypes of the SC resemble existential constructions and include an NP that
resembles the e-subject (example 18 below). Another subtype consists of a mere
verb (19), and yet another of a verb accompanied by a locative (20).
(18) Helsingi-ssä on lämmin-tä. (FI)

Helsinki-INE be.PRS .3 SG warm-PAR
‘It is warm in Helsinki.’
(19) Sata-a. (FI)

rain-PRS .3 SG
‘It is raining.’
(20) Mere-llä myrskyä-ä. (FI)

sea-ADE storm-PRS .3 SG
‘It is storming at sea.’
What unites such different subtypes of SCs is thus meaning rather than form,
which is quite variable. Thus an approach that emphasizes formal unity of puta-
tive constructions might actually find the category of the SC too incoherent to be
useful. From the point of view of existentials and partitive subjects it makes
sense to concentrate on those SCs that resemble the existential construction
and have a possible subject candidate.
In addition to expressions of weather and other natural circumstances, there
are SCs that express physical or psychological states. This subtype of state
clauses thus closely resembles the Estonian category of experiencer clauses
discussed above (examples 16 and 17). In these Finnish expressions, like the
Estonian ones, the clause-initial locative introduces an animate (typically
human) experiencer, whereas the psychological or physiological state of this
experiencer is indicated by a clause-final NP. This final NP resembles the exis-
tential subject and is usually in the nominative (example 21), sometimes in the
partitive (22), while the initial NP that indicates the experiencer takes the ades-
sive case. The Finnish adessive (basic spatial meaning ‘at/on’) is a productive
means for the indication of many kinds of relationships with an animate refer-
ence-point (for the term, see Langacker 1993); like in Estonian, it also indicates
the possessor in canonical possessive constructions (23).
(21) Opettaja-lla on kylmä. (FI)

teacher-ADE be.PRS .3 SG cold.NOM
‘The teacher is cold.’
(22) Opettaja-lla on tylsä-ä. (FI)

teacher-ADE be.PRS .3 SG boring-PAR
‘The teacher is bored.’
(23) (Canonical possessive)

Opettaja-lla on kirja ~ kirjo-j-a. (FI)
teacher-ADE be.PRS .3 SG book.NOM ~ book-PL- PAR
‘The teacher has a book ~ [some] books.’
Finnish grammars (e.g., Hakulinen & Karlsson 1979) have often classified posses-
sive constructions as a subtype of existentials. Indeed, they conform to the con-
structional scheme of existentials, the main difference being that the clause-
initial locative indicates a possessor. In contrast, state clauses and possessive–
existential costructions are kept apart as distinct clause types by the compre-
hensive grammar of Hakulinen et al. (2004), in spite of SCs that formally resem-
ble possessives and existentials. It seems that this solution is at least partially
motivated by the properties of the clause-final NPs of state clauses, which are
even less subject-like than existential subjects in general. They are often adjec-
tival phrases (as in 22) and in most cases non-referential. They also differ from
typical existential subjects in that they do not turn into the partitive under nega-
tion; consider (24):
(24) Opettaja-lla ei ole kylmä. (FI)

teacher-ADE NEG .3 PERS be.CNG cold.NOM
‘The teacher is not cold.’
Another construction type close to existentials in Finnish is that of result clauses

(cf. the Estonian example 14 above). Result clauses resemble existential clauses
but differ from their prototypical instances both semantically and morphosyn-
tactically. In the same way as its Estonian counterpart, the Finnish result clause
predicates a quality acquired by an entity that undergoes a change. Its general
schematic meaning is ‘X becomes Y’. The undergoer of the change is indicated
by the clause-initial locative NP that carries the elative ‘out of’ case, which also
reflects the dynamic conceptualization of the change – the elative can be under-
stood as indicating the initial stage of the undergoer of the change as a meta-
phoric source. The quality or role acquired by the undergoer is indicated by the
clause-final element (an AP or an NP) that grammatically resembles the E-NP.
Examples are provided below.
(25) Liisa-sta tul-i opettaja. (FI)

name-ELA become-PST.3 SG teacher.NOM
‘Liisa became a teacher.’
(26) Kahvi-sta tul-i vahva-a. (FI)

coffee-ELA become-PST.3 SG strong-PAR
‘The coffee became strong.’
Both examples predicate an incipient quality of the undergoer introduced by the

elative NP. The construction is idiomatic and strongly associated with the verb
tulla ‘become’ (also meaning ‘come’), though it also allows other verbs of change
as in (27) and (28):
(27) Pennu-sta kasvo-i iso koira. (FI)

puppy-ELA grow-PST.3 SG big.NOM dog.NOM
‘The puppy grew into a big dog.’
(28) Touka-sta kehitty-y perhonen. (FI)

caterpillar-ELA develop-PRS .3 SG butterfly.NOM
‘The caterpillar develops into a butterfly.’
The final NP of a result clause resembles the E-NP in that its case marking varies
between the nominative and the partitive. A difference from experiencer clauses
is that negation turns the final element of result clauses into the partitive, which
is a feature shared by the e-subject (cf. 29):
(29) Pennu-sta ei kasva-nut iso-a koira-a. (FI)

puppy-ELA NEG .3 SG grow-PTC big-PAR dog-PAR
‘The puppy did not grow into a big dog.’
On the basis of the construction types one can observe that there are differences
related to the range of uses of the partitive subject between Finnish and Estonian.
In Finnish existentials, not only mass nouns but most plural subjects are in the
partitive – the indefiniteness of the existential subject in terms of its discourse-
new status as well as the unboundedness of the quantity of its referent motivate
this. In Finnish there are only few instances with plural existential subjects in
the nominative case.2 In contrast, Estonian uses the nominative plural subject
that triggers verb agreement quite extensively even in its existentials, without
losing the existential (presentative) meaning of the construction. Though similar
constructions are grammatical in Finnish (cf. example 11 above), these are not
conceived as existentials but rather as non-existential sentences with a marked
XVS word order.
The use of the partitive subject is not limited only to existentials but it is
distributed over many slightly different clause (construction) types in both lan-
guages. Common features of these clause types are inverted (XVS) word order
and the rhematic position of the subject. While comparing Estonian and Finnish,
2 Such e-subjects resemble pluralia tanta expressions and indicate entities consisting of two
pair-like components (scissors, trousers, pairs of boots, eyeglasses), or closed sets (chess-
pieces, a brood of animals, etc) and behave like singular count-noun e-subjects: Pöydä-llä
on shakkinappula-t [table-ADE is chess-piece-PL.NOM] ‘Tere is [a set of] chess-pieces on the
table’vs. Pöydä-llä ei ole shakkinappulo-i-ta [table-ADE is NEG chess-piece-PL-PAR] ‘There are
no chess pieces on the table’; note that the negated form can as well be the counterpart of an
affirmative partitive, in which case the sense of a set is missing.
one can see that the Finnish partitive subject is more widely used than its
Estonian counterpart. In Estonian, it is used more often in “pure” existentials
and in possessive clauses (but not as systematically as in Finnish), but its
use is restricted mainly to negatives in experiential clauses, result clauses and
cannot be used in state clauses.
3 Semantic and pragmatic implications

of the partitive subject: negated and
interrogative clauses
In this section we take a closer look at the uses of the partitive in negated and
interrogative clauses in the two languages. Both negated and interrogative
clauses serve a clearly distinguishable pragmatic function. In earlier literature
it has been argued that existential subjects under negation are non-specific.
However, there are instances where the partitive of negation is used to mark
subjects that are clearly specific, in particular in Estonian. What makes these
expressions remarkable is that their affirmative counterparts would not count
as existential clauses – thus there seem to be pairs of affirmative vs. negative
clauses that cut across the borderlines of classical clause types. The same accounts
for interrogatives as well.
A connection between negated and interrogative clauses is their non-fact
modality. Givón (2001) distinguishes four types of epistemic modality: presuppo-
sition, realis assertion, irrealis assertion and negative assertion. Among these,
irrealis and negative assertion have various similarities and semantic-pragmatic
connections and thus can be grouped into a supermodality of non-fact (Givón
2001: 301–303). Yes-no interrogatives represent irrealis modality due to their
low epistemic certainty (Givón 2001: 312), in connection to the existentials and
related constructions they express the possible non-existence or non-presence
of the subject’s referent. Negative clauses represent negative modality and in
existentials, they express the non-existence of the subject’s referent. Thus, the
common feature of both clause types is the doubt in the presence or existence
of the subject’s referent. Before we turn to negative and interrogative existen-
tials, we will discuss briefly the basic principles governing the alternation of
nominative and partitive E-NPs in Estonian.
As Estonian existentials allow both nominative and partitive E-NPs, the
interplay between the nominative and partitive has been of interest for many re-
searchers, and the main focus of the research has been the semantic conditions
Figure 1: Distribution of the usages of nominative and partitive subjects in Estonian existentials
(Rannut 1964)
which determine the use of the nominative and the partitive in marking the
e-subject (cf. Rannut 1964, Mihkla et al. 1974, EKG II, Nemvalts 2000, Metslang
2012, Metslang, this volume, etc). A more comprehensive overview on different
usages of the partitive subject in Estonian is given by Nemvalts (2000), and
more recently Metslang (2012), here we introduce briefly only some basic usages.
The classical schema in Figure 1 (taken from Rannut 1964) summarizes the dis-
tribution of the (subject-marking) partitive and nominative in Estonian.
The partitive subject is prototypically used in negative existentials (30) or
existentials expressing doubt (31), and in affirmative existentials where the
referent of the subject is distributable and quantitatively unbounded (32).
(30) Ta-l ei ole kodu-s raamatu-i-d. (ES)

3 SG -ADE neg be.CNG home-INE book-PL- PAR
‘S/he does not have books at home.’
(31) Väheusutav, et ta-l kodu-s raamatu-i-d on. (ES)

unlikely.NOM that 3 SG -ADE home-INE book-PL- PAR be.PRS .3 SG
‘[It is] unlikely that s/he has any books at home.’
(32) Laua peal on leiba. (ES)

table.GEN on be.PRS .3 SG bread.PAR
‘There is [some] bread on the table.’
The partitive also often indicates that only a part of a larger potential group is
involved, and rather often this part is relatively small. If we compare (33) and
(34), we can observe that the main difference in meaning is in the way of repre-
senting the quantity of the schoolchildren: in (33), the quantity (number) of the
schoolchildren is not specified and it is not clear whether there are other people
involved in the situation as well, in (34) it appears that there are only some
schoolchildren among other people.
(33) Saali taga.osa-s istu-vad kooli.lapse-d. (ES)

room.GEN back.part-INE sit-PRS .3 PL school.child-PL . NOM
‘In the back of the room sit schoolchildren; There are schoolchildren
sitting at the back of the room.’
(34) Saali taga.osa-s istu-b kooli.lapsi. (ES)

room.GEN back.part-INE sit- PRS .3 SG school.child-PL . PAR
‘In the back of the room sit some schoolchildren; There are some
schoolchildren sitting at the back of the room.’
Note that in Finnish only the counterpart of (34) could be interpreted as an exis-
tential clause, whereas the structure represented by (33), with a verb showing
agreement, would be interpreted as a non-existential sentence with a postverbal
subject. In Finnish, such an expression would be natural in a slightly marked
discourse context where the emphasis is on the different kinds of people who
occupy different parts of the room; for instance: the children sit in the back
whereas their teachers or parents sit in the front (see Vilkuna 1989 for a detailed
analysis on postverbal subjects in Finnish). The function of this structure is thus
very different in the two languages compared.
3.1 The partitive in negated clauses

Both in Finnish and in Estonian, the core arguments under the scope of nega-
tion typically take partitive case-marking. This concerns both e-subjects and
objects. In both languages, the existential subject under the scope of negation
is marked with the partitive, and the sentence expresses the absence of the sub-
ject’s denotant in a given context. The classical Finnish definition of existentials
as clauses that allow the so-called partitive subject is intended to cover even
expressions where the subject either is or could be in the partitive case, given
that the conditions were the right ones. This is meant to justify the classification
of even expressions with a nominative e-subject (i.e. phrases headed by singular
count nouns) in the category, on the argument that the replacement of the
subject with a mass noun or a plural form, and especially in the negation of
the sentence, would result in the partitive case.
Somewhat simplified, then, the rule that distinguishes existential subjects
from their non-existential counterparts is such that a nominative subject that
under negation turns into the partitive indicates that the clause is to be classi-
fied as an existential, whereas a nominative subject that maintains its case mark-
ing even under negation indicates that the clause is not existential, even in
cases where the subject follows the verb. However, there are problematic uses
of the partitive of negation in subject marking, where the affirmative counterpart
is a singular 3rd person nominative subject preceding the verb, which means
that the affirmative counterpart sentence is not existential, according to the pre-
vailing classification. In fact, it is possible to negate examples such as (35) in
two ways: by maintaining the nominative marking of the subject (36), as ex-
pected, or by replacing it with the partitive of negation (37), which results in a
situation where a non-existential intransitive affirmative clause has a negated
counterpart that according to the classical definition would be an existential
clause:
(35) Laukku on pöydä-llä. (FI)

bag be.PRS .3 SG table-ADE
‘The bag is on the table.’
(36) Laukku ei ole pöydä-llä. (FI)

bag NEG .3 SG be.CNG table-ADE
‘The bag is not on the table.’
(37) Laukku-a ei ole pöydä-llä. (FI)

bag-PAR NEG .3 SG be.CNG table-ADE
‘The bag is not on the table / There is no bag on the table.’
As the alternative English translations of (37) show, ‘the bag’ can be understood
either as specific (‘the bag’), which is the case in (36) as well, or as non-specific
(‘no bag’), as is typical in particular when the e-subject follows the verb. The
difference between (36) and (37) is that (36) selects ‘the bag’ as its topic, or the
starting point of the predication (in the sense of Langacker 1991 or Chafe 1994),
and predicates something about its locative relationship with the world, main-
taining (in classical terminology) the existential presupposition – i.e. that we are
talking about a specific, existing bag that is situated somewhere outside the
location of the table, where its presence is denied by this example. Example
(37), with its partitive e-subject, sets its viewpoint on the location (for more
detailed arguments supporting this feature of Finnish existentials, see Huumo
2003), and predicates that the location is empty with respect to a particular bag
(the specific reading) or any bag (the non-specific reading) – as the focus is
limited to the indicated location, there is no presupposition about the existence
of the bag elsewhere.
In Estonian, it has been argued that a crucial difference appears in regards
to the position of the partitive subject in a negative clause. A partitive subject
that occurs in the (clause-initial) topic position (38) implicates that the subject
referent exists (it has a specific referent); what is under the scope of negation
here is only the location of the subject referent. In (39), it is the (postverbal) sub-
ject itself that is under the scope of negation, making it understood as non-
specific (Sang 1983: 95–96).
(38) Orava-t ei ole puuri-s. (ES)

Squirrel-PAR NEG be.CNG cage-INE
‘The squirrel is not in the cage.’
(39) Puuri-s ei ole orava-t. (ES)

cage-INE not be.CNG squirrel-PAR
‘In the cage there is no squirrel.’
Thus like the Finnish example (37), the Estonian (38) functions as the negated
counterpart of a clause that is not existential, and maintains the specific reading
for its subject. This shows that not all partitive subjects under negation are refer-
entially nonspecific; there are instances in both languages where a specific sub-
ject receives the partitive marking as a result of negation.
In Estonian, the use of the partitive of negation has reached striking extents.
It is quite natural and common to use the partitive when negating the presence
of highly topical participants (such as discourse participants, even the speaker)
in locations by using the partitive instead of the nominative. In such instances,
the partitive is specific and its referent is definite. The function of such a clause
is to deny the presence of the participant in a certain location or in certain
circumstances; cf. (40) where the speaker’s future presence at work is denied.
Example (41) also deviates from typical existentials, since the focus of negation
is on the clause-final locatives, whereas the referent of the initial partitive is
highly topical and discourse-old. Thus, as far as information structure is con-

cerned, clauses like (40) and (41) resemble typical normal clauses rather than
existentials.
(40) Min-d ei ole homme töö-l. (ES)

1 SG - PAR NEG be.CNG tomorrow work-ADE
‘Tomorrow I’ll not be at work’ [lit. ‘There will be no me at work tomorrow’].
(41) Te-da ei ol-nud ei kodu-s ega töö juures. (ES)

3 SG - PAR NEG be-PTC NEG home-INE nor work.GEN POSTP
‘S/he was neither home nor at work’ [lit. ‘There was no him/her at home
nor at work’].
As is the case in Finnish, Estonian examples like (40) and (41) have counterparts
where the subject maintains its nominative case in spite of negation; examples
(42) and (43) are canonical negated normal clauses.
(42) Ma ei ole homme töö-l. (ES)

1 SG . NOM NEG be.CNG tomorrow work-ADE
‘Tomorrow I’ll not be at work’
(43) Ta ei ol-nud ei kodu-s ega töö juures. (ES)

3 SG NEG be-PTC NEG home-INE nor work.GEN POSTP
‘S/he was neither at home nor at work’
The affirmative counterparts of both (40) vs. (42) and (41) vs. (43) are typical
normal clauses with a clause-initial nominative subject triggering verb agree-
ment (44 and 45).
(44) Ma ole-n homme töö-l. (ES)

1 SG . NOM be-PRES .1 SG tomorrow work-ADE
‘Tomorrow I’ll be at work.’
(45) Ta ol-i kas kodu-s või töö juures. (ES)

3 SG . NOM be-PST.3 SG either home-INE or work.GEN POSTP
‘S/he was either at home or at work.’
The question to be asked is, what motivates the nominative vs. partitive alterna-
tion in negated counterparts of affirmative clauses such as (44) and (45), where
only the nominative is possible, and whether there are meaning differences
brought up by the case marking of the subject. When we compare pairs like
(40) vs. (42) or (41) vs. (43), we can observe slight meaning differences between
the versions with the partitive and the nominative. The partitive subject, as else-
where, represents its referent as less agentive and less volitional than the nomi-
native subject. This has the result that the versions with the partitive imply that
the absence of the subject referent from the location or situation is not volitional
but perhaps due to circumstances not in the subject referent’s control. In con-
trast, the referent of the nominative subject may be conceived as more volitional
decision-maker as regards it presence or absence in the location.
As the partitive subject in clauses such as (40) and (41) are highly topical
and express specific and discourse-old referents, such clauses do not belong to
typical existentials (see also Sang 1983: 94–97). In this respect, the use of the
partitive in examples such as these seems to exceed the boundaries of existen-
tials. Compared with Finnish, which sometimes allows the use of the partitive
subject as a negative counterpart of a non-existential nominative in examples
like (37), it can be pointed out that such uses are more common and widespread
in Estonian, especially in instances where the subject is a speech act participant
referred to by a personal pronoun. Interestlingly, similar usages have been ob-
served also in Northern Russian, Lithuanian and Latvian, which use the partitive
genitive instead of the nominative in clauses similar to (40) and (41) (see Seržant,
forthcoming a; forthcoming b). Seržant explains these usages with the expand-
ing of the non-referentiality reading (which is one of the determiners of using
genitive-partitive in these languages) from the NP to the whole situation. Thus,
our Estonian data is in this respect closer to Russian and Latvian, the closest
contact languages of Estonian, than to Finnish, a close cognate.
3.2 Interrogatives
Another context where the partitive subject seems to exceed the boundaries of
existentials, especially in Estonian and to a more moderate extent in Finnish,
are interrogative clauses that ask for the presence or existence of the subject
referent. In yes-no interrogatives, typically both the nominative and the partitive
are possible, even in cases where the subject is a singular count noun. For
instance, in (46), the referent of the subject is not distributable, but the subject
is an indefinite count noun (‘phone’). Indefiniteness may thus be one factor
contributing to the use of the partitive, but since in general indefiniteness alone
is not sufficient to trigger the partitive in singular count noun subjects, there
might be other factors at work as well – more precisely, factors related to (poten-
tial) negation. By this we mean that example (46) leaves open the possibility
that the subject referent is non-existent, if the addressee does not have a tele-
phone, and that this possible non-existence then motivates the partitive. Similar
uses can be observed in Finnish (47).
(46) Kas su-l telefoni on? (ES)

Q 2 SG -ADE telephone.PAR be.PRES .3 SG ?
‘Do you have a telephone?’
(47) Katso, tule-e-ko sie-ltä auto-a. (FI)

look.IMP come-PRS .3 SG - Q there-ABL car-PAR
‘Check if there is a car coming!’
As the general rule goes, in declarative existentials the partitive is not used with
singular count nouns – thus, in addition to negated clauses, interrogatives are
another category where the partitive seems to be gaining a wider terrain of
usage. The declarative counterparts of (46) and (47) would be with nominative
case marking, and the partitive would make them ungrammatical; cf. (48) and
(49) respectively:
(48) Mu-l on telefon (~ *telefoni). (ES)

1 SG -ADE be.PRES .3SG telephone.NOM (*PAR )
‘I have a telephone.’
(49) Sie-ltä tule-e auto (~ *auto-a). (FI)

there-ABL come-PRS .3 SG car.NOM (*PAR )
‘There comes a car.’
In Estonian, in the same way as in negated clauses, interrogatives also allow the
use of the partitive in subjects with definite and topical referents such as proper
names (50) and pronouns (51). As an alternative, the nominative is also possible
in these examples (52).
(50) Kas Renate-t on seal? (ES)

Q name-PAR be.PRES .3 SG there
‘Is Renate there?’
(51) Kas sin-d homme on liikve-l? (ES)

Q 2 SG - PAR tomorrow be.PRES .3 SG moving-ADE
‘Will you be around tomorrow?’
(52) Kas sa homme ole-d liikve-l? (ES)

Q 2 SG . NOM tomorrow be-PRS .2 SG moving-ADE
‘Will you be around tomorrow?’
Again, the declarative counterparts of (50) and (51) would only be grammatical
with a nominative subject. Thus the use of the partitive seems to be exceeding the
boundaries of existentials not only in negated but also in interrogative expres-
sions, and in Estonian more productively than in Finnish, where at least the
counterpart of (51) would be ungrammatical; on the other hand, the counterpart
of (50) is marginally possible (e.g., in a telephone conversation). Similarly to the
negated instances discussed above, the interrogative examples with the partitive
subject seem to reduce the conceived volitionality and agentivity of the subject.
A question arises why the use of the partitive has expanded in negative and
interrogative clauses which do not fulfil the criteria of existentials, while the use
of the partitive has reduced in other contexts in Estonian. As it was already
pointed out above, a connection between negated and interrogative clauses is
their non-fact modality (Givón 2001) which covers both irrealis and negation.
The non-fact modality seems to be the common denominator between the uses
of partitive subjects in Estonian negative and interrogative clauses: in both
cases, the existence or presence (in a location) of the referent of the subject is
not taken for granted. Irrealis modality is a typical context for non-canonically
marked subjects in many languages, as well as existential constructions (Onishi
2001). Thus it is not surprising that in irrealis contexts the use of partitive sub-
jects has prevailed and even widened so that also highly topical referents can be
marked with the partitive. Motivating the extensive use of the partitive subject in
Estonian with non-fact modality would also explain why it can be used for
highly topical entities. In such contexts, the partitive marking reflects the re-
duced agentivity of its referent, as it expresses a lower degree of volitionality
than the nominative.
4 Conclusions
In sum, our study shows that both in Finnish and in Estonian the so-called
partitive subject is typical in existential clauses, though its use as a definitive
criterion for the clause type is not without problems: in both languages there
are uses where the partitive subject apparently steps outside the category of
existentials. In particular, the partitive triggered by negation seems to be in use
even in instances where the corresponding affirmative clause would not count
as an existential clause. In affirmative existentials, the partitive reflects indefi-
niteness and unboundedness of the quantity the existential subject refers to,
and this function bears a clear resemblance with the use of the partitive as an
object marker, where it marks these same factors, and, in addition, unbounded-
ness of aspect.
When comparing the two languages under scrutiny with each other, it is
easy to see that the range of uses of the partitive is not equal in them. First, exis-
tential subjects in the plural are more frequently marked with the nominative in
Estonian than in Finnish, which favors the partitive. In Estonian, these plural
postverbal subjects also trigger verb agreement, while the (structurally correspond-
ing) clauses in Finnish would not count as existentials. On the other hand, the
use of the partitive in negated and interrogative clauses seems to be more wide-
spread in Estonian than it is in Finnish. It is remarkable that in Estonian even
definite, topical clause-initial subjects may receive partitive marking in negated
and interogative clauses, and extreme instance being the use of partitive sub-
jects indicating speech act participants. In Finnish such uses seem to be far
less productive. This can be explained by the non-fact modality: in negative and
interrogative clauses, the partitive marking indicates that the subject referent’s
existence or presence is not taken for granted, although it is referential and
definite. It also reduces the agentivity of the subject’s referent making it less
volitional.
In general, in Estonian the use of the partitive case as the marker of the sub-
ject in existential clauses seems to be reducing, at least if one assumes Finnish
to be more conservative in this respect – another possibility is of course that the
Estonian partitive has never been in such a systematic use in plural existential
subjects as in Finnish. Instead, the Estonian partitive subject has developed
towards a marker of subjects in contexts of non-fact modality, while its Finnish
counterpart shows only weak traces of such a development.
References
Chafe, Wallace L. 1994. Discourse, consciousness, and time. The flow and displacement of
conscious experience in speaking and writing. Chicago: University of Chicago Press.
Chesterman, Andrew. 1991. On Definiteness: A Study with Special Reference to English and
Finnish. Cambridge Studies in Linguistics 56. Cambridge: Cambridge University Press.
Givón, Talmy. 2001. Syntax: An Introduction. Volume I. Amsterdam: Benjamins.
EKG II = Erelt, Mati, Reet Kasik, Helle Metslang, Henno Rajandi, Kristiina Ross, Henn Saari, Kaja
Tael & Silvi Vare. 1993. Eesti keele grammatika II. Süntaks. Lisa: kiri. Tallinn: Eesti Teaduste
Akadeemia Keele ja Kirjanduse Instituut.
Erelt, Mati. 1978. Märkmeid partsiaalsubjekti kohta eesti keeles. Eesti keele grammatika
küsimusi. Keel ja Struktuur X. Tartu: Tartu Riiklik Ülikool. 3–9.
Erelt, Mati & Helle Metslang. 2006. Estonian Clause Patterns – from Finno-Ugric to Standard
Average European. Linguistica Uralica XLII (4). 254–266.
Hakanen, Aimo. 1972. Normaalilause ja eksistentiaalilause. Sananjalka 14. 36–76.
Hakanen, Aimo. 1978. Kontrastiivista lauseanalyysia: Eksistentiaalilauseet. Turun yliopiston

suomalaisen ja yleisen kielitieteen laitoksen julkaisuja 8. Turku: Turun yliopisto.
Hakanen, Aimo. 1980. Existential clauses. Introduction. In Osmo Ikola (ed.), Congressus
Quintus Internationalis Fenno-Ugristarum. Pars III. Turku: University of Turku.
Hakulinen, Auli & Fred Karlsson. 1979. Nykysuomen lauseoppia. Helsinki: Suomalaisen
Kirjallisuuden Seura.
Hakulinen, Auli, Maria Vilkuna, Riitta Korhonen, Vesa Koivisto, Tarja Riitta Heinonen & Irja Alho.
2004. Iso suomen kielioppi. Helsinki: Suomalaisen Kirjallisuuden Seura.
Helasvuo, Marja-Liisa. 2001. Syntax in the making: The emergence of syntactic units in Finnish
conversation. Studies in discourse and grammar 9. Amsterdam: Benjamins.
Huumo, Tuomas & Marja-Liisa Helasvuo (forthcoming). On the subject of subject in Finnish. In
Marja-Liisa Helasvuo & Tuomas Huumo (eds.), Canonical and non-canonical subjects in
constructions. Amsterdam: Benjamins.
Langacker, Ronald W. 1991. Concept, image, and symbol: The cognitive basis of grammar.
Berlin: Mouton de Gruyter.
Langacker, Ronald W. 1993. Reference-point constructions. Cognitive Linguistics 4(1). 1–38.
Metslang, Helena. 2012. On the Case-Marking of Existential Subjects in Estonian. SKY Journal
of Linguistics, 25. 151–204.
Mihkla et al. 1974 = Karl Mihkla, Lehte Rannut, Elli Riikoja & Aino Admann. 1974. Eesti keele
lauseõpetuse põhijooned 1. Lihtlause. Tallinn: Valgus.
Nemvalts, Peep. 1978. Eksistentsiaallauseist. Eesti keele grammatika küsimusi. Keel ja Struktuur
X. Tartu: Tartu Riiklik Ülikool. 10–41.
Nemvalts, Peep. 2000. Aluse sisu ja vorm. Alusfraasi käändevaheldus tänapäeva eesti
kirjakeeles. Tallinn: Eesti Keele Sihtasutus.
Onishi, Masayuki. 2001. Non-canonically marked subjects and objects: Parameters and properties.
In Alexandra Y. Aikhenvald, R.M.W. Dixon & Masayuki Onishi (eds.), Non-canonical marking
of subjects and objects. Typological Studies in Language 46, 1–51. Amsterdam: Benjamins.
Rannut, Lehte. 1964. Täis- ja osaalus tänapäeva eesti kirjakeeles. Keel ja Kirjandus 1. 32–39.
Sang, Joel. 1983. Eitus eesti keeles. Eesti NSV Teaduste Akadeemia Keele ja Kirjanduse
Instituut. Tallinn: Valgus.
Seržant, Ilja A., forthcoming a. Independent (bare) partitive genitive in Russian and North
Russian. To be published in Seržant, I. A. and B. Wiemer (eds.), Contemporary Approaches
to Dialectology: The area of North, Northwest Russian and Belarusian vernaculars. Slavica
Bergensia 13. Bergen: Department of Foreign Languages, University of Bergen.
Seržant, Ilja A., forthcoming b. Denotational properties of the independent partitive genitive in
Lithuanian. To be published in Holvoet, Axel and Nicole Nau (eds.), Grammatical Functions
and their Non-canonical Coding in Baltic. Amsterdam: Benjamins.
Acknowledgments
This study was supported by Estonian Science Foundation (project PUT90).
Helena Metslang
5 Partitive noun phrases in the Estonian
core argument system
Case is an inflectional category that is typically used for marking a noun’s relation
to other parts of the sentence. It is used for identifying grammatical relations
such as subjects and objects but also for signalling a variety of meanings (Butt
2008: 27). This study discusses the role of Estonian partitive case use in various
kinds of argument marking. The partitive-permitting arguments – the object and
the argument of the existential construction (e-NP) – will be analyzed and con-
trasted against each other and they will also be compared with the transitive
and intransitive clause subjects. The distributions of semantic, pragmatic and
coding properties among partitive arguments will be compared with the properties
of arguments with different coding. The paper also analyzes how the typologically
controversial Referential hierarchy distinguishes the arguments and case-uses. A
new comparison is made of the Estonian object’s and e-NP’s differential case-
marking systems. The salience of each factor is measured on the empirical
data. The study discusses the subjecthood status of the e-NP and the informa-
tion structurally based fluid intransitivity manifested in its marking.1
Keywords: grammatical relations, differential object marking, partitive,

referential properties, message packaging
1 Introduction
Case as one of the formal coding means is an integral part of a language’s argu-
ment realization system. It is used for identifying grammatical relations and also
for signalling a variety of meanings: referential and tense/aspect differences,
topicality, focus and modalities (Butt 2008: 27). However, several of these func-
tions can also be expressed by other means of coding as well as by pragmatic
‘message packaging’ (for example by agreement and the referent’s activeness in
the discourse respectively). Chafe (1976: 28) uses the term ‘message packaging’
1 This study was funded by Deutscher Akademischer Austauschdienst, Alfred Kordelin’s Foun-
dation’s and European Social Fund’s DoRa scholarships and the research grant SF0180084s08
Morphosyntactic structure and development of Estonian. I thank Merilin Miljan for comments
and feedback. Naturally, the responsibility for any shortcomings is entirely my own.
178 Helena Metslang
to refer to a speaker’s use of a particular syntactic form to serve a particular

pragmatic function. Roughly speaking, the speaker uses ‘packaging’ to present
the hearer instructions on how to manipulate the message and to integrate it
to his current knowledge. Case is always tied to some linguistic element. This
article discusses case in connection with argument types.
The aim of this article is to determine the distribution and functions of the
partitive arguments in the Estonian core arguments’ system and to compare
them with differently coded arguments: arguments denoted by zero-anaphora,
NPs in other cases, non-finite arguments and other clausal arguments. Estonian
partitive case is a heterogeneous category that is used to express a range of
meanings relating to indefinite quantity and lower transitivity, fulfilling most of
the functions attested cross-linguistically (see the semantic map of the partitive
in Luraghi and Kittilä, this volume, Figure 2). I will study the semantic nature of
partitive arguments and analyze which referential properties Estonian opens its
subject, object and e-NP positions to.
Following the Finnish example (Helasvuo 2001) I use the term ‘e-NP’ to
denote the nominative or partitive marked NPs in existential constructions (ECs
for short).2 As discussed for example in Nemvalts (2000), ECs are clause level
constructions that introduce new referents in the discourse. They prototypically
include a topical locative phrase, a predicate verb and a post-verbal non-topical
subject-like argument. The predicate verbs of ECs are intransitive verbs with an
existential meaning or verbs used in the way that the existential meaning com-
ponent is foregrounded (see examples (9) and (10) in Section 3). There are no
verbs that are exclusively only used in ECs − the same verbs can occur in an
ordinary intransitive clause environment (Metslang 2012: 154−155; Nemvalts
2000: 20, 44; see Huumo 1999: 41 for closely related Finnish).
Thus far, Estonian core arguments have been widely studied from the view-
point of specific features: for example, the complex case-marking rules (one of
the more popular topics of debate in Finnic linguistics), semantics, word order,
etc. In this study, a more general view is employed, and a larger number of
interrelated phenomena are approached together and checked against corpus
data:
– The partitive-permitting arguments’ role (semantics and message packag-
ing) among the core arguments will be analyzed.
– The role of partitive marking will be juxtaposed with the roles of other
coding possibilities.
2 In this paper, I use the abbreviation HM to indicate the author’s examples. Note that S –
the sole argument of an intransitive verb must be understood, in the citations of Tael (1988)
and Huumo (1993), as transitive or intransitive subject or an e-NP. INF – infinitive refers to
da-infinitive and ma-infinitive with its case-forms.
Partitive noun phrases in the Estonian core argument system 179
– The motivations of the Estonian equivalents of differential object marking

(DOM) and differential subject marking (DSM) systems will be compared
and, on the basis of empirical data, the relative importance of each case-
marking factor will be demonstrated (what is called ‘Estonian equivalent of
DSM’ in this article is e-NP’s case alternation).
– The e-NP’s subjecthood, as well as the related question of the fluid intransi-
tivity alignment type will be briefly discussed (the latter is attested for example
in French impersonal constructions largely similar to ECs; Creissels 2008).
In addition to analyzing how partitive-marked arguments are distinct from the

ones that are realized differently I also give brief descriptions of particular coding
and meaning-related properties in their own right (for example, I will look at how
the number category combines with different referential properties and case-
assignment motivations) in order to provide broader background information.
In Section 2 I will define the main notions of this study. Section 3 outlines
the Estonian clause types and argument types under consideration. In Section
4 I will give an overview of the Estonian DOM and e-NP case rules, relying on
the findings on dependent-marking factors in Estonian linguistics and typology.
Section 5 tests these case-marking systems against empirical data and discusses
the subjecthood status of e-NP on the basis of this analysis. An interim summary
of the coverage of all O and e-NP case-assignment factors in the corpus is made
in Section 5.5. The meaning-related properties of each argument type will be
studied in Section 6. In Section 7 I will analyze the arguments’ message packag-
ing. In Section 8 I will compare all argument types and parameters with each
other and discuss the reasons for O’s and e-NP’s closeness and the potential
fluid intransitivity system in Estonian. Section 9 summarizes the main outcomes
of the study.
The analysis is based on seven 2000 word narrative extracts from the fiction
sub-corpus of the Corpus of Estonian Literary Language.3 I looked at 130 intran-
sitive clauses, 130 transitive clauses and 130 ECs. These fictional texts reflected
upon various events and states of affairs in people’s lives.4 The example sentences
of this paper are from this corpus, unless otherwise marked. The sentences have
been shortened when necessary. I chose a relatively small dataset and a limited
number of texts in order to be able to include textual information for pragmatic
3 Corpus of Estonian Literary Language, http://www.cl.ut.ee/korpused/baaskorpus/. (30 July

2013.)
4 In the used corpus texts, the narratives were told from the point of view of the first or third
person.
180 Helena Metslang
analysis in the study. The benefit of this method is the possibility of also getting
a preliminary picture of the partitive argument’s pragmatic role in discourse; its
drawback is the potential for the small dataset to over-represent less frequent or
text type-dependent phenomena. In order to establish more general properties
of Estonian partitive arguments, a study on a larger dataset and on a variety of
text types and communicative situations is necessary.
2 Background
In order to discuss the grammatical realization, pragmatic role and meaning of
the partitive arguments, we need to define the notions construction and argument
role in a way that guarantees reliable comparisons. In the discussion of argument
realization, the closely related notion of grammatical relation is also needed. In
the following, I will provide the definitions in line with Construction Grammar.
Constructions are form-meaning correspondences that are symbolic units of
the speakers’ linguistic knowledge. In Construction Grammar no strict division is
assumed between the lexicon and syntax (Goldberg 1995: 7; Croft 2001: 58; see
also Langacker 1987: 54). Therefore the notion construction is used in a broad
sense here. The inventory of constructions in a language contains for example
case and agreement constructions, lexical items, idioms, control constructions,
non-finite constructions, argument structure constructions, word order and sen-
tence type constructions (cf. Bickel 2010; Croft 2001; Goldberg 1995). In the
framework of Radical Construction Grammar (Croft 2001: 25−28), constructions
form a taxonomic network. The nodes of the network are constructions which
are partial structures with different levels of schematicity. Croft states that any
construction with idiosyncratic morphological, lexical, syntactic, etc. properties
must be represented as an independent node in the network. The taxonomy is
hierarchical and any construction can have multiple parents. Croft (ibid.) illus-
trates the taxonomy model with the following constructions from English:
– [SUBJECT VERB OBJECT] − the transitive construction (a wholly schematic
construction)
– [SUBJECT kick OBJECT] – a verb-specific construction
– [SUBJECT kick the bucket] − a specific idiom construction
– [SUBJECT AUXILIARY-n’t VERB] – a negation construction (the use of a direct
object is not specified)
Of particular interest for determining argument types are Goldberg’s argument

structure constructions (basic clause types) depicting scenes basic to human
experience, e.g. moving or doing something to someone (Goldberg 1995: 5; see
also Section 3 for the main views on clause types in Estonian linguistics).
In discussing argument roles, several authors find that they are not primi-
tives, but are instead derived from richer semantic structures. Constructions as
well as verbs specify which roles are profiled (Goldberg 1995: 49). Hence, argu-
ments are semantic elements that are defined by their relation to the predicate
or construction (e.g. the agent role) (Bickel 2010; Goldberg 1995: 49). Goldberg
defines argument roles as semantically constrained relational slots in the dynamic
scene associated with the construction or the verb. In addition, arguments can be
defined by their referential type (e.g. animate; Bickel 2010). This is a completely
semantic view on argumenthood that is regarded as a more stable and reliable
basis for comparison of the morphosyntactic argument properties than the syn-
tactic perspective on argumenthood that relies on for example case and passivi-
zation properties (see e.g. Witzlack-Makarevich 2011: 43−45 for discussion).
I assigned one of the following roles to every core argument in the dataset:
the sole argument of an intransitive verb (occurring in a non-existential con-
struction, abbreviated as S), the most actor-like argument of a transitive verb
(A) and the ‘not most actor-like argument’ of a transitive verb (O; otherwise
also abbreviated as P by several authors) (cf. Bickel 2010). In addition, I look
separately at e-NP, the sole argument of the intransitive verb in an EC. Similarly
to expressions denoting entities, expressions denoting propositions may also
serve as arguments of a predicate (Lambrecht 1994: 74). Drawing upon this
view, I also included non-finite and other clausal arguments in the S, A, O and
e-NP roles in the study. I found that the most suitable method in identifying
such arguments is pronominalization.5 I only picked examples that were clearly
deductable to a verbal or constructional argument and discarded unclear cases.
For example:
(1) Üksnes Renke ütle-s „Tunne-n kaasa“ / se-da.

only Renke.NOM say-PST.3SG sympathize-1SG with / this-PAR
‘Only Renke said, “I am sorry”’/ ‘Only Renke said this.’ (clausal O,
replaceable by a pronoun
5 Pronominalization has also been used in identifying Estonian subjects and direct objects
in Erelt et al. (1993: 39, 46), see also Croft (2001: 187–188) on pronominalization and other
argument defining criteria and on the potential drawbacks of their usage.
182 Helena Metslang
(2) Esialgu Ø vaja vaada-ta, et mingi peavarju

first be.3 necessary look-INF that some.GEN shelter.GEN
püsti saa-ks / se-da.
up get-COND / this-PAR
‘First (we) have to see that (we) get some shelter up.’ / ‘First we need this.’
Lit. ‘First (it is) necessary to see that would get some shelter up / this.’
(clausal e-NP, replaceable by a pronoun)
Although the argument of the EC can (in rare cases) occur as a clause, I will still
use the term e-NP (existential noun phrase) for the sake of convenience.
Once single arguments have been identified semantically, it is possible to
build grammatical relations of arguments on the basis of their morphosyntactic
properties (Witzlack-Makarevich 2011: 61). A number of linguists hold the view
that grammatical relations are language-specific and construction-specific (e.g.
Dryer 1997; Croft 2001; Witzlack-Makarevich 2011). According to the construction-
specific view to grammatical relations, they are relations between arguments and
their specific constructions. According to Bickel (2010) grammatical relations are
equivalence sets: arguments are “treated the same way by some construction in a
language, for example, being assigned the same case in a language, or triggering
the same kind of agreement”. Also syntactic behaviour constructions can distin-
guish between representatives of what might in general be considered the same
global, i.e. cross-constructional grammatical relation. For example, although the
Estonian active clause subject can occur as the antecedent of a reflexive pro-
noun (i.e. participate in the reflexivization construction), the passive subject in
general cannot:
(3) Ma ju usu-n enda-sse.

1 SG .NOM MDA believe-1SG self-ILL
‘But I do believe in myself’ (active intransitive subject as an antecedent of
a reflexive pronoun)
(4) Politsei töö kergendamise-ks ol-i-d (ta-li)

police.GEN work.GEN simplifying-TRAN be-PST-3PL 3 SG -ADE
endai/*j kõrvale jäe-tud juhiloa-dj .
self.GEN beside leave-PASS.PST.PTCP driving.licence-NOM.PL
‘For making police’s work easier (his) driving licence was left next to him
(by himself). Lit. ‘(To himi,) the driving licencej was left beside selfi/*j to
simplify the police’s work.’ (passive subject, the general impossibility of
anteceding reflexive pronouns)
Therefore the Estonian reflexive construction defines active and passive subject
as separate grammatical relations. The nominative case construction defines the
A and O arguments as one grammatical relation because they bear the same
case. However, the partitive and genitive-marked O are different grammatical
relations from the point of view of these case constructions (cf. Van Valin 2005:
89ff.). With respect to the agreement construction, A and S occur as one gram-
matical relation (the verb agrees with both of them). Estonian O (marked by
any object case, the partitive, genitive or nominative) bears a different grammat-
ical relation to the agreement construction, as the verb does not agree with it.
As mentioned above, among the most relevant constructions for determin-
ing grammatical relations in this paper are argument structure constructions.
When I use the cover term ‘subject’ in this paper, I am referring to two different
grammatical relations: (i) the grammatical relation that an A argument bears to
a transitive argument structure construction and (ii) the grammatical relation an
S argument bears to an intransitive (non-existential) argument structure con-
struction. I use the term ‘object’ for denoting the grammatical relation occurring
between the O argument and a transitive argument structure construction. For
example:
(5) Ma istu-si-n.
1 SG .NOM sit-PST-1SG
‘I was sitting.’ (HM, intransitive clause)
(6) Naabri-d ehita-si-d suvila-t.

neighbour-NOM.PL build-PST-3PL summer.cottage-PAR
‘The neighbours were building a summer cottage.’ (HM, transitive clause)
Throughout the article, in general, I do not include e-NP in the subject category,
unless explicitly described otherwise (above all it proves useful in the dis-
cussion of fluid intransitivity and the Estonian version of DSM). Grammatical
relations that are determined via verbs or argument structure constructions are
global, cross-constructional arguments (as for instance different objects can vary
in the use of case, zero-anaphora and word order).
It is also necessary to stipulate what is meant by argument realization in this
article. Argument realization concerns the ways how arguments can be expressed
(selection in the subject or object role, case, position, syntactic behaviour, etc.;
Levin & Rappaport Hovav 2005; see also Witzlack-Makarevich 2011: 114–118).
Although some influential approaches have a narrow view on this term (Levin
and Rappaport Hovav (2005) only include these ways of argument expression
in the definition that are triggered by a verb’s argument structure), a broader
184 Helena Metslang
view is adopted here. I use argument realization as an umbrella term for all
means of expressing arguments. I do not delimit the term to merely lexically
conditioned phenomena but include all possible factors that determine the formal
properties of an argument.
As one of the purposes of this study is to compare the realization of different
argument types and establish the argument types’ overall proximity or distance
with respect to each other, we must choose the factors for the analysis and deter-
mine how they delimit or favor certain equivalence sets, i.e. alignment types (cf.
Bickel & Nichols 2008: 305−306) of the four arguments under scrutiny. Several
authors (e.g. Croft 2001: 41) find that there is no justified a priori way of choos-
ing one criterion over the other for this purpose – a principled analysis rather
takes all relevant criteria into account. Therefore, in the process of characteriz-
ing Estonian partitive arguments, a large set of properties is considered in this
paper.
Sections 4 and 5 compare the O and e-NP case-marking systems. Two addi-
tional notions that are relevant for this discussion are inclusiveness and aspect. I
will use Lyons’ (1999: 2−13) definition of inclusiveness that relates to the broader
notion of definiteness. Definiteness is comprised of qualitative and quantitative
definiteness (inclusiveness). In Estonian, the arguments’ case-marking can depend
on the latter. If an NP is definite due to inclusiveness, the reference is to the totality
of the objects or mass in the context which satisfies the description. For example,
in the sentence Beware of the dogs, ‘the dogs’ is definite because it refers to
all the dogs, i.e. inclusive amount relevant in this context (for example in a
particular house). Quantitative definiteness also involves uniqueness (the Sun)
but Lyons subsumes it under inclusiveness.
Aspect is a universal semantic category that has a multitude of grammatical
and lexical means of expression in the world’s languages. In the analysis I rely
on Comrie’s (1976: 16−24) delineation of aspectual categories. According to him
the perfective is used when the speaker wishes to denote a complete situation,
with a beginning, middle, and end. A situation depicted by the perfective aspect
can have parts and internal complexity but it should still be looked at as a
single event with clearly circumscribed limits. The use of the imperfective pays
special attention to the internal structure of the situation. In this case the situa-
tion is viewed from within. For example:
(7) Saat-si-n üh-t nais-t koju.

escort-PST-1SG one-PAR woman-PAR home.ILL
‘I was escorting a woman home.’ (imperfective aspect, the action is
ongoing at the moment the speaker is referring to)
(8) Saat-si-n ühe naise koju.

escort-PST-1SG one.GEN woman.GEN home.ILL
‘I escorted a woman home.’ (HM) (perfective aspect, the single event is
complete in the speaker’s depiction)
3 Estonian argument types and clause types

Thorough research has been done earlier on Estonian core arguments. The
monographs and articles that have been published more recently cover their
role, referential properties, coding, information structure and word order (e.g.
Erelt et. al 1993; Nemvalts 1996 and 2000; Lindström et al. 2008; Hiietam 2003;
Huumo 1993 and 2002; Lindström 2005), their function of expressing aspectual
alternation (Erelt et. al 1993; Tamm 2007, 2009 and 2012; Vaiss 2004) and voice
distinctions (Erelt et. al 1993; Torn-Leesik 2009).
In Estonian linguistics, argument realization has been largely discussed in
connection with simple sentence clause types (clause types for short; Erelt &
Metslang 2006; see also Erelt et al. 1993; Huumo 1993; Erelt 2005; Nemvalts
2000). Clause types can be analyzed as Construction Grammar’s argument struc-
ture construction types (cf. Goldberg 1995; Croft 2001). See Metslang (2013) for a
discussion on the semantic and morphosyntactic grounds of the Estonian clause
type notion.
Two main simple clause types have been proposed in Estonian: marked and
unmarked clauses (Erelt et al. 1993, Erelt & Metslang 2006; see also Huumo
1993). Unmarked (multifunctional) clauses are highly schematic constructions
and are used with a wide range of verbs and argument structures (see examples
(5) and (6) above). An example of the marked (monofunctional) clause types is
the EC that includes the possessive clause as a subtype (see examples (9) and
(10)). Erelt and Metslang define the unmarked and marked clauses by a set of
properties that includes topicality, event type, argument roles and argument
coding. In unmarked clauses the topic is predominantly the clause-initial subject,
while in marked clauses it is the clause-initial oblique NP (Erelt & Metslang 2006).
(9) Prantsusmaa-l on hea kliima.

France-ADE be.3 good.NOM climate.NOM
‘There is a good climate in France.’ (HM, EC)
(10) Ta-l on suvila.

3 SG -ADE be.3 summer.cottage.NOM
‘S/he has a summer cottage.’ (HM, possessive clause)
186 Helena Metslang
This article only focuses on unmarked clauses and one type of marked clauses,
the EC. To characterize the coding of these clause types, the final part of this
section briefly outlines the case and agreement properties of A, S, O and e-NP.
To indicate the salience of the different coding options I will illustrate this with
some preliminary data on corpus frequencies.
The arguments of Estonian unmarked clauses and ECs occur in three gram-
matical cases: the nominative, genitive and partitive. Table 1 presents the fre-
quencies of all case-uses in the studied data (zero-anaphora and clausal argu-
ments bare the value ’irrelevant’) (cf. the methodology overview in Section 1).
Table 1: Distribution of cases among the arguments in the corpus (%)
Case/Argument A S O e-NP
nominative 32 34 3 31
genitive 0 0 100 0
partitive 0 0 58 42
irrelevant 40 37 20 3
Total (abs. No) 130 130 130 130
In Estonian, the verb agrees with the subject in number and person (see again
examples (5) and (6) in Section 2). Unlike in Finnish, in Estonian agreement
is explicit in the case of both speech act participants (SAP) and third person
subjects. In negation and the conditional and quotative moods, the verb con-
jugation paradigm lacks subject-verb agreement. For example:
(11) Ma / te ei istu.
1 SG . NOM / 2 SG .NOM NEG sit
‘I am not sitting. / You (PL) are not sitting.’ (negation in indicative mood)
As mentioned above, the verb does not agree with the direct object. It agrees
with the e-NP in the affirmative clause:
(12) Siis järgne-s hommikune udu.

then follow-PST.3SG morning.NOM fog.NOM
‘Then morning fog followed. (The next activity that followed our boat trip
was walking in the morning fog.)’
(13) Pilti-de-l ol-i-d kodulooma-d.

picture-PL-ADE be-PST-3PL domestic.animal-NOM.PL
‘There were domestic animals in the pictures.’
Plural partitive e-NPs are an exception to this rule: they trigger neither number
nor person agreement:
(14) Pilti-de-l ol-i koduloom-i.

picture-PL-ADE be-PST.3SG domestic.animal-PAR.PL
‘There were domestic animals in the pictures.’ (nominative as an
unmarked case of EC)
The agreement frequencies in the corpus are summarized in Table 2 (the clausal
arguments and ECs with omitted predicates are marked with the value ‘irrelevant’).
Table 2: Agreement with the predicate in person and number (absolute numbers)
Feature value A S O e-NP

agrees with the verb 129 128 0 114
no agreement 0 0 104 9
irrelevant 1 2 26 7
Total 130 130 130 130
We can see that the coding of e-NPs has considerable similarities to the overt
coding of unmarked clause subjects: 57% of e-NPs are in the subject case, while
88% formally agree with the verb. It can be argued though that in the ECs, the
third person singular is the verb’s unmarked form. Therefore it is not clear
whether the verb predominantly agrees with the e-NP. Only 14% of the e-NPs
are clear instances of verbal agreement. In other cases the predicate verb is in
the unmarked form or its forms have been neutralized in the paradigm (negative
verb forms or conditional mood, the present tense paradigm of the verb olema
‘to be’) (see example (12) for the default verb form and (9) for the use of the
verb olema).
4 Introduction to case-alternation rules of

O and e-NP
Both the Estonian object and e-NP have differential case-marking systems that
are multifactor phenomena expressing a variety of fine-grained distinctions.
This is characteristic of languages with relatively rich case systems (de Hoop &
de Swart 2009: 5). In this section I will outline the earlier findings on the factors
influencing the Estonian object’s and e-NP’s case. The data studied in this paper
188 Helena Metslang
suggests that it is possible to consider e-NP as a non-canonical intransitive sub-

ject (see Sections 5.2 and 8). Therefore it is interesting to compare the case-
alternation system of e-NP with the theoretical views on DSM. I will start with
some definitions that reflect that DSM and DOM are conditioned by similar factors.
In their introduction to the optimality theoretic collection on DSM, de Hoop
and de Swart summarize the possible factors that can influence subject-marking.
They find that these factors can belong to any level of grammar:
DSM is a cross-modular phenomenon that is not triggered or constrained by semantic or

pragmatic features in the input alone. Rather, it is the optimal outcome of a conflict
between certain rules, which can be syntactic, semantic, pragmatic, morphological or pho-
nological in nature. (de Hoop & de Swart 2009: 5.)
Remarkably, Estonian e-NP case-assignment factors include almost all of these

levels and therefore I will treat e-NP’s case alternation as the Estonian equiva-
lent of DSM. In addition, also lexical factors that cannot always be reduced to
semantic features play a dominant role in differential marking of the e-NP accord-
ing to the treatment used in this study (which is based on Metslang 2012).
As Bickel (2010) describes, DOM often involves the situation when, “O argu-
ments are mapped into different GR [grammatical relations] (. . .) for some con-
struction, depending, mostly in a statistical rather than categorical way, on
such referential properties as animacy, humanness, definiteness, specificity or
more general notions of saliency.” Hence even just the O argument’s referent can
influence DOM in a multifaceted way. In addition to the referential properties,
Dixon (1994) also discusses other factor groups that determine alignment of O
or any other argument into the same or different grammatical relations (lexical
predicates and clausal factors). He presents a comprehensive system of align-
ment split factors consisting of referential, verb-related and clausal factors. In
Dixon’s (1994: 70−110) treatment:
1. the main referential factors that influence argument coding are animacy
and humanness, definiteness, specificity, lexical class, person and number;
2. the second type of factors that determine a language’s argument coding
involves lexical predicates and generalized predicate classes;
3. clausal conditions include tense, aspect and mood, morphological form of
the predicate, main vs. subordinate clause, polarity and other arguments in
the clause.
(See also Witzlack-Makarevich 2011: 73−157.) To summarize, there are similar and
extremely varied sets of possible argument realization factors that apply to any
argument type: almost anything can potentially influence an argument’s coding.
As the following subsections show, Estonian provides ample evidence of this.
In the following sections I will outline how, in Estonian, two ways of O’s
and e-NP’s case assignment are especially important. They regard all three levels
in Dixon’s system. Estonian arguments’ case can be determined lexically – by
particular verbs (especially in the case of Os; strategy (2) above) or nouns (espe-
cially in the case of e-NPs; strategy (1)). Alternatively, it can follow the situation-
ally triggered fluid strategy that is independent of lexical constraints (cf. (7) and
(8) in Section 2). In this case the case-marking reflects the semantics of each
particular instance of use (strategy (3); see also Dixon 1994: 78 and Section 8 in
this paper).
4.1 Direct object

Estonian direct objects’ case-marking depends on all the above-mentioned factor
types. As mentioned above, particular lexical predicates are especially influential
in Estonian DOM. There are also two kinds of factors that are related to the object-
NP’s referential properties (personal pronouns and nominal inclusiveness) and
two kinds of clause level factors (aspect and the morphological form of the
predicate).
Concurrence of inclusiveness and aspect

The object’s case depends on the totality-partiality system that, among other
functions, indexes distinctions in the NP’s inclusiveness (quantity) and the situa-
tion’s aspect. In the case of the direct object, the total cases are the genitive and
nominative. These two cases are used when the quantity of the object referent is
inclusive and the aspect of the clause is perfective. The default total case is the
genitive. A common use of the nominative is marking plural NPs (see example
(15)). The partitive is frequency-wise the unmarked case of the direct object
and it is used when the aspect is imperfective or the nominal quantity is non-
inclusive (Erelt 2009: 9) (see examples (16)−(18)).
(15) Ost-si-me oma tehnika / masina-d Austria-st.

buy-PST-1PL our.GEN technology.GEN / machine-NOM.PL Austria-ELA
‘We bought our technology / machines from Austria.’ (HM) (perfective
aspect and inclusive quantity)
(16) Osta-me oma tehnika-t / masina-i-d Austria-st.

buy-1PL our.GEN technology-PAR / machine-PL-PAR Austria-ELA
‘We are buying our technology / machines from Austria.’ (HM)
(imperfective aspect and non-inclusive quantity)
190 Helena Metslang
(17) Ta jõ-i vett ja hakka-s siis söö-ma.

3 SG .NOM drink-PST.3SG water.PAR and start-PST.3SG then eat-INF
‘He drank some water and then started to eat.’ (Erelt 2009: 9) (perfective
aspect, non-inclusive quantity)
I will repeat examples (7) and (8) from Section 2 for the sake of convenience.
Object case-alternation signifies the aspectual distinction here as the referent is
inclusive in both clauses.
(18) Saat-si-n üh-t nais-t koju.

escort-PST-1SG one-PAR woman-PAR home.ILL
‘I was escorting a woman home.’ (imperfective aspect, inclusive quantity)
(19) Saat-si-n ühe naise koju.

escort-PST-1SG one.GEN woman.GEN home.ILL
‘I escorted a woman home.’ (HM) (perfective aspect, inclusive quantity)
Lexical predicates
Object case is often governed by particular aspect-related verb classes. Vaiss
(2004) analyzed 495 simple verbs and divided them into four groups as follows:
– perfective verbs (9% of all simple verbs, e.g. tapma ‘to kill’, leidma ‘to find’)
can mostly be used with the total case object and denote perfective aspect,
(20);
– aspectual verbs (16%, e.g. parandama ‘to repair’) can be used with both the
total cases or the partitive. They denote, depending on the case, perfective
or imperfective aspect, see (21) and (22);
– verbs only taking a partitive object (46%; often mental and cognition verbs)
denote depending on the verbal meaning, tense and sentential context
either imperfective or perfective aspect (23);
– verbs with several aspectual meanings constitute 29% of simple verbs (e.g.
verbs with a general meaning like ajama ‘to drive’, andma ‘to give’).
(20) Leid-si-n väljapääsu.

find-PST-1SG exit.GEN
‘I found an exit.’
(21) Isa paranda-b jalgratas-t.

father.NOM repair-3SG bike-PAR
‘Father is repairing the bike.’ (Vaiss 2004: 44; imperfective aspect)
(22) Isa paranda-b jalgratta ära.

father.NOM repair-3SG bike.GEN up
‘Father will repair the bike.’ (Vaiss 2004: 44; perfective aspect)
(23) Kohta-si-n ainult üh-t inimes-t.

meet-PST-1SG only one-PAR person-PAR
‘I only met one person.’
Vaiss also shows that out of 253 phrasal verbs, about 80% are perfective verbs
(2004: 65, 83).
Speech act participant pronouns

As in many other languages, the Estonian personal pronoun system has excep-
tional coding. Semantically, personal pronouns are singular count nouns (ma ‘I’,
sa ‘you.SG’) or are often used to denote inclusive quantity groups (me ‘we’ and
te ‘you.PL’). Therefore, in combination with the perfective aspect they should
take a total case in the object position. However, in the object position the
personal pronouns denoting SAPs tend to take the partitive even when the
aspect of the situation is perfective and the quantity is inclusive (Erelt et al.
1993: 53),6 compare (24) and (25).
(24) Komandant kirjuta-s Peetri sisse.

housemaster.NOM register-PST.3SG Peeter.GEN in
‘The housemaster registered Peeter.’ (HM; perfective aspect, inclusive
quantity referent)
(25) Komandant kirjuta-s min-d sisse.

housemaster.NOM register-PST.3SG I-PAR in
‘The housemaster registered me.’ (perfective aspect, inclusive quantity
referent)
Hence a certain case is preferred for distinguishing referentially high Os from

As. This relates to the two argument coding strategies that are used in lan-
guages: discrimination and indexing. Comrie (1978) outlines the discriminatory
view on argument marking by showing that in transitive sentences, as there are
two core arguments, A and O, there is a need to differentiate which element is
which. In accusative languages it is common to distinguish A from O by overtly
6 This is also the case with the reflexive pronoun enese/enda ‘oneself.GEN/PAR’.
192 Helena Metslang
marking O. O marking is in the majority of accusative languages more complex

than the nominative (i.e. the case of A) (ibid.).7 A different view on argument
marking strategies suggests indexing or direct marking. In the case of direct
marking, each instance of use of a verb is dealt with separately depending on
the semantic functions of the argument NPs (cf. Dixon 1994: 24). The exceptional
preference for increased overt marking of Estonian SAP Os (by the use of the
partitive) can be analyzed as a discriminatory argument marking tendency that
tends to override the other case-assignment rules (see below for discrimination
and indexing).
Morphological form of the predicate

The morphological form of the predicate can be classified under the clausal
factors of argument realization in Dixon’s (1994) approach. In the Estonian
negative clause only the partitive object can be used:
(26) Me ei osta oma tehnika-t / masina-i-d Austria-st.

1 PL .NOM NEG buy our.GEN technology-PAR / machine-PL-PAR Austria-ELA
‘We do not buy our technology / machines from Austria.’ (HM)
Secondly, as already indicated above, in general total O occurs in the genitive in

Estonian, see example (27). This case use (instead of the unmarked nominative)
may be necessary for serving the discriminatory function – for distinguishing O
from A in the transitive clause. The second object case rule of this section con-
cerns these transitive clauses where the subject is not overtly used. Namely, the
predicate’s form influences the total object’s case in imperative, impersonal and
some infinitival constructions (Erelt et al. 1993: 49). In these (predominantly)
subjectless constructions only nominative total O is used – the genitive is un-
grammatical here, see (29) and (30). For distinguishing O from A, the genitive
total object is unnecessary, hence the discriminatory function does not stipulate
O’s case use. In Estonian subjectless clauses, the object case seems to have the
function of indexation of inclusiveness and aspect instead (see also Helasvuo
2001: 43−44 on the same phenomenon in Finnish).8 In the case of Finnish im-
peratives, Bickel (2010) describes this phenomenon as scenario (argument coding
depends on how two or three arguments interact with each other, i.e. define a
scenario).
7 Comrie uses the abbreviation P instead of O.

8 In the clauses with an overt A argument, total Os are still in the nominative if they are in the
plural. As A arguments are predominantly singular (see Section 6) then, in the cases when O is
in the plural A and O are distinguished rather by the number than case category. See also
Tamm (2004) on the role of total cases in Estonian.
(27) Mari sõ-i koogi ära.

Mari.NOM eat-PST.3SG cake.GEN up
‘Mari ate up the cake.’ (HM)
(28) *Mari sõ-i kook ära.

Mari.NOM eat-PST.3SG cake.NOM up
Intended meaning: ‘Mari ate up the cake.’ (HM)
(29) Söö kook ära!

eat.IMP.2SG cake.NOM up
‘Eat up the cake!’ (Erelt et al. 1993: 53)
(30) *Söö koogi ära!

eat.IMP2SG cake.GEN up
Intended meaning: ‘Eat up the cake!’ (HM)
4.2 Existential noun phrase

Similarly to objects, e-NP’s case-alternation shows a totality-partiality system:
it alternates between the partitive and the nominative (to make the e-NP data
comparable with O, I call the nominative marking on e-NP the ‘total case’). The
e-NP’s case-marking predominantly depends on lexical and clausal properties
(see the introduction of Section 4). It can be seen that the e-NP lacks the main
object case-marking properties, as:
– the e-NP cannot be in the genitive;
– singular count nouns and other inclusive quantity NPs cannot occur in the
partitive in affirmative ECs;
– with e-NP, aspect is never the sole trigger of e-NP’s case-marking – it is only
an optional reading, a side-effect of inclusiveness (Metslang 2012, see also
Nemvalts 1996).
Section 5 compares the e-NP’s and O’s case-marking systems in detail. The
following overview and the successive analysis in Section 5 of the e-NP are
based on the set of rules from Metslang (2012) and on the findings of Nemvalts
(1996).
Inclusiveness (lexical and situational)

The nominative is frequency-wise the unmarked case of the e-NP. In affirmative
clauses, the e-NP usually occurs in this form (31). The partitive is commonly used
194 Helena Metslang
when the speaker chooses to emphasize the referent’s non-inclusive quantity in a

particular situation under consideration (when both the e-NP’s head noun and
the predicate verb allow the use of both the nominative and partitive; compare
example (32) with (31)) or the noun lexeme is an ‘existential partitive’ – a noun
obligatorily marked by the partitive in (sometimes frozen) ECs (see (33)). Exis-
tential partitives usually denote referents with a non-inclusive quantity. In
some cases this semantic feature is not evident and the partitive use is lexical-
ized (for the latter reason this case choice factor is uniformly treated as a lexical
phenomenon here and not as a semantic feature; cf. Metslang 2012).
(31) Selle-l kase-l on juba lehe-d.

this-ADE birch-ADE be.3 already leaf-NOM.PL
‘This birch has leaves already.’ (adapted from Vilkuna 1992: 61)
(32) Selle-l kase-l on juba leht-i.

this-ADE birch-ADE be.3 already leaf-PAR.PL
‘This birch has some leaves already.’ Lit. ‘On this birch is already leaves.’
(adapted from Vilkuna 1992: 61; partitive denoting non-inclusive quantity
of an e-NP)
(33) Enda=l=gi Ø ruumi vaevalt ringi pööramise-ks.

self=ADE=CL (be.3) space.PAR merely around turning-TRAN
‘We ourselves (have) only just (enough) space for turning around.’
(existential partitive noun)
The nominative is often used on divisible e-NPs (mass nouns and plural count
nouns). Similarly to the object’s total case, it can mark inclusive quantity if there
is a contextual boundary to the referent. For example, in (31) the contextual
boundary is the leafage of the whole tree. In this example, the nominative is
used because the speaker is referring to the whole leafage of the tree. When
there is no contextual boundary to the referent, i.e. the referent is not part of
a bounded whole (see also Koptjevskaja-Tamm & Wälchli 2001: 665 on this
semantic feature) and the inclusiveness of the referent is irrelevant (34) then
the use of the nominative is semantically unmarked. The role of the EC is merely
stating the existence of the subject referent in the given location and not its
inclusive quantity.
(34) Pilti-de-l ol-i-d kodulooma-d.

picture-PL-ADE be-PST-3PL domestic.animal-NOM.PL
‘There were domestic animals in the pictures.’ (nominative as the
unmarked case of the e-NP)
Hence, EC’s nominative-partitive alternation can express the distinction of

just identifying the class of the entity vs. referring to the entity’s quantity (cf.
Nemvalts 2000: 150). In this volume Paykin describes a similar alternation in
Russian unaccusative and transitive constructions (partitive genitive – nominative
alternation and partitive genitive – accusative alternation respectively).
As mentioned above, the nominative is often caused by the lexical properties
of the Estonian e-NP’s phrasal head. The ‘existential nominatives’ category in-
cludes nouns that are inherently unmarked for inclusiveness (35) or have inher-
ently inclusive quantity (this mainly includes singular count nouns, example
(36)).
(35) Siis järgne-s hommikune udu.

then follow-PST.3SG morning.NOM fog.NOM
‘Then the morning fog followed. (The next activity that followed our boat
trip was walking in the morning fog)’
(36) Ja korraga torka-s mu-lle pähe veider mõte.

and suddenly strike-PST.3SG I-ALL head.ILL strange.NOM thought.NOM
‘And suddenly I got this strange idea.’ Lit. ‘And suddenly stroke into my
head a strange idea.’
Hence, the nominative marking of e-NPs denotes either inclusive or unmarked

quantity both in the case of situationally and lexically determined case-assignment.
Morphological form of the predicate

In neutral, non-contrasted negative clauses, e-NPs take the partitive:
(37) Peenra-l ei kasva lill-i.

flowerbed-ADE NEG grow flower-PAR.PL
‘There are no flowers growing in the flowerbed.’
Predicate verbs and constructions

The case choice of the e-NP peripherally depends on predicate verbs and clausal
constructions (cf. Rätsep 1978; Nemvalts 2000; Metslang 2012). Although the
impact of verbs and constructions on e-NP’s case is rare, it is necessary to
discuss it here for the purpose of giving a full picture of the case system. Rätsep
(1978) describes the impact of clausal constructions on argument marking via
verb-governed sentence patterns. Rätsep’s sentence pattern is a generalized abstrac-
tion that links a set of verbs with a set of grammatically similar simple sentences
196 Helena Metslang
that share the number, coding and order of the verb’s arguments and obliques.9
In Metslang (2012) I specify the distinction between constructions and lexical
predicates as e-NP case-factors. There are verbs that determine their e-NP’s
case. However, a vast majority of the EC verbs and constructions permit the use
of both cases. Examples (38) and (39) present ECs whose e-NP’s case is deter-
mined by the whole construction or by a specific verb.
(38) Meistri-t jätku-s iga-le poole.

master-PAR suffice-PST.3SG any-ALL towards
‘The master could help out everywhere.’ Lit. ‘Master sufficed/was
everywhere.’ (Rätsep 1978: 154; ‘partitive e-NP only’ construction:
exceptional use of singular count nouns in the partitive)
(39) Lähene-si-d valimise-d.

approach-PST-3PL election-NOM.PL
‘The elections were approaching’ (HM) (‘nominative e-NP only’ verb)
The verb lähenema only occurs in constructions taking a nominative argument.

The verb jätkuma also takes nominative subjects in other constructions.
5 Corpus study: Comparison of the O’s and e-NP’s

case-marking
In this section I will examine corpus data on Dixon’s and Witzlack-Makarevich’s
argument realization determinants (cf. the beginning of Section 4). I will analyse
how referential, predicate-related and clausal case-assignment factors affect O’s
and e-NP’s case and show how they interact with each other. On the basis of the
analysis of 104 transitive clauses and 125 ECs (that contain case-marked Os
and e-NPs), I will weigh the importance of all factors and compare their usage
frequency in O’s and e-NP’s case assignment. Sections 5.1 and 5.2 deal with the
referential factors influencing the O’s and e-NPs’ case. Section 5.1 compares the
formation of O’s and e-NP’s case from the viewpoint of more clear-cut referen-
tiality rules and suggests a new integrated account of quantification-related
hierarchies. In Section 5.2 I will compare how the Referential hierarchy affects
these arguments’ case. I will show that there are some statistical tendencies
that can be overridden by the rules described in Section 5.1. I also discuss in
9 The elements’ order in the clause is determined on the basis of context free sentences,
therefore the word order criterion has been ignored in this corpus-based study.
this section what the Referential hierarchy tells us about the subjecthood of the
e-NP. Sections 5.3 and 5.4 study how the predicates and clausal context affect
O’s and e-NP’s case, sometimes intertwining with the referentiality rules. Section
5.5 summarizes the arguments’ case-formulation principles on the basis of the
corpus frequencies and discusses the general subjecthood status of the e-NP.
5.1 Referential restrictions on argument case

In this subsection I will integrate the accounts of O’s and e-NP’s referential case-
assignment rules outlined in Sections 4.1 and 4.2. It will be discussed to what
extent O and e-NP share referential case choice factors in the corpus and to
what extent they are different. Both lexically determined as well as situationally
determined (fluid) nominal properties will be discussed. More specifically, two
quantification-related hierarchies of semantic case-assignment factors will be
proposed. Their impact is measured on the corpus data along with the impact
of two lexical factors. Table 3 summarizes the corpus findings that show how
frequent the application of each referential rule is in the data.
Table 3: Referential rules affecting case-marked Os’ and e-NPs’ case in the data (%, n = 229)
The factor’s level in Referential case- Proportion of all Os Proportion of all e-NPs
Dixon’s (1994) system assignment factor whose case is trig- whose case is trig-
gered by this factor gered by this factor
Situational Quantification-related 54 30
hierarchies
Lexical Personal pronouns 4 0
(SAPs)
Inclusiveness 0 47
semantics of noun
lexemes
As can be seen, on the referential level, an important factor that influences both
O’s and e-NP’s case is the situational (fluid) quantification. Lexicalized case
choices also play a dominant role in e-NPs’ marking.
Lexically inherent properties represent two distinct phenomena in the
case of the direct object and e-NP. The pronominal direct objects referring to
SAPs tend to take the partitive case even in environments where otherwise total
case object is used (see Section 4.1) and form a small minority of O arguments in
the corpus. The lexical properties affecting the e-NP’s case marking are similar
to the semantics of quantification-related hierarchies (see below). The largest
group of e-NPs have their case triggered by this factor in the corpus (see Figure
2 in Section 5.5).
198 Helena Metslang
In the corpus of 229 sentences the largest proportion of Os and e-NPs have
their case assigned on the basis of situationally triggered quantification (in-
clusiveness) parameters. I suggest that this prominent part of O’s and e-NP’s
case-marking systems relies on identical principles. I see a need for a unified
approach to account for the overlaps in both arguments’ case-marking systems.
In the following I will propose two quantification-related hierarchies for it. These
are the Quantitative markedness hierarchy and the Inclusiveness hierarchy and
they concern the plural and mass noun Os and e-NPs. The establishment of the
hypothesis of these hierarchies allows for fine-grained empirical comparisons
between the marking of O and e-NP.
The application of these hierarchies depends above all on whether the
referent’s inclusiveness is relevant or irrelevant for the speaker. I will first intro-
duce the Quantitative markedness hierarchy.
(40) Quantitative markedness hierarchy

UNMARKED QUANTITY > MARKED QUANTITY
Quantitative markedness depends on whether the speaker refers to the referent’s

inclusiveness or not. The argument takes the total case if it has a semantically
unmarked quantity (it is not specified whether the quantity of the referent in
question is inclusive or non-inclusive, or whether it participates in the situation
totally or only part of it does) (see example (34) in Section 4.2). The e-NPs’ quan-
tity can be either marked or unmarked. The Os’ quantity is always marked in the
studied corpus, however it is possible to find examples of Os with a referent
with unmarked inclusiveness:
(41) Linnaosavalitus-te-s moodusta-takse lasteringi-d.

borough-PL-INE form-IMPS children’s.group-NOM.PL
‘Children’s groups will be formed in boroughs.’ (adapted from Erelt et al.
1993: 51)
(42) Nii kurb lugu, et aja-b pisara-d silma.

so sad.NOM story.NOM that drive-3SG tear-NOM.PL eye.ILL
‘It is such a sad story that it makes me cry.’ Lit. ‘So sad story that drives
tears in eye.’ (Vaiss 2004: 103)
Hence most of the Os and some of the e-NPs need their case to be assigned else-
where.
If an argument’s quantity is relevant to the speaker (i.e. it is overtly deter-

mined whether the referent has inclusive or non-inclusive quantity) then the
Quantitative markedness hierarchy is not sufficient for accounting for the argu-
ment’s case. In this case another (nested) hierarchy determines its case-marking.
I call it the Inclusiveness hierarchy (I use two separate hierarchies because they
concern different level phenomena).
(43) Inclusiveness hierarchy

INCLUSIVE QUANTITY > NON - INCLUSIVE QUANTITY
If the argument has a non-inclusive quantity, it takes the partitive and if it has
inclusive quantity it has the opportunity of taking the total case. As mentioned
in Section 4.1, in the case of object, also the situational aspect is involved in the
final case designation of inclusive/non-inclusive O arguments. No further factors
influence such e-NPs’ case. The distinction can be seen in examples (44) and
(45) for e-NP and (15) and (16) for O, repeated here as (46) and (47):
(44) . . . kui maja-l on sarika-d peal.

when house-ADE be.3 rafter-NOM.PL on
‘. . . once when the house has rafters (built) on top.’
(e-NP, inclusive quantity)
(45) (Raputa-si-n põõsas-t, nii et)

shake-PST-1SG bush-PAR so that
must-i marj-u ema-le pähe sada-s.
black-PAR.PL berry-PAR.PL mother-ALL head.ILL fall-PST.3SG
‘(I shook the bush so that) black berries kept falling on my mother’s
head.’ (e-NP, non-inclusive quantity)
(46) Ost-si-me oma tehnika / masina-d Austria-st.

buy-PST-1PL our.GEN technology.GEN / machine-NOM.PL Austria-ELA
‘We bought our technology / machines from Austria.’ (HM)
(O, inclusive quantity)
(47) Osta-me oma tehnika-t / masina-i-d Austria-st.

buy-1PL our.GEN technology-PAR / machine-PL-PAR Austria-ELA
‘We are buying our technology / machines from Austria.’ (HM)
(O, non-inclusive quantity)
200 Helena Metslang
The semantic difference between unmarked quantity and non-inclusive quantity

reference is that the former makes no reference to quantity at all (it just identi-
fies the referent) whereas the latter states that the quantity is not determined,
unbounded.
The hierarchies treat e-NP and O similarly in the sense of both influence on
the case and statistical tendencies. This is a sign of the proximity of these two
arguments. The features on the left hand side of each hierarchy contribute to the
total case marking of e-NP and O. The corpus data suggest that e-NP’s preferences
are not very strongly biased towards any of the options of the two hierarchies
(see Figure 2 and Table 6 in Section 5.5 and Metslang 2012: 196). As regards the
Inclusiveness hierarchy, the frequency distinctions of the inclusive and non-
inclusive O are also relatively small. The only strong distinction is O’s clear
preference for marked quantity over unmarked quantity in the Quantitative
markedness hierarchy and this makes it different from the e-NP. Larger quantita-
tive studies are needed to confirm e-NP’s and O’s dispositions in these hierar-
chies in order to state which semantic options and respective case-uses are
more frequent among each argument type.
5.2 Referential hierarchy and Estonian argument realization

It has been typologically observed that arguments’ realization (including the
selection of arguments into grammatical relations like subject or object) often
depends on their referential type. They can for example be defined by their
‘social importance’, organized by referential hierarchies (cf. Bickel 2010). For
instance, subjects tend to be highly referential. This section studies whether the
typologically described referential hierarchies have any impact on Estonian O
and e-NP realization.
Bickel (2010: 410) summarizes this phenomenon as follows: the grammatical
relations of many languages rather favor animates than inanimates, known than
unknown referents, etc. These properties can determine for example inclusion/
exclusion from the subject, object or some other category (e.g. only the semanti-
cally higher, more agent-like entity can occur in the subject position), case-
assignment (e.g. animate O receives dative marking and inanimate O receives
nominative marking), agreement rules, etc. (see also Givón 2001: 200, 220−221).
These preferences are described by various referential hierarchies:
(48) A . SPEECH ACT PARTICIPANT > KIN / NAME > HUMAN > ANIMATE > INANIMATE > MASS
B . SPECIFIC > NONSPECIFIC REFERENTIAL > GENERIC / NONREFERENTIAL
C . KNOWN / TOPICAL / THEMATIC / DEFINITE > NEW/ FOCAL / RHEMATIC / INDEFINITE
D. SINGULAR > PLURAL
(Bickel 2010: 410)
It has been suggested that in many languages, O arguments that are higher on
hierarchies of prominence, animacy, definiteness and the like receive different
case marking from the O arguments lower on the hierarchy (Comrie 1989).10 It
is often predicted that higher O arguments should carry overt (‘accusative’)
case marking in contrast to lower O arguments, which should carry no overt
case marking (i.e. be in the unmarked ‘nominative’) (cf. Witzlack-Makarevich
2011: 76). Overt marking of A is expected if it occurs on the right hand side of
the hierarchy, whereas As with zero-marking are more likely to occur on the
left. In split intransitive systems, distinctive S marking is used in different ends
of the hierarchy (cf. Dixon 1994: 85ff.). In the world’s languages there are studies
that both support and counter the claims about universal hierarchy effects (de
Hoop & de Swart 2009: 2−5; Bickel, Witzlack-Makarevich & Zakharko, to appear).
Thematic importance reflected by these hierarchies is an aspect of topicality that
also statistically determines whether the referent remains topical in the subse-
quent discourse or will not be mentioned again (Givón 2001: 198−199, 455−456).
I have adopted a rather detailed version of the hierarchies in order to better
describe the lower, less agent-like referents that are common among Os and
e-NPs in the corpus, see (49).
(49) The version of the Referential hierarchy used in this analysis

SPEECH ACT PARTICIPANT > HUMAN > CONCRETE > ABSTRACT > EVENT
> NON - REFERENTIAL
Hence, in the following I will only look at these topicality-related semantic prop-
erties that are typologically commonly used in referential scales, and will not try
to integrate this approach with the quantity-related case factors described in
Section 5.1. The quantification-related hierarchies belong to a different domain
of grammar because they probably do not represent topicality oppositions.
Furthermore, at least the Inclusiveness hierarchy does not demonstrate a strong
ability to distinguish between the argument types in the corpus.
Non-referential elements are the ones that do not have nominal reference.
The set of non-referential subjects, objects and e-NPs involves propositions like
example (1) in Section 2, events where the potential nominal reference is not
foregrounded (50) and nominal parts of more or less opaque multi-word verbs (51).
(50) Seal pol-nud kombe-ks ilma põhjuse-ta naer-da.

there NEG.be-PST.PTCP custom-TRAN without reason-ABE laugh-INF
‘It was not the custom there to laugh without a reason.’ (e-NP depicting
a non-referential event)
10 Again, Comrie uses the abbreviation P instead of O.

202 Helena Metslang
Figure 1: Mapping of the referential categories to argument positions (%)
(51) Ehk tule-b me-i-l isa-ga se-l teema-l juttu.

perhaps come-3SG 1PL-ADE father-COM this-ADE topic-ADE talk.PAR
‘Perhaps this topic will come to discussion between my father and me.’
(e-NP as a part of a multi-word verb juttu tulema ‘to come to discussion’)
I also distinguish a separate ‘event’ category that has become referential via
nominalization (see (68) and (69) in Section 6.3). Also situations and states have
been classified as events here.
The 520 corpus examples show some clear distinctions in how different ele-
ments of the hierarchy are mapped to A, S, O and e-NP in Estonian (see Figure 1).
There is no categorical distribution of the semantic hierarchy properties
among different arguments. All these properties are distributed statistically. In
general, almost all the arguments can take referents with several meanings (SAP,
abstract, etc.).
In the data, the Referential hierarchy gets divided between humans and
inanimate entities. The referents to the left of the hierarchy tend to occur in
the S and A positions, and to the right as O and e-NP. A and S align together:
reference to SAPs and people naturally takes place in the A and S positions
(85% of the A and S arguments depict SAPs or other people), and hardly ever
in the O and e-NP positions. This supports the grammatical relations based
Referential hierarchy hypothesis discussed in (Givón 2001 and Bickel 2008). The
hypothesis suggests that the rank of an argument in the Referential hierarchy
correlates positively with access to grammatical relations as higher arguments
are more likely to be topical. Most of the remaining S and A arguments denote
processes happening with abstract or inanimate entities or occur in equative,
attributive or locative clauses. For example:
(52) Üllatusefekt meeskonna-s ol-i suur.

surprise.effect.NOM team-INE be-PST.3SG big.NOM
‘The element of surprise in the team was big.’ (S denoting an event)
With respect to the Referential hierarchy, S is a more heterogeneous category

than A. O is the most heterogeneous category of the four and it mainly expresses
lower entities. The O and e-NP positions show considerable similarities as re-
gards the hierarchy: abstract entities and concrete inanimate objects primarily
occur in these positions (78% of the abstract and 79% of concrete referents
of the corpus). The prevalence of mapping non-human referents in object and
e-NP relations is also linked to the referent’s activeness status in the discourse,
the phrase’s heaviness and word order (see Helasvuo (2001: 56) on this in Finn-
ish). NPs referring to humans (subjects) are preferably preverbal in Estonian
(Lindström 2002: 102; see also Section 7.2 of this article). The expression of non-
referential content among the core arguments is largely restricted to the object
relation in Estonian. A total of 79% of the non-referential elements occur at the
object position (mainly infinitival constructions and clausal constructions,
including direct and reported speech, as in example (1) in Section 2). This is
deeply linked to the fact that the objects are often the foci of the sentence and
the focus does not have to be referential (cf. Lambrecht 1994: 336). Due to the
special role of the EC, 96% of e-NPs are referential although they are usually
foci too (cf Section 1; this is also characteristic of Finnish e-NPs; Helasvuo 2001:
100).
One may ask whether the adopted version of the Referential hierarchy explains
the statistical distribution of O’s and e-NP’s case. However, in the corpus the
arguments’ case ultimately depends on the rules outlined in Section 4. For
example, despite the fact that concrete e-NPs preferentially take the nominative,
concrete referents in negative clauses are still marked with the partitive, see (37)
in Section 4.2. Yet, in the corpus there is some partial evidence for hierarchy
effects being suitable for explaining DOM in Estonian. The fact that SAPs – the
semantically high Os – tend to be expressed by the overtly case-marked partitive
and much less by the total case (i.e. the cases without an overt marker) supports
the Referential hierarchy explanation (see Table 4 and example (25) repeated
here as (53)). However, the fact that also lower entities – abstract notions and
events – are more frequently denoted by the partitive rather than total cases,
example (54), contradicts it.
204 Helena Metslang

housemaster.NOM register-PST.3SG 1 SG -PAR in
‘The housemaster registered me.’ (referentially high O with overt marking)
(54) Ihka-si-n uus-i aeg-u.

yearn-PST-1SG new-PAR.PL time-PAR.PL
‘I desired that the times would change.’ Lit ‘I desired new times.’
(referentially low O with overt marking)
Table 4: Distribution of the Referential hierarchy properties among differently coded elements
(% of all Os and all e-NPs in the corpus)11
Argument Case SAP human concrete abstract event non-ref Total %

(abs. No)
O (130) Total 0 2 10 11 2 0 25
Partitive 6 5 8 26 10 0 55
Irrelevant 0 0 0 0 0 20 20
e-NP (127) Nom 0 5 17 31 3 0 56

Partitive 1 2 5 29 3 0 40
Irrelevant 0 0 0 0 0 4 4
As indicated above, the Referential hierarchy approach suggests that if there are
splits in the coding of the intransitive subject then semantically low subjects are
preferably overtly marked, whereas semantically high subjects are not. When
evaluating the Referential hierarchy effect on the e-NP, it is in principle possible
to analyze two alternative viewpoints. We can regard both nominative and partitive
e-NPs as low S-arguments or we can only regard partitive e-NPs as low S-arguments
and consider nominative e-NPs high S-arguments. The latter approach, i.e. looking
at the e-NP internal case-alternation is the general focus of analysis in the rest
of Section 5 and it would be consistent to follow this approach also here. The
problem with it is that even nominative e-NPs are clause-final and non-topical
and therefore not so different from the partitive e-NPs. As topicality is a notion
very closely tied to subjecthood, it is difficult to consider even nominative e-NPs
high intransitive subjects for the purposes of this analysis. The current dataset is
also too small for comparing the distribution of high and low referents among
total case and partitive e-NPs. In the present corpus it is hard to find any prefer-
11 Due to their low frequency, the arguments with non-human animate referents have not been
included in the table.
ences supporting the hierarchy (e.g. for higher e-NP referents to be marked by
the nominative and lower e-NP referents by the partitive), rather the contrary.
Only all the non-referential case-marked e-NPs in the corpus (9) have the parti-
tive case. In the following I will proceed with the first option.
I checked in the corpus whether the Referential hierarchy effects can explain
the case-marking distinction between A-like S marking (subject of the intran-
sitive unmarked clause that is always in the nominative) and O-like S marking
(e-NP with object-like case-alternation). In general the Referential hierarchy
applies: SAPs and humans are expressed by canonical intransitive subjects
(unmarked case: the nominative) and inanimate and abstract referents by e-NPs
(marked case: the nominative-partitive alternation).
To summarize, A and e-NP are the referentially more homogenous categories
whereas S and O are used for expressing a variety of meanings. The distribution
of split object marking only partly corresponds to the Referential hierarchy. The
statistical distribution of differential subject-marking (if we consider all e-NPs
non-canonical, O-like subjects) supports the prediction of the Referential hierarchy.
The intransitive subjects of unmarked clauses have high referents and receive A-
like marking, while e-NPs have low referents and receive O-like marking (often
baring overtly marked case and lacking verbal agreement and prototypically
having post-verbal position). The fact that the coding distinction between the
unmarked clause S and existential construction’s e-NP corresponds to the pre-
dictions of the Referential hierarchy is another argument for analyzing the
whole e-NP as a (non-canonical) representative of the intransitive subject cate-
gory (see also Section 8).
5.3 Lexical predicate as a case-assignment factor

As is typical in the languages of the world, lexical predicates and predicate
classes have a significant affect on Estonian DOM. A total of 32% of the case-
baring Os in the corpus have their case assigned by the predicate (see examples
in Section 4.1). In the data the predicate only triggers the partitive and not the
total case – this supports Vaiss’ (2004) finding that ‘partitive verbs’ are the
dominant verb class in Estonian. In the rest of the data the predicate permits
both object cases and at the same time, the object case depends on other factors.
Among case-marked e-NPs, 2% get their (nominative) case from the predicate
verb, for example:
206 Helena Metslang
(55) Ristmiku poole lähene-s suur kollane kõuts.

junction.GEN towards approach-PST.3SG big.NOM yellow.NOM tomcat.NOM
‘A big yellow tomcat was approaching the junction.’
5.4 Clausal properties influencing the argument case

Clausal properties influencing O’s and e-NP’s case involve the predicate’s mor-
phological form, clausal construction and aspect. This subsection discusses the
corpus occurrences of all these factors. In the treatment of aspect, the com-
binations of how clausal and an argument’s referential properties intertwine in
object case assignment will also be discussed.
Aspect is a prominent influence on the object case, but not on the e-NP’s
case. The interplay of the Inclusiveness hierarchy and aspect in the determina-
tion of the object case is illustrated in examples (15)–(19) in Section 4.1. In the
corpus there are 54% of objects whose case is determined by the co-influence
of these two factors. Out of these, two thirds of clauses have perfective aspect
and only one third have imperfective aspect. The following table details how
aspect and nominal inclusiveness co-determine the direct object case in the
corpus. As it can be seen, the most frequent result of all these combinations is
the total case object. The use of the partitive object is often determined by other
factors in the data.
Table 5: Aspect and quantity combinations determining the object case in the corpus (absolute
numbers)
Aspect and Perfective aspect Perfective aspect Imperfective Imperfective

quantity Inclusive quantity Non-inclusive aspect aspect Non-
quantity Inclusive quantity inclusive quantity
Examples 33 (total case) 4 (partitive) 11 (partitive) 8 (partitive)
Negation is a case-assignment factor that causes partitive use in the case of

both objects and e-NPs, (see example (26) in Section 4.1 and (37) in Section
4.2). In the corpus, negative clauses only constitute 10% of the sentences con-
taining case-marked Os and 20% of the sentences containing case-marked e-NPs.
The corpus does not contain other examples where the predicate verb’s form
determines the argument’s case. It was not possible to study the effect of the
impersonal (that requires total Os to be nominative) on the object case – the
impersonal was not included in the study because it is often difficult to differen-
tiate between the impersonal object and passive subject.
Rätsep (1978) developed a set of verb-governed sentence patterns that, among

others, contains clausal constructions that affect the case of e-NPs. The list of
constructions determining e-NP’s case was examined by Nemvalts (1996) and
reviewed by Metslang (2012). The influence of this factor on particular e-NPs’
marking is very rare, and one has to be cautious when distinguishing it from
the influence of lexical predicates. There was only one e-NP in the data whose
case was determined by a clausal construction:
(56) Lauda-s looma jaoks ruumi jätku-b.

stable-INE animal.GEN for space.PAR be.enough-3SG
‘There is enough space for the animal (cow) in the stable.’
(partitive object determined by the construction)
Although also the use of an ‘existential partitive’ noun affects the e-NP case here
(see Section 4.2), I regard the construction type the primary factor, as it would
also determine the e-NP’s case if it was not an existential partitive noun (cf.
example (57) and Metslang 2012).
(57) Lauda-s looma jaoks heina jätku-b.

stable-INE animal.GEN for hay.PAR be.enough-3SG
‘There is enough hay for the animal (cow) in the stable.’
(partitive object determined by the construction)
5.5 Interim conclusion: O’s and e-NP’s case-determining

factors
To conclude this section, I will present the comparison of O’s and e-NP’s case-
marking motivation on the basis of the studied corpus material. Figure 2 shows
the frequencies of the primary case-motivations of O and e-NP in the corpus
(only case-marked arguments are considered).
In the figure, the factors ‘inclusiveness unmarked’ and ‘Inclusiveness hierar-
chy’ both refer to the quantification-related hierarchies suggested in Section 5.1.
The plural or mass noun argument’s case depends on whether the NP referent’s
quantity is inclusive or non-inclusive (then the argument’s total or partitive
marking depends on the Inclusiveness hierarchy) or is irrelevant (then the
argument receives the total case from the Quantitative markedness hierarchy).
Employing these hierarchies is useful because it allows us to compare and gen-
eralize over the case-assignment motivations of O and e-NP. These hierarchies
208 Helena Metslang
Figure 2: Frequencies of O’s and e-NP’s case-assignment motivations in the corpus (absolute
numbers, n = 229)
differ from the typologically widely used Referential hierarchy in that they are
not (indirectly) based on the topical – non-topical opposition.
Among the case-marking criteria discussed in Section 5, the most frequent
case-choice factor of e-NP is the NP’s lexical properties. O’s most common case
factor is the Inclusiveness hierarchy. As the more frequent case of O is the
partitive, it is noteworthy that it is not the Inclusiveness hierarchy that most
commonly determines O’s partitive marking in the data but the lexical predi-
cates. In addition, in the case of O, marked quantity is always somehow com-
bined with aspect. In the figure, the Inclusiveness hierarchy works together
with aspect in the case of O. The factor ‘NP’s lexical properties’ involves the
noun types ‘Existential nominatives’ and ‘Existential partitives’ in the case of
e-NP and personal pronouns in the case of O. Among Os the partitive has a
much more varied set of functions than the total cases, but among e-NPs both
the nominative and the partitive bare several functions.
Table 6 compares the significance of each case-determining factor for O
and e-NP. It does not specify whether the factor determines the total or partitive
case (this was already discussed above). In the case of both O and e-NP, there is
usually an overlap of various case-marking motivations. The table only mea-
sures the factors that dominate over others, i.e. trigger case-marking. Only situa-
tional aspect and nominal quantity are treated as having equal influence on
the object’s case and are presented as two parallel overlapping factors. When
looking at this table, it must be kept in mind that the crucial precondition for
the e-NP case-alternation possibility is the semantico-pragmatic EC clause type
itself. I have not added ‘clause type’ in the table because it does not exactly
specify which case-form the NP takes (the factor ‘constructions’ in the table is a
narrower, a more specific category, cf. Sections 4.2 and 5.4).
The table shows that there are some case motivations in the corpus that O
and e-NP share and some features that they differ in. It is possible to draw the
following conclusions on the basis of Table 6:
– The prominent factors influencing the case of O are lexical predicates and
the combination of aspect and the Inclusiveness hierarchy. The main case-
factors of e-NP are lexically inherent inclusiveness of nouns and also
negation.
– The main case-assignment factors that are shared by the e-NP and O are
negation and the Inclusiveness hierarchy. However, the former is more fre-
quent in determining e-NP’s case, and the latter in determining O’s case.
– There are factors that can determine only one argument’s case but (at least
in most cases) not the other’s: the influence of lexical predicates and un-
marked inclusiveness (the Quantitative markedness hierarchy).
Table 6: Comparison of the dominating factors influencing object’s and e-NP’s case in the data
(n = 229)
Level No Decisive case factor O e-NP

% prominence % prominence
Referential 1. inclusiveness unmarked (Quantitative 0 absent 16 significant
properties markedness hierarchy)
2. Inclusiveness hierarchy 54 prominent 14 significant
3. inclusiveness of noun lexemes 0 absent 47 prominent
4. personal pronouns (SAPs) 4 insignificant 0 absent
Verb’s 5. lexical predicates 32 prominent 2 insignificant
properties
Clausal 6. aspect 54 prominent 0 absent
properties
7. negation 10 significant 20 prominent
8. constructions (e.g. Rätsep 1978) 0 insignificant 1 insignificant
To put it simply, an important overall property influencing both the object’s and
e-NP’s case is referential – the general quantitative inclusiveness – but it is tied
to different grammatical levels and feature combinations (inherent in the lexical
semantics of the noun or the verb or the situation’s/construction’s meaning; see
factors 1−3, 5 and 8 in Table 6). O’s total case is caused by NP referent’s inclusive
quantity that is situationally motivated (the same noun can occur expressing
inclusive or non-inclusive quantity in different situations, even when comple-
menting the same verb). Situationally motivated quantification is also a signifi-
cant case-determining factor of e-NPs but lexically motivated inclusiveness is
more prominent.
210 Helena Metslang
To summarize, the object and e-NP largely depend on the same case-marking
system that includes all the main argument realization factor levels (Dixon 1994;
Witzlack-Makarevich 2011): referential and clausal factors as well as lexical
predicates. However, their application has some significant differences between
these arguments. In Section 5.2 I discussed the possibility of regarding e-NP as a
referentially low intransitive subject (a non-canonical S argument). This idea is
supported by the predictions of the Referential hierarchy hypothesis that has
been suggested in typology. Section 5 has shown that e-NP’s (that could be
regarded as a non-canonical intransitive subject, low S) case-marking greatly
resembles that of O. This is confirmed by both the set of rules in Figure 2 and
Table 6 and it also goes well with the Referential hierarchy’s predictions (I will
get back to this issue in Section 8). The case-marking of O can be explained by
the Referential hierarchy partly. As the hierarchy correctly predicts, there is a
preference for referentially high Os (i.e. pronouns denoting SAPs) to get overt
marking (the partitive) in the contexts where lower O arguments get the total
case that lacks an overt morphological marker (Erelt et al. 1993: 53). However,
in the case of DOM, the hierarchy predicts that the low Os are more likely to
have zero-marking, but this is not true in the case of Estonian.
6 Semantics and content-related properties of

partitive arguments
In this section I will discuss the meaning properties of different argument types.
The properties under scrutiny are discourse importance, person, number and
the situation type the arguments occur in. The semantic and content-related
properties’ influence on the O’s and e-NP’s case-marking will also be described.
None of these features is strictly attached to specific arguments or case uses –
they only characterize them statistically.
6.1 Discourse importance

Language speakers express entities’ referential importance in the discourse via
grammatical means (Chafe 1994: 89−91). Discourse importance concerns the
number of times an entity is referred to in the text. In the following I analyze
entities as discourse prominent if the text is about them and they reoccur in a
protagonist role throughout the text. I regard a referent as salient, if it is referred
to at least four times (often there is at least one passage about them in the texts
studied here). They tend to be less important human participants and more
central inanimate objects. Trivial participants are people, objects and abstract
phenomena mentioned in the text only once or twice.
Referential importance is linked to activeness: discourse inactive entities
tend to have trivial importance (Chafe 1994: 89−91). Prominent participants are
tracked throughout the discourse, therefore they tend to have active status (see
Section 7.3). However, first they need to be introduced in the text as new entities.
Sometimes trivial entities are packaged as active when they are referred to more
than once. Also events can sometimes be evaluated on the scale of discourse
importance (for example, in the corpus text about family relationships, getting
married is a salient event).
In the present corpus, the majority of discourse prominent referents occur
in the unmarked clause subject (S/A) role. Salient referents are less common
in the texts. Trivial discourse referents occur predominantly as Os and e-NPs
(almost 69% of Os and e-NPs are trivial). In the latter Estonian is similar to
Finnish where the EC’s role is also to introduce new entities in the discourse –
entities that are later rarely mentioned again in the text (Helasvuo 2001: 99).
Table 7 presents the distribution of discourse importance among the arguments
in the corpus (the category is usually evaluated as ‘irrelevant’ in the case of
clausal arguments and when the semantics of the noun is bleaching – the NP
and verb are in the process of forming one semantic unit).
Table 7: Distribution of discourse importance in the corpus (%)
Feature value A S O e-NP Total (abs. No)

Prominent 75 52 13 7 190
Salient 3 13 8 14 50
Trivial 22 35 65 73 253
Irrelevant 0 0 14 6 27
Total (abs. No) 130 130 130 130 520
Among the objects and e-NPs, case-marking does not have a considerable corre-
lation with discourse importance (Figure 3).
Discourse prominent Os are often SAP pronouns or nouns denoting pro-
tagonists and are marked with the partitive (see (53) in Section 5.2).
The unmarked way of introducing new participants in the discourse is using
a nominative e-NP. Although the participants introduced by an e-NP are usually
of trivial importance, a significant proportion of e-NPs are also used to mention
more salient participants. An example of a discourse prominent partitive e-NP is
the following (cf. Huumo and Lindström, this volume, on the comparison of
such constructions in Estonian and Finnish):
212 Helena Metslang
(58) Kui min-d kodu-s ei ol-nud . . .

when 1 SG -PAR home-INE NEG be-PST.PTCP
‘When I was not at home. . .’ (e-NP)
Figure 3: Distribution of discourse salience with respect to case (%, n = 260)
An example of a nominative e-NP denoting a salient referent (that will be talked

about in the whole paragraph) is (55), repeated here as (59):
(59) Ristmiku poole lähene-s suur kollane kõuts.

Junction.GEN towards approach-PST.3SG big.NOM yellow.NOM tomcat.NOM
6.2 Person
In the data, SAPs are naturally expressed by the subjects of unmarked clauses,
especially by A-arguments (Figure 4). Again, O shows a slightly greater variation
in the use of person category than the e-NP: among Os there also occurs refer-
ence to SAPs and Os also have a higher proportion of unspecified person refer-
ence due to the high frequency of clausal objects.
Person reference does not have a clear influence on the arguments’ case-
marking (Table 8).
Figure 4: Distribution of the person category in the corpus (% n = 520)
Table 8: Distribution of the person category among case uses of O and e-NP (%)
Argument Case 1st 2nd 3rd Unspecified Total (abs. No)

O total case 0 0 100 0 33
partitive 9 1 90 0 71
irrelevant 0 0 0 100 26
e-NP total case 0 0 100 0 74
partitive 0 0 100 0 51
irrelevant 0 0 0 100 5
Total (abs. No) 112 12 362 34 520
6.3 Number
A and S are more uniform than the partitive-permitting arguments in their pref-
erence for singular NPs (Table 9). It is possible that O and e-NP fit better with
the non-singular semantics because they are often used for referring to various
trivial, inanimate entities or non-referential entities that are not singular actors
of highly transitive actions like A and S arguments tend to be.
Table 9: Distribution of the number category among the arguments (%)
Feature value A S O e-NP Total (abs. No)

singular 92 83 62 68 398
plural 6 15 16 28 84
unspecified 2 2 22 4 38
Total (abs. No) 130 130 130 130 520
214 Helena Metslang
Figure 5: Distribution of number and referent types among A, S, O and e-NP (%, n = 464)
Figure 6: Number-marking of total case and partitive Os and e-NPs (%, n = 229)
When looking at A, S, O and e-NP together, the singular constitutes about 77%
and plural about 16% of all arguments. The plural is most commonly used with
inanimate concrete entities (see Figure 5).
The more frequent case among both singular and plural Os is the partitive.
Among singular and plural e-NPs it is the nominative (Figure 6). Hence number
marking follows the frequency-wise unmarked case of both arguments.
Table 10 presents how the number category combines with the referential
properties among Os and e-NPs in the data.
Table 10: Distribution of referential properties among singular and plural arguments
(n = 258; % of all Os and % of all e-NPs)12
Argument Coding SAP person con-crete abstract event non- Total

referential (abs. No)
O singular 6 7 13 24 8 4 81
plural 1 4 10 1 20
e-NP singular 1 2 9 42 5 7 87
plural 5 12 9 1 36
irrelevant 4 5
The table gives the percentage of all referent types among Os and e-NPs. For
example, 6% of Os are singular SAPs. The cells with the value 0% have been
left empty.
In the data, reference to concrete and abstract entities is common in the
singular and plural of both arguments. The following examples show the use of
concrete Os and e-NPs in the singular and plural (I repeat examples (20) from
Section 4.1 and (31) from Section 4.2 as (60) and (63) here).
(60) Leid-si-n väljapääsu.

find-PST-1SG exit.GEN
‘I found an exit.’ (concrete singular O)
(61) Kraaving võtt-is prillid ja kissita-s

Kraaving.NOM take-PST.3SG glasses.NOM.PL and squint-PST.3SG
min-d lühinägelikult.
I-PAR short-sightedly
‘Kraaving took his glasses (off) and squinted at me short-sightedly.’
(concrete plural O)
(62) Silla taha kogune-s rämpsu.

bridge.GEN behind gather-PST.3SG litter.PAR
‘Litter was gathering behind the bridge.’ (concrete singular e-NP)
(63) Selle-l kase-l on juba lehe-d.

This-ADE birch-ADE be.3 already leaf-NOM.PL
‘This birch has leaves already.’ (adapted from Vilkuna 1992: 61.)
(concrete plural e-NP)
12 The category ‘non-human animates’ has been left out due to a very small number of
examples.
216 Helena Metslang
The following set exemplifies how abstract Os and e-NPs are used in the singu-
lar and plural (I repeat examples (54) from Section 5.2 and (36) from Section 4.2
as (65) and (66) here).
(64) Silvia sa-i kuul-da meie aja tavaloo.

Silvia.NOM get-PST.3SG hear-INF our.GEN time.GEN typical.story.GEN
‘Silvia was told a typical story of our time.’ Lit. ‘Silvia got to hear a
typical story of our time.’ (abstract singular O)
(65) Ihka-si-n uus-i aeg-u.

yearn-PST-1SG new-PAR.PL time-PAR.PL
‘I desired that the times would change.’ Lit ‘I desired new times.’
(abstract plural O)
(66) Ja korraga torka-s mu-lle pähe veider mõte.

and suddenly strike-PST.3SG 1 SG -ALL head.ILL strange.NOM thought.NOM
‘And suddenly I got this strange idea.’ Lit. ‘And suddenly stroke into my
head a strange idea.’ (abstract singular e-NP)
(67) Korrapealt hakka-vad vaimukuse-d ta suu-st lenda-ma.

immediately start-3PL retort-NOM.PL 3 SG .GEN mouth-ELA fly-INF
‘Immediately, retorts start flying from his mouth.’ (abstract plural e-NP)
Event reference occurs more with singular Os and e-NPs:
(68) Näg-i-n selle-s märguanne-t.

see-PST-1SG this-INE message-PAR
‘I saw a message in it.’ (singular O denoting an event)
(69) Tagatoa-st kost-is köhimis-t.

back.room-ELA sound-PST.3SG coughing-PAR
‘One could hear coughing in the back room.’ Lit. ‘There sounded
coughing in the back room.’ (singular e-NP denoting an event)
Non-SAP person reference tends to occur as singular O and as plural e-NP, see
example (7) in Section 2 and (120) in Section 7.4 respectively. To conclude, the
selection of referent types is extremely heterogeneous among both argument
types. The variation is larger in the singular than plural. No case-specific conclu-
sions can be made on the use of the singular and plural among referent types
because the number of representatives in each group is too small.
Figure 7: The relationship between the number category and O’s case motivations (%, n = 104)
Figure 8: The relationship between the number category and e-NP’s case motivations
(%, n = 125)
Case use motivations of the partitive-permitting argument types (O and

e-NP) as outlined in Sections 4 and 5, are also more varied among singular Os
and e-NPs than among the plural ones (see Figures 7 and 8). This is possibly
due to the higher frequency of singular arguments in the corpus. However, there
are also notable differences in the proportions of how different case motivations
between the singular and plural Os and e-NPs are distributed. Plural Os’ case-
marking is determined by the Inclusiveness hierarchy and lexical predicates.
The case of singular Os also depends in addition to these factors on negation
and noun lexemes.
Singular e-NPs’ case mainly depends on the inclusiveness of noun lexemes
and negation. The main factors determining the case-marking of plural e-NPs
are the Quantitative markedness hierarchy (i.e. the NPs are unmarked for inclu-
siveness and therefore in the nominative) and the Inclusiveness hierarchy.
218 Helena Metslang
6.4 Situation type

To find out whether partitive-permitting arguments and partitive NPs have any
preferences for situation types, I evaluated every sentence in the data against
the Vendlerian verb classes (based on Vendler 1967). The four basic verb classes
are determined as follows.
States depict static situations which are inherently temporally unbounded (atelic), and
both achievements and accomplishments express changes of state, which are inherently
temporally bounded (telic): achievements are instantaneous, while accomplishments are
not. Activities are dynamic, inherently temporally unbounded (atelic), states of affairs
(Van Valin 2005: 32).
On the basis of Van Valin (2005), I have also included two additional event types
in the analysis: semelfactives that are punctual events which have no result
state and active accomplishments that are the telic (aspectually bounded, per-
fective) use of activity verbs. The difference between accomplishments and
active accomplishments is that active accomplishments can be modified by
adverbs like violently, vigorously, actively, strongly and energetically but accom-
plishments cannot (Van Valin 2005: 33). Instead of only analyzing the lexical
meaning of the verb, I have looked at the whole situation depicted in the clause.
The aspectual interpretation of the clause is the composition of verb semantics
on the one hand and contextual material on the other (cf. discussion on this
topic in Dowty 1979: 61ff. and in Huumo 2010). This is particularly the case in
Estonian – a language that lacks overt aspect marking on the verb and where
object case and various perfectivizing or durative adverbials and particles are
used for marking the aspect of the situation (Erelt et al. 1993: 25; Vaiss 2004:
12–14).
In the following I will measure the arguments’ occurrence in the situation
types across the corpus. The examples are meant to clarify how the sentences
have been categorized. Figure 9 shows the distribution of the situation types
among the unmarked intransitive and transitive clauses and the ECs in the
corpus. When looking at the figure and the rest of the data of this subsection,
caution must be applied as these results are characteristic of the particular corpus
texts studied. Other texts, genres and modes of communication are likely to give
different results.
In the corpus, intransitive clauses and ECs depict all situation types and
also transitive clauses occur with all situation types but accomplishments. Most
transitive and intransitive clauses express achievements, activities and states in
the corpus, ECs tend to denote states and activities where the existential mean-
ing component is foregrounded (see Table 11 and examples at the end of this
subsection).
Figure 9: Distribution of situation types among clause types (% of all transitive and intransitive
clauses and ECs, n = 390)
Table 11: Distribution of situation types in the corpus (%, n = 390)
Situation type Transitive Intransitive EC Example

semelfactive 73 23 4 Only Renke said “I am sorry”. (1)
achievement 48 37 15 I pulled myself together. (85)
active 48 43 9 The housemaster registered Peeter. (24)
accomplishment
activity 40 35 25 I was thinking my own thoughts. (84)
state 15 28 57 These have interested me before. (71)
accomplishment 0 53 47 (My) fitness stabilized. (70)
(70) Mänguvorm stabiliseeru-s.

fitness.NOM stabilize-PST.3SG
‘(My) fitness stabilized.’ (intransitive clause expressing an accomplishment)
Total case Os primarily occur in achievement situations (55% of all total case
Os) and partitive case Os in activities (46% of all partitive Os). Nominative and
partitive e-NPs mostly appear in states (50% and 76% respectively). Nominative
e-NPs – the larger and more versatile e-NP group – often also occur in activities
(30%), achievements (12%) and accomplishments (8%).
One possible explanation for this distribution among both Os and e-NPs is
in Hopper and Thompson’s (1980) transitivity parameters. In their set of transi-
220 Helena Metslang
Figure 10: Combinations of the most common referent and situation types (%, n = 517)13
tivity parameters, among other properties, the following parameters are corre-
lated: aspect (one of the semantic bases of situation types), kinesis (action vs.
non-action) and affectedness (marked by the total case – partitive alternation in
Estonian). Also in the current corpus, in the case of O total affectedness (marked
with a total case) correlates with perfective aspect (achievement situations). These
situations can be regarded as highly transitive. In the context of lower transitivity,
partial affectedness (non-inclusive quantity, marked with the partitive case)
correlates with imperfective aspect (activities). It is also possible to draw some
parallels with ECs, although ECs differ in many respects from the transitive
clause and from the A and O arguments discussed by Hopper and Thompson.
e-NPs that are only partially affected by or involved in the situation (non-inclusive
quantity, marked by the partitive) occur more often in states (the non-action
situation type). In ECs the situation type is also correlated with the polarity
parameter touched upon by Hopper and Thompson. Affirmative clauses are higher
on the transitivity scale than negative clauses (ibid.). In the data, 84% of negative
clause e-NPs and only 48% of affirmative clause e-NPs occur in states.
The most common referent types (SAPs and other humans, concrete and
abstract entities) occur most frequently in the most common event types –
states, activities and achievements (see Figure 10).
States as low transitivity situations are strongly dominated by abstract refer-
ents. Possibly due to the content of the corpus texts (frequent discussions about
emotional, mental and cognitive states), reference to SAPs occurs in states more
often than reference to non-SAP humans. In the rest of this section I will concen-
13 To keep the figure simpler, the 3 occurrences of non-human animates have not been
indicated.
trate on states, activities and achievements that either contain O or e-NP and
compare the use of these arguments with the unmarked clause subjects. About
74% of the core argument referents in these situation types are SAPs, humans
and abstract entities. Other referent types are distributed more diversely through-
out the corpus.
States. The most common referent types in states are abstract entities (49%
of 165 examples) and SAPs (16%). More typical uses of the partitive SAPs involve
cognitive and emotional states where the experiencer is mapped to the object
relation (71). SAPs also occur in negative existential constructions, see (58) in
Section 6.1, repeated here as (72).
(71) Nee-d ol-i-d min-d varem=gi huvita-nud.

these-NOM.PL be-PST-3PL 1 SG -PAR earlier=CL interest-PST.PTCP
‘These had interested me before.’ (O, experiential construction)
(72) Kui min-d kodu-s ei ol-nud . . .

when 1 SG -PAR home-INE NEG be-PST.PTCP
‘When I was not at home. . .’ (e-NP)
Abstract O, e-NP and A are commonly present as stimuli of mental, cogni-

tive states. O and e-NP can occur both in the total and the partitive case, see
example (65) in Section 6.3 and (73)−(74).
(73) Su-l on õigus.

2SG-ADE be.3 right.NOM
‘You are right.’ (e-NP)
(74) See teg-i ta-lle mure-t.

this.NOM do-PST.3SG 3 SG -ALL worry-PAR
‘This worried him.’ (A)
Abstract entities typically appear in equative and attributional constructions

(in O and S relations). See example (75) and also example (52) in Section 5.2,
repeated here as (76).
(75) Nähtavasti kujuta-n enda-st teis-t äärmus-t.

seemingly represent-1SG self-ELA other-PAR extreme-PAR
‘Seemingly I represent another extreme. Lit. ‘Seemingly I represent from
myself another extreme.’ (O)
222 Helena Metslang

‘The surprise effect in the team was big.’ (S)
Abstract e-NP referents are often found in both locative and possessive construc-
tions ((77) and (78) respectively), S and e-NP are also frequent in various other
abstract relations (see (79) for an S).
(77) . . . ilma et ta näo-s kübe-t=ki kadedus-t ole-ks.

without that 3 SG .GEN face-INE particle-PAR=CL envy-PAR be-COND
‘. . .without him having a hint of envy in his face.’ (e-NP)
(78) Aga see viga on mu-l juba taadi-lt.

but this.NOM flaw.NOM be.3 1 SG -ADE already old.man-ABL
‘But I already have this flaw from my old man.’ (e-NP)14
(79) See ei tähenda.

this.NOM NEG matter
‘This does not matter.’ (S)
e-NPs can also appear in states conceptualized as (pseudo) events:
(80) Ärevus-t pid-i kusagilt mujalt kiirga-ma.

anxiety-PAR must-PST.3SG from.somewhere from.elsewhere radiate-INF
‘Anxiety had to be coming (radiating) from elsewhere.’ (e-NP, metaphoric
use of a predicate with emanation semantics)
14 I consider this sentence as a marginal EC because the referent of the e-NP is one of the two
referents introduced in the narrative by this sentence. Although the actions/situations that
reflect this character trait that the NP viga (flaw) denotes were discussed in the preceding
context, they are referred to as events (an activity and a state: a row and not knowing some-
thing) there and not as to properties. The speaker’s judgment that these activities reflect this
particular character trait, is new in the discourse. Cf. the extract from a novel depicting 19th
century Estonian village life and the relations between wealthy estate-owner overlords and
village people: The guest told them the story about his family – the old man and his sisters.
The family rented one house after another in different manor estates but did not stay anywhere
for long. He said, every time ended with a row between them and the estate owner. − “Probably
you just didn’t know how to live your life then, “Leena said. − “Probably not. We probably
didn’t know how to bow down deep enough before the lords although this was expected of us.
But I already have this flaw from my old man and probably it cannot be cured anymore.”
(Translation − HM)
(81) Tema pilgu-st voola-s armastus-t.

3 SG .GEN gaze-ELA flow-PST.3SG love-PAR
‘There was love flowing from his gaze.’ (e-NP)
Activities. The most frequent NP referents in activities are non-SAP humans

(28% of 145 activity clauses in the corpus, see for instance A in (82) and O in
example (7) in Section 2) and SAPs (27%, see e.g. A in example (7) in Section 2
and S in (83)). 20% of activities involve abstract referents, see example (82).
(82) Relli teosta-s oma poliitika-t vankumatult.

Relli.NOM implement-PST.3SG own.GEN policy.PAR steadily
‘Relli implemented his policy steadily.’ (activity with a non-SAP human A
and an abstract O)
(83) Sina treeni-d, kui mõni teine ei treeni.

2 SG .NOM practice-2SG when some.NOM other.NOM NEG practice
‘You practice when some others do not.’ (activity with a SAP S)
Abstract referents often occur in cognitive and emotional activities – in O and

e-NP positions:
(84) Mõtle-si-n om-i mõtte-i-d.

think-PST-1SG own-PAR.PL thought-PL-PAR
‘I was thinking my own thoughts.’ (O)
Achievements. Referents high in the Referential hierarchy – SAPs and other

humans – often occur in events high in transitivity: they constitute 31% and 23%
of all achievements respectively.
(85) Võt-si-n ennas-t kokku.

take-PST-1SG oneself-PAR together
‘I pulled myself together.’ (SAP in the O position)
(86) Treener kisku-s valu-st kõvera-sse.

coach.NOM double.up-PST.3SG pain-ELA doubled.up.position-ILL
‘The coach doubled up in pain.’ (a human referent in the S position)
Abstract entities constitute 24% of achievements. The abstract entities that

are mapped in S, O and e-NP positions in achievements tend to occur in cogni-
tive and emotional processes.
224 Helena Metslang
(87) Tema jutu-st läk-s mu-l enamik kaotsi.

3 SG .GEN talk-ELA go-PST.3SG 1 SG -ADE majority.NOM missing
‘The majority of his talk was lost on me.’ (S)
(88) Siis tekki-s ähmane lootuskiir.

then occur-PST.3SG hazy.NOM ray.of.hope.NOM
‘Then a hazy ray of hope appeared.’ (e-NP)
(89) See ol-i avalda-nud ta-lle mulje-t.

this.NOM be-PST.3SG make-PST.PTCP 3 SG -ALL impression-PAR
‘This had impressed him.’ (O)
There are no examples of the use of partitive abstract e-NPs in achievements. In

sentences with infinitival constructions it is often the lighter finite verb that
gives the situation the achievement meaning, see examples (64) and (67) in
Section 6.3.
7 Message packaging of core arguments

The message packaging phenomena include the following methods of formulat-
ing the meaning: the given-new opposition, focus of contrast, definite-indefinite
opposition, subject and topic. These categories are used for representing the
individual whose point of view the speaker is taking while speaking or with
whom the speaker empathizes (Chafe 1976: 28), see also Section 1. I have selected
for discussion the distribution of four pragmatics-related properties: phrase weight
and word order (as means of expression that strongly correlate with topicality;
cf. Lindström 2005 on Estonian), discourse activeness (a notion preferred to given-
ness, also called discourse activation) and definiteness. I will also compare the
message packaging features of the partitive arguments to the arguments in the
total case.
7.1 Phrase weight

Syntactic weight means the complexity of clausal elements and it shows in their
length; it is one of the factors co-influencing Estonian word order (Lindström
2005: 23−24). This subsection points out general tendencies how phrase weight
is distributed among different arguments, cases and referent types in the corpus.
In the analysis, the arguments are ordered as follows:
(90) ZERO -ANAPHORA > PRONOUN > BARE NOUN > FULL NP > HEAVY PHRASE
Figure 11: Distribution of phrase weight among argument types (% of all A, S, O and e-NP
arguments, n = 520)
Zero anaphora is the wide-spread tendency across languages to leave out topical
arguments (Bickel 2010: 420). In the following, zero-anaphora is regarded as
the lightest possible way of coding an argument whereas heavy phrase is the
heaviest and most complex option. Phrases consisting of more than two words
are regarded as heavy here. They include NPs, VPs and clauses (mainly direct
speech, reported speech or relative clauses). Medium weight elements of the
clause – phrases consisting of one or two words where usually the (head) noun
is preceded by an adjective, determiner or a quantifier – are analyzed as full NPs.
Most often the arguments in the corpus have medium weight coding (bare
nouns and full NPs, 46% of 520 arguments), 33% of the arguments have prono-
minal coding (zero anaphora, pronoun) and 21% heavy coding. Among A and S
arguments 84% and 80% are respectively marked by zero anaphora, pronouns
or bare nouns. It is less common among O (45%) and e-NPs arguments (40%),
see Figure 11.
The typical phrase weights of O and e-NP – bare nouns and full NPs – are
expectedly more prone to partitive use in the corpus than NPs of other weight
(Table 12).
Table 12: Case-marking of different weight A, S, O and e-NP (%)
Phrase weight Total case Partitive Irrelevant

pronoun 75 25 0
bare noun 65 35 0
full NP 65 35 0
heavy phrase 53 18 29
Total 64 29 7
226 Helena Metslang
Figure 12: Distribution of phrase weight between differently marked arguments (%, n = 260)
Among both Os and e-NPs heavier arguments occur more often as non-nominal
arguments or in the argument’s frequency-wise unmarked case (Figure 12).
Only A and S are coded by zero anaphora in the corpus, see example (103)
below. I did not find any clear examples of zero marking among objects and
e-NPs in the data.
Pronouns always occur in the nominative in the unmarked clause subject
position (91) and if in the object position, they are mostly in the partitive (17
out of 19 occurrences, see (92) and (85) in Section 6.4). Hence the personal pro-
nouns demonstrate clear-cut accusative alignment in the corpus (see Helasvuo
2001: 40−64 on Finnish). Among e-NPs the there was only one pronominal use
in the corpus, see example (72) in Section 6.4.
(91) Mina aga imesta-si-n, mis. . .

1 SG .NOM but wonder-PST-1SG what
‘But I was wondering, what. . .’ (pronoun as S)

commandant.NOM write-PST.3SG 1 SG -PAR in
‘The commandant registered me.’ (pronoun as O)
Bare nouns are especially common among e-NPs and also among Ss and
Os. They occur slightly more often in the partitive than in the total case both as
an O and an e-NP (61% of all bare noun Os and 56% of all bare noun e-NPs). See
(93) for a total case bare noun O, (89) in Section 6.4 for a partitive bare noun O
and (73) and (80) from Section 6.4 for a for a nominative and partitive bare noun
e-NP, repeated here as (94) and (95).
(93) Ta murd-is jalaluu.

3 SG .NOM break-PST.3SG leg.bone.GEN
‘S/he broke his leg bone.’ (bare noun O in the total case)
(94) Su-l on õigus.

2SG-ADE be.3 right.NOM
‘You are right.’ (bare noun e-NP in the total case)
(95) Ärevus-t pid-i kusagilt mujalt kiirga-ma.

anxiety-PAR must-PST.3SG from.somewhere from.elsewhere radiate-INF
‘Anxiety had to be coming (radiating) from somewhere else.’
(bare noun e-NP in the partitive)
Why the case distribution of bare noun e-NPs does not reflect e-NP’s general
preference for the nominative has possibly to do with word order and polarity
issues. The number of clause-final nominative and partitive bare noun e-NPs is
similar in the corpus (20 and 19 occurrences respectively). However, the pre-
verbal position (i.e. the non-neutral position for e-NP), as in example (95), is
favored by the partitive bare nouns (10 partitive vs. 2 nominative occurrences).
The higher frequency of partitive bare noun e-NPs is mainly caused by the
focusing strategy of the predicate verb in negative ECs:
(96) Antti-t pol-nud näh-a.

Antti-PAR NEG.be-PST.PTCP see-INF
‘(I) couldn’t see Antti.’ Lit. ‘Antti was not visible.’ (preverbal bare noun
e-NP in a negative clause)
There is also a small number of different frequent EC-like constructions with

frozen bare-noun e-NPs:
(97) Tegemis-t on huvitava isiksuse-ga.

dealing-PAR be.3 interesting.GEN personality-COM
‘Here we are dealing with an interesting personality
(He is an interesting personality).’
Full NPs are less common among A and S and more common among O and
e-NP. Above all, full NPs occur as total case Os; the distribution of full NPs
228 Helena Metslang
among the nominative and partitive e-NPs is equal in the corpus, see the follow-
ing examples.
(98) Vastu kevade-t otsi-s Kraaving mu-lle

before spring-PAR search-PST.3SG Kraaving.NOM 1 SG -ALL
omaette toa.
private room.GEN
‘Before spring, Kraaving found me a private room.’ (full NP, total case O)
(99) Vaene papa keera-b haua-s teis-t külge.

poor.NOM grandfather.NOM turn-3SG grave-INE other-PAR side.PAR
‘Poor grandfather is turning (the other side) in his grave.’
(full NP, partitive O)
(100) Mingi-d võime-d ta-l igatahes ol-i-d.

some-NOM.PL skill-NOM.PL 3 SG -ADE anyway be-PST-3PL
‘He definitely did have some (supernatural) skills.’
(full NP, total case e-NP)
(101) . . .ilma et ta näo-s kübe-t=ki kadedus-t ole-ks.

without that 3 SG .GEN face-INE particle-PAR=CL envy-PAR be-COND
‘. . . without him having a hint of envy in his face.’ (full NP, partitive e-NP)
Heavy phrases have a strong tendency to occur as O and e-NP (see (106)
in Section 7.2 and (103) for a total case and a partitive O and (104) for a total
case e-NP and (105) for a partitive e-NP). Heavy phrases are more commonly
nominative e-NPs and non-nominal Os that have no case-marking.

Silvia.NOM get-PST.3SG hear-INF 1 PL .GEN time.GEN typical.story.GEN
typical story of our time.’ (total case heavy O)
(103) Tund-si-n vaid üha kasva-va-t pinge-t,

feel-PST-1SG just progressively grow-PTCP-PAR tension-PAR
ärevus-t, hirmu=gi.
anxiety-PAR fear.PAR=CL
‘I felt progressively growing tension, anxiety and even fear.’
(partitive heavy O)
(104) Ristmiku poole lähene-s suur kollane

junction.GEN towards approach-PST.3SG big.NOM yellow.NOM
kõuts.
tomcat.NOM
(total case heavy e-NP)
(105) Mu-l pol-nud erilis-t usku, et

1 SG -ADE NEG.be-PST.PTCP particular-PAR belief.PAR that
see ilmu-b.
it.NOM come.out-3SG
‘I did not have particular belief in it coming out (getting published).’
(partitive heavy e-NP)
Unlike in the case of other phrase weights, the proportion of nominative heavy
phrases among e-NPs (as opposed to partitive heavy phrase e-NPs) is higher than
the general proportion of nominative uses among e-NPs. This might be due to
the partitive’s general function of marking non-inclusive quantity. If an EC is
used for presenting a complex referent in the discourse (in the form a long
multiword phrase), adding more complexity to the clause by the use of gram-
matical quantity reference (non-inclusive quantity) may be undesirable. Also,
most of the nouns belonging to the group ‘existential partitives’ (see Section
4.2) occur as bare nouns in the data.
The lower the referent is in the Referential hierarchy the more likely it is to
be marked by a heavy phrase in the corpus (Table 13; to keep the table simpler,
the few non-human animate examples have been left out).
Table 13: Distribution of Referential hierarchy properties between phrases of different weight
(%, n = 517)
Weight/Semantics SAP person concrete abstract event non-referential

zero-anaphora 70 10 0 0 0 0
pronoun 29 23 0 6 22 0
bare noun 0 38 62 35 25 29
full NP 1 15 25 32 28 0
heavy phrase 0 14 13 27 25 71
230 Helena Metslang
7.2 Word order

The most frequent word orders in written Estonian are SVX and XVS (25% and
24% of clauses respectively; Tael 1988: 6).15 One of the most important factors
influencing Estonian word order is information structure, i.e. givenness of clause
constituents and also the degree of focus (Lindström 2005: 185). This is espe-
cially the case in the second half of the clause whereas in the beginning of the
clause syntactic factors also play an important role (Tael 1990: 37). Namely,
Estonian word order is strongly influenced by the V2-rule (i.e. the verb comes
second rule; Huumo 2002: 502; Tael 1988: 40) which determines the subject
position and overrides the tendencies based on the inherent properties of the
subject that may suggest some other position in the clause; it thus sets a frame
and inside it pragmatic tendencies can work (Huumo 2002: 502). Omission of the
subject-verb inversion causes ungrammatical or strongly emphatic word order
(Huumo 1993: 152). Huumo also summarizes the two textual bases of subject-
verb inversion (XVS word order) caused by the V2-rule:
– inversion is caused by the subject’s properties: untypical, heavy subjects are
postponed;
– some verbal complement or adjunct is topicalized in the beginning of the
clause, the verb remains in the second position and the subject’s position
is post-verbal (inversion caused by topicalization). The topicalized entity could
be an adverbial, object or predicate nominal.
The neutral word order in ECs is XVS. Direct order (SV) is usually non-neutral in
ECs, being caused by emphatic or contrastive stress or by text linking. In rare
cases, direct order is the neutral order of ECs, see example (96) in Section 7.1.
This subsection compares the arguments under discussion according to
their word order preferences. It will:
– look at the distribution of word order types among A, S, O and e-NP;
– look at the distribution of word order types among total case and partitive
Os and e-NPs;
– discuss the possible correlations that word order might show with other
argument properties.
On the basis of Huumo (1993) the following categories were formulated for ana-
lyzing the arguments’ position.
– Neutral direct order – SV order dominating in the unmarked clause type (see
Section 3). The clause is neutral and no topicalization or focusing is taking
place, see examples (93) in Section 7.1 for an unmarked clause and (96) for
an EC.
15 In Tael’s treatment, S (subject) involves A, S and e-NP.

– Neutral inversion – VS order occurring in typical ECs. The clause is neutral

and no topicalization or focusing is taking place, see (88) in Section 6.4.
– Inversion (topicalization) – VS order occurring in the unmarked clause type
in the corpus (see examples below). The subject is located after the verb
due to the transition of an adverbial, object or predicative from its neutral
position to the sentence-initial position. The clause can be neutral or
emphatic.
– Inversion (subject postponing) – VS order occurring in the unmarked clause
type. The subject occurs after the verb because it is heavy or unprototypical.
The clause can be neutral or emphatic.
– Direct order (topicalization) – SV order occurring in ECs and unmarked
clauses. The subject is before the verb due to the transition of an e-NP or
unmarked clause subject from its neutral position to the sentence-initial
position. The clause can be neutral or emphatic.
– Direct order (focusing) – SV order occurring in ECs. The verb (phrase) or
other element has been given prominence by moving it to the clause-final
position. The clause can be neutral or emphatic.
In the corpus, in about 80% of the cases the sentences follow the neutral word
order where the clause has no topicalization or focusing. Information structural
alternations in the subject-predicate order are relatively rare. Table 14 presents
the comparison of word order choices in the three clause types in the corpus.
Table 14: Distribution of subject-verb order in intransitive, transitive and existential clauses (%)16
Feature value Intransitive Transitive Existential Total (abs. No)

neutral direct order 84 80 5 219
inversion (topicalization) 15 18 0 41
inversion (subject postponing) 1 1 0 4
neutral inversion 0 0 83 108
direct order (topicalization) 0 1 8 12
direct order (focusing) 0 0 4 6
Total (abs. No) 130 130 130
About 80% of Os and e-NPs occur in sentences with different word order con-
figurations. In the case of O this is direct and in the case of e-NP, inverse order.
The neutral word order of the clause type dominates with both total and parti-
tive case O and e-NP. The size of the corpus does not allow for more detailed
16 The clauses with zero-anaphora are marked as having neutral word order.
232 Helena Metslang
analysis of the other word order types. In the following, I will present examples
of the topicalization and focusing in the data.
When time adverbials are topicalized, V2-rule causes inversion in transi-
tive and intransitive clauses, see example (98) in Section 7.1.
Heavy A and S arguments cause inversion due to subject postponing.
(106) Sõna-d luge-s Ø peale keskealine

word-NOM.PL read-PST.3SG (we.ALL) on middle-aged.NOM
tüse naisterahvas.
voluptuous.NOM woman.NOM
‘A middle-aged voluptuous woman did the reading (in the wedding
ceremony). Lit. ‘A middle-aged voluptuous woman read the words on
(us).’ (transitive clause, subject postponing)
When an e-NP is topicalized, non-neutral direct order is common. In the

corpus, the topicalized e-NPs usually have abstract referents, see example (100)
in Section 7.1. Also in negative ECs the word order is often direct: e-NP – verb –
locative phrase. These sentences support Vilkuna’s (1989: 164) claims (on Finnish
but also applicable to Estonian) that although the predicates of ECs usually do
not bare new information of the clause they sometimes do bare polarity-new
information. I analyze these examples as direct word order with focusing,
see example (72) in Section 6.4.
In the data, almost all referent and situation types and even phrase weight
types occur with the most frequent word order types. The less frequent word
order types are too rare in the current corpus to show their possible correlations
with phrase weight or other properties.
7.3 Discourse activeness

One of the features chosen for the pragmatic analysis of the argument types
under scrutiny is discourse activeness which relates to the notions of ‘given/
old’ and ‘new’ information. According to Lambrecht (1994), activeness and iden-
tifiability are pragmatic categories that can be applied to (nominal) referents,
unlike for example topic that is a relation of aboutness between a proposition
and a discourse entity. Topicality is often not uniformly retraceable to single
lexical items and therefore harder to take into account in this study.
In Lambrecht’s treatment (1994: 76–115) activeness concerns discourse refer-
ents that the speaker assumes being identifiable (stored in the addressee’s
mind). The speaker assesses these referents being in one of the three activation
states at the time of the speech act: active, inactive or accessible (semi-active).
Active referents are the ones that are in a person’s focus of consciousness at a
given moment in time. He finds that the clearest evidence for assumed active-
ness is pronominal coding (pronouns, inflectional or zero coding). Inactive refer-
ents are usually coded by phonologically accented lexical phrases. Accessibility
means potential for easy activation due to text-internal or text-external factors
(situational or inferential factors). When referents are presented as accessible,
the sentence’s presupposition structure conveys a request to the addressee to
draw certain inferences, which are necessary to arrive at the correct interpreta-
tion of the referent. Phrases with accessible referents do not have direct phono-
logical or morphological coding devices but they may be indirectly expressed by
syntactic means (ibid.). Cross-linguistically, some coding devices, for example
zero anaphora, signal maximal referential continuity ((topical) rementions in
the discourse). Others, like stressed pronouns, full lexical nouns, signal referential
discontinuity (Givón 2001: 463). Some grammatical relations are better suited than
others for expressing certain activation states, for example for introducing new
referents into discourse (e.g. Helasvuo 2001: 90−93).
Evaluating the corpus from the viewpoint of discourse activeness is a diffi-
cult and inevitably somewhat subjective task as sets of contextual word (and
phrase) meanings are far from uniform. Some of the principles that I used for
making decisions are the following. The analysis of activeness only focuses on
referential arguments. The activation status of propositions, events and nominal
parts of some more or less opaque multi-word verbs has not been analyzed (the
value ‘irrelevant’, see also Section 5.2). For example:
(107) Mõni nädal hiljem sa-i komandant

a.few.NOM week.NOM later get-PST.3SG commandant.NOM
haisu ninna.
odor.GEN nose.ILL
‘A few weeks later the commandant found out about it.’ Lit. ‘A few
weeks later the commandant got the odor (of the secret) in his nose.’
(the activeness status of O is irrelevant)
How long an entity stays activated in the discourse depends on several factors,
for example on how many sentences there have been in the text after the last
mention of the entity (Chafe 1994: 33). I consider entities that were last men-
tioned no more than five sentences earlier ‘active’.17 I analyze a referent (inferen-
tially) accessible when it is part of the ‘schema’ prompted by an active referent
17 See also (Huumo 2002) on this kind of analysis on Finnish and Estonian written language.
234 Helena Metslang
(c.f. e.g. Lambrecht 1994: 99). For example in a text where a steam engine has
been mentioned, the referent steam is accessible. In a text where a participant
(person) has been mentioned, his clothes and body parts are accessible, see
example (99) in Section 7.1. I do not regard mental states as accessible –
although they can also be seen as a part of a human being, they are too numer-
ous to be readily accessible when a human referent is activated in the dis-
course), see (76) in Section 6.4.
I treat a subgroup of referents, the generics, together with other unspecific
entities under ‘reference to a category’ classification (see the next subsection).
Generics can be both new and old information (Erteschik-Shir 2007: 9), see
examples (108) and (109) for inactive generic or other ‘category reference’.
(108) Alati on süüdi naine.

always be.3 guilty woman.NOM
‘The woman is always guilty.’ (generic S, inactive in the discourse,
category reference)
(109) Jutu vahele mäng-is neli rauka viiuli-t.

talking.GEN between play-PST.3SG four.NOM elderly.person.PAR violin-PAR
‘Between the speeches four elderly people played violin.’ (unspecific O,
inactive in the discourse, category reference)
In the data, the subject position is rather used for referring to active entities in
the discourse (see Table 15). O is used for referring to a variety of referents; how-
ever, most Os have referents that are inactive in the discourse. The primary func-
tion of the EC is presenting new referents in the discourse and therefore other
activeness states are quite rare in this position. Referring to accessible referents
is the least common strategy and this kind of entities can occur in several roles.
Table 15: Referents’ state of activeness (%, n = 520).
Feature value A S O e-NP

active 80 67 16 4
accessible 6 10 14 8
inactive 12 21 45 77
Total (abs. No) 130 130 130 130
For example, active arguments were used in the S and O positions in the data as
follows (I repeat example (83) of Section 6.4 as (110) here).

‘You practice when some others do not.’ (active S)
(111) “Ma saan kõigest aru,” ütles treener pika vaikimise järel. (. . .)
Kraaving teg-i uuesti mõtlemispausi.
Kraaving.NOM make-PST.3SG again thinking.pause.GEN
‘“I fully understand,” the coach said after a long silence. (. . .) Kraaving
paused again for thinking.’ Lit. ‘. . . Kraaving made again a thinking
pause.’ (active O)
In the earlier context of (111) (four sentences earlier), the character was already
said to have made a pause in the discussion. The transitive clause under scrutiny
refers to him making a thinking pause again. The type reference to a pause as
such is still active although the particular referent (the second pause) is new.
Active e-NPs occurred in negation or doubt contexts in the corpus (I will repeat
example (96) of Section 7.1 here).
(112) Antti-t pol-nud näh-a.

Antti-PAR NEG.be-PST.PTCP see-INF
‘(I) couldn’t see Antti.’ Lit. ‘Antti was not visible.’ (active e-NP)
Most of the accessible entities are in the S and O role and they are inanimate.
Sometimes A and e-NP are also packaged this way, see examples (113)−(114), as
well as (99) from Section 7.1 and (87) from Section 6.4, repeated as (115) and (116)
here.
(113) Tema hääl väljenda-s siiras-t imestus-t.

3 SG .GEN voice.NOM express-PST.3SG genuine-PAR surprise-PAR
‘His voice was expressing genuine surprise.’ (inferentially accessible A)
(114) Seintest ja põrandast imbusid minuni avaldused teistest eludest.

Viienda-lt korruse-lt kost-is kumeda-i-d helikatke-i-d.
fifth-ELA floor-ELA sound-PST.3SG dull-PL-PAR sound.fragment-PL-PAR
‘There were expressions of other lives filtering to me through the walls
and the floor. From the fifth floor there could be heard some dull sound
fragments.’ (textually accessible e-NP)
236 Helena Metslang
Figure 13: Discourse activeness status of different referential properties (%, n = 464)
(115) Vaene papa keera-b haua-s teis-t külge.

poor.NOM grandfather.NOM turn-3SG grave-INE other-PAR side.PAR
‘Poor grandfather is turning (the other side) in his grave.’ (inferentially
accessible O)
(116) Tema jutu-st läk-s mu-l enamik kaotsi.

3 SG .GEN talk-ELA go-PST.3SG 1 SG -ADE majority.NOM missing
‘The majority of his talk was lost on me.’ (inferentially accessible S)
From the point of view of reference to inactive entities in the discourse, O and
e-NP radically differ from A and S, as it is their most common activeness status.
See examples (67) from Section 6.3 and (98) from Section 7.1 for an inactive e-NP
and O.
Among all the arguments in the corpus, the higher the referent is in the
Referential hierarchy the more likely it is discourse active. Most of human refer-
ents in the corpus are active. Concrete and abstract entities are usually inactive.
There is also a higher proportion of inactive than active events (Figure 13; com-
pare with Figure 1 in Section 5.2 for the allocation of reference types among
argument roles).
The differences between the use of the total and the partitive case especially
concern the activeness status of events and non-SAP humans. Event reference
tends to get inactive status far more often in the discourse when the NP is in
the partitive: with event reference 71% of partitive and 39% of total case NPs
are inactive. Person reference gets active status in the discourse more commonly
Figure 14: Distribution of activeness statuses throughout differently case-marked arguments

(%, n = 260)
when the NP is in a total case (with person reference, 71% of total case and only
50% of partitive NPs are active).
With both O and e-NP, the case distribution of the most frequent activeness
state (inactive) largely reflects the frequency-wise unmarked case of the argument
(Figure 14). However, partitive e-NPs seem to show a slightly greater versatility
in their discourse status.
One of the most common functions of the partitive case – signalling non-
inclusive quantity – is rather suited with indefiniteness and new discourse refer-
ents. Only 9% of all case-marked Os and e-NPs in the corpus are in the partitive
and at the same time active. Predictably, among them there is no argument
whose partitive marking is caused by non-inclusive quantity; their case is rather
determined by negation, imperfective aspect, influence of a predicate verb or the
occurrence of a SAP pronoun as an O.
In the data, the heavier the argument is, the more likely it is discourse in-
active (Figure 15). It has been suggested by Lambrecht (1994: 96) that languages
differ widely with respect to the tolerance for non-pronominal coding of active
referents. In the present corpus such coding is quite common: 28% of active
referents are not marked by pronouns or zero-anaphora but by heavier phrases.
238 Helena Metslang
Figure 15: Distribution of phrase weight among arguments with a different activeness status
(%, n = 520)
7.4 Definiteness
Estonian does not have grammatical means exclusively devoted for marking
definiteness. Therefore I treat definiteness as a purely meaning-based parameter
here that links pragmatics and semantics.
Definiteness is a category that is closely related to topicality (e.g. Givón
2001: 472−473) and includes the interplay of the following factors: familiarity or
identifiability on the one hand (notions relating to the qualitative definiteness of
entities) and uniqueness or inclusiveness on the other (notions relating to the
quantity of entities) (Lyons 1999: 2–13). In this study I only analyze an entity as
‘definite’ if it is both quality and quantity-wise definite: an entity or a group of
entities that is uniquely identifiable for the listener and has an inclusive quan-
tity. Definiteness of an expression can come from situational or associative use
(association of an entity with a particular situation or the preceding discourse),
the interlocutors’ general knowledge, anaphora or the referent’s uniqueness in
the active context (ibid.). ‘Indefinite’ means that the entity is not referred to as
‘definite’. I have adopted from Vilkuna (1992: 106–125) an additional definiteness-
related property ‘category reference’, a phenomenon close to genericity. When
the reference is to a category, the focus is on the description – which is more
important for the communicative purpose than referring to a specific individual.
For example: Now you are asking the wrong person! The sentence would lose its
communicative function if the NP was replaced with a specific one: Now you are
asking me! Also other unspecific referents have been assigned the ‘category
reference’ like viiulit mängima ‘to play violin.’ In many cases the definiteness
distinction is semantically not relevant (this has been suggested about Finnish
by Vilkuna 1992: 177).
Definiteness is a category closely linked with discourse activeness. However, it
is necessary to treat these notions separately in this study because various com-
binations of activeness status and definiteness values of referents are possible.
For instance unique referents can be definite but inactive arguments: ‘I dis-
covered the notion ‘humanness’ for myself’ and ‘I lost the spectators’ sympathy’.
Reference to a category is possible with both active and inactive referents.
As topics must be in the possession of the hearer (Erteschik-Shir 2007: 15),
the typical topics – A and S – usually have definite referents in the corpus (see
Table 16).
Table 16: Distribution of definiteness among the arguments (%, n = 520)
Argument Definite Indefinite Category Irrelevant Total (abs. No)

A 94 1 2 3 130
S 87 2 3 8 130
O 34 27 4 35 130
e-NP 11 60 5 24 130
Again, S is also definiteness-wise more varied than A in the corpus and seems to
be a bit more of a suitable category for phrases whose definiteness is not speci-
fied. Cross-linguistically objects are not prototypical topics and therefore their
definiteness value is unpredictable (Givón 2001: 472−473). In the data, O and
e-NP demonstrate more variation in the definiteness parameter than A and S and
also more than O and e-NP do with regard to the discourse activeness parameter.
Table 16 gives support to the view that definiteness and discourse activeness
should indeed be studied separately when comparing the message packaging
properties of argument types. The distinctions between O and e-NP particularly
come to light when looking the definiteness feature value distributions (espe-
cially as O’s referent is often definite), to a lesser degree in the case of discourse
activeness (similarly to e-NP, O is seldom active in discourse). Further, the pro-
portion of Os and e-NPs in the corpus, for which the property definiteness is
irrelevant is higher that of those for which the property discourse activeness is
irrelevant.
Regarding the case-marking of Os and e-NPs in the corpus, the frequency-
wise unmarked case of the argument dominates among definite and indefinite
240 Helena Metslang
Figure 16: Distribution of definiteness among the case uses of arguments (absolute numbers,
n = 260)
arguments. Only when an e-NP’s referent is neither definite nor indefinite does it
rather tend to be in the partitive than nominative (see Figure 16).
As mentioned in Sections 4 and 5, definite O and e-NP (i.e. arguments that
are at the same time identifiable and with inclusive quantity) can be marked
with both total and partitive case. The case-marking rules determine O’s and
e-NP’s case as follows:
– definite e-NPs in affirmative clauses are in the nominative;
– definite e-NPs in negative clauses are in the partitive;
– definite Os in perfective affirmative clauses are in a total case (with the
exception of SAP pronouns);
– definite Os in imperfective affirmative clauses are in the partitive;
– definite Os in negative clauses are in the partitive.18
See example (117) for a partitive definite O and (118) for a total case definite O.
Although e-NPs are mainly used for referring to indefinite inactive referents,
e-NP can also denote other kinds of referents. See example (119) for a nomina-
tive definite e-NP and (112) in 7.3 for a partitive definite e-NP in a negative clause.
18 Table 5 in Section 5.4 outlines how inclusiveness and aspect work together in the process
of object case-assignment.
(117) Võt-si-n ennas-t kokku.

take-PST-1SG oneself-PAR together
‘I pulled myself together.’ (partitive definite O)
(118) Meeskond võtt-is teate noore abiellumise-st

team.NOM take-PST.3SG message.GEN youth.GEN getting.married-ELA
vastu soliidselt.
for respectfully
‘The team received the message of the youth getting married
respectfully.’ (total case definite O)
(119) Köögi-s surista-s aeg-ajalt külmkapp.

kitchen-INE buzz-PST.3SG from.time.to.time fridge.NOM
‘In the kitchen, the fridge was buzzing from time to time.’ (total case
definite e-NP)
In example (118), I regard the abstract entity ‘definite’ because it is referential

and the sentence uniquely specifies its identity.
See (18) and (19) in Section 4.1 for examples of indefinite O. In both sentences
the O referent is quantitatively definite (a singular count noun) but qualitatively
indefinite (unfamiliar). Therefore the overall definiteness of the phrase is indefinite.
The examples (100) in Section 7.1 and (120) have nominative and partitive in-
definite e-NPs.
(120) Õue-s mängi-b laps-i.

outside-INE play-3SG child-PAR.PL
‘There are some children playing outside.’ (translated from Vilkuna: 1992:
47; indefinite partitive e-NP)
Category reference is rare in the corpus, see examples (109) in Section 7.3 and
(121) for O and e-NP.
(121) On ju elukutse-i-d nagu tuletõrjuja ja autojuht,

be.3 MDA profession-PL-PAR like fire-fighter.NOM and driver.NOM
mille-ga laps on varase-st ea-st tuttav.
that-COM child.NOM be.3 early-ELA age-ELA familiar.NOM
‘Obviously, there are professions like fire-fighter and driver that children
are familiar with from an early age.’ (e-NP, reference to a category)
242 Helena Metslang
In many cases, nominal or clausal expressions (including propositions like direct

speech and also mental processes and thoughts) cannot be analyzed from the
point of view of definiteness as they are not referential. I regard the definiteness
category ‘irrelevant’ for them (see (107) in Section 7.3 for an example of O).
The activeness status is distributed among the arguments on the definite-
ness scale as follows:
– Active referents tend to be definite As and Ss in the corpus (48% and 40%
respectively of all active entities in the corpus) and, to some extent, definite
Os (10%).
– Accessible referents are definite Os, Ss and As (27%, 24% and 12% respec-
tively of all accessible entities) and indefinite e-NPs (12%).
– Inactive referents are frequently indefinite e-NPs (36%) and Os (15%).
In the following I will compare the overall definiteness, activeness status and
the Referential hierarchy properties data in the corpus. The kind of reference
that dominates in the studied 390 clauses from fiction texts is active reference
to people and inactive reference to abstract entities. Also non-referential phrases
are common. The most frequent reference types in the corpus are the following:
(a) definite discourse active SAPs (24%, see example (122));
(b) definite discourse active other humans (13%, see (86), repeated as (123)
here);
(c) indefinite inactive abstract entities (11%, see example (124));
(d) inactive abstract entities in whose case definiteness is not relevant (7%, see
(70) repeated here as (125));
(e) non-referential elements in whose case both categories, activeness and
definiteness are irrelevant (7%, see (1) in Section 2).

‘You practice when some others do not.’ (S)
(123) Treener kisku-s valu-st kõvera-sse.

coach.NOM double.up-PST.3SG pain-ELA doubled.up.position-ILL
‘The coach doubled up in pain.’ (S)

Silvia.NOM get-PST.3SG hear-INF our.GEN time.GEN typical.story.GEN
typical story of our time.’ (O)
(125) Mänguvorm stabiliseeru-s.

fitness.NOM stabilize-PST.3SG
‘(My) fitness stabilized.’ (S)
The types (a), (c), (d) and (e) also prevail among the most common reference
types of partitive NPs. Partitive NPs often also refer to indefinite events that are
inactive in discourse. I repeat example (69) from Section 6.3 here:
(126) Tagatoa-st kost-is köhimis-t.

back.room-ELA sound-PST.3SG coughing-PAR
‘One could hear coughing in the back room.’ Lit. ‘There sounded
coughing in the back room.’ (singular e-NP denoting an event)
There is a considerable group of NPs that have abstract referents of trivial dis-
course importance in whose case definiteness rather seems to be an irrelevant
category. Their activeness can still be evaluated as they are referential. They are
mostly inactive e-NPs, Os and Ss:
(127) Enda-l=gi Ø ruumi vaevalt ringi pööramise-ks.

self-ADE=CL (be.3) space.PAR merely around turning-TRAN
‘We ourselves (have) only just (enough) space for turning around.’
(e-NP belonging to the ‘existential partitives’ group)
(128) Tema pilgu-st voola-s armastus-t.

3 SG .GEN gaze-ELA flow-PST.3SG love-PAR
‘There was love flowing from his gaze.’ (e-NP)

‘The surprise effect was big in the team.’ (S)
8 Discussion
In this paper, 130 examples of each of four Estonian core arguments (A, S, O and
e-NP) were compared from the point of view of 11 properties:
– coding (case and agreement);
– semantics (number, person, the Referential hierarchy, discourse importance
and the situation type the argument is participating in);
– message packaging (phrase weight, word order, discourse activeness and
definiteness).
244 Helena Metslang
This section draws generalizations across all these properties. To assess the
proximity of the arguments to each other I compared, in the case of each argu-
ment, the values of all the 11 properties to the values of a hypothetical ideal
A-argument (a discourse salient SAP who is volitionally in control of a highly
transitive action). To evaluate the arguments with respect to each property I chose
the value each argument demonstrated most frequently in these 130 examples.
For instance, when picking which group of Os is most representative of the O
argument in the comparison of the ‘activeness’ feature, it is relevant that out of
130 Os the largest group (45%) is discourse inactive. Therefore, when evaluating
the property ‘activeness’ I assigned the O argument the value ‘inactive’. After
comparison to the ideal A (that is in most cases active in discourse), O’s value
‘inactive’ only gets the numeric value 0.33 out of 1.0 (see Tables 17 and 18
below).
When ordering and rating the feature values numerically (Table 17) I con-
sidered the scales presented in this article, transitivity properties (following
Hopper & Thompson 1980: 252) and also the frequency data of the A argument
in the corpus. As the values of the largest argument groups differed from each
other significantly (e.g. the discourse activeness value of the largest group of
As and of the largest group of Os are in different ends of the activeness con-
tinuum: ‘active’ and ‘inactive’ respectively) this method proved robust enough
for determining the arguments’ positioning with respect to each other.
Most of the 11 criteria studied in this paper were suitable for showing dis-
tinctions between Estonian core arguments (Figure 17). The only feature that
does not distinguish the arguments is number and it is therefore regarded less
relevant for this task. The remaining 10 criteria clearly define the A and S cate-
gories and outline the borders of the direct object. The analysis also makes
it evident that e-NP is a separate category that has similarities especially with
the object and to a lesser degree with the subject (see also Helasvuo 2001 on
a similar result in Finnish). The greatest division in the Estonian core argument-
system according to this study is between subjects and partitive-permitting
arguments.
A and S map together in 8 criteria out of 10 that include semantic, coding
and packaging properties. The features that A and S do not share in the data,
can be classified as semantic: person preference and the situation type where
the argument occurs in (intransitive clause has a more varied set of uses than
transitive clause, e.g. the static equative, attributional and locative clauses).
In 6 criteria out of 10 the largest groups of the direct object and e-NP are
similar. O and e-NP have in common both semantic (person, the Referential hier-
archy and discourse importance) and message packaging features (activeness,
phrase weight and word order). The largest groups of Os differ from the largest
Table 17: Relative ordering of feature values (a higher ranking means ‘more like an ideal A’ or ‘with a higher degree of transitivity’)
Case Agreement Person Ref. hierarchy Discourse imp. Situation type Phrase weight Word order Definiteness Activeness
total case 1 agr with verb 1 SAP 1 SAP 1 prominent 1 achievement 1 zero-anaphora 1 preverbal 1 definite 1 active 1
partitive 0 unmarked 0.5 third 0.5 human 0.8 salient 0.67 active accompl. 0.8 pronoun 0.75 postverbal 0 indefinite 0.67 accessible 0.67
verb form
lack of agr 0 unspecified 0 animate 0.7 trivial 0.33 semel-factive 0.6 bare noun 0.5 irrelevant 0.33 inactive 0.33
concrete 0.5 irrelevant 0 activity 0.4 full NP 0.25 category 0 irrelevant 0
abstract 0.3 accomplishment 0.2 heavy phrase 0
event 0.2 state 0
irrelevant 0
Table 18: Ratings of most frequent argument groups.
Argument Case Agreement Person Ref. Discourse Situation Phrase Word Definiteness Activeness
hierarchy imp. type weight order
A 1.00 1.00 1.00 1.00 1.00 0.40 1.00 1.00 1.00 1.00
S 1.00 1.00 0.50 1.00 1.00 0.00 1.00 1.00 1.00 1.00
O 0.00 0.00 0.50 0.30 0.33 0.40 0.50 0.00 0.33 0.33
Partitive noun phrases in the Estonian core argument system
e-NP 1.00 0.50 0.50 0.30 0.33 0.00 0.50 0.00 0.67 0.33
245
246 Helena Metslang
Figure 17: Argument ratings according to their largest feature value groups
groups of e-NPs in definiteness, the situation type they occur in, coding and also
in that O is a category that has a high degree of variation.19 As pointed out
above, O and e-NP also have in general a lot in common in terms of both differ-
ential case-marking systems and agreement. The similarity between O and e-NP
can be explained by focality and the semantic bonding between these argu-
ments and the predicate verb (see below).
The most clear-cut difference between subjects and partitive-permitting
arguments is in message packaging: they cluster differently in all four criteria.
Also two interrelated semantic criteria distinguish partitive-permitting arguments
from subjects: the Referential hierarchy and discourse importance. Subjects are
19 In the case of four properties (person, situation type, phrase weight and definiteness) some
arguments’ largest and second largest groups are very close in size. Hypothetically, if the
second largest group was the largest in the case of these properties in the corpus, the results
in Figure 17 would be different. In the discussion of the status of the e-NP (whether it is closer
to A/S or O) none of the four properties would make a difference in the number of features e-NP
shares with O and A/S.
usually high in the Referential hierarchy and discourse prominent. Partitive-

permitting arguments on the other hand, are low and of trivial importance. The
intransitive subject and partitive-permitting arguments share the preference for
third person in the corpus.
e-NP and the subject (S or both S and A) are similar in only 3−4 criteria out
of 10 in the corpus. They share their preferred case use and agreement (although
the agreement between e-NP and the unmarked form of the verb is not a strong
indicator of e-NP − subject similarity; see Section 3). The properties the largest
group of e-NPs share with the intransitive subjects are third person reference
and the preference for occurrence in states. However, if we also look at the
general distribution of all situation types of the S argument, it rather resembles
transitive clauses (see Figure 9 in Section 6.4).
When comparing different case uses of O and e-NP only statistically smaller
biases appear in the corpus:
– The main result of case comparisons is the finding that the category ‘situa-
tion type’ distinguishes between total case and partitive Os whose referents
either tend to occur in achievements or activities respectively.
– When an e-NP’s referent is neither definite nor indefinite (it is indifferent
towards this meaning distinction) then it rather tends to be in the partitive.
– Partitive e-NPs have a slightly larger proportion of discourse active referents
than the nominative ones. Also the e-NPs that cannot be analyzed with
respect to this feature (denoting propositions, events and parts of opaque
multi-word verbs) are more often in the partitive.
– Bare noun e-NPs occur slightly more in the partitive than in the nominative.
Similarity of O and e-NP – a result of semantic bonding with the verb?

The main factor causing O and e-NP similarity seems to be their common topi-
cality status. The influence of topicality on argument realization (selection to
grammatical relations, coding and message packaging) and its relation with
arguments’ semantic properties has been confirmed by all the aspects of this
study. In addition to topicality, also another feature has been pointed out that
unites O and e-NP – the argument’s semantic bonding with the verb.20
The dependence of the object referent and its particular characteristics (e.g.
affectedness) on the predicate verb is well-known. Vilkuna (1989: 163, 175, 181)
describes the e-NP’s and verb’s semantic bonding as follows. The predicate
verb of the EC expresses existence and also the manner of being located in a
place. Usually there is a particular manner of location that suits with particular
20 I thank Marja-Liisa Helasvuo for some illuminating ideas on the roles of both of these
factors.
248 Helena Metslang
referents that often depends on the referent’s high or low individuation.21 This
causes the e-NP’s referential dependence on the verb. She brings the following
examples from Finnish:
(130) Katosta riippui hämähäkinseittejä.

ceiling-EL was-hanging cobwebs-PAR
‘Cobwebs were hanging from the ceiling.’ (Vilkuna 1989: 163;
e-NP low in individuation)
(131) Toimistossa työskentelee naisia.

office-INE works women-PAR
‘Women work in the office.’ (Vilkuna 1989: 163; e-NP high in individuation)
These sentences have the same structure in Estonian:
(132) Lae-st rippu-s ämblikuvõrk-e.

ceiling-ELA hang-PST.3SG cobweb-PAR.PL
‘Cobwebs were hanging from the ceiling.’
According to Vilkuna the EC’s predicate verb’s function of expressing the manner
of location is the basis for the stronger semantic bonding between the verb and
the e-NP. In the case of many idiomatic expressions, the frame contains a fixed
predicate verb compound with a post-verbal object or e-NP and the topic posi-
tion can be filled freely:
(133) (Ta-l) on vesi ahju-s.

3 SG -ADE be.3 water.NOM oven-INE
‘(He) is in trouble.’ Lit. ‘He has water in the oven.’ (compound of
the verb and e-NP)
This makes these collocational units similar to the ones illustrated in (130)–
(132).
Is the e-NP’s case-alternation a manifestation of fluid intransitivity?

Alignment holds between sets (usually pairs) of argument roles that have the same
formal treatment (e.g. case) in some context. Alignment patterns define language-
specific and construction-specific grammatical relations (Bickel & Nichols 2008:
21 The term individuation is used in the sense of Hopper & Thompson (1980).
305). In accusative alignment the coding and/or behaviour constructions treat

the S and A argument in one way and the O argument in another way. In erga-
tive alignment, S and O are treated in the same way and A differently. The lan-
guages where accusative case alignment dominates have pragmatically-oriented
case-marking that is designed for coding the grammatical subject and direct
object, regardless of semantic roles or transitivity (Givón 2001: 203). In such lan-
guages the subject and topic role are greatly merged, the subject role is prefer-
entially filled by higher positions of the Referential hierarchy that are inherently
more likely to be topical (cf. Bickel: 2010; Givón 2001). The Estonian A, S and O
arguments largely comply with this description.
When analyzing the potential alignment split in Estonian subject marking
we must first discuss the subjecthood-status of e-NP. Onishi (2001) shows that
typologically non-canonical arguments often differ quite considerably from the
canonical ones (from the point of view of case, agreement, behaviour and the
clause’s degree of transitivity). Due to the shared properties with S/A (incl.
the largest group of e-NPs having nominative marking and verbal agreement;
see above in this section) it is possible to treat it as a ‘non-canonical subject’ or
‘subject-like argument’ (the term used by Barðdal (2006)). This is supported by
the fact that the category that results from the combination of the intransitive
subject (as the referentially high S) and e-NP (as the referentially low S) supports
the Referential hierarchy’s predictions on subject case alternation (see Section
5.2).
In the following I will look at e-NP’s alignment with the coding properties of
the core arguments (A/S vs. O). As the partitive is a case that clearly cannot
mark prototypical subjects, I take its occurrence as a valid argument for align-
ment judgements. The other e-NP’s case option, the nominative, can be regarded
as a sign of subject-like marking. Unfortunately it is a weaker proof of e-NP bear-
ing a subject case as also O occurs frequently in total cases, including the
nominative. It is also possible to regard e-NP’s case alternation as a whole as an
object-like property as differential case-marking is in general impossible with
unmarked clause S and A.
In Estonian, the differential marking (split system) depends on information
structure: the alternative object-like marking of S takes place in non-topical exis-
tential arguments (Erelt 2008: 71−76). As pointed out above and in Section 5.2,
one can choose between two alternative analyses: considering both nominative
and partitive e-NPs ‘SO’ (instransitive subject marked like O) or only regarding
partitive e-NPs as ‘SO’. In Section 5.2 we saw that the first alternative neatly dis-
tinguishes e-NPs as semantically low subjects whose overt marking is in line
with the Referential hierarchy’s predictions. The first option is also more viable
due to the fact that the nominative and partitive e-NP have similar results with
250 Helena Metslang
respect to the semantic and message packaging properties studied in this paper –
the nominative e-NP is not closer to the intransitive subject than the partitive
e-NP. Topicality is a stronger basis for describing the split in S marking than
just case.
If we regard both nominative and partitive e-NPs as ‘SO’ then SA occurs in
unmarked constructions and SO in ECs. It is relevant that SO’s object-like coding
is then characterized by object-like case alternation between the total and parti-
tive cases, often also by object-like lack of agreement (if we include unmarked
third person inflection), as well as the post-verbal position.22 Under this analysis,
SO marking is suitable with a majority of intransitive verb lexemes that have
existential meaning or an existential meaning component. This conforms to the
typologically attested fluid intransitivity system – that is when each intransitive
verb has the possibility of two kinds of marking for its core NPs – SA and SO.
Fluid argument coding usually reflects conceptual and constructional properties
(argument’s marking depends on control, volitionality, etc.; Dixon 1994: 78−83;
Witzlack-Makarevich 2011: 131−136).
A similar phenomenon of O-like non-topical S has been described in French
(Creissels 2008). The argument occurs in presentational impersonal construc-
tions that denote existence or coming into existence at a location:
(134) Il est entré trois garçons.

AG3SGM AUX.PRS.3SG enter.PTCP.SGM three boy.PL
‘Three boys entered.’ (Creissels 2008)
In these constructions the verb does not inflect for person, S is clause-final and
aligns in most ways with O. Creissels explains this non-canonical construction
by the need to mark discourse new S as non-topical. He finds that the occurrence
of fluid intransitivity is motivated by pragmatics:
A is typically more topical than P, and new referents are typically introduced in P position;
consequently, in a language in which accusative alignment predominates, it is natural to
de-topicalize S by means of a construction in which S is aligned with P.23 (ibid.)
Hence the fluid intransitive system in Estonian resembles in most respects the
one proposed for French.
22 Splits based on word order have been considered typologically rare but still existent in
several languages (Donohue 2008: 27−28).
23 When using the abbreviaton P the author refers to the same argument as the one denoted
by O in the present paper.
9 Conclusion
This paper discusses how close the Estonian partitive-permitting arguments –
object and e-NP (existential clause noun phrase) – are to each other and to the
transitive and intransitive subject. I analyze 390 sentences and look at 11 para-
meters: coding (case and agreement); semantics (number, person, the Referential
hierarchy, discourse importance and the situation type the argument participates
in); message packaging (phrase weight, word order, discourse activeness and
definiteness). When comparing the statistically largest group of each parameter
in the case of each argument type, 10 criteria distinguish the four argument
types whereas number marking (singular/plural) gets the same results for all of
them.
From the point of view of these criteria, the greatest division in the Estonian
core argument system is between unmarked clause subjects and partitive-
permitting arguments (O and e-NP). Topicality-related message packaging fea-
tures show a strong correlation with semantic properties (especially with the
Referential hierarchy and discourse importance): from the point of view of these
features S clusters together with A and O patterns with e-NP. Coding also corre-
lates with these content properties. Hence the results of this paper illustrate the
major impact that topicality has on Estonian argument realization and usage.
Only two semantic properties, situation type and person reference do not sup-
port the formulation of these two, otherwise similar, argument groups.
The largest groups of the direct object and e-NP are similar in six criteria out
of 10. O and e-NP share both semantic and message packaging features. They
also have overlaps in case and agreement. O and e-NP have largely the same
differential case-marking system which depends on referential and clausal
case-assignment factors as well as lexical predicates. However, as shown in the
corpus analysis, the application of the case-marking rules has some significant
differences between these arguments. In Estonian, case can be assigned to O
and e-NP in two ways. It can be determined lexically – by the verb (especially
in the case of Os) or by the noun (especially in the case of e-NPs). Sometimes
case-marking follows the fluid strategy where case depends on use/situational
semantics (in the case of both arguments, especially Os). The main factor influ-
encing both object’s and e-NP’s case is referential – inclusiveness – but it occurs
in different connections and combinations with O and e-NP. In the paper, two
hierarchies were suggested to integrate O’s and e-NP’s case-assignment: the
Quantitative markedness hierarchy and the Inclusiveness hierarchy. The former
determines the case on the basis of whether the argument is semantically marked
for quantification (inclusiveness) at all or whether inclusiveness is unmarked.
252 Helena Metslang
The latter determines the argument’s case on the basis of whether the referent’s
quantity is inclusive or non-inclusive.
Out of 10 properties, e-NP and subject (S or both S and A) are similar in 3−4
semantic and coding criteria. The paper suggests that it is possible to treat e-NP
as a non-canonical S and then the coding distinctions between S and e-NP can
be regarded as a manifestation of fluid intransitivity.
The study also analyzes whether there are any statistical preferences for the
object’s and e-NP’s case across different semantic, coding and message-packaging
properties. However only minor biases are identified, the main one being that
O’s case reflects (aspect-based) Vendlerian situation type distinctions.
References
Barðdal, Jóhanna. 2006. Construction-specific properties of syntactic subjects in Icelandic and
German. Cognitive Linguistics 17(1). 39–106.
Bickel, Balthasar. 2008. On the scope of the referential hierarchy in the typology of gram-
matical relations. In Greville G. Corbett & Michael Noonan (eds.), Case and Grammatical
Relations, 191–210. Amsterdam: Benjamins.
Bickel, Balthasar. 2010. Grammatical relations typology. In Jae Jung Song (ed.), The Oxford
Handbook of Language Typology, 399−444. Oxford: Oxford University Press.
Bickel, Balthasar & Johanna Nichols. 2008. Case marking and alignment. In Andrej Malchukov
& Andrew Spencer (eds.), The Oxford Handbook of Case, 304−321. Oxford: Oxford Univer-
sity Press.
Bickel, Balthasar, Witzlack-Makarevich, Alena & Zakharko, Taras (to appear). Typological
evidence against universal effects of referential scales on case alignment. In Ina Bornkessel-
Schlesewsky, Andrej Malchukov & Marc Richards (eds.), Scales: A cross-disciplinary
perspective. Berlin: Mouton de Gruyter.
Butt, Miriam. 2008. Modern approaches to case: An overview. In Andrej Malchukov & Andrew
Spencer (eds.), The Oxford handbook of case, 27−43. Oxford: Oxford University Press.
Chafe, W. L. 1976. Givenness, contrastiveness, definiteness, subjects, topics, and point of view.
In Charles N. Li (ed.), Subject and topic, 25−56. New York: Academic Press.
Chafe, W. L. 1994. Discourse, consciousness and time: The flow and displacement of conscious
experience in speaking and writing. Chicago: University of Chicago Press.
Comrie, Bernard. 1976. Aspect (Cambridge Textbooks in Linguistics). Cambridge: Cambridge
University Press.
Comrie, Bernard. 1978. Ergativity. In Wilfred P. Lehmann (ed.), Syntactic Typology, 329–394.
Sussex: The Harvester Press.
Comrie, Bernard. 1989. Language universals and linguistic typology, 2nd revised edition.
Chicago: University of Chicago Press.
Creissels, Denis. 2008. Remarks on split intransitivity and fluid intransitivity. In Olivier Bonami
& Patricia Cabredo Hofherr (eds.), Empirical issues in syntax and semantics 7. Colloque de
Syntaxe et Sémantique à Paris (CSSP) 2007, 139−168.
Croft, William. 2001. Radical Construction Grammar: Syntactic theory in typological perspective.
Oxford: Oxford University Press.
Dixon, R. M. W. 1994. Ergativity. Cambridge: Cambridge University Press.
Donohue, Mark. 2008. Semantic alignment systems: what’s what, and what’s not. In Mark
Donohue & Søren Wichmann (eds.), The typology of semantic alignment, 24–76. Oxford:
Oxford University Press.
Dowty, David R. 1979. Word meaning and montague grammar. Dordrecht: Reidel.
Dryer, Matthew S. 1997. Are grammatical relations universal? In Joan Bybee, John Haiman &
Sandra A. Thompson (eds.), Essays on language function and language type. Dedicated
to T. Givón, 117–143. Amsterdam: Benjamins.
Erelt, Mati. 2005. Source-marking resultatives in Estonian. Linguistica Uralica 1. 20–29.
Erelt, Mati. 2008. Eesti keele tüübist ja keelekorraldusest [Estonian from a typological
perspective and language planning]. Emakeelne Eesti, emakeelne Euroopa, 67−79. Tallinn:
Estonian Language Foundation.
Erelt, Mati. 2009. Typological overview of Estonian syntax. Sprachtypologie und Universalien-
forschung. Akademie Verlag 62(1–2). 6–28.
Erelt, Mati, Reet Kasik, Helle Metslang, Henno Rajandi, Kristiina Ross, Henn Saari, Kaja Tael &
Silvi Vare. 1993. Eesti keele grammatika II. Süntaks, lisa: kiri [Estonian grammar II. syntax
(Appendix: Script)]. Tallinn: Estonian Academy of Sciences, Institute of Language and
Literature.
Erelt, Mati & Helle Metslang. 2006. Estonian clause patterns – from Finno-Ugric to standard
average European. Linguistica Uralica 42(4). 254–266.
Erteschik-Shir, Nomi. 2007. Information structure. The syntax-discourse interface. Oxford: Oxford
University Press.
Givón, Talmy. 2001. Syntax, volume 1. Amsterdam: Benjamins.
Goldberg, Adele E. 1995. Constructions: A construction grammar approach to argument
structure. Chicago: University of Chicago Press.
Helasvuo, Marja-Liisa. 2001. Syntax in the making. The emergence of syntactic units in Finnish
conversation. Amsterdam: Benjamins.
Hiietam, Katrin. 2003. Definiteness and grammatical relations in Estonian. Manchester:
University of Manchester doctoral thesis.
Hoop, Helen de & Peter de Swart. 2009. Cross-linguistic variation in differential subject
marking. In Helen de Hoop and Peter de Swart (eds.), Differential subject marking.
(Studies in Natural Language and Linguistic Theory 72), 1−16. Dordrecht: Springer-Verlag
New York Inc.
56. 251–299.
Huumo, Tuomas. 1993. Suomen ja viron kontrastiivista sanajärjestysvertailua [Contrastive word
order comparison of Finnish and Estonian]. In Valma Yli-Vakkuri (ed.), Tutkimuksia syntaksin
ja pragmasyntaksi alalta. (Studia Comparativa Linguarum Orbis Maris Baltici 1), 97−158.
Turku: University of Turku.
Huumo, Tuomas. 1999. Kieltolauseen partitiivisubjektin semantiikkaa [The semantics of
partitive subjects in negative clauses]. Sananjalka 41. 21–42.
Huumo, Tuomas. 2002. Syntax or discourse pragmatics: A contrastive analysis on Finnish and
Estonian word order. In Luis I. Rábade & Susana M. Doval Suárez (eds.), Studies in Contras-
tive Linguistics. Proceedings of the 2nd International Contrastive Linguistics Conference,
Santiago de Compostela; October, 2001, 495−502. Santiago de Compostela: Universidade
de Santiago de Compostela.
254 Helena Metslang
Journal of Linguistics 46(1). 83−125.
Koptjevskaja-Tamm, Maria & Bernhard Wälchli. 2001. The Circum-Baltic languages. An areal-
typological approach. In Östen Dahl & Maria Koptjevskaja-Tamm (eds.), Circum-Baltic
Languages, volume 2: Grammar and Typology (Studies in Language Companion Series
55), 615−750. Amsterdam: Benjamins.
Lambrecht, Knud. 1994. Information Structure and Sentence Form. Topic, Focus, and the Mental
Representations of Discourse Referents. Cambridge: Cambridge University Press.
Langacker, Ronald W. 1987. Foundations of Cognitive Grammar, volume 1 (Theoretical Pre-
requisites). Stanford: Stanford University Press.
Levin, Beth & Malka Rappaport Hovav. 2005. Argument realization. Cambridge: Cambridge
University Press.
Lindström, Liina. 2002. Veel kord subjekti ja predikaadi vastastikusest asendist laiendi
järel [Once more about subject’s and predicate’s reciprocal position after the adjunct].
Emakeele Seltsi aastaraamat 47. 87−106. Tallinn: Teaduste Akadeemia Kirjastus.
Lindström, Liina. 2005. Finiitverbi asend lauses. Sõnajärg ja seda mõjutavad tegurid suulises
eesti keeles [The position of the finite verb in the clause: word order and the factors affect-
ing it in spoken Estonian]. University of Tartu doctoral thesis. Tartu: Tartu University Press.
Lindström, Liina, Mervi Kalmus, Anneliis Klaus, Liisi Bakhoff & Karl Pajusalu. 2008. Ainsuse 1.
isikule viitamine Eesti murretes [Reference to first person singular in Estonian dialects].
Emakeele Seltsi aastaraamat 54. 159−185. Tallinn: Teaduste Akadeemia Kirjastus.
Metslang, Helena. 2012. On the case-marking of existential subjects in Estonian. SKY Journal of
Linguistics 25. 151−204.
Metslang, Helena. 2013. Coding and behaviour of Estonian subjects. Journal of Estonian and
Finno-Ugric Linguistics 4(2). 217−293.
Nemvalts, Peep. 1996. Case marking of subject phrases in modern standard Estonian. Studia
Uralica Uppsalaensia 25. Acta Universitatis Upsaliensis. Uppsala: Uppsala University
doctoral thesis.
Nemvalts, Peep. 2000. Aluse sisu ja vorm: alusfraasi käändevaheldus tänapäeva eesti
kirjakeeles [The subject’s meaning and form: Subject case alternation in contemporary
written Estonian]. Tallinn: Estonian Language Foundation.
Onishi, Masayuki 2001. Introduction: Non-canonically marked subjects and objects: Parameters
and properties. In R. M. W. Dixon, Alexandra Aikhenvald & Masayuki Onishi (eds.), Non-
canonical Marking of Subjects and Objects, 1–51. Amsterdam/Philadephia: John Bejamins
Publishing Company.
Rätsep, Huno. 1978. Eesti keele lihtlause tüübid [Estonian Simple Clause Types]. ENSV TA
Emakeele Seltsi toimetised 12. Tallinn: Valgus.
Tael, Kaja. 1988. Sõnajärjemallid eesti keeles (võrrelduna soome keelega) [Word Order Patterns
in Estonian (in Comparison with Finnish)] (Preprint KKI-56). Tallinn: Academy of Sciences
of the Estonian SSR. Section of Social Sciences.
Tael, Kaja. 1990. An Approach to word order problems in Estonian (Preprint KKI-66). Tallinn:
Estonian Academy of Sciences. Division of Humanities and Social Sciences.
Tamm, Anne. 2004. Relations between Estonian Verbs, Aspect, and Case. Doctoral Dissertation.
Budapest: ELTE BTK.
Tamm, Anne. 2007. Estonian transitive verbs and object case. In Miriam Butt & Tracy Holloway
King (eds.), Proceedings of the LFG06 Conference, Universität Konstanz, 485−504. Stanford:
CSLI Publications.
Tamm, Anne. 2009. The Estonian partitive evidential: Some notes on the semantic parallels
between the aspect and evidential categories. In Lotte Hogeweg, Helen de Hoop & Andrej
Malchukov (eds.), Cross-linguistic Semantics of Tense, Aspect, and Modality, 365−401.
Tamm, Anne. 2012. Scalar verb classes. scalarity, thematic roles, and arguments in the
Estonian aspectual lexicon. Firenze: Firenze University Press.
Torn-Leesik, Reeli. 2009. The voice system of Estonian. Sprachtypologie und Universalienfor-
schung 62(1−2). 72−90.
Vaiss, Natalia. 2004. Eesti keele aspekti väljendusvõimalusi vene keele taustal [Expressing
aspect in Estonian (on the background of Russian)]. Tallinn: Tallinn University MA thesis.
Van Valin, Robert D. 2005. Exploring the Syntax-Semantics Interface. Cambridge: Cambridge
University Press.
Vendler, Zeno. 1967. Verbs and times. In Zeno Vendler (ed.), Linguistics in Philosophy, 97–121.
Vilkuna, Maria. 1992. Referenssi ja määräisyys suomenkielisten tekstien tulkinnassa [Reference
and definiteness in the interpretation of Finnish texts]. Suomi. Helsinki: Suomalaisen
Kirjallisuuden Seura.
Witzlack-Makarevich, Alena. 2011. Typological variation in grammatical relations. Leipzig:
Universität Leipzig doctoral thesis.
M. M. Jocelyne Fernandez-Vest
6 Finnish Partitive and resultativity in
translation(s): a discourse-cognitive
approach
As the result of a study of the Finnish partitive (FIP) in a discourse-cognitive per-

spective, centered on partitive (PAR) as a mark of the O(bject), and excluding its
automatic use in a negative/dubitive sentence (Fernandez-Vest 2010a), I compare
the main criteria for choosing the FIP: context/cotext, situation, or simply the
inherent aspect of the verb? The verbs of the corpus (oral/written, narratives
and dialogues of fiction) have been classified into 3 categories – RES (resulta-
tive), IRS (irresultative) and RES-IRS (resultative-irresultative), according to their
most typical aspectual meaning. This repartition, attempting to refine the tradi-
tional conception of resultativity in Finnish grammars (i.e. that the V leads to a
result, ISK 1431) is based on /+ and –values/ (FIP O = verb [–decisive change]
and situation [–end point], to be combined with the aspect of the verb (RES/
IRS/RES-IRS) and its internal variation (derivation suffixes)), shows the great
number of RES-IRS verbs (nearly 50%) and the ground difference betwen activity
verbs and mental verbs (nearly 70% of the second take a FIP O – Askonen 2001).
The discussion of the respective choice criteria is nourished by a comparison
between translations of Finnish texts into another Finno-Ugric language (Northern
Sami, which has no partitive case) – and neighboring Indo-European (Scandina-
vian) languages, and vice-versa. The role of this special case for the information
structuring of the sentence and discourse is apprehended through the variations
of translation: how the translator 1. transfers a selection of the combined fea-
tures of FIP into a partitive lacking language, 2. conversely, makes explicit with
FIP a polarity which was only latent. The 1st section settles the theoretical back-
ground, the 2nd section presents the corpus and analyzes some selected excerpts,
the 3rd section is reserved to a brief evaluating synthesis and some conclusions.
Notions borrowed from the Role and Reference Grammar and Cognitive Linguis-
tics are referred to.
Keywords: Finnish aspect, object case alternation, translation, interpretation

cues, discourse foregrounding
258 M. M. Jocelyne Fernandez-Vest
1 The theoretical background

1.1 From grammatical aspect to the representation of events
An internationally known linguist, the founder and main theoretician of the
Role and Reference Grammar (RRG), wrote recently:
A fundamental issue dividing theories of the syntax-semantics interface is whether the

semantic representation of clauses is projected from the lexical representation of the verb
which determines to a large extent the syntactic structure of the clause or whether it is
constructed or composed based on the NPs and PPs co-occurring with the verb in a clause;
in the latter view, the verb has a very general or underspecified meaning. The empirical
problem underlying this dispute concerns the ability of a single verb to occur in a variety
of morphosyntactic contexts, as illustrated with the English verb shatter (The window
shattered / The burglar shattered the window / The burglar shattered the window with a
crowbar / The crowbar shattered the window / *The window shattered with a crowbar.)
(Van Valin 2011)
The two theories in question are the projectionist vs. constructivist views
of meaning. The projectionist view, launched by Foley & Van Valin (1984, follow-
ing Dowty 1979), and developed in Van Valin (1993, 2005), Van Valin & LaPolla
(1997), was initially inspired by studies of the Russian and Georgian languages; it
stipulates that the interpretation of the object NP depends on properties of the
verb. The constructivist or constructionalist view, originally based on English and
championed by Goldberg (1995) and the following versions of construction gram-
mars, stipulates that the interpretation of a verb is related to properties of the
object NP or the choice of PP.
The study presented here has definitely to do with the syntax-semantics
interface, and could be tempted to adopt a constructionist view, a choice counter-
balanced by the knowledge I have of the primary importance of the verbs in the
debate about partitive object in Finnish. Fortunately, the author of the first quo-
tation has also recognized elsewhere that, from a processing perspective, both
views are valid and necessary:
1. Based on the Levelt (1989) model of a speaker, the speaker formulates a
semantic representation of the message that is to be conveyed, and this
semantic representation determines the morphosyntactic coding of the utter-
ance; in terms of logical structures, the choice of an activity or active accom-
plishment logical structure would determine the form of the object or the
nature of the PP.
2. The hearer, on the other hand, is necessarily constructivist, since the mor-
phosyntactic coding of the utterance contains the cues to its interpretation;
the only way a hearer can tell whether a verb is being used as an activity or
Finnish Partitive and resultativity in translation(s) 259
active accomplishment in a language like English is with reference to the

coding of the object or the choice of PP. (Van Valin 2010)
This second quotation is doubly relevant for my purpose: not only because
it exempts me from choosing between the two theories, but also because it sets
the right frame for my personal choice of investigating the relation between par-
titive and resultativity via translations. The translator is namely both a “speaker”
(/writer) and a “hearer” (/reader): in order to select the appropriate morphosyn-
tactic coding for the target utterance, he must first catch the interpretation cues
contained in the original author’s utterance.
The following analysis of excerpts from my corpus should therefore in prin-
ciple make it possible to verify whether the “cues” have been rightly caught and
interpreted, then transferred to the other language. Given the specificities of the
partitive (PAR) case, already analyzed and theorized by several generations
of Finnists/Finno-Ugrists/general linguists, this verification cannot be so easy
when translating from Finnish to other languages: no criterion (apart from the
negativity of the sentence, Fernandez-Vest 2010a) has been shown to be abso-
lutely constraining for the use of the partitive associated to most verbs. Con-
versely, the choice of the PAR is not automatically induced by an original Sami
or Norwegian text, since in most cases the FIP does not correspond any obvious
morphosyntactic feature (except for indefinite articles for the object NP in
Norwegian – at least in certain settings). In other words, Finnish is evidently
more suitable a language than English for evaluating the respective construc-
tionist / projectionist approaches . . . provided one does not restrict the scope of
observation to the VPs and the NPs, as the intrusion of translated texts in the
debate will emphasize. What should be scrutinized? This will be the topic of
sections 2 and 3.
1.2 Finnic Aspect and resultativity

In the Finnish research on aspect, resultativity has been a central notion, studied
in particular in relation with the rules for choosing the case of the object.
One has discussed in this respect the category of “resultative verbs” (e.g. löytää
‘find’, tappaa ‘kill’, unohtaa ‘forget’, and the corresponding resultative sentences)
has been discussed. The opposed category of irresultative verbs is composed of
verbs which do not normally express the pursuit of a result, e.g. verbs of feeling
(rakastaa ‘love’) or verbs describing an alternative movement (e.g. heiluttaa ‘to
wave’).
One can also distinguish a category of “quasi-resultative” verbs, i.e. verbs
which feature an ongoing state and can occur in a sentence expressing duration
in the same way as irresultative but take nevertheless a total object like resul-
tative verbs. This category includes cognitive verbs like tietää ‘know’, muistaa
‘remember’ as well as verbs denoting relations like sisältää ‘content’, omistaa
‘possess’:
Tiedän asia/n
know:PRS .1 SG thing:ACC . SG
‘I am aware of the thing.’
Kuori sisältää vain lasku/n

envelope:NOM . SG contain:PRS .3 SG only bill:ACC . SG
‘The envelope contains only a bill.’
(T. Itkonen 1976, Leino 1991).
The notion of Aktionsart is also used in several different meanings. It refers

often to the aspectual characteristic of verbal lexems. In Finnic research, the
types of action have been often related to the verbal suffixes (e.g. Penttilä 1963:
533–534).
This classification relies mainly on the four aspectual classes defined by
several theoreticians, e.g. Vendler (1967): state (have, possess, want, like, love. . .),
activity (inherently unbounded – run, swim, push. . .), achievement (punctual
change of state or onset of an activity – recognize, find, cross the border),
accomplishment (non punctual change of state, inherently bounded process –
paint a picture, write a novel, build a house. . .). Such a categorization relying
on the internal actual characteristics of the verbs can be considered funda-
mental for the analysis of aspect in Finnish (Heinämäki 1984).
The most usual pairs telic /atelic, resultative /non resultative, dynamic /static
are generally applied to verbs, sentences and situations, although one can choose
to reserve “telic/atelic” to situations, and “bounded/unbounded” to verbal fea-
tures (ISK: 1430–1432) – see also Depraetere (1995).
In parallel with the numerous studies on FIP, other in-depth investigations
have been conducted regarding the other main Finnic language, Estonian, both
in a grammatical internal and a general linguistic perspective (see Rajandi &
Metslang 1979, Metslang 1994, Tamm 2004a).
Starting from the observation that although grammatical aspect in Estonian
has not developed into a consistent grammatical category – as typologists do not
consider Estonian as a language with fully grammaticalized aspect (Metslang &
Tommola 1995: 300–301) – aspect emerges in the object case alternation, Anne
Tamm has thus recently dedicated several ground studies to the understanding
and representation of the object case alternation for some classes of transitive
verbs. An “outer aspect” and an “inner aspect” are distinguished as the basis
for testing different hypotheses: the perfective-imperfective hypothesis, the re-

sultativity one and the boundedness one for the outer aspect, the telicity hypo-
thesis for the inner aspect. None of the tests gives exclusive definitive results,
which shows that the aspectual interpretation of sentences and lexical aspects
are both involved, and leads to a proposal of analysis in terms of potential telicity
and boundability.
We can retain from this demonstration, partly obscured by the heavy formal-
ism of the Lexical Functional Grammar Framework, the possibility to distinguish
maximally bounded sentences (they describe an event with clear boundaries and
that cannot be continued) from minimally bounded (encoded by a partitive
object) but also that “a boundable transitive verb does not occur without aspect
and aspectual case marking on its object” (Tamm 2006).
A previous study of the same author, starting from the idea of the composi-
tionality of aspect (Verkuyl 1989, 1993; Kiparsky 1998, 2001), had also shown
that object alternation in Estonian – even though Estonian atelic verbs (stative
and activity verbs) cannot more clearly that Finnish ones be divided into “hard
partitive verbs” and “soft partitive verbs” (Klaas 1999) – cannot be regarded as
being dependent on verbal classification only (Tamm 2003).
The ambition of most investigations conducted by Finno-Ugrists in this par-
ticular chapter of syntax-semantics interface – the complex relation of aspect
and the object case alternation – was, at least until the introduction of cognitive
studies (Huumo 2006, 2009), to find criteria for the classification of verbs.
The theoretical basis for this lexical categorization of aspect was clearly settled
in the 1970s in Hakulinen and Karlsson’s reference handbook (Hakulinen &
Karlsson 1979): as opposed to the time of the action expressed by the tempus of
the verb, which belongs to the domain of deixis, aspect has no relation with the
moment or the duration of the action but with its resultativity. Generally con-
sidered a syntactic-semantic category, aspect is mainly related with the nature
of the action expressed by the verb (German Aktionsart), which is a lexical
feature (see Comrie 1976, Dahl 1985) – Hakulinen & Karlsson (1979: 183–185).
Later, fennists have dedicated several studies to the puzzling problem of the
verbs which seem to challenge the norms of perfectiveness vs. imperfectiveness
in correlation with the inherent meaning of the verbs and the aktionsart carried
by the sentence (Kangasmaa-Minn 1984). A classical example is the difference
between Pidän porttia (FIP) lukittuna and Pidän portin (acc.) lukittuna ‘I hold
the gate locked’ is not simply a question of aspect: the first one has a connota-
tion of, that habitualness.
Another debated example is Kannoin koria (FIP) ‘I carried / was carrying the
basket’, that normally – different from Kannoin korin (acc.) ‘I carried the basket
(back, home etc.)’ – descriptive or reserved to a continuous/simultaneous action
(see the French imparfait), but can in a precise context express a perfective even
though durative action, e.g. Kannoin koria loppuun asti ‘I carried the basket untill
the end (helping somebody else)’ – see Kangasmaa-Minn (1993).
Let me add two further angles on the question, which have proved to be
essential for my concerns, related to enunciation theories on the one hand, to
text linguistics on the other.
The first one was launched by L. G. Larsson, an outstanding Scandinavian
Finno-Ugrist who has steadfastly investigated the aspectual role of FIP. His novel
ideas, after comparing the case of the O in the Finnic and Baltic languages, and
in Mordvin, was that one should clearly separate the lexical from the contextual
level, and reserve ‘aspect’ to the contextual level (Larsson 1983: 28ff.). Studying
the role of the Baltic influence on the aspectual system of Finnish (Larsson 1984),
he compares the role played by verb-prefixes in Lithuanian and some particles
in Finnish (which has no prefixes) and specially in Estonian, where terminativity
is expressed by ära ‘away’ (see also Metslang 2001, Tamm 2004b):
(1) Pool oli ju ära söödud

half was:3SG already away eat:PST.PTCP
‘Man hatte schon zur Hälfte abgespeist.’
(from Wiedemann 1973; entry sööma ‘essen’ [to eat])
An equivalent particle is used in Finnish to emphasize the lexical termina-

tivity of some V – pois ‘away’:
(2) Suuryritykset pyrkivät syömään

big-company:NOM.PL endeavor:PRS.3PL eat:INF3
pois pienyrittäjät
away small contractors:ACC.PL
‘The big companies endeavor to swallow the small entrepreneurs.’
Opposing two types of verbs called “irresultative”, Larsson argues that this
term hides an essential difference between the two partitive objects: with a V
like matkia ‘imitate’, the O is unlikely to occur in any case other than the PAR,
whereas with a V like kantaa ‘carry’ the PAR denotes that the circumstances
of the process of carrying are commented upon, and does not make explicit
the end or the result of the act: Kannan taakka/a ‘I am carrying the burden’,
different from Kannan taaka/n (acc.) ‘I carry the burden’ that merely states the
act of carrying. The term “aspect”, restricted to the contextual level, should
in the description be included in a fixed hierarchy (//lexical government of
the V / expression of indefinite number of the O / Aspect//).
In this approach, the FIP is seen as enlightening the subjective perspective

of the speaker/writer, who “describes”, “comments upon” through simply choos-
ing this morphological case.
Another point of view, related to the components of narration, is developed
by M. Leinonen.
She refers to the Slavonic origin of the studies of aspect, initially considered
as characteristic of languages where all the verbs can be divided into two
classes – perfective vs. imperfective. Related to the universal semantic compo-
nents of the action (dynamic/static, telic/atelic. . .), the Russian aspect can thus
be defined as a binary privative opposition (Maslov 1973, 1980): the perfective
aspect signals that the action is taken as an indivisible whole, or as an action
that has reached its limits, whereas the imperfective aspect signals the absence
of these properties (Leinonen 1982). In comparison, the Finnish system has a
much greater flexibility, as even basically durative activities can be made totally
closed by adding a directional adverbial which indicates the end state (see also
Dahl & Karlsson 1975). “There is hardly any limit to the possible modification
from durative action to telic total events, if one can imagine a new state of the
participant produced by the action” (Leinonen 1984: 248). In this perspective,
the non-natural distinctions produced by aspectual means deal with the narra-
tive sequence vs. non sequence, or foreground vs. background. The definition of
perfectivity as “a view of events as a whole, whose completion is a necessary
prerequisite to a subsequent event” and its opposition, the use of non total
forms “which leads to a two way action meaning” (Leinonen 1984: 250) gives a
new insight in the narrative function of FIP (see also Hopper 1979).
1.3 The Finnish Partitive (FIP)

The Finnish Partitive (FIP) has long been considered a core subject of analysis and
debate within Finno-Ugric studies (Sadeniemi 1929, Denison 1957, Itkonen 1975,
Moravcsik 1978, Leino 1991 among others), also in interlanguages (Fernandez-
Vest 1982). Accurately so, as it appears as the pivot of the whole grammatical
structure of the language: although clearly identified as a separate case ending,
it resists attempts to be assigned a syntactic or semantic case function. Accord-
ing to the modern Comprehensive grammar of Finnish, the main characteristic of
FIP is its meaning of unboundedness, a notion which covers different elements:
an indefinite quantity, number or matter, or the fact that the situation is pre-
sented as lacking an end point. Its use as the case of the object is subject to
three alternative conditions:
1. the sentence is morphosyntactically negative, or negative by inter-pretation

(restrictive, dubitative)
2. the sentence is aspectually unbounded
3. the referent of the object is indivisible and is quantitatively indefinite.
Some further remarks, in connection with these three conditions:

– the negativity of the sentence is the strongest criterion, since both aspect
and definiteness are neutralized in it, which in its turn implies that the PAR
is possible with all the verbs, since they can all be denied;
– there are no verbs getting only a total object, but there are many verbs that
get only a PAR object;
– it is impossible to inventory all verbs with a PAR object, since their un-
bounded aspect can be modified in certain circumstances: if the sentence
is positive, the PAR can be motivated by the unbounded aspect of sentence,
or by the quantitative indefiniteness of the NP. Besides, the aspect and the
quantitative indefiniteness cannot in most cases be separated: they are
entangled in many ways.
(see ISK: 887–891, 1538–1540)
In sum, the partitive object of my present corpus will be analyzed primarily

from the point of views of text linguistics (cotext and context) and (speaker’s
and hearer’s) subjectivity of the utterance. I do not intend to propose a new clas-
sification of the Finnish verbs, but I still need the verbs as a primary element of
inventory and selection, and will therefore use the classification available which
best meets my requirements of enunciative criteria beyond the morphosyntactic
features.
2 The aspectual choice of partitive object in

translated texts: Finnish – Sami – Norwegian
2.1 The methodology
The most appropriate consistent work on the choice of the object in Finnish
turned out, among the many studies available, to be a PhD conducted at the
University of Oulu ten years ago. It was adequate by its main goal:
[after placing each verb of the corpus with a nominal object into one of the three aspectual
verb categories /resultative, irresultative or resultative-irresultative/]
“to find out whether the sentence context, the aspect of the sentence and the situation
being described affect the choice of the objet’s case or whether this choice is governed by
the inherent aspect of the verb” (Askonen 2001: 482)
which I will, for my own purposes, reformulate as following:
– To find out whether the choice of the object’s case is governed by the inherent aspect of the
verb, or whether this choice is equally – or primarily – affected by the context, the aspect of
the sentence and/or the situation.
I will leave out some other goals of this PhD, for instance to count the verbs and
objects in the different categories: my own study is not quantitative, which
would require statistics on a larger corpus than the one collected. But this work
was also adequate thanks to its method of classification of the verbs, which was
preferred for three main reasons:
1. the Oulu corpus consisted of written and spoken texts produced in real
language use situations, which guaranteed that many different styles were
represented (informal spoken language, reporting, radio-commentaries. . . ;
written language of fiction and non-fiction). This corpus was large enough
to also guarantee some validity to the classification criteria and the con-
sistency of the quantitative results: the material included 28,076 nominal
objects and 1447 verbs.
2. the analysis was primarily syntactic-semantic, but also involved (morphologi-
cal and) pragmatic aspects. The comparison of a given verb and its object in
different situations showed that aspect in Finnish is a situational feature
that conforms to situational variability, and that aspect is ultimately a prag-
matic concept implicit in the verb, determined by the semantic factors
underlying each situation, which leaded the author to consider that “the
speaker’s or writer’s personal intention is an underlying factor of this kind”
(Askonen 2001: 484).
3. Consequently the classification was based on + and – values:
a. According to the basic rule for the choice of a partitive object, the object
is in a partitive form when the verb is [–decisive change] and the situa-
tion is [–end point]. An inherently irresultative verb causes the object to
be in the partitive case, but an inherently irresultative verb may also have
a partitive object whenever the verb expresses an ongoing activity or is
part of an irresultative infinitive phrase. (This rule also applies to the
choice of the object’s case in a negative sentence.)
b. According to the second rule for the choice of a partitive object, the
object is in the partitive whenever the resultative verb is [+decisive change]
and the situation is [–end point]. The object’s referent in this case is
countable. This is an instance of the quantifying role of resultative verbs.
c. According to the third rule for the choice of a partitive object, the object
of a momentary verb is in the partitive whenever the verb is [–decisive
change] and the situation is [+end point]. (Askonen 2001: 484).
2.2 A corpus between oral and written language
2.2.1 Criteria for the selection of the corpus
2.2.1.1 Comparison of translated and edited oral discourses?

Translated oral discourses were excluded from the first stage: it is practically
impossible to find reliable translations of oral conversations, apart from situa-
tions of (professional or semi-professional) interpretation either simultaneous
or consecutive, where the comparison comes up against many obstacles (Miller
& Fernandez-Vest 2006).
Another possible comparison which was shortly considered was one of
internal contrastivity, i.e. between an oral and written version of Finnish. But
after testing a random sampling of excerpts from an exceptional corpus –
recorded interviews first published in a book by the Finnish Literature Society,
then retranscribed by myself from the original record – which I have used for
other purposes (Fernandez-Vest 1995), this track was abandoned. That comparison
namely turned out to be unsuccessful: the oral vs. written functionality of partitive
objects could not be investigated, as partitives were nearly nonexistent in the
oral version. In general terms, the oral register is characterized in Finnish,
besides specific constructions like initial and final Detachments (Fernandez-
Vest 2004, 2014 in press; Amon 2009), by the use of numerous local cases
instead of the grammatical ones.
Investigating the spatial dynamics of discourse, one can oppose the real,
present or memorized movement “towards” or “away from” the deictic hic et
nunc of the utterance, the “fictive embodied” movement which refers to the
imaginative capacity of cognition, for which the model of “general fictivity
pattern” (GFP, Talmy 1995) was initially elaborated. Finnish is particularly favor-
able for expressing movements //towards / or / from// the human body as a
deictic center: the language plays on the inchoative (movement towards) and
terminative (separation or departure) rections, e.g.
(3) [What had you liked to do / if you had been in good health?]
[ORAL] ehkä tota noin . . . mulla olis ollu . . . tullu aivoihini
(ILL.PL) ajatus mennä teatterikouluun (ILL.SG)
‘Perhaps you know . . . I should have had . . . would have come to my brain
(ILL) the thought to go to a theater school’.
[WRITTEN] olisin ehkä ajatellut kaydä teatterikoulua (PAR)

‘I should perhaps have thought to attend a theater school (PAR).’
(Fernandez-Vest 1995)
The speaker evokes the arrival of an idea into her brain (to come + illative)
in a thetical order (V–SBJ) motivated by the event-character of the process. The
edited version reverses the movement of the thought (with the speaker as the
source – syntactically the subject – of the verb ‘to think’).
From this results that the number of partitive objects is much smaller in
impromptu speech than in written texts; most of them occur in negative sentences.
This observation of an oral corpus thus supports the proposal that a localist
interpretation does justice to the Finnic (and Samic) case systems in oral Finnish
(not to speak of the “automatic” local rections, like kysyä + ablative ‘ask some-
body’, pitaä + elative ‘like something/somebody’) and suggests another type of
hierarchy than the official one: local cases, cognitively more salient, are neither
more pheripheric nor less “structural” than grammatical ones (Cienki 1995, Blake
2001, Fernandez-Vest 2010b). But the mere fact that partitives were very rare in
the oral version disqualified the corpus for this present study. In the authentic
oral conversation, partitive objects occurred namely over 90 % in negative utter-
ances, which fall outside the scope of this study.
The following tests concerned external contrastivity.
2.2.1.2 A preliminary study: translations Finnish-Sami and Sami-Finnish

The verbs of the examples chosen have been placed in one of the three aspec-
tual categories:
RES (Resultative)
IRS (Irresultative)
RES-IRS (Resultative-Irresultative).
Their description is based on the semantic classes defined by Askonen 2001:

Activity V, sub-categories Action V, Moving or Motion V; Mental V, sub-categories
Perception V, Reaction V (Physical and psychic emotion, the subject’s attitude),
V of thought, Communication V.
As usual in text linguistics, I gloss only the parts of the examples relevant
for my demonstration.
The corpus of translations was first intended to be equally bilateral, starting
from a novel originally written in Finnish.
From Finnish to Sami

Corpus: Mukka 1974, 2008.
A. The translator simply drops the nuances brought about by the FIP, ex.
(4) Fi. Hän istui aivan liikkumatta, mutta kun olin usean minuutin ajan
tuijottanut häntä ikkunasta, hän käänsi päätään ja katsoi suoraan minuun.
‘She sat completely motionless, but when I had several minutes stared at
her (PAR.SG) from the window, she turned her head (PAR.SG) and looked
directly at me.’
Sa. Son čohkkái áibbas lihkaskeahttá, muhto go ledjen máŋggaid minuhtaid

gaiván su lássaráigge, de son jorggihii ja geahčastii munnje.
• Finnish: The first transitive V is IRS, Mental V (Perception), PAR O.

The 2nd transitive V is a RES Activity V (Action), a classical example studied by
Fennists: ACC O would be normal (usual), but PAR introduces an impression of
carelessness or improvisation – not a decisive movement (see Kangasmaa-Minn
1978).
• Sami: ‘she turned’, intransitive (= no O, no mention of ‘‘head’’ or ‘‘back’’).
B. The aspectual values are rendered by a different construction, for instance

the duration by a progressive form, ex.
(5) Fi. Join monta pannullista laihaa

drink:PST.1SG many:PAR.SG pot:PAR.SG thin:PAR.SG
kahvia ja sulatin suussa/ni
coffee:PAR.SG and melt:PST.1SG mouth:INE.SG.POSS
vain yhden sokeri/palan kuppia kohti.
only one:ACC.SG sugar/lump:ACC.SG cup:PAR.SG for
‘I drank many pots of thin coffee and melted in my mouth only one
sugarlump per cup.’
Kirjoitin romaani/ni viimeistä

write:PST.1SG novel:GEN.SG.POSS last:PAR.SG
lukua, kohtaa jossa
chapter:PAR.SG passage:PAR.SG which:INE.SG
käsitellään Verneri Krookin
deal:PRS.PASS Verner:NOM.SG Krook:GEN.SG
murhaan liittyviä
murder:ILL.SG be.connected:PRS.PTCP.PAR.PL
asioita ja Heikki Väkkäräistä
thing:PAR.PL and Heikki:NOM.SG Vähkäräinen:ELA.SG
joka tuon teon teki.

who:NOM.SG that:DEM.ACC.SG action:ACC.SG do:PST.3SG
‘I was writing the last chapter of my novel, the passage where one deals
with things related to the murder of Verner Krook and with Heikki
Väkkäräinen who did that action.’
• Finnish: juoda ‘to drink’, RES-IRS, here IRS as the O is partly bounded (pannul-
linen, the content of a pot is not a count name), sulattaa ‘to melt’ here RES with
a restricted / delimited result: only one sugar lump.
2nd sentence, kirjoittaa ‘to write’ RES-IRS here IRS: the chapter is a work in
progress, relative clause V käsitellä ‘to treat, deal with’, IRS. Final V tehdä ‘to
do’, here RES.
• Sami: identical construction in the 1st sentence, but personal form in the 2nd ‘I
dealt with’.
2nd sentence, progressive form (Gerund II) where there was a PAR in Finnish.
The main clause of the second sentence can be compared with another one
in a following chapter:
(50 ) Fi. Kirjoitin puolen tunnin ajan

write:PST.1SG half:GEN.SG hour:GEN.SG time:ACC.SG
ja sain valmiiksi liuskan yhden
and get:PST.1SG ready:TRAN.SG one:ACC.SG sheet:ACC.SG
‘I wrote half an hour and I got ready one sheet’
Sa. Čállen su birra diibmobeale, oktiibuot ovtta árkka
• Finnish: the first V, kirjoittaa ‘to write’ RES-IRS is here RES: there was a result,
expressed by the second V (RES, Activity V, Moving) + predicative adjective in
the translative, ACC O.
• Sami: shorter, more synthetic formulation: ‘(I wrote about her for half an
hour), together one sheet’.
From Sami to Finnish

The choice was still more restricted: I selected the prose texts of a bilingual
anthology (Skabmatolak, 1973), completed by a novel written three decades later
(Vest 2005, 2006).
• Two regular types of transfer devices were detected:
1. A specific morphological form of the Sami V is rendered by the FIP:

(6) Sa. Máhte-Máhtte jugistii káffegohpa ja ráhkkanišgođii Pentte maŋimuš

hášiid rádjat. () Dalle heastaáigge Máhtte lávii rádjat maŋimuš hášiid
borgemánu gaskkamuttos, dál áiggui geargat vel suoidnemánu bealde.
(Vest 2005)
Fi. Matti hörppäsi kahvikupin tyhjäksi ja lähti korjaamaan Pentan viimeisiä

haasioita. () Silloin hevosaikaan Matilla oli tapana korjata viimeiset haasiat
elokuun puolivälissä. (Vest 2006b)
‘Matti tossed off the cup of coffee (ACC.SG) empty (TRAN.SG.) and went out
to gather Pentti’s last hayricks (PAR.PL). () At the time of horses Matti used
to gather the last (ACC.PL) hayricks (ACC.PL) in mid-August’.
• Finnish: hörpätä ‘to empty quickly’, korjata ‘to gather, pick up, take away’, V
of Action-movement, RES-IRS, used here with the two different constructions:
1. Matti went out in order to gather + PAR (an objective, but no guaranted
result); 2. [remembering the past] Matti used to gather + ACC (global, accom-
plished perspective)
• Sami: 1. jugistit ‘to drink quickly’, derivative V of juhkat ‘to drink’, the descrip-
tive aspect is transferred in Finnish on an additional noun (adjective) in the
TRANslative; 2. no difference of construction of the V rádjat ‘to gather’ in the
two sentences, apart from what precedes the 1st ‘gathering’ – an inchoative V
(ráhkkanišgoahtit ‘to begin to prepare’), ‘he began to prepare/equip himself in
order to gather’, rendered in Finnish by a simpler ‘he went out and gathered’ +
PAR O. The inchoative aspect is thus transferred on the PAR in Finnish.
2. A specific syntactic form attached to the Sami verb, i.e. the durative value of
the Sami progressive form (Gerund II), is rendered by a PAR:
(7) Sa. Piera goavrái návetuvssa ovddas. Čoarverieban lei buviheamen su. (. . .)
Dávvet čuoččui dobbelaččas ja fuoikkui. Son lei čuovvumin, mo biro borai su
kránnjá.
(Paltto 1971, Skabmatolak 1973)
Fi. Piera makasi navetan oven edessä. Sarvikettu oli kuristamassa häntä.
(. . .) Taavetti seisoi vähän kauempana ja voihki. Hän seurasi kun paholainen
söi hänen naapuriaan.
‘Piera lay in front of the cowshed door. The horned fox was eating (INF3.
INE) him (PAR.SG). () T. stood a little further away and groaned. He
followed when the devil was eating his neighbor (PAR.SG)”
• Finnish: kuristaa ‘to strangle” V of Action, RES; syödä “to eat” V of prehension,
RES-IRS
• Sami: in the two sentences (1) and (4) there is a progressive form (Gerund 2).
3. A third device consisted in applying strictly the peripheral rules attached to

negation in Finnish, partly depending on the interpretation, e.g. the diffi-
culty of a task conveyed by an adjective, before an infinitive clause:
(8) [In Heaika’s eyes, Ivvár was a bit strange]

Sa. Das lei viehka váttis fáhtet gitta. Sáhtii leat dasto áigebotta jaska, dušše
geahčai apmasit sutnje ja ain moddjii () de čilgii vuđolaččat uhcimušge ášši.
(Vest 2005)
Fi. Hänestä oli vaikea saada otetta. () Saattoi sitten olla minuuttikaupalla
vaiti, katseli vain omituisesti vierastaan ja hymyili (), niin selitti perusteelli-
sesti pienimmän+kin asian. (Vest 2006b)
‘On him was difficult to get a grip (PAR.SG). He could then be for minutes
silent, watched only strangely his guest (PAR.SG) and smiled (), he
explained thoroughly the smallest (ACC.SG) thing even (ACC.SG.+DIP).’
• Fi. saada ‘to get’ RES but here uncompleted action, negative adjective – PAR;
katsella ‘to watch’, diminutive-frequentative, PAR; selittää ‘to explain’, RES-IRS,
resultative in this cotext, ACC.
• Sa. all Os are normally in the ACC-GEN, apart from the O of geahčai ‘looked at’
which is in the directive (‘looked towards him’).
The first Sami sentence relativates the difficulty – viehka váttis ‘rather diffi-
cult’ –, and this could have motivated the choice of an ACC O.
In sum, the comparison Finnish-Sami gave a relatively poor result: either
the nuances introduced by the FIP were simply dropped (4), or the aspectual
values were nearly automatically rendered by a construction of equivalent
value, e.g. the duration by a progressive form (5). In the first case, the PAR was
neglected, in the lack of a direct equivalent, in the second one it was replaced
by a structure of equivalent aspectual value.
The comparison Sami-Finnish produced a wider range of results: the FIP
was used as the equivalent of specific Sami morphological (6) or syntactic (7)
forms, but it also reflected the subjective interpretation of the translator, who
could choose to apply the secondary rules attached to Finnish negation (8).
The selection of the main corpus was rationalized according to these obser-
vations: it was considered more productive to concentrate exclusively on the
translation from Sami to Finnish, as this was hypothesized to give the translator
larger possibilities of choice regarding the PAR. Besides, the conditions of the
translation process were more strictly taken into account. Translation by two
individual translators, native speakers of Finnish and Norwegian and experi-
enced translators were preferred to a collective translation, and an entire novel
was preferred to anthology excerpts.
The analysis will be divided into two parts:
– a comparison in different contexts of two types of verbs which have a spe-
cial relevance for the functionality of PAR. Two verbs are RES and allow to
see whether any leeway is left to the translator, two verbs are RES-IRS, and
allow therefore to analyze whether or not the translator uses his room for
manœuvre, and in what conditions;
– an evaluation of the narrative function of PAR vs. ACC in two excerpts of
texts.
2.2.2 The main corpus: translation into Finnish (and Norwegian) from Sami
(Vest 1988, 1989, 1990)
2.2.2.1 Two prototypically resultative verbs in various contexts and cotexts
• MUISTAA ‘to remember’, Mental V, V of thought, RES
(9) – Alat/han sinä/kin oppia kun muistat isän neuvot.

Begin (PRS.2SG/DIP) you (NOM.SG)/too (DIP) learn (INF I) when remember
(PRS.2SG) father (GEN.SG) advice (ACC.PL)
‘Why, you begin you too learning when you remember your father’s advice!’
(16.77)
The verb has here its full function of resultativity: the remembered advice
referred to are precisely the ones which the father gave his son before sending
him on a second selling trip (of home-made Sami shoes to Norwegian shop-
keepers): ‘You should always haggle, bargain!’. But the rendering of the O as
total/definite is not automatic: there are no special cues in the Sami sentence
inducing this interpretation of definiteness. Only the knowledge the reader/trans-
lator has of the previous context – the father’s anger after the first (failed) selling
expedition and the authoritative instructions he gave before the second – can
lead him to choose an ACC: Sa. go muittát áhči rávvagiid ‘when you remember
your father’s advice’ could, without this longer context (the whole chapter), be
interpreted vaguely as ‘some of your father’s advice’. The Norwegian translation
expresses also clearly this restrictive object through putting the verb ‘to listen to’
in place of the verb ‘to remember’: bare du høre på far sine råd ‘only you listen
to your father’s advice’.
(10) Muistan myös niitä aikoja, jolloin maailman kovuus alkoi rikkoa lapsuuden
kauniita kuvia. (1.8)
‘I remember also those (PAR.PL) times (PAR.PL) when the world’s hardness
began breaking my childhood’s beautiful (PAR.PL) pictures (PAR.PL).’
The Sami sentence in itself gives no special cues: Muittán maiddái daid
áiggiid, goas could mean (I remember (just) those times when. . .), pointing pre-
cisely at a distinct period. But the co-text is decisive in Sami and Finnish: this
sentence is integrated in a series (list), which begins with ‘to the mind rose
the happy times of my earliest childhood’ (first sentence of the paragraph). The
Norwegian translator has even chosen to emphasize this series by repeating
the thetic construction in this second sentence: Fram steig også tida da livets
hardhet (forward stepped also times (indefinite) when the hardness of life () ).
This choice is ultimately validated by the immediately following cotext: no
mention of any precise event which would justify understanding that definite
times or periods were in question.
• SAADA ‘to get, obtain, receive’, Activity V, Moving or Motion V, RES
(11) Neljäs poika sai nimekseen Toivo Veijo Tapio. (4.17)

‘The fourth son got for-his-name (TRAN) Toivo Veijo Tapio (NOM)’
Basic construction of this RES V: the O is in its bare form (NOM), as is the
rule with a proper name in Finnish. Sami has an identical construction: oaččui
namman (essive, the equivalent of the Finnish translative). The Norw. SUBJ is the
agent of an other RES V: han kallte sin søn for TVT ‘he called his son (as) TVT’.
(12) Olemme yrittäneet ongella saada kalaa (14.68)

‘We have tried with a hook and line to get some fish (PAR)’
The Finnish lexical structure and word order reflect faithfully the Sami sen-
tence: in both languages, the instrument is emphasized by its projection in front
of the V. Norwegian moves the instrument to the end of the sentence, as the
main rheme, and changes the indefinite O into definite: ta fisken på sluk (take
the fish with (/on) a hook and line).
(13) Ompeleminen todella sai nyt uutta vauhtia. (16.73)

‘The sewing really got now a new (PAR) speed (PAR)’
The PAR O expresses an abstract element, uncountable. The translation from

Sami is literal. The Norwegian VP has an indefinite O without an article, in fact a
lexis ( fikk fart (got speed)).
(14) Yhä vieläkin minua on vaikea saada Tenolle. (2.9)

‘Still even today me (PAR) is difficult to get onto the Teno’.
The PAR O is due to a semantically negative predicative (adjective ‘diffi-

cult’); identical Sami construction: mu lea váttis oažžut Deanu ala. Norwegian,
more explicit as to the exact meaning of the sentence, replaces the place-name
by a noun: vanskelig å få meg med på elva ‘difficult to get me [with] onto the
river’. The context was about a son whose father’s authoritative demands of
round-the-clock (salmon) fishing eventually put him righ off fishing.
2.2.2.2 Two resultative-irresultative verbs: what triggers off the choice?
• TEHDÄ ‘to do, make’, Activity V, Action V, RES-IRS

Due to its broad scope, this V occurs in many sentences. The two following
examples are intended to illustrate its two main aspectual values.
(15) Kaiken mitä hän teki, hän teki parhaan kykynsä mukaan.
‘All (ACC) what (PAR) he did, he did the best way he could.’ (2.9)
Finnish has an ACC O expressing the accomplishment of the verb. Sami and
Norwegian have an identical construction.
(16) – Eivät ihmiset koskaan kysy kauanko teit sitä, mutta sitä ne kyllä ruukaavat
kysyä miksi tuo on noin ja tämä näin. (21.105)
‘People never ask for how long you did it (PAR), but this they indeed use to
ask why that is like that and this is like this.’
The Sami original selects another verb, bargat ‘to work’ – guhkágo barget
dieinna ‘(for how long) you worked with this (comitative)’. The Norwegian trans-
lation transfers the meaning of this V to the subject NP of a depersonalized
sentence: kor lang tid et arbeid tar ‘how long a work takes’ – and elevates thus
the (indirect) interrogative clause to the level of a general judgement. The use of
the FIP for expressing the duration of the ‘deed’ has the advantage to maintain
the original V, and thereby preserve the enunciative tonality of the original Sami
sentence: even though general in its scope, this remark does have an addressee
in the second person, and is part of the father and son’s (labored) dialogue.
• TÄYTTÄÄ ‘to fill’, Activity V, Action V, RES-IRS

Although more specialized than the preceding V, the action of “filling” can also
be considered with two different aspects.
(17) Seudun ihmiset tapasivat tulla täyttämään papereitaan isälle. Isä oli
viitseliäs täyttämään ihmisten hakemuksia. (22.109)
‘The people of the neighborhood used to come to fill (inf.3. ILL) their
papers (PAR.PL) to father. Father was assiduous to fill people’s applications
(PAR.PL)’.
The V is different in the two sentences from the point of view of Actance,
semantically factitive/causative in the first, simply active in the second, but it
has exactly the same form in both, and it takes identically a PAR O, which re-
flects the progressive unaccomplished process. The Sami original has a different
form: morphologically (suffixed) factitive in the first – deavddi/hit ‘to get filled’
—, basic in the second: deaddit ‘to fill’. Norwegian distinguishes also two voices:
passive in the first (papirer de skulle ha utfyllt ‘papers they should get filled’),
active in the second (å fylle dem ut ‘to fill them’).
(18) Tunsin miten viha alkoi vähitellen täyttää minut. (14.65)

‘I felt how hatred began gradually to fill me (ACC)
The Sami sentence has a derived inchoative V and a different word order in the
rhematic part: Dovden mo vašši deavdi/gođii mu gulul gulul ‘I felt how hatred
began to fill me little by little’. This final adverbial, intensive intensified by
repetition as well as by its placement (the main rheme), gives the sentence, in
spite of its progressive meaning, a connotation of accomplishment: the filling
of the speaker/main figure by hatred towards a father who constantly comments
upon his hard work with negative and despising remarks will take some time,
but it is doomed to come to an end point. This is rendered in Finnish by the
ACC O, unexpected for a double reason: the presence of an auxiliary ‘begin’
and of an adverb of progression. The Norwegian translator has perceived this
announced accomplishment in the Sami word order, and rendered it with a set
expression: (I felt) hvordan jeg sakte men sikkert fyltes av hat ‘how I slowly but
surely get filled of hatred’.
2.2.2.3 The narrative function of FIP

This function can be observed in two types of texts, which differ by their respec-
tive degree of description vs. narration.
• Description of mental activity
(19) Mitä lienee mies ajatellut, sitä on paha mennä sanomaan.

Ehkä hän mietiskeli tulevia toimia.
Ehkä häntä oli jo alkanut painaa suvun raskas luonne.
Hetket jolloin oli joutunut kierimään pohjamudissa, ajatus pyyhkäisi
muistojen joukosta pois. (26.140–141)
‘What (PAR.SG) may the man have thought, that (PAR.SG) is difficult to tell.
Maybe he meditated future (PAR.PL) actions (PAR.PL).
Maybe him (PAR.SG) had already begun-to press of the family (GEN.SG) the
heavy (NOM.SG) nature (NOM.SG).
The moments (ACC.PL) when he had to roll in the bottom-mud, the thought
(NOM.SG) wiped-out from the rest of the memories.’
The verbs of the first three sentences are all IRS: it is consequently normal
that each of them has a PAR O, even if it could be explained also by their
morphology – ajate/lla ‘to think’ and mietiske/llä ‘to meditate’ have both a
diminutive-frequentative suffix – or by their semantic value: both are mental
Vs, and it has been shown on the basis of the huge Oulu corpus that “mental
activity rarely reaches an end point” (Askonen 2001: 488). The IRS nature of the
V of the third sentence is also strengthened by the inchoative auxiliary alkaa, ‘to
begin’: the aspectual case of the O of an infinitive V is primarily determined by
independent criteria, but is has also some connection with the general aspect of
the sentence – why infinitives subordinated to a verb expressing the beginning
of the action or an endeavor, e.g. yrittää ‘to try’, usually have a PAR O, as the
action is unbounded (see ISK: 890).
Interesting is however that the sequence of sentences with a PAR O is sud-
denly interrupted by a longer sentence with an ACC O, also marked by a strongly
thematized word order: the fact that the O NP, determined by a long subordi-
nated clause, is fronted, emphasizes the action expressed by the V of the main
clause. In principle, pyyhkäistä ‘to wipe’ belongs minimally to the RES-IRS class
like the basic V pyyhkiä ‘to wipe, strike’, and maximally to the IRS class like
other derived momentaneous Vs (e.g. kysäistä (<kysyä) ‘to ask in passing’) but
is here bounded by the particle pois ‘away’, which selects its RES feature. String-
ing together all three transitive V sentences, where a series of IRS constructions
ends up in an abrupt RES construction with a thematized O, elaborates a textual
strategy: the alternating interplay of PAR O and ACC O is a decisive mechanism.

In this part of the paragraph, the PAR Os are dominating – they contribute to
build up a descriptive background of loose mental activity suddenly interrupted
by an Action V, even though expressing the activity of “the thought”. In the
Sami original, the effect of hesitation and wandering of the mind is insured, as
in Finnish, by the morphology of the V (potential mood – Maid leš jurddašan
(What may (potentially) have thought) – a device unavailable in Norwegian –
and the repetition of the adverb “perhaps” (ganske, with variants in Norwegian).
The sudden interruption and change from an uncertain mental activity to a
decisive activity of the thought is marked in the same way by the fronting of
the O: Sa. Bottuid – jurdda sihkasti eret (The moments – the thought wiped
out); but this fronting of the O turns the Norwegian sentence into a passive
one (a diathetic process common to non-flexional Indo-European languages):
Øyeblikkene – ble visket ut (The moments – were wiped out). Apart from the
greater syntactic flexibility of the two Finno-Ugric languages, one can notice
that this strategy of foregrounding is also signaled/announced by the persisting
role of the PAR O in the previous sentences.
Is the strategy as pregnant in a text where the narration of events is primary?
• Narration of a hunting story
(20) [During a cloudberry trip, my father began to tell us about a fellow]

1. joka löysi ahman jäljet ja lähti hiihtämään niiden perään.
‘who came across the tracks (ACC.PL) of a wolverine and set off on his
skis in pursuit’
2. Kun ahma huomasi ihmisen seuraavan, se paransi vauhtia/an.
‘When the wolverine discovered that a human was after him, he increased
his speed (PAR.SG/POSS).’
3–4. Ei tiennyt millaisen pedon hän oli nyt saanut peräänsä. Ei ollut mies ensim-
mäistä kertaa ahmaa hiihtämässä.
‘He didn’t know what kind of (ACC.SG) predator (ACC.SG) he had got
after him (ILL.SG.POSS). This man was not for the first time wolverine
(PAR.SG) hunting (INF III. INE.SG)’
5. [To the children] – () silloin yritän kertoa sen loppuun.
‘Then I’ll try to tell it (ACC.SG) to the end (ILL.SG).’
6. [The youngest children] – Nyt sinun pitää jatkaa sitä kertomista.
‘Now you have to continue that (PAR.SG) story telling (PAR.SG).’
7–8. Isä pysähtyy siihen ja alkaa kääriä rauhassa sätkää. Odotamme jatkoa.
‘Father stops there and begins to roll calmly a cigarette (PAR.SG.). () We
wait for the continuation (PAR.SG).’
[The man has just begun to reach the wolverine, but the animal dis-
appears into the mountain]
9–10. Lopulta ahmaparka tekee virheen. Se jättää turvallisen vaaran.
‘Finally, the poor wolverine makes a mistake (ACC.SG). He leaves the safe
(ACC.SG) mountain (ACC.SG).’
11. Kertomus saa uutta vauhtia.
‘The story gets a new (PAR.SG) speed (PAR.SG).’
12. [The wolverine switches back behind a mountain peak.]
13. [An eagle has set its claws into the wolverine] ja yrittää saada sen
ilmaan.
‘and tries to get him (ACC.SG) into the air (ILL.SG).’
[The man fires off a couple of shots, and both animals fall to the ground.]
14. Sillä miehellä sitä tuuria, ihmettelee Toivo.
‘That guy had that (PAR.SG) great luck (PAR.SG), wonders Toivo.’ (= What
great luck that guy had!) (17. 84–87)
General commentary: this story – a modern variant of the old orally trans-
mitted Sami tales – is mixed up with a description of the storyteller’s and his
children’s attitudes – which is cut off here, apart from 7. And 8.
Some remarks about the ordinarily puzzling verbs of this story:
a. the RES Vs followed by a PAR O.
The unique occurrence is 11., where the O ‘speed’ is not assimilated to a
general term or integrated in a lexis (see (13)) but indivualized by an adjec-
tive; it corresponds to the indefinite NP of the Indo-European languages.
b. the RES-IRS Vs and the condition of choice of the O case.
In 2. parantaa ‘to improve, increase’ is obviously unbounded: the improve-
ment of the speed is not decisive, and will have to be repeated several times
before the unexpected end of the chase.
c. What about the IRS Vs? They are all normally (/normatively) followed by
a PAR: inherently atelic Vs (6. jatkaa ‘to continue’, 8. odottaa ‘to wait’) or
involving a duration or a descriptive movement (7. kääriä ‘to roll, wrap’).
From the point of view of syntax, one can notice that the bounding operated
by the adjunction of a local adverbial is determinant: it is automatic in 5. where
the RES V is indeed followed by a lexeme of terminativity with an internal local
case of direction (loppuun ‘to the end’ ILL), but even in 13., where the dominant
RESultativity of the infinitive V could have been counterbalanced by the un-

boundedness of the introducing V yrittää (see (19) above), as it was in (14) by
the negativity of the adjective vaikea ‘difficult’.
The sole at first sight puzzling sentence could be 3. The unboundedness of
the main V tietää, an IRS (mental) V of thought, is in principle strengthened by
the negative V (ei + past participle ‘he did not’), but this negativity is obliterated
by the boundedness of the V saada ‘to get’ in the subordinate clause: /‘what
kind of’ (adjective) + ‘predator’ (noun)/ is not the O of the main clause but
of the complement clause, the V of which generally determines the O case in
Finnish. Saada can in certain circumstances get a PAR O (as we saw in (12)–
(14)), but this possibility is cancelled by the presence of a local adverbial: the
ACC O is the O of the subordinate clause introduced by an interrogative adjective
(millainen ‘what kind of, what sort of’) and bounded by a local adverbial ( perä
‘rear, back, end’ > perä/än(sä) ‘after, following’), actually a postposition formed
of a still productive lexeme (and a possessive suffix), which makes it a stronger
spatial marker than a usual adposition (see Hagège 2009, 2010).
2.2.2.4 PAR Os vs. ACC Os as a discourse strategy

In other words, the occurrences and functions of the PAR Os in this text are not
more unusual than the ones we saw in previous examples. The main interest of
these PAR Os is their structuring power, beyond the sentences, in the globality
of the text. Two main episodes of the hunting adventure (on skis!) are marked
by an initial action of the man (1. He discovers the tracks of the animal), and
by the wolverine’s fatal error (9–10. He makes a mistake, and leaves the safety
of the mountain). In both cases, the verbs have a total O: 1. löytää ‘to find’, RES
Motion V, 10. tehdä ‘to do’, RES-IRS Action V. Between these two episodes, the
verbs are IRreSultative and have PAR Os. The suspense of the story is thus
supported and conveyed ahead by the longer, somewhat hesitating descriptive
sentences, which have in common their aspectual IRreSultativity signaled by a
PAR O.
On the whole, PAR Os are consequently more numerous than ACC Os,
which indicate a decisive turn in the core events related – admittedly assisted
in this task by a few intransitive verbs (see Note 13). And to conclude the story,
the PAR has the last word: it marks in 14. an exclamation, which is also one
of the functions of the PAR, especially in connection with greetings as Os of an
implied verb of wish.
What cues in the Sami original?

There are very few cues apt to motivate the choice of a PAR O vs. ACC O. The
structuring and sequencing of the narrative episodes are mainly indicated by
Discourse Particles – still numerous although declining as other spatial cogni-
tive markers in the typological evolution from oral to written Sami (Fernandez-
Vest 2005a, 2009a, 2010c), and some adverbs: 2. (When we saw the man) de dat
dieđusge buoridii ‘then (DIP)+ he of course (adv.) improved’. In fact, the most
obvious contrastive devices for distinguishing different rythms and tempos are
syntactic (short sentences of happening vs. longer descriptions), and above
all lexical: the sudden and fatal break in the otherwise repetitive wolverine’s
run is expressed by a short intransitive verb: 9. Geatki riehpu feile ‘The poor
wolverine failed’ – V feilet, less usual than feailla váldit ‘mistake-to-take’. Actu-
ally, the Sami’s lexical preference is still more visible in sentences where Finnish
has an ACC O, justified (/imposed upon) by a bounding adverbial: 13. ja viggá
fidnet ‘and endeavours / makes an effort to get him’, where Sami could have
used the weaker geahččalit ‘to try’ (as in 5.). The same lexical tendency is
noticed in Norwegian: 13. strever med å løfte jerven ‘endeavors (with) lifting the
wolverine’. This tendency is probably strengthened by the need to compensate
the stiffness of adpositions in place of (Finno-Ugric) local cases: 5. prøve å ta
resten ‘try to take the rest (of the story)’, instead of ‘to tell until the end (Fi. ILL)’.
This necessary “lexicalization” affects indeed more directly the sentences
whose equivalents require an ACC O in Finnish, but the global result is that the
devices used by non-PAR languages – both a Finno-Ugric (Sami) and an Indo-
European (Norwegian) – for structuring the text are more punctual, less inbuilt
in the information structuring and therefore stylistically less subtle than the
various possibilities offered in Finnish by the Object case alternation.
3 Finnish Partitive Object in the light of

translations
The starting point of this study was not strictly grammatical, but rather an updated
discussion about the new insights that a semantic-cognitive approach brings to
the conception of aspect. Furthermore, the main work which was referred to for
borrowing an aspectual classification of Finnish verbs had clearly demonstrated
that although primarily syntactic-semantic, the aspectual choice of Object case
alternation also involves pragmatic features. In addition to that, I insisted on
considering the role of Object partitives within the structuring of texts/discourses.

And on top of all, the different approaches should be evaluated through trans-
lations.
No need to say that, taking into account all these ingredients did not guaran-
tee to obtain a hand-to-mouth recipe. In other words, I am not sure that this
study adds any decisive evidence to Askonen’s optimistic conclusion: “When
the activity expressed by the verb causes or has caused a change in the referent
or location of the object and the situation described by the sentence has reached
an end point from which it cannot continue unchanged, the object is in the
accusative case, while otherwise it is in the partitive” (Askonen 2001: 485).
But one can definitely consider that aspect is in Finnish a pragmatic concept
implicit in the verb, which is determined specifically by the semantic factors
underlying each situation – and by the morphosyntactic options intentionally
exploited by the speaker. I claim thus that this study brings some new light on
which cognitive, situational and textual criteria limit the flexibility of Finnish
regarding its Object case alternation. The speaker/writer’s subjectivity put forward
by some researchers (see Larsson above) could be obscured by the fact that a
second subjectivity is superimposed on it: the translator’s. But the translator’s
perception and interpretation of the speaker’s intentions are essential underly-
ing factors. Besides, the translator’s choice helps to bring out the potentialities
offered by the FIP. In this sense, translations (sometimes excluded from com-
parative/contrastive linguistics as “subjective” – but who could pretend that
language use in general is totally “objective”?), enriches the knowledge we had
of FIP and its multiple dimensions. Another controversial bias traditionally con-
veyed by syntacticians (and some typologists) which this study challenges is
the following: “what is important in a language is what the speaker cannot be
without saying / is obliged to say”. Partitive O is, on the contrary, interesting
when it is not obligatory – even in positive sentences – i.e. specially with RES-
IRS verbs that allow for case variation, which implies an active choice from the
speaker, taking into account the context and co-text. Both verbs with a broad
and a narrow scope are concerned, and can behave differently according to the
context and/or co-text (ex. (15)–(18)). Another significant angle is the text/
discourse strategy, independent of the case most often taken by the verbs: can
one go so far as hypothesizing that the choice of verbs most frequently allowing
a privileged PAR vs. ACC complements could have been primary in the elabora-
tion of a text strategy in (18) and (19), prior to the unitary lexical meaning of the
verbs? At least, FIP has indeed proved a useful means for structuring the text,
superior to many other less flexible devices.
Last but not least, the relevance of the notion of “resultativity” traditionally
attached to FIP can also be questioned: other notions put forward by Cognitive
linguistics could be referred to in this context, for instance the progression via
“phase meanings” (Tommola 1984), Talmy’s notions of synoptic perspectival
mode (with global scope of attention), as opposed to a sequential perspectival
mode (with local scope of attention) – see Huumo (2009) – or the relation
between language and thought, as reflected by the morphosyntactic vs. lexical
expression of semantic categories in neighboring languages (Fernandez-Vest
2009b). This opens another perspective, based on information structuring, which
I have only tackled here (hinting at the change of word order – thematized or
rhematized objects or adverbials – in several examples), and which will be the
next stage of study of Finnish Partitives in translated texts (see Fernandez-Vest
2012).
References
Amon, Marri. 2009. Les détachements finaux en estonien oral: Productivité et fonctionnement.
In M.M. Jocelyne Fernandez-Vest & Danh Thành Do-Hurinville (eds.), Plurilinguisme et
traduction, des enjeux pour l’Europe – Multilingualism and translation, challenges for
Europe (Grammaire & Cognition 6), 151–164. Paris: Editions L’Harmattan.
Askonen, Ebba. 2001. Objektin aspektuaalinen sijanvalinta [Aspectual choice of the object
case] (Suomen ja saamen kielen ja logopedian laitoksen julkaisuja – Publications of the
Department of Finnish, Saami and Logopedics 19). Oulu: Oulun yliopisto.
Blake Barry J. 2001. Case. 2nd edn. Cambridge: Cambridge University Press.
Cienki, Alan. 1995. 19th and 20th century theories of case. A comparison of localist and cogni-
tive approaches (Historiographia linguistica 22), 1–2, 123–162. Amsterdam: Benjamins.
Comrie, Bernard. 1976. Aspect. An introduction to the study of verbal aspect and related
problems. Cambridge: Cambridge University Press.
Dahl, Östen. 1985. Tense and aspect systems. Blackwell: Oxford.
Dahl, Östen & Fred Karlsson. 1975. Verbal aspects and object marking: a comparison between
Finnish and Russian (Logical Grammar Reports 17). University of Göteborg: Department of
Linguistics.
Dahl Östen & Maria Koptjevskaja-Tamm (eds.). 2001. The Circum-Baltic languages. Their
typology and contacts: Vol. 2. Grammar and typology (Studies in Language Companion
Series (SLCS) 55). Amsterdam: Benjamins.
Denison, Norman. 1957. The Partitive in Finnish (Annales Academiae scientiarum fennicae B
108). Helsinki: Suomalainen Tiedeakatemia.
Depraetere, Ilse. 1995. On the necessity of distinguishing between (un)boundedness and
(a)telicity. Linguistics and philosophy 18(1). 1–19.
Dowty, David. 1979. Word meaning and Montague Grammar. Dordrecht: Reidel.
Fernandez-Vest, M.M. Jocelyne. 1982 [1977]. Le finnois parlé par les Sames bilingues d’Utsjoki-
Ohcejohka (Laponie finlandaise). Structures contrastives, syntaxiques, discursives [Finnish
spoken by bilingual Sami in Utsjoki-Ohcejohka (Finnish Lapland). Contrastive, syntactic,
discourse structures], (L’Europe de Tradition Orale 1). Paris: SELAF.
Fernandez-Vest, M.M. Jocelyne. 1995. Morphogenèse orale du sens: De l’espace des langues
aux objets de discours. In M.M. Jocelyne Fernandez-Vest (ed.), Oralité et cognition:
invariants énonciatifs et diversité des langues. Intellectica 1(20). 7–53.
Fernandez-Vest, M.M. Jocelyne. 2004. Mnémème, Antitopic – Le post-Rhème, de l‘énoncé au
texte. In M.M. Jocelyne Fernandez-Vest & Shirley Carter-Thomas (eds.), Structure Informa-
tionnelle et Particules Enonciatives – essai de typologie, (Grammaire & Cognition 1–2),
65–104. Paris: Editions L’Har-mattan.
Fernandez-Vest, M.M. Jocelyne. 2005a. Information structure and typological change: Northern
Sami challenged by Indo-European models. In M.M. Jocelyne Fernandez-Vest (ed.),
Les langues ouraliennes aujourd’hui: approche linguistique et cognitive – The Uralic
Languages today: a linguistic and cognitive approach, Préface de Claude Hagège (Biblio-
thèque de l’Ecole des Hautes Etudes 340), 563–576. Paris: Editions Honoré Champion.
Fernandez-Vest, M.M. Jocelyne. 2005b. Morphogenèse du sens: de l’espace au theme. In M.M.
Jocelyne Fernandez-Vest (ed.), Oralité et cognition – incarnée ou située? Orality and
cognition – embodied or situated? (Grammaire & cognition 3), 79–110. Paris: Editions
L’Harmattan.
Fernandez-Vest, M.M. Jocelyne. 2009a. Typological evolution of Northern Sami: spatial cogni-
tion and Information Structuring. In Jussi Ylikoski (ed.), The Quasquicentennial of the
Finno-Ugrian Society (SUST 258), 33–55. Helsinki: Suomalainen Seura.
Fernandez-Vest, M.M. Jocelyne. 2009b. Bilingual knowledge in a perspective of cognitive
semantics: evidence from Northern Sami and Californian Finnish. In M.M. Jocelyne
Fernandez-Vest & D.T. Do-Hurinville (ed.), Plurilinguisme et traduction, des enjeux pour
l’Europe – Multilingualism and translation, challenges for Europe (Grammaire & Cognition
6), 65–78. Paris: Editions L’Harmattan.
Fernandez-Vest, M.M. Jocelyne. 2010a. Typologie de la négation dans quelques langues
ouraliennes nord-occidentales: degrés de phrasticité, degrés d’oralité. In Franck Floricic
& Renée Lambert-Brétière (eds.), Typo 4 – La négation et les énoncés non susceptibles
d’être niés, 177–196. Paris: Presses du CNRS.
Fernandez-Vest, M.M. Jocelyne. 2010b. Uralic localism in a discourse perspective: a comparison
of Local Cases in Samic and Finnic languages. Paper presented at the “Uralic Case
Systems Workshop” of the 14th International Morphology Meeting, Budapest, 13–16
May, 2010.
Fernandez-Vest, M. M. Jocelyne. 2010c. Typological evolution of information grammar in Uralic.
In Csúcs Sándor (ed.), Papers from the XIth International Congress of Finno-Ugric studies
08–15 August 2010, Piliscsaba. Budapest: Reguly Társaság.
Fernandez-Vest. M.M. Jocelyne. 2011. Typological change of information grammar in Uralic. In
Csúcs Sándor (ed.), Papers from the XIth International Congress of Finno-Ugric studies
08-15 August 2010. Pilisosaba. Budapest: Reguly Társaság.
Fernandez-Vest, M.M. Jocelyne. 2012. Information structuring aspects of object case alternation:
Finnish translated from Sami. In Hans-Hermann Bartens, Cornelius Hasselblatt & Eberhard
Winkler (eds.), Festschrift to Lars-Gunnar Larsson on his 65th birthday, Veröffentlichungen
der Societas Uralo-Altaica, 55–62.
Fernandez-Vest, M.M. Jocelyne. 2014 in press. Detachments for cohesion. Toward an informa-
tion grammar of oral language. Berlin: Mouton de Gruyter.
Foley, Wiliam & Robert D. Van Valin Jr. 1984. Functional syntax and universal grammar.
Goldberg, Adele. 1995. Constructions: A Construction Grammar Approach to Argument Structure.
University of Chicago Press.
Groot, Casper de & Hannu Tommola (eds.). 1984. Aspect bound. A voyage into the realm
of Germanic, Slavonic and Finno-Ugrian aspectology. 2. Verb, sentence and aspect.
Dordrecht: Foris Publication.
Hagège, Claude. 2009. Verbs vs adpositions as markers of spatial meaning, Paper presented at
the Association for Linguistic Typology 8th Biennal Meeting (ALT8), 23–26 July, 2009.
University of California, Berkeley.
Hagège, Claude. 2010. Adpositions. Function-marking in human languages (Oxford studies in
typology and linguistic theory). Oxford: Oxford Univ. Press.
Hakulinen Auli & Fred Karlsson. 1979. Nykysuomen lauseoppia (SKST 350) Helsinki: Suoma-
laisen Kirjallisuuden Seura.
Heinämäki, Orvokki. 1984. Aspect in Finnish. In Casper de Groot & Hannu Tommola (eds.), Aspect
bound. A voyage into the realm of Germanic, Slavonic and Finno-Ugrian aspectology. 2.
Verb, sentence and aspec, 153–178. Dordrecht: Foris Publication.
Hopper, Paul J. 1979. Aspect and foregrounding in discourse. In Tom Givón (ed.), Syntax and
semantics. Discourse and syntax, 213–242. New York: Academic Press.
tion. Cognitive Linguistics 20(1). 43–70.
ISK = Iso suomen kielioppi [The comprehensive grammar of Finnish]. 2004. Auli Hakulinen,
Maria Vilkuna, Riitta Korhonen, Vesa Koivisto, Tarja-Riitta Heinonen & Irja Alho. Helsinki:
SKS.
Itkonen, Terho. 1975. Erään sijamuodon ongelmia. Opuscula Instituti linguae fennicae, 53. 173–
217. Helsinki: Helsingin ylipisto.
Kangasmaa-Minn, Eeva. 1978. On aspect variations of Finnish underived verbs. Turku: Suomen
kielitieteellisen yhdistyksen julkaisuja 2.
Kangasmaa-Minn, Eeva. 1984. Tense, aspect and Aktionsart in Finno-Ugrian. In Casper de Groot
& Hannu Tommola (eds.), Aspect Bound, 77–93. Dordrecht: Foris Publication.
Kangasmaa-Minn, Eeva. 1993. Aspektista ja sen sukulaisilmiöistä suomalais-ugrilaisissa
kielissä. In Valma Yli-Vakkuri (ed.), Studia comparativa linguarum orbis Maris Baltici 1.
13–23.
Kiparsky, Paul. 1998. Partitive case and aspect. In Miriam Butt & Willem Geuder (eds.), The pro-
jection of arguments, 265–307. Stanford: CSLI Publications.
Kiparsky, Paul. 2001. Structural case in Finnish. Lingua 111. 315–376.
Klaas. Birute. 1999. Dependence of the object case on the semantics of the verb in Estonian,
Finnish, and Lithuanian. In Mati Erelt (ed.), Estonian. Typological studies. III (Publications
of the Department of Estonian of the University of Tartu 11), 47–83. Tartu: University of
Tartu.
Larsson, Lars-Gunnar. 1983. Studien zum Partitivgebrauch in den ostseefinnischen Sprachen.
Uppsala: Finsk-ugriska Institutionen.
Larsson, Lars-Gunnar. 1984. The role of Baltic influence in the aspectual system of Finnish.
In Casper de Groot & Hannu Tommola (eds.), Aspect bound. A voyage into the realm of
Germanic, Slavonic and Finno-Ugrian aspectology. 2. Verb, sentence and aspect, 97–109.
Dordrecht: Foris Publication.
Leino, Pentti. 1991. Lauseet ja tilanteet: Suomen objektin ongelmia [The sentences and situa-
tions: Problems of the Finnish object], (Suomi 160), Helsinki: SKS.
Leinonen, Marja. 1982. Russian Aspect, “temporal’naja lokalizacija’ and Definitness / Indefinitness.
Helsinki: Neuvostoliitto-instituutin vuosikirja 27.
Leinonen, Marja. 1984. Narrative implications of aspect in Russian and in Finnish. In Casper
de Groot & Hannu Tommola (eds.), Aspect bound. A voyage into the realm of Germanic,
Slavonic and Finno-Ugrian aspectology. 2. Verb, sentence and aspect, 239–255. Aspect
bound.
Levelt, Willem Johannes Maria. 1989. Speaking. Cambridge, MA: MIT Press.
Maslov, Jurij S. 1973. Universal’nye semantičeskie komponenty v soderžánii grammatičeskoj
kategorii soveršennogo nesoveršennogo vida. Sovetskoe slavjanovedenie 4. 72–83.
Maslov, Jurij S. 1980. Struktura povestvovatel’nogo teksta i tipologija slavjanskix vidovremen-
nyx system. Svantevit Årgång VI (1). 43–70.
Maslov, Jurij S. 1988 [1983]. Chapter 2. Resultative, perfect and aspect. In Vladimir P. Nedjalkov
(ed.), Typology of resultative constructions. Translated from the original Russian edition,
Leningrad, Nayka, 1983. English translation edited by Bernard Comrie, 63–84. Amsterdam:
Benjamins.
Metslang, Helle. 1994. Temporal relations in the predicate and the grammatical system of
Estonian and Finnish (Oulun yliopiston Suomen ja saamen kielen laitoksen tutkimusrap-
portteja 39). Oulu: Oulun yliopisto.
Metslang, Helle. 2001. On the developments of the Estonian aspect: the verbal particle ära.
In Östen Dahl & Maria Koptjevskaja-Tamm (eds.), The Circum-Baltic languages. Their
typology and contacts: Vol. 2. Grammar and typology (Studies in Language Companion
Series (SLCS) 55), 443–479. Amsterdam: Benjamins.
Metslang, Helle & Hannu Tommola. 1995. Zum tempusystem des Etnischen [On the tense
system of Estonian]. In Rolf Thieroff (ed.), Tense systems in European languages. II.
(Linguistische Arbeiten 338), 299–326. Tübingen: Niemeyer.
Miller, Jim & M.M. Jocelyne Fernandez-Vest. 2006. Spoken and written language. In Giuliano
Bernini & Marcia L. Schwartz (eds.), Pragmatic organization of discourse, (Empirical
approaches to language typology, Eurotyp 20–8), 9–64. Berlin: Mouton de Gruyter.
Nedjalkov, Vladimir P. (ed.). 1988 [1983]. Typology of resultative constructions. Translated
from the original Russian edition, Leningrad, Nayka, 1983. English translation edited by
Bernard Comrie. Amsterdam: Benjamins.
Penttilä, Aarni. 1963. Suomen kielioppi [Finnish grammar]. 2. Tarkistettu pianos. Porvoo: WSOY.
Rajandi, Henno & Helle Metslang. 1979. Määrämata ja määratud object [Indefinite and definite
object]. Tallinn: Eesti NSV Teaduste Akadeemia, Keele ja Kirjanduse Instituut.
Sadeniemi, Matti. 1929. Objektin totaalisuudesta ja partiaalisuudesta. Virittäjä 33.
Talmy, Leonard. 1995. Fictive motion in language and “ception”. In Toward a Cognitive Semantics,
Vol. 1, MIT Press, 99–175.
Tamm, Anne. 2003. Estonian transitive verb classes, aspect case, and the progressive. In
Anne Dahl & Peter Svenonius (eds.), Proceedings of SCL Working papers of Language
and Linguistics, Nordlyd 31(4).
Tamm, Anne. 2004a. Relations between Estonian Aspect, verbs and object case. Budapest: ELTE
dissertation.
Tamm, Anne. 2004b. On the grammaticalization of the Estonian perfective particles. Acta
Linguistica Hungarica 51(1–2). 143–169.
Tamm, Anne. 2006. Estonian transitive verbs and object case. In Miriam Butt & Tracy Holloway
King (eds.), Proceedings of the LFG06 Conference Universität Konstanz. Stanford, CA: CSLI,
http://ccsli-publications-stanford.edu/LFG/11/lfg06.pdf
Tommola, Hannu. 1984. On the significance of ‘phase meanings’’, in Casper de Groot & Hannu
Tommola (eds.), Aspect bound. . . , 111–132.
Van Valin Robert D., Jr. 1993. A synopsis of Role and Reference Grammar. In Robert Van Valin
(ed.), Advances in Role and Reference Grammar, 1–164. Amsterdam: Benjamins.
Van Valin Robert D., Jr. 2005. Exploring the syntax-semantics interface. Cambridge: Cambridge
University Press.
Van Valin Robert D., Jr. 2010. Achievements and accomplishments. Paper presented at the
International Conference Representation of events, Paris III-Sorbonne Nouvelle, 28–30
October, 2010.
Van Valin Robert D., Jr. 2011. Lexical representation, co-composition, and linking syntax and
semantics. In James Pustejovsky et al. (eds.), New developments in the Generative Lexicon,
Dordrecht, Kluwer. http://linguistics.buffalo.edu/ research/rrg.html
Van Valin Robert D., Jr. & Randy LaPolla. 1997. Syntax: structure, meaning and function,
Vendler Zeno. 1957 [1967]. Verbs and times, The Philosophical review, 6. 143–60. Reprinted in
Z. Vendler, 1967, Linguistics in philosophy. Ithaca: Cornell University Press.
Verkuyl Henk. 1989. Aspectual classes and aspectual composition. Linguistics and philosophy
12(1). 39–94.
Verkuyl Henk. 1993. A theory of aspectuality. The interaction between temporal and atemporal
structure. Cambridge: Cambridge University Press.
Wiedemann Ferdinand J. 1973. Estnisch-deutsches Worterbuch, 4:ter. Tallinn: Druck.
Corpus of translations
Mukka Timo K. 1974. Laulu Sipirjan lapsista. Jyväskylä: Gummerus.
Mukka Timo K. 2008. Sipirjá. Sami translation by Jovnna-Ánde Vest. Kárášjohka: Davvi Girji.
Skabmatolak. Sabmelaš kirjjalašvuođa antologiija – Tulia kaamoksessa. Saamelaisen kirjalli-
suuden antologia [Fires in the polar night. An anthology of Sami literature], 1973, Samuli
Aikio, Erkki Itkonen & Pekka Sammallahti (toaim.), (Aikio Sulo, 1974 [1968], Čierastallam
[Mäenlaskua], 295–300; Paltto Kirsti, 1971, Soagŋu, Bijadat [Painajainen], 320–333).
Helsinki: Otava.
Vest Jovnna-Ánde. 1988 [2nd 1989]. Čáhcegáddái nohká boazobálggis. Kárášjohka: Davvi Girji.
Vest Jovnna-Ánde. 1989. Reintråket ender ved bredden. Norwegian translation by Laila Stien.
Oslo: Aschehoug.
Vest Jovnna-Ánde. 1990. Poropolku sammaloituu. Finnish translation by Eino Kuokkanen. Oulu:
Pohjoinen.
Vest Jovnna-Ánde. 1996. The cloudberry trip. In Harald Gaski (ed.). In the shadow of the
Midnight sun. Contempory Sami prose and poetry, 167–178. Kárášjohka: Davvi Girji.
Vest Jovnna-Ánde. 2005. Árbbolaččat. III. Kárášjohka: Davvi Girji.
Vest Jovnna-Ánde. 2006a. Arvingarna. III. Swedish translation by Riitta Taipale. Kárášjohka:
Davvi Girji.
Vest Jovnna-Ánde. 2006b. Perilliset. III. Finnish translation by Jovnna-Ánde Vest. Kárášjohka:
Davvi Girji.
III Basque
Urtzi Etxeberria
7 The definite article and the partitive
particle in Basque: dialectal variation*
The aim of this paper is to make an as thorough as possible a description of both

the Basque so-called definite article and the partitive particle [-rik] across its
various dialects. The reason why I concentrate on these two elements is because
they are clearly related to each other; the partitive particle [-rik] is taken to be
the morpheme that is used in polarity contexts instead of the existential inter-
pretation of the definite article [-a(k)]. The paper does not aim at giving neither
a syntactic nor a semantic analysis of these two morphemes, and it should only
be taken as an initial approximation to the topic.
Keywords: definite article, partitive, bare nouns, existential interpretation,

dialectal variation.
1 Introduction
The main aim of this paper is quite modest: as the title itself claims, the paper
aims at investigating the relationship between the so-called Basque definite article
(both in its singular and plural form) and the Basque partitive particle. In other
words, this paper will try to provide a description of the behaviour of these two
elements. These elements show dialectal variation: the most significant variation
is found between the western-central Basque and the eastern Basque, and it is
* The research conducing to this paper has benefited from the Basque Government project
GIC07/144-IT-210-07 and Hm-2008-1-10, from the project FR2559 from Fèderation Typologie et
Universaux Linguistiques, from the MCE projects FFI2008-00240, FFI2011-29218, FFI2011-
26906, FFI2011-23356, from the UPV/EHU project UFI11/14, from the Aquitaine-Euskadi project
HM-2012, as well as from the ANR projects TSABL (ANR-07-CORP-033), ISQI (ANR 2011-JSH2-
004-1), and from the Franco-German ANR/DFG project Vers une typologie des pronoms
impersonnels humains (TypoImp). Many thanks to Beñat Oihartzabal, Aurelia Arcocha, Xarles
Bidegain, Battittu Coyos, Maia Duguine, Isabelle Duguine, Oihana Larrandaburu, Marylin
Recalt, and Maider Bedaxagar for help with the eastern Basque data. Thanks also to Beñat
Oihartzabal, Jasone Salaberri and Irantzu Epelde for sharing with me data they collected for the
project Norantz (http://www.norantz.org/web/en/sarrera). Thanks also to the two anonymous
reviewers for this volume. I am very grateful to Silvia Luraghi and Tuomas Huumo for inviting
me to write this paper, as well as for their patience. Usual disclaimers apply.
292 Urtzi Etxeberria
Figure 1: Map of Basque Dialects (Zuazo 2008)
exactly here where this paper wants to put its attention. The final aim of this
paper is not to provide a specific syntactic analysis of these elements, however,
I do expect that this initial step will make it possible to do so in the near future
(cf. Etxeberria in prep). Thus, let us take this paper (together with Txillardegi
1977, Irigoien 1985, Artiagoitia 2002, 2006, Etxeberria 2005, 2010, to appear,
Manterola 2008, Santazilia 2009, Etxeberria & Etxepare 2009, Etxeberria in
prep, among many others) as a starting point of a thorough analysis of the func-
tional structure of the Basque nominal phrase across its various dialects.
Before we move on to the next section, I would like to make it clear that
when I say western-central or eastern Basque I’m talking in a vague way, since
I don’t know where exactly the linguistic borders are situated. Now, this
paper makes a two-side division: when we talk about the definite article [-a(k)],
on the one hand, we will have the Bizkaian, the Gipuzkoan, and the Lapurdian
dialects, and on the other, we will have the Souletin (from Zuberoa, the most
eastern dialect of Basque); cf. Figure 1. I assume that Bizkaian and Gipuzkoan
behave alike since, apart from the possibility of having a demonstrative and the
definite article in the same noun phrase in Bizkaian: a gizon-a ‘that man-the’,
there are no differences between them. I will not consider the dialect from
Low Navarrese in this paper and I will assume that this dialect behaves just
The definite article and the partitive particle in Basque: dialectal variation 293
like Souletin when it comes to the behaviour of the definite article.1 Note also
that the dialect from Navarre is not mentioned in the paper; the reason why
this is so is due to the fact that I have not been able to collect data from this
dialect. However, I do believe that the behaviour of this dialect concerning the
use of the definite article basically parallels Gipuzkoan. While considering the
partitive [-rik], the division is the following: (i) Bizkaian and Gipuzkoan (and
Navarrese), (ii) Lapurdian, and Souletin.
The paper is divided into two sections: the first section makes an as thorough
as possible description of the various uses of the definite article across dialects.
The second section concentrates on the partitive particle [-rik] and its dialectal
variation.
2 The definite article [-a(k)]

As it will be shown below, the Basque definite article is used with more semantic
meanings than that of languages surrounding it (cf. among other Artiagoitia 2002,
2004, 2012, Eguren 2006, 2012, Etxeberria 2005, 2008, 2010, 2012a, to appear).
A quote by Trask (2003: 119) is in order here: “The label ‘definite article’ is
misleading, since this article is of a much broader use than the English definite
article.”
In this quote, Trask compares the use of the Basque definite article [-a] with
the use of the English definite article. We could easily make the same claim if we
compared the Basque definite article to the definite article of languages like
French or Spanish. And I agree with Trask’s claim, as long as we only consider
western-central dialects (or Standard Basque) where the definite article appears
in positions where other languages use bare nouns (BNs). In fact, we could be
led to think that the Basque definite article is not really definite. However, this
paper assumes (in line with Etxeberria 2005, 2010, to appear) that, despite
its various semantic uses (see below), the Basque morpheme [-a] is always the
definite article and as such it always base-generates at the same syntactic posi-
tion. In fact, what this paper assumes is that the definite article and the number
markers ([-Ø] for singulars and [-k] for plurals) are base-generated in different
syntactic positions: [-a] in the head of Determiner Phrase (DP), number markers
in the head of Number Phrase (NumP). However, it is not the aim of this paper to
1 There exists a difference between these two dialects: while Souletin makes use of the
definite article on predicative use (cf. Manterola 2008; Santazilia 2009; cf. also Etxeberria
to appear, in prep), this is not the case in Low-Navarrese. Be that as it may, this paper will
not consider the predicative use of the definite article. (cf. among others Zabala 1993, 2003,
Artiagoitia 1997, Eguren 2005, 2006, 2012, Matushansky 2005 for possible analyses)
provide arguments in favour of this proposal and I assume it as correct. The

reader is referred to Etxeberria (2005, 2010, to appear); cf. Manterola (2006,
2009, 2012), Etxeberria (to appear) for a diachronic analysis of the Basque
definite article.
In what follows, the paper concentrates on making a general description of
the behaviour of the morpheme [-a], i.e. the definite article. Thus, the Basque
definite article is a bound morpheme that takes the phonetic forms [-a] (when
singular) and [-ak] (when plural).2
(1) a. gizon-a
man-DET. SG
‘the man’
b. gizon-ak
man-DET. PL
‘the men’
In western-central varieties there is also a proximate plural definite article [-ok].3
(2) gizon-ok
man-DET. PROX . PL
The article appears in the final position of the nominal phrase, attached to
the noun as in the example in (1) or attached to an adjective:
(3) a. liburu urdin-a

book blue-DET. SG
‘the blue book’
b. liburu urdin txiki-a

book blue small-DET. SG
‘the small blue book’
2 Some authors argue that the plural form of the Basque definite article [-ak] is a single
element (cf. Goenaga 1978, 1991, Euskaltzaindia 1993, Ticio 1996, Artiagoitia 1997, 2002,
2004, 2012, Rodriguez 2003, Trask 2003). Based on Etxeberria (2005), I defend that singular
and plural markers and the definite article [-a] are base-generated in different syntactic posi-
tions; see also Eguren 2006, 2012; see §2.1. However, for ease of exposition, I will refer to [-a]
and [-ak] as the singular and the plural determiner respectively.
3 Although there is no singular proximate singular in modern Basque, -ori, -or, -au or o are
attested in early texts; it is still possible to find -o in actual Bizkaian in hemen berton ‘right
here’, along with hemen bertan ‘right here’.
c. liburu urdin txiki polit-a

book blue small nice-DET. SG
‘the nice small blue book’
In general terms, the properties of the definite article [-a] are those pre-
sented in the previous paragraphs.4 Let us move now to see the behaviour and
use of this element in different dialects.
2.1 Bizkaian, Gipuzkoan, and Lapurdian (western-central)

It’s been assumed (cf. among others, Laka 1993, Artiagoitia 1997, 1998, 2002,
2004, 2012, Etxeberria 2005, 2010, 2012a) that BNs cannot be used in argument
position in Basque and that the use of the definite article is necessary if sentences
are going to be grammatical.5,6 This is actually one of the most characteristic
4 There is another very interesting use of the definite article that takes place in all varieties:
the presence of the morpheme [-a(k)] is obligatory with so-called strong quantifiers, cf. (i)
below, where the definite article has been argued to be the overt contextual domain restrictor
of the quantifier (cf. Etxeberria 2005, 2007, 2009, 2012b, Etxeberria & Giannakidou 2009,
et seq.).
(i) Ikasle guzti*(-ak) berandu etorri ziren.
student all-DET. PL late come AUX : PL
5 The presence of an indefinite determiner (ia) or a weak quantifier (ib) (cf. Etxeberria, 2005,
2008, 2012a, in prep) also makes the sentence grammatical.
(i) a. Mutil bat berandu iritsi zen.
boy one late arrive AUX . PAST
‘A boy arrived late.’
b. Mutil asko berandu iritsi ziren.
boy many late arrive AUX . PAST
‘Many boys arrived late.’
Note that the morpheme [-a] cannot be combined with the indefinite determiner or with the
weak quantifiers in the examples in (i); this is also a property that applies to all varieties of
Basque (cf. Etxeberria 2008, 2012a, in prep for a more general presentation of this data and
for a possible analysis).
6 As already mentioned in the introduction, this paper will only concentrate on the argumental
use of the definite article (and the partitive [-rik]). However, this is not the only possible usage
of this element since it can also have predicative uses: Jon irakaslea da ‘Jon is a teacher’ vs.
Jon irakasle dago ‘Jon is working as a teacher’. In predicative uses, the article plays the
role of the participle or of individual-level predicates (cf. among others Zabala 1993, 2003,
Artiagoitia 1997, Eguren 2005, 2006, 2012, Matushansky 2005 for possible analyses). There is
also variation (western-eastern) in these uses of the definite article (cf. Manterola 2008, 2012,
Santazilia 2009).
properties of the Basque definite article [-a] in these varieties (also in Standard
Basque).7
Subject position:
(4) a. Irakasle*(-a) berandu etorri zen.
teacher-DET. SG late come AUX : SG
‘The teacher came late.’
b. Irakasle*(-ak) berandu etorri ziren.

teacher-DET. PL late come AUX : PL
‘The teachers came late.’
Object position:
(5) a. Kepa-k baloi*(-a) hartu zuen.
Kepa-ERG ball-DET. SG take AUX : SG
‘Kepa took the ball.’
b. Kepa-k baloi*(-ak) hartu zituen.

Kepa-ERG ball-DET. PL take AUX : PL
‘Kepa took (the) balls.’
Indirect Object position:

(6) a. Kepa-k gizon*(-a-)ri baloi-a eman zion.
Kepa-ERG man-DET. SG . DAT ball-DET. SG . ABS give AUX : SG
‘Kepa gave the ball to the man.’
b. Kepa-k gizon*(-e-)i baloi-a eman zien.

Kepa-ERG man-DET. PL . DAT ball-DET. SG . ABS give AUX : PL
‘Kepa gave the ball to the men.’
If BNs cannot appear in argument position in these dialects of Basque,

the question that could come to our mind is how these varieties of Basque
express what other languages express by means of BNs.8 For example, English
7 In fact, everything that we will say about the behaviour of the definite article in these
varieties also applies to the behaviour of the definite article in Standard Basque.
8 In this paper, apart from Basque, the languages that appear the most are English, Spanish,
and French. There are also languages without articles: Russian, Korean, Japanese, Chinese,
Dene Suline, among others. These languages make use of different means to express
(in)definiteness: by word order, by context, by case etc. These languages will not be treated in
this paper.
and other Germanic languages can use bare plurals and mass terms without D
to express both the kind reading (7) as well as the existential reading (8).9
(7) a. Fishes appeared 390 million years ago.

b. Silver has the atomic number 47.
(8) a. Garazi has eaten olives.

b. Moles are ruining my parents’ vegetable garden.
As soon as we approach the phenomenon we notice that the Basque definite

article in these varieties is of a much broader use than the definite article of
languages like English or Romance languages: in addition to the usual referential
interpretation that we get both in (2) and (3) the definite article also appears in
contexts where other languages typically present BNs, e.g. Spanish or English –
not French, where BNs are not accepted but in coordination contexts; cf.
Roodenberg (2004). To begin with, when the Basque [NP+DET] sequence com-
bines with kind level predicates (e.g. evolve, become extinct, be common, etc.;
cf. Carlson 1977; cf. also Krifka et al. 1995), the usual referential interpretation
(i.e. the definite extensional interpretation) disappears and it adopts a kind
reading where the DP makes reference to the species as a whole (creating an
intensional interpretation, which makes reference to the biggest plurality of the
set denoted by the NP in all possible worlds and situations). This comes as
no surprise as many other European languages also make use of the definite
determiner to express the kind interpretation, e.g. Romance languages, or Greek
(cf. Kleiber 1990, Krifka et al 1995, Chierchia 1998, Zamparelli 1998, Fara 2001,
Dayal 2004, etc.).
(9) a. Dinosauru-ak aspaldi desagertu ziren.

dinosaur-DET. PL long time ago disappear AUX
‘Dinosaurs disappeared a long time ago.’
b. Nitrogeno-a ugaria da gure unibertsoa-n.

nitrogen-DET. SG abundant is our universe-INE
‘Nitrogen is abundant in our universe.’
In the examples in (9), the DPs dinosauruak ‘dinosaur-DET. PL ’ and nitrogenoa

‘nitrogen-DET. SG ’ do not make reference to a specific set of dinosaurs or to a
9 I assume (in line with Chierchia 1998, 2009) that mass terms are atomic (cf. (26)), although
their atomic nature is vague.
specific quantity of nitrogen, but to the species dinosaurs and to the species
nitrogen.
Now, when Basque definite DPs fill the direct object slot, the definite DP
can but need not obtain the referential interpretation and can get the so-called
existential interpretation (cf. Carlson 1977, among others), (plurals and masses;
singular definite DPs can only be interpreted existentially in some very specific
contexts, see examples (11–12)). In other words, in the examples in (10) we need
not be talking about a specific set of candies or a specific quantity of wine.10
(10) a. Kepa-k goxoki-ak jan zituen.

Kepa-ERG candy-DET. PL . ABS eat AUX : PL
‘Kepa ate (the) candies.’
b. Bartolo-k garagardo-a edan zuen.

Bartolo-ERG beer-DET. SG . ABS drink AUX : SG
‘Bartolo drank (the) beer.’
Note that in the examples in (10) the object DPs cannot make reference to the
whole species denoted by the NP. However, a referential interpretation is possible
for both the object DPs in (10); that is, if we were to offer English translations
(10a) and (10b) would be ambiguous: (10a) ‘Kepa ate the candies’ or ‘Kepa ate
candies’; (10b) ‘Bartolo drank the beer’ or ‘Bartolo drank beer’.11,12
10 Romance languages make use of different strategies to obtain this existential interpreta-
tion. Both Spanish and Italian are able to use BNs (just like English or other Germanic lan-
guages). On the other hand, French makes use of the so-called partitive determiner des/du
and no BNs are allowed (Italian also has a partitive determiner). See Chierchia (1998), Zamparelli
(2000, 2002a, 2002b), Kleiber (1990), Bosveld-de Smet (1997), Heyd (2003), Bosque (1996),
Laca (1996).
Sp.: Juan ha bebido [café].
Juan has drunk [coffee]
‘Juan has drunk coffee.’
Fr.: Pierre a mange [des sucreries].

Pierre has eaten [of-the sweets]
‘Pierre has eaten sweets.’
11 There are different analysis that have tried to account for the existential reading of Basque
definite DPs. Due to lack of space, I will not get into explaining these approaches. The reader is
referred to Artiagoitia (2002, 2004, 2012), Eguren (2006), Etxeberria (2005, 2010, to appear).
12 The existential reading could also be called indefinite; however, there are some important
differences between the existential reading that the Basque definite article (or BNs in English
or Spanish) can obtain and the existential interpretation that real indefinites (e.g. a) get. When
Apart from these cases, some singular definite DPs can also get the so-called
existential-like reading as shown by the example in (11). This paper will not try
to provide an account for this use; cf. Rodriguez (2003), Etxeberria (2005, in
prep) or Eguren (2006) for a possible analysis; cf. Manterola (2009, 2012) for a
diachronic analysis.
(11) Joseba-k auto-a erosi zuen.

Joseba.ERG car-DET. SG . ABS buy AUX
‘Jon bought (the/a) car.’
This sentence is also ambiguous: in one of the readings Jon has bought a spe-
cific car, e.g. the one that he mentioned he was going to buy, a Citroën 2CV; in
the other reading the sentence in (11) is taken to be more or less parallel to
something like ‘Jon has bought a car’ where we don’t know which car we are
talking about, hence parallel to the non-specific reading of a car.
However, the sequence [count N + singular DET] in (11) can only be inter-
preted existentially in very specific contexts: so-called stereotypical contexts
(such as buying a car, having a wife/husband, having a baby, wearing a hat,
etc.). Furthermore, all the examples in (12) have a clear sense of possession,
that is, once you buy a car/house, you become the possessor, having something
is also closely related to possession, as it is wearing something (e.g. hat). It is
obvious then that singular definite DPs in Basque do not get the existential
interpretation as easily as plurals or mass terms do.
existentially interpreted, the Basque definites DPs do not behave like usual indefinites and
must always take narrow scope (pace the Ambiguity approach), just like BNs in English.
(i) a. #Nere aita-k bi sator hil ditu ordubetez.
my father-ERG two mole kill AUX hour-FOR
‘My father has killed two moles for an hour.’
b. Nere aita-k satorr-ak hil ditu ordubetez.
my father-ERG mole-DET. PL kill AUX hour-FOR
‘My father has killed moles for an hour.’
The sentence in (ia) can only be interpreted with the indefinite bi sator ‘two moles’ having wide
scope over the atelic adverbial [bi sator > adv.] and asserts that the same two moles have been
killed again and again; a rather strange state of affairs. The sentence in (ib) on the other hand
is completely grammatical. The reading we get is one where my father has killed different
moles and the definite DP must necessarily take narrow scope below the adverbial
[adv. > satorrak].
(12) a. auto-a/etxe-a erosi

car-DET. SG /house-DET. SG buy
b. senarr-a/emazte-a eduki
husband-DET. SG /wife.DET. SG have
c. txapel-a eraman
hat-DET. SG bring
Romance languages (e.g. Spanish, French and Romanian), in order to

express what the examples in (12) express, make use of singular BNs (cf. Bosque
1996 for an extensive presentation of Spanish data; cf. Dobrovie-Sorin, Bleam &
Espinal 2005, Espinal & McNally 2007, Espinal 2010 for possible analyses).
(13) comprar coche/casa, tener marido/mujer, llevar sombrero, etc.

buy car/house have husband/wife bring hat
Note that normally Basque [count N + singular DET] sequences that appear in
object position of object-level predicates can only get referential interpretations,
in contrast with what we previously saw in the examples in (12). In other words,
in the examples in (14), we are necessarily talking about a specific book, a specific
boy, and a specific magazine, respectively; and there is no way we can get an
existential-like reading.
(14) a. liburu-a erosi

book-DET. SG buy
‘buy the book’
b. mutil-a ikusi
boy-DET. SG see
‘see the boy’
c. aldizkari-a irakurri
magazine-DET. SG read
‘read the magazine’
There is a difference between the Basque spoken in Bizkaia and Gipuzkoa

and the one spoken in Lapurdi. In existential sentences (with neutral intona-
tion), the nominal expression can appear combined with the definite article
(when mass or plural) in the three varieties, no matter the construction we use
to create existential sentences, (15a) or (15b) (cf. Etxeberria 2012a for extensive
discussion on Basque existential sentences).13 However, when we use them as

an exclamation, it is possible to use BNs with no article in existential sentences
in the Lapurdian dialect, as can be seen in (16a); in the varieties of Bizkaia and
Gipuzkoa, on the other hand, BNs are not allowed, (16b).
(15) a. Bada ardo-a hotzgailua-n.

yes-is wine-DET. SG fridge-INE
‘There is wine in the fridge.’
b. Hotzgailua-n ardo-a dago.

fridge-INE wine-DET. SG be-egon.SG 14
‘There is wine in the fridge.’
13 We could discuss whether the sentences in (15) are existential sentences or locative
sentences. In English, for example, existential contexts do not admit definite articles whereas
locative contexts do admit them (so-called definiteness effect or definiteness restriction,
Milsark 1977; cf. Freeze 1992, Bresnan 1994; cf. Etxeberria to appear for Basque).
(i) a. There is a book on the table. [existential]
b. The book is on the table. [locative]
Note that instead of using the mass term that we use in the examples (15) and (16), we would
use a count noun, the resulting sentence would be completely ungrammatical. An indefinite
article would rescue the sentence (iic–iid).
(ii) a. *Bada gizona gela-n.
yes.is man-DET. SG room.- NE
b. *Badira gizonak gela-n.
yes.are man-DET. PL room-INE
c. Bada ikasle bat gela-n.
yes.is student one room-INE
‘There is a student in the room.’
d. Badira ikasle batzuk gela-n.
yes.are student some room-INE
‘There are some students in the room.’
The sentence in (15b) would not be ungrammatical with a count term like gizon ‘man’, but
the interpretation would be completely locative. In other words, the English translation of the
sentence in (iii) would be the man is in the room, not there is a man in the room.
(iii) Gelan gizona dago.
room-INE man-DET. SG be-egon.SG
14 Basque, like Spanish (Lujan 1981, Schmitt 1992, Fernández Leborans 1999), distinguishes
between a locative copula and a characterizing one (Etxepare, 2003). Intuitively, the locative
(16) a. Bada jende hemen!!

yes-is people here
b. Bada jendea hemen!!

yes-is people-DET. SG here
‘There is people in here!’
The sentence in (16a), despite using the BN jende ‘people’, is interpreted as

making reference to a big quantity of people; i.e., the sentence in (16a) is equal
to something like bada jende asko/pila bat hemen ‘there is a lot of people here’.
This interpretation is impossible for the sentences in (15).
2.2 Souletin or when the article is less used

In Souletin, the behavior of the definite determiner is that of a ‘well-behaved’
definite determiner; i.e. it forces a referential interpretation in episodic contexts
no matter what syntactic position it appears in. However, as a general rule,
Souletin makes less use of the morpheme [-a], and in this dialect it is possible
to use BNs in some specific syntactic positions (something impossible in the
western-central varieties, as we saw above) – cf. Txillardegi (1977), Coyos (1999),
Casenave-Harigile (2006), Etxebarne (2006) for a description of the use of BNs in
Souletin; cf. also Etxeberria (to appear) for a possible analysis of the data.15
In Souletin, it is possible to use BNs in direct object position as the follow-
ing examples show:
(17) a. Bortü-a-n ikusi dut behi, ardi eta mando.

mountain-DET. SG - INE see aux cow sheep and mule
‘I have seen cows, sheep, and mules in the mountain.’
(Coyos 1999)
copula egon ascribes a temporary property to the subject of predication (ib), whereas the char-
acterizing copula izan introduces an inherent property of the subject (ia).
(i) a. Jon oso barregarri-a da
Jon very funny-DET is
‘Jon is a very funny guy.’
b. Jon oso barregarri dago (mozorro horrekin)
Jon very funny is-LOC costume that-with
‘Jon is very funny (in that costume).’
15 Thanks to Battittu Coyos, Oihana Larrandaburu, Marylin Recalt, and especially Maider
Bedaxagar, for help with the Souletin data.
b. Dembora da (. . .) içan deçada-n diru.

time is be have-COMP money
‘It is time for me to have money.’
(Bourciez 1895)
c. Manexe-k hur edan dizü. Peio-k ogi jan dizü.

Manex-ERG water drink AUX Peio-ERG bread eat AUX
‘Manex has drunk water. Peio has eaten bread.’
(Norantz)
These BNs are interpreted existentially, just as the DPs in object position of the
sentences in (10). To be interpreted existentially means that in the examples in
(17) we are not talking about a specific set of cows, sheep, or mules, or a specific
quantity of money, water, or bread. Now, if we would add the definite article to
these BNs, they would necessarily get a specific (referential) interpretation.
BNs in Souletin can also appear in existential contexts.
(18) a. Badüzü etüdiant arrua-n.

yes-you.have student street-INE
‘There are students in the street.’
b. Badüzü hur godaleti-n.

yes-you.have water glass-INE
‘There is water in the glass.’
Note that BNs are allowed neither in subject position (with ergative or absolu-
tive case) nor in indirect object position as the following examples clearly show.
Subject
(19) a. Ergative: (i) *Ikasle-k hori egin dü.
student-ERG that.ABS do AUX
(ii) Ikasle-ek hori egin düe.
student-DET. PL . ERG that.ABS do AUX
‘The students did that.’
b. Absolutive: (i) *Ikasle jin da.

student-ABS come AUX
(ii) Ikasle-ak jin dira.
student-DET. PL . ABS come AUX
‘The students came.’
Indirect object:
(20) a. *Ikasle-ri librü eman deiot.
student-DAT book-ABS give AUX
b. Ikasle-ei librü eman deiet.

student-DET. PL . DAT book-ABS give AUX
‘I gave the student books.’
One other property of BNs is that they can not be combined with kind-level
predicates (see (9)), where the BN would be making reference to the species as a
whole, and the presence of the definite article is necessary for the sentence to be
grammatical.
(21) a. Lehu*(-ak) desagertze-ra dira.

lion-DET. PL . ABS disappear-ALL aux
‘Lions are about to disappear.’
b. Nitrojeno*(-a) paketa da gure lürraldea-n.

nitrogen-DET. SG . ABS abundant is our country-INE
‘Nitrogen is abundant in our country.’
They cannot either appear in generic sentences (cf. Carlson & Pelletier 1995).
In these cases too, the presence of the definite article is obligatory.
(22) a. Gatü*(-a) intelijent da.

cat-DET. SG intelligent is
‘The cat is intelligent.’
b. Ni-k errespeta-tzen dütüt erakasle*(-ak).

I-ERG respect-PROG AUX teacher-DET. PL
‘I respect teachers.’
Let us now go back to the contexts where BNs are accepted in Souletin, i.e.
to the examples in (17). One could think that the existential reading of BNs is
related to plurality. In other words, one could be led to think that whenever we
make use of a BN in Souletin, we necessarily make reference to a set with more
than one member. And this does make sense considering that: (i) Spanish or
English BNs (which can be interpreted existentially) always appear with the
plural number marker [-s] (except for the mass terms); (ii) in western-central
Basque, in order to obtain the existential reading we make use of the plural
definite article [-ak], cf. example (10) (with mass terms we would use [-a]; note
that I assume – in line with Etxeberria 2005, 2010 – that mass terms are number
neutral and that they do not have any kind of number morphology; cf. also
Delfitto & Schroten 1991, Doetjes 1997, Dayal 2004, Krifka 2004).
However, this way of thinking is not correct. What is important in the exis-
tential interpretation of the BNs in (17) is not whether they make reference to a
singularity or plurality, but rather, to make non-specific reference to what the
noun denotes in the real world. Evidence in favour of this idea comes from the
following examples: in Souletin, the translation of the sentence in (23) would be
realised by means of a BN, as in (24).
(23) Do you have children.
(24) Badüzü haur?16

yes-you.have child
The answer that a Souletin speaker would give to a question like (24) would be
positive in a situation where the answerer has a single child. If the BN denoted a
plurality, the response to the question in (24) should be negative (in this very
same context), but it is not.
We get an equal effect in examples such as (25). Consider the following
scenario: the Souletin speaker of the previous example goes to the National
Health Service office and sees a sign with the sentence in (25) in it. No doubt,
despite having a single child, the Souletin speaker would wait in the left queue.
Again, this should not be what the Souletin speaker would do in case the BN
denoted a plurality.
(25) Haur badüzü, jar zite eskerre-ko herroka-n.

child yes-you.have put AUX left-GEN queue-INE
‘If you have children, wait in the left queue.’
What these examples come to show is that BNs in Souletin denote the whole
lattice, and that it does not matter whether they make reference to a single ele-
ment, i.e. an atom, or to a plurality. In other words, BNs in Souletin are number
neutral; that is, they contain no number specifications at all and they can make
reference to any number of objects. (cf. Jespersen 1924, Chierchia 1998, Corbett
2000, Dayal 2004, Rullman & You 2006, Wilhelm 2008, etc.)
16 It is possible to use the partitive [-rik] in this context, i.e. badüzü haurrik?. However, there is
a preference for the BN. Cf. 3.2.
(26)
Further evidence for this idea comes from the possibility of having a BN in predi-
cative position which can be used to predicate of a singularity or of a plurality.
As shown in (27a) and (27b) respectively.17
(27) a. Jon erakasle da.

Jon teacher is
‘Jon is a teacher.’
b. Jon eta Miren erakasle dira.

Jon and Miren teacher are
‘Jon and Miren are teachers.’
The behaviour and distribution of BNs in Souletin reminds us of the behav-

iour and distribution of BNs in (at least some) Romance languages. In Spanish,
for example, BNs are possible in direct object position as shown in (28a,b) – or
in postverbal subject position, (28c), cf. Bosque 1996.
(28) a. Juan ha comido patatas.

Juan AUX eaten potatoes
‘Juan has eaten potatoes.’
b. Mikel ha bebido café.

Mikel AUX drunk coffee
‘Mikel has drunk coffee.’
c. Llegaron estudiantes.
arrived students
‘Students arrived.’
The BNs in the examples above can only obtain the so-called existential inter-
pretation, as is the case with Souletin BNs.
Now, in subject position (except for examples such as those in (28c)), the
presence of the definite article is necessary if the sentence is going to be gram-
matical. See Bosque (1996: p. 173).
17 For more discussion on this and for a possible analysis, cf. Etxeberria (to appear, in prep).
(29) a. *(Los) médicos trataron de salvar al niño.

DET. PL doctors treat of save to-DET. SG child
‘The doctors tried to save the child.’
b. *(Los) estudiantes comieron patatas.

DET. PL students eat potatoes
‘The students ate potatoes.’
c. *(Los) dinosaurios están extintos.

DET. PL dinosaurs are extinct
‘Dinosaurs are extinct.’
Despite similarities, there is a significant difference between Souletin BNs

and those of Romance languages, and this difference concerns number mark-
ing. Spanish plural BNs need to appear with the plural marker [-s] – not mass
terms – as can be seen in (30a), whereas Souletin BNs are real BNs and appear
without any kind of number marking, (30b).18,19
(30) a. Spanish BNs:

patata-s, leone-s, estudiante-s, (vino). . .
potato-pl lion-pl student-pl wine
b. Souletin BNs:
behi, ardi, mando, (arno). . .
cow sheep mule wine
Summarizing, this section has first provided a general description of the use
of the Basque definite article to then present its behaviour in various dialects.
From what we have seen, the more we move to the eastern provinces of the
Basque County, the less we use the definite article. Whereas western-central dia-
lects (cf. §2.1) need the presence of the definite article in all argumental posi-
tions, this is not so in Souletin where BNs are allowed in direct object position
(cf. §2.2).
The following section concentrates on the use of the partitive [-rik] in different
Basque dialects. But before we do that, we will make a general (non-exhaustive)
description of this particle.
18 Cf. Etxeberria & Etxepare (2009) for a possible analysis of this property.
19 The behaviour of Souletin BNs appears to be similar to the BNs of languages such as
Korean, Japanese, Dene Suline, i.e. to articleless languages. There is a significant different
among these languages and Souletin: Souletin does have a definite [-a(k)] as well as an
indefinite article bat ‘one’. For more on this, cf. Etxeberria (to appear, in prep).
3 The partitive particle [-rik]

Many authors have treated the partitive particle [-rik] as an article (cf. Larramendi
1927; Azkue 1905, 1923 among others; cf. (see de Rijk) 1972 for historical refer-
ences; cf. also de Rijk 2008). If this is really the case (and it is true that syntac-
tically it behaves as an article in polarity contexts; however, interesting as this
discussion may be, I won’t get into it) the partitive must be a special kind of an
article since it does not accept case markers (in opposition to what happens
with any other Basque article, e.g. the definite article). In this paper, we will
use the term ‘partitive’ to refer to the particle [-rik] (in line with other traditional
grammars; cf. Lafitte 1962).20
The partitive only attaches to transitive direct objects (31) and to intransitive
subject (32) and it behaves as a “polarity element” in that it only appears in
polarity contexts.
(31) a. Kepa-k ez du baloi-rik ekarri.

Kepa-ERG no AUX ball-PAR bring
‘Kepa has not brought (any) ball.’
b. Maia-k ez du ardo-rik edan.

Maia-ERG no AUX wine-PAR drink
‘Maia has not drunk wine.’
c. *Katu-rik ez du jan sagu-rik.

cat-PAR no AUX eat mouse-PAR
(32) a. Mendia-n ez da hildako animalia-rik azaldu.

mountain-INE no AUX dead animal-PAR appear
‘No dead animal appeared in the mountain.’
b. Bilera-ra ez da irakasle-rik etorri.

meeting-ALL no AUX teacher-PAR come
‘No teacher has come to the meeting.’
20 The origin of the partitive is related to the ablative (see de Rijk 1996; although see
Ariztimuño this volume for arguments in favour of the idea that the partitives’ origin is related
to the dative case). In eastern dialects the form [-rik] is sometimes used instead of the ablative
[-tik]. Furthermore, [-rik] is used with an ablative sense in all Basque dialects: mendi-rik mendi
‘lit.: mountain-from mountain’, aho-rik aho ‘mouth-from mouth’, ate-rik ate ‘door-from door’,
etc; cf. also de Rijk (2008).
Concerning the meaning of the partitive, it makes completely unspecific

reference to what the nominal expression denotes (cf. also de Rijk 1996, Etxepare
2003). In other words, what the speaker knows is that there are no members (or
quantity) in the set denoted by the noun. Thus, it would seem that the partitive
NP makes reference to the whole lattice since it does not matter whether they
make reference to a single element, i.e. an atom, or to a plurality (cf. example
(26)), as was the case with the BNs in Souletin (cf. §2.2).
Some of the syntactic environments allowing the partitive are the following
(cf. de Rijk 1972, 1996; Trask 2003; Etxepare 2003 for a complete description): (i)
negative sentences (31–32); (ii) existential sentences (33a); (iii) yes/no questions
(33b); (iv) protasis of conditional (33c); (v) before clauses (33d); (vi) without
clauses (33e); (vii) superlative (33f); (viii) with some quantifiers (33g).
(33) a. Bada atzerritar-rik Donostia-n!21

yes-is foreigner-PAR Donostia-INE
‘There are foreigners in Donostia!’
b. Goxoki-rik nahi al duzu?

candy-PAR want Q AUX
c. Taxi-rik lortu nahi baduzu, hobe duzu

taxi-par get want yes-you.have better aux
ilara honetatik ez mugitu.
queue this-from no move
‘If you want to get a taxi, you better not move from this line.’
d. Tontakeria-rik egin baino lehen, joan zaitez etxe-ra.

silly thing-PAR do than before go AUX home-ALL
‘Go home before you do silly things.’
e. Diru-rik gabe atera naiz etxe-tik.

money-PAR without leave AUX home-from
‘I left home without money.’
21 Constructions such as those in (33a) are only possible as an exclamation; see de Rijk
(1972). The interpretation that these kinds of sentences get is parallel to bada atzerritar asko
‘there are many foreigners’; cf. example (16). If we add an adjective to the noun, the sentence
need not be an exclamation: ardo onik badute taberna honetan ‘they have good wine in this
bar’, gizon onik bada Euskal Herrian ‘there are good men in the Basque Country’ (cf. de
Rijk 1972: 178; cf. also 30th law of the Academy of the Basque Language, Euskaltzaindia –
http://www.euskaltzaindia.net/dok/arauak/Araua_0030.pdf).
f. Jostailu-rik polit-ena ni-k ekarri dut gaur.

toy-PAR nice-SUP I-ERG bring AUX today
‘I brought the nicest toy today.’
g. Zeresan-ik asko / gizon-ik aski

gossip-PAR many man-PAR enough
lagun-ik franko / Esker-rik asko!22
friend-PAR many thank-PAR many
‘many gossiping / enough men / many friends / many thanks’
This paper will only concentrate on the first three uses of those we just
mentioned. Note also that the paper will not consider either the use of the parti-
tive with participles – what de Rijk (1972) calls stative [-rik] – (Jon gaixorik dago
‘Jon is sick’) or the use of the partitive with the complementizer [-en] (ez dut uste
azalduko direnik ‘I don’t think they will come’ Cf. Laka 1990, Uribe-Etxebarria
1994).
The partitive particle can be argued to be the negative form of the exis-
tential interpretation (in absolutive case) of the Basque definite article [-a(k)]
(cf. Irigoien 1985, de Rijk 1972). Before I proceed, let me make a clarification
note on de Rijk (1972: 140): de Rijk argues that the English translation of the
Basque sentence in (34a) is (34b) – but see the glosses. He proposes (34c) as
the correct negative form of the sentence (34a); (34d) on the other, would not be
the correct negative form of (34a) since the article [-a] would only get a definite
interpretation.23
(34) a. Ijito-a ikusi degu. (de Rijk 1972: (6a))

gipsy-DET. SG see aux
‘We have seen the gipsy.’
b. We have seen a gipsy. (de Rijk 1972: (6a))
c. Ez degu ijito-rik ikusi. (de Rijk 1972: (7a))

no AUX gipsy-PAR see
‘We have not seen any gipsy.’
d. Ez degu ijito-a ikusi. (de Rijk 1972: (8a))

no AUX gipsy-DET. SG see
‘We have not seen the gipsy.’
22 This use can be said to be nowadays lost, except for the fossilized eskerrik asko ‘many
thanks’.
23 de Rijk (2008: 292) uses an example with a mass term instead of a count term as in (34a),
which is parallel to the sentence in (36) below.
As shown in section 2.1, the singular definite article [-a] can only be inter-
preted existentially in very specific contexts (so-called stereotypical contexts,
which are clearly related to possession) – examples (11–12) – and the example
in (34a) is not such a context. As a consequence, if we would translate the
sentence in (34a) to a language that contains a definite determiner, we would
be forced to use the definite determiner due to the fact that the only possible
interpretation of ijitoa ‘gipsy-DET. SG ’ in (34a) is definite and referential. Thus,
the correct English translation is the one we have in the glosses, i.e. we have
seen the gipsy (and not the one in (34b), for which Basque has a perfect transla-
tion: ijito bat ikusi dugu ‘lit.: gipsy one see aux’). And the negative form of (34a)
would be (34d); in both cases we are making reference to a specific gipsy.
Then, it is clear from the examples above that it is not possible to use the
partitive [-rik] as the negative form of elements that force a definite and specific
reading (as is the case with the article [-a] when combined with count terms). In
fact, we get exactly the same effect with the plural version of the definite article
[-ak] in (35a): if the sequence [noun+plural article] is interpreted as definite, its
negative form will also make use of the definite article [-ak] (35b). On the other
hand, if the sequence [noun+plural article] is interpreted existentially (remember
that this interpretation is only allowed in direct object position, cf. section 2, cf.
also Artiagoitia 2002, 2004; Etxeberria 2005, 2010, to appear), its negative form
will make use of the partitive [-rik], as shown in (35c) – note also that we use the
singular form of the auxiliary.
(35) a. Kepa-k ikasle-ak ikusi ditu. [definite / existential]

Kepa-ERG student-DET. PL see AUX : PL
‘Kepa has seen (the) students.’
b. Kepa-k ez ditu ikasle-ak ikusi. [definite / *existential]
Kepa-ERG no AUX : PL student-DET. PL see
‘Kepa has not seen the students.’
c. Kepa-k ez du ikasle-rik ikusi. [*definite / existential]
Kepa-ERG no AUX : SG student-PAR see
‘Kepa has not seen (any) students.’
We would observe exactly the same behaviour if we used the definite article
[-a] with mass terms (cf. example (10b)).24 Thus, the sentence in (36a) is ambiguous
24 Cf. section 2.2, where we briefly argue that mass terms are number neutral. It is due to
this number neutrality that the definite article [-a] that combines with mass terms in western-
central Basque cannot be considered a singular number marker (Etxeberria 2005, 2010 argues,
pace Artiagoitia 2004). Cf. Etxeberria (2005, 2010) for discussion on this and for arguments
in favour of the idea that [-a] is a definite determiner in every context, but very flexible in its
ability to type-shift, a property that allows us to account for the various interpretations that it
forces.
between a definite and an existential interpretation of the direct object: in the

definite interpretation, we would be talking about a specific cognac, e.g. one
that has been mentioned before in the conversation; in the existential inter-
pretation on the other hand, we would not be talking neither about a specific
cognac nor about a specific quantity of cognac. The negative form of the definite
interpretation is the one in (36b), whereas the negative form of the existential
interpretation will make use of the partitive particle [-rik] as in (36c).
(36) a. Kepa-k cognac-a edan du. [definite / existential]

Kepa-ERG cognac-DET. SG drink AUX
‘Kepa has drunk (the) cognac.’
b. Kepa-k ez du cognac-a edan. [definite / *existential]

Kepa-ERG no AUX cognac-DET. SG drink
‘Kepa has not drunk the cognac.’
c. Kepa-k ez du cognac-ik edan. [*definite / existential]

Kepa-ERG no AUX cognac-PAR drink
‘Kepa has not drunk any cognac.’
Summarizing, the properties of the Basque partitive particle [-rik] in general

are the following: (i) the partitive is used as the negative form of the existential
interpretation of the definite article (in some contexts, which vary depending
on the variety, as we will see below); (ii) the partitive makes reference to a non-
specific quantity of what the nominal expression denotes.
In what follows, we will observe the behaviour and use of the partitive [-rik]
in different dialects; recall that we will only concentrate on three of the uses
that we mentioned in (33): negative sentences, yes/no questions, protasis of
conditionals.
3.1 [-rik] in Bizkaia and Gipuzkoa

It is possible to argue that the use of the partitive in these varieties is parallel
to the use we described in the previous section. Thus, the partitive [-rik] would
be used as the substitute of the existential interpretation of the definite article
(plurals and masses) in polarity contexts.
(37) a. Ez dugu baloi-rik/ardo-rik erosi.

no AUX ball-PAR /wine-PAR buy
‘We haven’t bought any ball(s)/any wine.’
b. Erosi al duzue baloi-rik/ardo-rik?

buy Q AUX ball-PAR /wine-PAR
‘Did you buy any ball(s)/any wine?’
c. Goxoki-rik/ardo-rik nahi baduzu, ez mugitu hemendik.

candy-PAR /wine-PAR want if-you.have no move here-from
‘If you want candies/wine, don’t move from here.’
As it’s been said before, the meaning of the partitive makes reference to a
non-specific quantity of what the nominal expression denotes. In other words,
what the speaker aims at expressing is that there are no elements from the set
denoted by the noun that have been bought (in the examples above); and it is
not important whether the set denoted by the noun is formed by one, ten or a
thousand members. This way, the partitive makes reference to the whole lattice
(just like BNs in Souletin – cf. section 2.2). It is possibly due to this property that
some authors have taken the partitive as the negative form of singular definite
forms (ijitoa ‘gipsy-DET. SG ’ in (34a)); a wrong conclusion, as it’s been shown.
(38)
If in these varieties of Basque the partitive makes reference to the whole

lattice, a question that one may ask is: is there any other means by which these
varieties can make reference to the whole lattice, e.g. as is the case with BNs in
Souletin? Will the plural version of the definite article [-ak] in its existential
interpretation be able to refer to the whole lattice?
Initially, it seems that the answer to these questions should be negative.
That is, if we were to say the sentences in (24) and (25) in these varieties, the first
sentences that would come to our minds would be the ones in (39) and (40),
where we use the partitive.
(39) Baduzu haur-rik?

yes-you.have child-PAR
‘Do you have (any) children?’
(40) Haur-rik baduzu, jar zaitez ezkerre-ko ilara-n.

child-PAR yes-you.have put AUX left-GEN queue-INE
Even if we were parents of a single child, we would answer to the question in

(39) with a ‘yes’ or we would wait in the left queue. Thus, we could maintain
that the partitive makes reference to the whole lattice, as we just mentioned.
Considering that the partitive is used as the negative form of the existential
interpretation of the definite article [-a(k)], one could think that in positive
sentences the existential interpretation of the definite article makes reference to
the whole lattice, and this appears to be correct. To begin with, some speakers of
these varieties do allow the definite article [-ak] instead of the partitive [-rik] in
the examples (39)–(40).25
(390 ) Badituzu haurr-ak?

yes-you.have child-DET. PL
‘Do you have children?’
(400 ) Haur-rak badituzu, jar zaitez ezkerre-ko ilara-n.

child-DET. PL yes-you.have put AUX left.GEN queue-INE
Further evidence in favour of this idea comes from the following examples,
where the plural definite DPs kruasanak ‘croissant-D.pl’ and ahateak ‘duck-D.pl’
do not necessarily make reference to a plurality of croissants and ducks, respec-
tively. Example (41a) from Artiagoitia (2002: (13))
(41) a. Pako-k kruasan-ak jaten ditu gosaltze-ko.

Pako-ERG croissant-DET. PL eat.PROG aux breakfast-FOR
‘Pako eats croissants for breakfast.’
b. Goazen ahateak ikustera.

go duck-DET. PL see-ALL
‘Let’s go see ducks.’
Thus, these sentences would show that the existential interpretation of the
definite article makes reference to the whole lattice.
The next section concentrates on the behaviour of the partitive particle in
the varieties of Lapurdi and Zuberoa (and Low Navarre).
25 The combination of the definite article [-a] with mass terms, e.g. ogia ‘bread-D.sg’ or ardoa
‘wine-D.sg’, also makes reference to the whole lattice. Recall that mass terms have been
argued to be number neutral (and atomic, although in a vague way). See fn.24.
3.2 [-rik] in Lapurdi and Zuberoa

The use that Lapurdian gives to the partitive [-rik] is more restricted than the use
described in the previous section. In Lapurdi, the partitive can only be used in
negative contexts (42a); in yes/no questions (42b) and in conditionals (42c) they
use the plural definite article instead.
(42) a. Ez dugu baloi-rik erosi.

no AUX ball-PAR buy
‘We have not bought any ball.’
b. Erosi al dituzue baloi-ak?

buy Q AUX ball-DET. PL
‘Did you buy balls?’
c. Goxoki-ak nahi badituzu, ez mugitu hemen-dik.

candy-DET. PL want yes-you.have no move here-from
‘If you want candies, don’t you move from here.’
As expected, the partitive makes reference to the whole lattice and it makes
non-specific reference to what the noun denotes. The plural definite article
shows a similar behaviour as the noun can make reference to a singularity or a
plurality and it denotes the whole lattice. This is more clearly seen in the follow-
ing examples.
(43) a. Haurr-ak badituzu?

child-DET. PL yes-you.have
b. Haurr-ak badituzu, jar zaitez ezkerre-ko ilara-n.26

child-DET. PL yes-you.have put AUX left-GEN queue-INE
Note that the DP haurr-ak ‘child-D.pl’ in the examples above can only get an
existential interpretation.
The use of the partitive in Lapurdian is obviously affected by French.
In French, the partitive ‘de négatif’ can only appear in negative contexts (and
26 Most of the speakers that I have interviewed would use the sentences in (43) to translate
the sentences in (23). However, there are some speakers that could use the partitive in these
contexts. It is also worth noting that there are some speakers that could use the partitive or the
in direct object position); cf. Abeille et al. (2004), Bartning (1996), Carlier (2004),
Englebert (1996), Heyd (2003), Kupferman (1996), Zribi-Hertz (2003).27
(44) Je n’ai pas d’enfants.

‘I don’t have (any) children.’
The use of the ‘de négatif’ is not correct neither in yes/no questions nor in
conditional sentences, where French would make use of the so-called partitive
determiners des/du as shown in the following examples.
(45) a. Est-ce que vous avez des enfants?

b. Est-ce que vous avez du vin?
‘Do you have wine?’
c. Si vous avez des enfants, mettez-vous sur la file de gauche.
Note that the existential reading of d’enfants in (44) and des enfants in (45)
(which is the only reading they can get, in opposition to what happens with the
Basque definite article; cf. section 2) makes reference to the whole lattice. The
partitive [-rik] and the plural form of the definite article [-a(k)] force these same
readings in Lapurdian. With this in mind, we could be led to think that the use
of the partitive in Lapurdian was more extended than what it is now (cf. Lafitte
1944, where it is argued that the partitive was used in yes/no questions and in
conditional sentences). Another noteworthy point is the following: the existen-
tially interpreted Basque definite article and the French des/du show a quite
similar behaviour since they are both (i) rejected as objects of generic sentences;
definite article in yes/no questions; these speakers find an interesting difference between the
sentences in (i) and (ii).
(i) Baduzue ogi-a?
yes-you.have bread-DET. SG
‘Do you have bread?’
(ii) Baduzue ogi-rik?

yes-you.have bread-PAR
‘Do you have (any) bread?’
They would use the partitive in (ii) while asking for bread in a bakery, where it is expected that
they will have some bread. However, the sentence in (i) would be more naturally used in e.g. a
petrol station, where it is not obvious that they will have bread for sale. In a neutral situation,
these speakers would use the sentence in (i). More research is needed to clarify this point.
27 The partitive de is also used with some weak quantifiers in French: beaucoup d’étudiant
‘lit.: many of student’, beaucoup de livre ‘lit.: many of book’, peu de femme ‘lit.: few of girl’,
etc. Cf. Doetjes (1997, 2002). Cf. example (33g) and footnote 23; cf. also Etxeberria in prep.
(ii) perfectly acceptable as objects of stage-level predicates; (iii) grammatical also

when combined with atelic adverbials (cf. Etxeberria 2005, 2010 for more on this).
The behaviour of the partitive [-rik] in Souletin is similar to the behaviour of
the partitive we just described for Lapurdian. The partitive can only be used in
negative contexts; in yes/no questions and in conditional sentences we would
use BNs as shown in (46b–c) (cf. also examples (24–25)).
(46) a. Ez dügü balu-rik erosi

no AUX ball-PAR buy
‘We have not bought any ball.’
b. Balu erosi düzüeia?

ball buy AUX
‘Did you buy balls?’
c. Huntto nahi badüzü, ez zitila hebentik igi

candy want if-you.have no AUX here-from move
‘If you want candies, don’t you move from here.’
As it’s been mentioned before (cf. section 1.2), in the examples in (46), the
partitive balurik ‘ball-part’, and the BNs balu ‘ball’ and huntto ‘candy’ make
reference to the whole lattice. Thus, the aim of using these nominal expressions
is to make non-specific reference to what the noun denotes in the real world.
However, it seems that not all Souletin speakers would agree with what
I said above: the grammars by Casenave-Harigile (2006) or Etxebarne (2006)
argue that the behaviour of the partitive in Souletin is similar to the behaviour
of the partitive in western-central dialects, and that it can be used in negative
contexts, in yes/no questions, and in conditional sentences (as well as in other
contexts, cf. (33)).
At this point, due to lack of data (and this is exclusively my fault), I’m
unable to argue in favour of one system (where the partitive [-rik] is used in
negative contexts only) or the other (where the partitive [-rik] is used in more
contexts).28 One possible scenario (specially in eastern dialects; although this
could also be assumed for western-central dialects, cf. section 3.1) is one where
both systems would coexist. If this scenario is plausible at all, one could think
that one of the systems will prevail over the other, and it seems that the system
28 I do not want anyone to doubt about the correctness of the data presented by Casenave-
Harigile (2006) and Etxebarne (2006). However, observing that the data I collected and the
data that they offer do not coincide (completely), I would like to make a deeper study of the
topic (cf. Etxeberria in prep).
that would prevail will be the more restricted one, i.e. the system where the
partitive can only be used in negative contexts (and which will parallel French).
Thus, the prediction is that the existentially interpreted plural definite DP (in
Lapurdian) and the BNs (in Souletin) will gradually take the place of the partitive
[-rik]. The projects Towards a Syntactic Atlas of the Basque Language (http://
www.iker.cnrs.fr/-tsabl-towards-a-syntactic-atlas-of-.html) and Syntactic micro-
variation in Basque: a theoretical and typological approach (http://basdisyn.net/),
which investigate the syntax of different Basque dialects, will hopefully shed
some more light on this.
References
Abeillé, Anne, Olivier Bonami, Danièle Godard & Jesse Tseng. 2004. The Syntax of French de-N’
phrases. In Stefan Müller (ed.), Proceedings of the 11th International Conference on Head-
Driven Phrase Structure Grammar, 6–26. Stanford, CA: CSLI Publications.
Artiagoitia, Xabier. 1997. DP predicates in Basque. In Alice Taff (ed.), University of Washington
Working Papers on Linguistics 15. 161–198.
Artiagoitia, Xabier. 1998. Determinatzaile Sintagmaren Hipotesia Euskal Gramatikan. Uztaro 27:
33–61.
Artiagoitia, Xabier. 2002. The functional structure of Basque noun phrases. In Xabier Artiagoitia
et al. (eds), Erramu Boneta: Festschrift for Rudolf P. G. de Rijk, 73–90. Vitoria-Gasteiz:
ASJU.
Artiagoitia, Xabier. 2004. Izen-sintagmaren birziklatzea: IS-tik izenaren inguruko funtzio-
buruetara. In Pablo Albizu & Beatriz Fernandez (eds.), Euskal Gramatika XXI Mendearen
Atarian: Arazo Zaharrak, Azterbide Berriak, 13–38. Vitoria-Gasteiz: UPV/EHU.
Artiagoitia, Xabier. 2006. Euskarazko izen sintagma: arkitektura eta egitura funtzionala.
Unpublished manuscript, UPV/EHU.
Artiagoitia, Xabier. 2012. The DP hypothesis in the grammar of Basque. In Urtzi Etxeberria et al.
(eds.), Noun Phrases and Nominalization in Basque: Syntax and Semantics, 21–78.
Azkue, Resurreción M. 1905. Diccionario vasco-español-francés. Bilbao.
Azkue, Resurreción M. 1923. Morfologia Vasca. Bilbao: La Gran Enciclopedia Vasca. [Edizio berria:
1969]
Bartning, Inge. 1996. Eléments pour une typologie des SN complexes en ‘de’ en français.
Langue française 109. 29–43.
Bosque, Ignacio. 1996. El sustantivo sin determinación. Madrid: Visor Libros.
Bosveld-de Smet, Leonie. 1997. On mass and plural quantification: The case of French des/
du-NPs. Groningen: University of Groningen dissertation.
Bresnan, Joan. 1994. Locative inversion and the architecture of Universal Grammar. Language
70: 72–131.
Carlier, Anne. 2004. Sur les premiers stades de développement de l’article partitif. Scolia 18:
117–147.
Carlson, Greg. 1977. Reference to kinds in English. Amherst, MA: UMass dissertation.
Carlson, Greg & Jefrey Pelletier. 1995. The generic book. Chicago: University of Chicago Press.
Casenave-Harigile, Junes. 2006. Xiberotar gramatika llaburra. Maule: Sü Azia.
Chierchia, Gennaro. 1998. Plurality of mass nouns and the notion of ‘semantic parameter’. In
Susan Rothstein (ed.), Events and Grammar, 53–103. Dordrecht: Kluwer.
Chierchia, Gennaro. 2009. Mass nouns, vagueness and semantic variation. Synthese 174. 99–
149.
Corbett, Greville G. 2000. Number. Cambridge: Cambridge University Press.
Coyos, Battittu. 1999. Le parler basque souletin des Arbailles: Une approche de l’ergativité.
Paris: L’Harmattan.
Dayal, Veneeta. 2004. Number marking and (in)definiteness in kind terms. Linguistics & Philos-
ophy 27. 393–450.
Delfitto, Denis & Jan Schroten 1991. Bare Plurals and the Number Affix in DP. Probus 3.2: 155–
186.
Dobrovie-Sorin, Carmen, Tonia Bleam & Maria Teresa Espinal. 2006. Bare nouns, number and
types of incorporation. In Svetlana Vogeleer & Liliana Tasmowski (eds.), Non-definiteness
and plurality, 51–79. Amsterdam: Benjamins & Linguistics Today.
Doetjes, Jenny. 1997. Quantifiers and Selection. On the distribution of quantifying expressions
in French, Dutch and English. University of Leiden dissertation.
Doetjes, Jenny. 2002. Beaucoup est Ailleurs. Expressions de degré et sous-spécification catégo-
rielle. In R. Bok-Bennema, B. de Jonge, B. Kampers-Manhe and A. Molendijk (eds.), Adver-
bial Modification, 125–138. Amsterdam: Rodopi.
Eguren, Luis. 2005. Marcas de predicación en vasco. In Beatriz Fernández & Itziar Laka (eds.),
Andolin gogoan: Essays in Honour of Professor Eguzkitza, 233–250. Bilbo: UPV/EHU.
Eguren, Luis. 2006. Non-canonical uses of the article in Basque. In Michael J. Hauser et al. (eds.),
Proceedings of the 32nd Annual Meeting of the Berkeley Linguistics Society. Berkeley.
Eguren, Luis. 2012. Predication markers in Basque. In Urtzi Etxeberria et al. (eds.), Noun Phrases
and Nominalization in Basque: Syntax and Semantics, 243–267. Amsterdam: Benjamins.
Englebert, Annick. 1996. L’article partitif: l’evolution des conditions d’emploi. Langue française
109. 9–28.
Espinal, Maria Teresa. 2010. Bare nominals in Catalan and Spanish: Their structure and
meaning. Lingua 120. 984–1009.
Espinal, Maria Teresa & Louise McNally. 2007. Bare singular nominals and incorporating verbs.
In Georg Kaiser & Manuel Leonetti (eds.), Proceedings of the Workshop. Definiteness,
specificity and animacy in Ibero-Romance Languages. Arbeitspapier 122. 45–62. Konstanz:
Universität Konstanz.
Etxebarne, Jüje. 2006. Xiberotar gramatika. Maule: Sü Azia.
Etxeberria, Urtzi. 2005. Quantification and domain restriction in Basque. UPV/EHU & HiTT
dissertation.
Etxeberria, Urtzi. 2007. On quantification in Basque and on how some languages restrict their
quantificational domain overtly. In Matthewson, L. (ed.), Quantification: A crosslinguistic
perspective, 225–276. London: Emerald.
Etxeberria, Urtzi. 2008. On quantification in Basque and on how some languages restrict their
quantificational domain overtly. In Lisa Matthewson (ed.), Quantification: A crosslinguistic
perspective, 225–276. London: Emerald.
Etxeberria, Urtzi. 2009. Contextually restricted quantification in Basque. In Anastasia Giannakidou
& Monica Rathert (eds.), QP Structure, Nominalizations, and the Role of DP, 76–107.
Oxford: Oxford University Press. Oxford Studies in Theoretical Linguistics Series.
Etxeberria, Urtzi. 2010. Making a definite be interpreted as an indefinite. In Roberta Pires &
Carmen Dobrovie-Sorin (eds.), Thematic Volume on Bare Nominals – Journal of Portuguese
Linguistics.
Etxeberria, Urtzi. 2012a. Quantification in Basque. In Ed Keenan & Denis Paperno (eds.), Hand-
book of Quantifiers in Natural Language, 83–164. Berlin: Springer, Studies in Linguistics
and Philosophy.
Etxeberria, Urtzi. 2012b. The way the definite article affects quantifiers in Basque – and
beyond. In Etxeberria, Urtzi, Ricardo Etxepare & Myriam Uribe-Etxebarria (eds.), Noun
Phrases and Nominalizations in Basque, 79–109. Amsterdam: Benjamins.
Etxeberria, Urtzi. To appear. Basque nominals: From a system with bare nouns to a system
without. In Ana Aguilar-Guevara, Bert Le Bruyn and Joost Zwarts (eds.), Weak Referentiality.
Amsterdam: Benjamins. Linguistics Today series.
Etxeberria, Urtzi. In preparation. Nominal expressions in Basque. IKER & CNRS thesis.
Etxeberria, Urtzi & Ricardo Etxepare. 2009. Zenbatzaileak komunztatzen ez direnean: Hiru
sistema. Lapurdum 13, UMR 5478 : CNRS (Université Michel de Montaigne Bordeaux III –
UPPA) Département interuniversitaire d’études basques de Bayonne.
Etxeberria, Urtzi & Anastasia Giannakidou. 2010. Contextual Restriction and the Definite
Determiner. In François Recanati et al. (eds.), Context-Dependence, Perspective and Relativity,
93–126. Berlin: Mouton de Gruyter. Mouton Series in Pragmatics 6.
Etxeberria, Urtzi, Ricardo Etxepare & Myriam Uribe-Etxebarria (eds.), Noun Phrases and
Nominalization in Basque: Syntax and Semantics. Amsterdam: Benjamins.
Etxepare, Ricardo. 2003. Negation. In Jose Ignacio Hualde & Jon Ortiz de Urbina (eds.), A
Grammar of Basque, 387–421. Berlin: Mouton de Gruyter.
Euskaltzaindia. 1993. Euskal Gramatika Laburra: Perpaus Bakuna, Bilbo: Euskaltzaindia.
Fara, Delia Graff. 2001. Descriptions as predicates. Philosophical Studies 102 (1): 1–42.
Fernandez-Leborans, María Jesús. 1999. La predicación: las oraciones copulativas. Gramática
Descriptiva de la Lengua Española. Madrid: RAE. Cap. 37.
Freeze, Ray. 1992. Existentials and Other Locatives. Language 68.3. 553–595.
Goenaga, Patxi. 1980. Gramatika Bideetan. Donostia: Erein.
Goenaga, Patxi. 1991. Izen Sintagmaren Egituraz. In Joseba Lakarra (ed.), Memoriae L. Mitxelena
Magistri Sacrum, 847–865. Donostia: UPV/EHU.
Heyd, Sophie. 2003. L’interprétation des Syntagmes Nominaux en “des” et “de” en Position Sujet
et Objet. Généricité, Habitualité et Incorporation Sémantique. Université Strasbourg II –
Marc Bloch dissertation.
Irigoien, Alfonso. 1985. Euskarako izen sintagma mugatzailerik gabekoez. Euskera XXX: 129–
139.
Jespersen, Otto. 1924. The philosophy of grammar. London: Allen & Unwin.
Kleiber, Georges. 1990. L’article LE générique. La généricité sur le mode massif. Genève-Paris:
Droz.
Krifka, Manfred. 2004. Bare NPs: Kind-referring, indefinites, both, or neither? In O. Bonami &
P. Cabredo-Hofherr, (eds.), Empirical Issues in Formal Syntax and Semantics 5; Selected
Papers from CSSP 2003. 111–132.
Krifka, Manfred, Francis Jeffry Pelletier, Gregory N. Carlson, Alice ter Meulen, Gennaro Chierchia
& Godehard Link. 1995. Genericity: An introduction. In Greg Carlson & Francis Jeffry Pelletier
(eds.), The Generic Book, 1–124. Chicago: University of Chicago Press.
Kupferman, Lucien. 1996. Un bien grand mot: de. De la préposition au mode de quantification.
Présentation. Langue française 109. 3–8.
Laca, Brenda. 1996. Acerca de la Semántica de los “Plurales Escuetos” del Español. In Bosque, I.
(ed.), El Sustantivo sin Determinación. La Ausencia de Determinante en la Lengua Española,
241–268. Madrid: Visor Libros.
Lafitte, Pierre. 1944. Grammaire Basque (Navarro-Labourind littéraire). Donostia: Elkar.
Lafitte, Pierre. 1962. Grammaire basque (navarro-labourdin littéraire). Revised edition, Bayonne.
Laka, Itziar. 1990. Negation in Syntax: On the Nature of Functional Categories and Projections,
MIT dissertation.
Laka, Itziar. 1993. Unergatives that assign ergative, unaccusative that assign accusative.
In Jonathan Bobaljik & Collin Phillips (eds.), Papers on Case and Agreement I, Cambridge,
MA: MIT Working Papers in Linguistics, 149–172.
Larramendi, Manuel. 1927. El impossible vencido. Arte de la lengua bascongada. Salamanca.
Luján, Marta. 1981. The Spanish copulas as aspectual indicators. Lingua 54: 165–210.
Manterola, Julen. 2006. -a euskal artikulu definituaren gainean zenbait ohar. In Joseba Lakarra &
Jose Ignacio Hualde (eds.) Studies in Basque and Historical Linguistics in Memory of R.L.
Trask – R.L. Trasken oroitzapenetan ikerketak euskalaritzaz eta hizkuntzalaritza historikoaz,
651–676. Gipuzkoako Foru Aldundia-EHU, Donostia-Bilbo. (Supplements of ASJU).
Manterola, Julen. 2008. -a morfemaren erabilera (eza) ekialdeko euskaretan. In Joaquin Gorrotxategi
et al. (eds.), Koldo Mitxeleran Katedraren II. Biltzarraren Aktak, Vitoria-Gasteiz: ASJU.
Manterola, Julen. 2009. -a and bat Basque articles and recent contact theories. Ms.: EHU.
Manterola, Julen. 2012. Synchronic ubiquity of the Basque article -a: a look from diachrony. In
Urtzi Etxeberria, Ricardo Etxepare & Myriam Uribe-Etxebarria (eds.), Noun Phrases and
Nominalization in Basque: Syntax and Semantics, 179–208. Amsterdam: Benjamins.
Matushansky, Ora. 2005. Iraqui head seeks arms: Are bare nouns created equal?. Paper
presented at the Bare Workshop, Utrecht.
Milsark, Gary. 1977. Existential Sentences in English. New York: Garland Publishing Inc.
Rijk, Rudolf P. G. de. 1972. Partitive assignment in Basque. In ASJU 6: 130–173. [Reprinted in
Rijk 1998].
Rijk, Rudolf P. G. de. 1996. On the origin of the partitive determiner. In ASJU 30 [Reprinted in de
Rijk 1998].
Rijk, Rudolf P. G. de. 1998. De lingua vasconum: Selected writings. EHU. (Supplements of
ASJU 43).
Rijk, Rudolf P. G. de. 2008. Standard Basque, A Progressive Grammar. Cambridge, MA: MIT Press.
Rodriguez, Sonia. 2003. Euskal artikuluaren sintaxiaz. Ms., UPV/EHU.
Roodenberg, Jasper. 2004. French bare arguments are not extinct: The case of coordinated bare
nouns. Linguistic Inquiry 35(2). 301–313.
Rullmann, Hotze & Aili You. 2006. General number and the semantics and pragmatics of in-
definite bare nouns in Mandarin Chinese. In Klaus von Heusinger et al. (eds.), Where
Semantics Meets Pragmatics, 175–196. Amsterdam: Elsevier.
Santazilia, Ekaitz. 2009. Artikuluaren 1895eko argazki bat VDM programaren bitartez. Ms.
UPV/EHU.
Schmitt, Cristina. 1992. Ser and estar: a matter of aspect. In K. Broderick (ed), Proceedings of
NELS 22, 411–426. MA: GLSA.
Ticio, Emma. 1996. El morfema -a en Vasco como marca de referencialidad. In Interlingüística
VI. AJL.
Trask, Robert Lawrence. 2003. The Noun Phrase: Nouns, Determiners, and Modifiers; Pronouns
and Names. In Jose Ignacio Hualde & Jon Ortiz de Urbina (eds.), A Grammar of Basque,
92–134. Dordrecht: Kluwer.
Txillardegi. 1977. L’emploi de l’indéfini en souletin. Fontes Linguae Vasconum 25. 29–54.
Uribe-Etxebarria, Myriam. 1994. Interface Licensing Conditions on Negative Polarity Items: A
Theory of Polarity and Tense Interactions, University of Connecticut dissertation.
Wilhelm, Andrea. 2008. Bare nouns and number in Dene Suline. Natural Language Semantics
16. 39–68.
Zabala, Igone. 1993. Predikazioaren Teoriak Gramatika Sortzailean. UPV/EHU dissertation.
Zabala, Igone. 2003. Nominal Predication: Copulative Sentences and Secondary Predication.
In Jose Ignacio Hualde & Jon Ortiz de Urbina (eds.), A Grammar of Basque, 324–339.
Dordrecht: Kluwer.
Zamparelli, Roberto. 1998. A Theory of Kinds, Partitives and OF/Z Possessives. In Alexiadou, A.
& C. Wilder (eds.), Possessives, Predicates and Movement in the Determiner Phrase, 259–
304. Amsterdam/Philadelphia: John Benjamins, Linguistic Today 22.
Zamparelli, Roberto. 2000. Layers in the Determiner Phrase. Rochester, NY: University of Rochester
dissertation.
Zamparelli, Roberto. 2002a. Definite and bare kind-denoting noun phrases. In Claire Beyssade
et al. (eds.), Romance Languages and Linguistic Theory 2000, 305–342. Amsterdam:
Benjamins.
Zamparelli, Roberto. 2002b. Dei ex machina: A note on plural/mass indefinite determiners.
Ms., Università di Bergamo.
Zribi-Hertz, Anne. 2003. Pour une analyse unitaire de ‘de’ partitif. In Francis Corblin et al.
(eds.), Indéfini et Prédication, 141–167. Paris: PUPS.
Zuazo, Koldo. 2008. Euskalkiak. Euskararen dialektoak. Donostia: Elkar.
Borja Ariztimuño López
8 The origin of the Basque partitive*
The aim of this paper is to offer a diachronic explanation of the Basque partitive
suffix -(r)ik (which is undoubtedly related to the old ablative -(r)ik; Mitxelena
[1961] 1977: 236). It also attempts to find out the etymological source of this
morpheme (for which the Proto-Basque verb *din ‘to come’ may be claimed),
and poses some evolutionary paths for the different uses that it has in Modern
Basque. All of that taking into account the cross-linguistic data, and drawing on
the grammaticalization theory.
Keywords: Basque partitive, grammaticalization, ablative, diachrony
1 Introduction
It is well-known that the partitive case of Basque (-rik after vowel, -ik after con-
sonant) is considered as such because of its expression of indefinite quantity,
and especially because it is (also) employed in negations and questions, as for
example the French de-négatif and partitive determiners des / du. However, as it
also expresses indefiniteness (it can never be used to refer to a definite object,
unlike, for example, the Finnish partitive), it is considered to be an indefinite
or polar determiner (at least in some of its uses), as opposed to the common def-
inite determiner or article – singular -a, plural -ak – of Basque (de Rijk 1972;
* This research is funded by a Predoctoral Grant of the Basque Government [BFI 2009-236],
and has been carried out within the Training and Research Unit UFI-11/14 of the University of
the Basque Country (UPV/EHU), the Consolidated Research Group “Historia de la lengua vasca
y lingüística histórico-comparada” (GIC10/83-IT-486-10) of the Basque Government, and the
research project “Monumenta Linguae Vasconum IV: Textos Arcaicos Vascos y euskera antiguo”
[FFI2012-37696] (MINECO).
I appreciate the chance that the SLE gave me to attend the workshop on partitives within its
43rd Annual Meeting, held in Vilnius in September 2010, and I am grateful to the editors Silvia
Luraghi and Tuomas Huumo for publishing my paper in this volume, which surely will serve
to clarify a little bit more the differences and similarities between the partitives among the lan-
guages of the world. I would also like to thank professors Joseba A. Lakarra, Ricardo Gómez
and Denis Creissels, for their remarks on a previous version of this paper, as well as Silvia
Luraghi, Tommaso Claudi and Julia Roses for their help in improving the readability of the text.
324 Borja Ariztimuño López
Laka 1996: §4.2; Trask 2003: 124–126). Thus, the impossibility of combining the
partitive with a determiner (unlike all the other cases) is interpreted as a result
of its very determiner nature (cf. Tab. 1); nevertheless, synchronic descriptions
need to make some sort of comparison between the partitive and the absolutive
case, in order to establish the correct use of the partitive.
Notably, however, a similar reasoning could be used conversely to defend
the casual nature of the partitive, since, like the other cases, and contrary to
the determiners, it does not allow any other case to be attached to it. Moreover,
it can be said that the partitive can only be attached to the indefinite form of a
word (cf. Tab. 2). In any way, a distinction must be made between the so-called
grammatical cases (to which the partitive is usually ascribed, mostly under the
category of the absolutive, cf. Hualde 2003: 185) and the peripheral cases (espe-
cially spatial ones), for they do not reflect the number / determinateness differ-
entiation in the same way (cf. Tab. 1–2); this irregularity will help us to clarify
the origin of the partitive (see also Tab. 3 and Fig. 4 in section 6).
Table 1: The Basque partitive as a determiner
INDF DET. SG DET. PL PAR
ABS -Ø-Ø -a-Ø -ak-Ø -(r)ik-Ø

ERG -Ø-k -a-k -e-k ―
DAT -Ø-(r)i -a-ri -e-i ―
ALL -ta-ra -Ø-ra -eta-ra ―
ABL -ta-tik -Ø-tik -eta-tik ―
LOC -ta-n -Ø-an -eta-n ―
Table 2: The Basque partitive as a case suffix
INDF DET. SG DET. PL
ABS -Ø-Ø -a-Ø -ak-Ø

ERG -Ø-k -a-k -e-k
DAT -Ø-(r)i -a-ri -e-i
PAR -Ø-(r)ik ― ―
ALL -ta-ra -Ø-ra -eta-ra
ABL -ta-tik -Ø-tik -eta-tik
LOC -ta-n -Ø-an -eta-n
However, I have attempted to give only a diachronic explanation here,

because the borderline between the determiner and the casual nature of the
partitive suffix in Modern Basque is sometimes quite thin, and it might be seen
The origin of the Basque partitive 325
as syntacticians’ or semanticians’ matter (for a discussion about the comple-

mentariness of the partitive and the definite article see Etxeberria 2010). In
short, this suffix can be used where it would be possible to use the absolutive,
that is, with objects of transitive verbs and with subjects of intransitive ones, but
unlike that case it doesn’t make any number or definiteness distinction, as if it
was a determiner itself.1
The paper is organized as follows. I start by showing briefly the main uses
and meanings of the partitive marker in the Basque language, beginning with the
current situation (section 2) and explaining later on some differences between
the modern language and the old written data (section 3). Then, in section 4, I
explain separately a particular usage of the partitive in superlative construc-
tions, followed by an approach to the “adverbial partitive” (section 5); in both
cases I present tentative reconstructions of their syntactic and semantic develop-
ment. To conclude, on the basis of the historical relationship – both formal and
semantic – between the partitive and the ablative in Basque, in section 6 I try
to state the etymology and a possible process of grammaticalization of the
morpheme itself. I leave aside another topic, that is, how the Basque language
expresses Partitive and Pseudo-Partitive Constructions (as described by Maria
Koptjevskaja-Tamm 2001) which, despite the name, are not always made by
means of the partitive suffix, but through the modern ablative case (remember
their genetic relationship), the genitive, or the mere juxtaposition.
2 The modern Basque partitive

In addition to the general constraint in the argument types that can be expressed
through a partitive-marked constituent, there are some syntactic environments
that license the use of this suffix. A description of the main ones is provided in
the following subsections (for the superlative-partitive see section 4).2
1 For a good semantic approach to the modern Basque partitive see Etxeberria (this volume).
2 For an account of all the partitive-licensing contexts see Ricardo Etxepare (2003: 549–554).
These contexts are basically some uses related to the main ones exposed in this paper or
variants of them.
2.1 Negation
One of the main conditions in Basque for a partitive-marked argument to appear
is that it falls under the scope of the negation particle ez (1–2). Below are
provided some such examples:3
Object of a transitive verb4

(1) a. Hemen ez dugu gazte-rik euskara aberats
here NEG we.have young-PAR Basque.language rich
bat-en jabe
one-GEN owner
‘We don’t have any young person owning a rich Basque language here.’
(Berria, 11-01-2004)
b. Mila milioi pertsona-k baino gehiago-k ez

thousand million person-ERG than more-ERG NEG
dute edateko ur-ik

they.have drinkable water-PAR
‘More than a thousand millions of people don’t have (any) drinkable
water.’ (Berria, 27-08-2004)
Subject of an intransitive verb

(2) a. Gorroto-rik ez da futbol-ean
hate-PAR NEG be football-LOC
‘There is not (any) hate in football.’
(Berria, 24-11-2004)
b. Eser zaitez, hemen ez dabil haize-rik

sit AUX .IMP here NEG it.moves wind-PAR
‘Sit down, the wind doesn’t blow here.’
(Anjel Lertxundi, Azkenaz beste, 2005)
3 Examples of Modern Basque have been obtained mostly from the Ereduzko Prosa Gaur
[Modern Exemplary Prose] corpus. Most historical or Old Basque data come from the corpus
of Klasikoen Gordailua [Depository of Classic Works] and the Basque General Dictionary of
Mitxelena & Sarasola (1987–2005). Citations of Lazarraga’s manuscript are, however, from
Bilbao et al. (2012).
4 I have aimed to state explicitly the two possible arguments with which the partitive can be
used, namely transitive objects and intransitive subjects, only in this subsection, because I
consider it to be too repetitive for the purpose.
The negation feature may also be due to the presence of other elements like
the postposition gabe ‘without’ (3) or the adverbial nekez ‘hardly’ (4).
(3) Bigarren itzuli-an partida-rik joka-tu gabe dago

second return-LOC match-PAR play-PTCP without he.remains
‘He keeps without playing any match in the second leg.’
(Berria, 23-02-2006)
(4) Txantxa-k nekez edirenen zuen haren mihi-an

joke-ERG hardly find:FUT aux.pst that:GEN tongue-LOC
mintzabide-rik
way.of.talking-PAR
‘The joke would hardly find place in his speech.’
(Joan Mari Irigoien, Lur bat haratago, 2000)
The negation is one of the licensing environments of the partitive also in

Finnish, with the difference that in this language the determined partitive (6)
and the affirmative partitive (7) exist as well. Remarkably, the French so-called
de-négatif (5a) has a similar function of the Basque partitive, as both are used
(mainly)5 in negative contexts and cannot take any determiner.6
French:
(5) a. Elle n’a pas mangé de carottes.
she has not eaten of carrots
‘She didn’t eat any carrots.’
(Hoeksema 1996: 16)
b. *Elle a mangé de carottes.

‘She has eaten of carrots.’
(Hoeksema 1996: 16)
5 Mainly, because the same preposition alone replaces the indefinite article in some affirma-
tive contexts (e.g. J’ai de belles fleurs ‘I have got beautiful flowers’).
6 Actually, in diachronical terms, the French partitive article is formed by the same preposition
de plus a definite article. It may be said that the de-négatif and the preposition of sentences
like that of the previous footnote is a partitive marker, or a preposition that functions as such.
For an extensive account of the evolution of the Latin ablatival preposition de into modern
Romance languages (such as French) see Carlier & Lamiroy (this volume), and for a cross-lin-
guistic overview of the relation between negation and partitive see Miestamo (this volume).
Finnish:
(6) En syönyt omen-a
NEG .1 SG eat:PTCP apple-PAR
‘I did not eat / was not eating an / the apple.’
(Koptjevskaja-Tamm & Wälchli 2001: 652)
(7) Matti osti olu-tta

Matti buy:IMPF.3 SG beer-PAR
‘Matti bought (some) beer.’
(Koptjevskaja-Tamm & Wälchli 2001: 652)
Thus, sentences like (5b) and (7) are nowadays totally ungrammatical in
Basque, as would be (6) if the determinate interpretation was chosen (‘the
apple’).
Hence, the partitive with the negation implies the “total” or, in other words,
“indefinite” negation. That is, it refers to the subject / object as a kind, not as a
specific one. This aspect is clearly seen in the more abstract (and less relevant to
speakers) semantic contrast between absolutive and partitive, where what is
being denied is something apparently specific and unique (as in the case of
proper names):7
(8) a. Ez duzu hor Peru-rik topatuko

NEG AUX there Peter-PAR find:FUT
‘You will not find Peter there.’ (emphatic)
‘You will not find any[body like] Peter there.’ (metaphoric)
b. Ez duzu hor Peru topatuko

NEG AUX there Peter find:FUT
‘You will not find Peter there.’
(9) a. ETA-rik ez ba-lego. . .

ETA-PAR NEG if-be:IRR .3 SG
‘If there were not ETA. . .’ (emphatic)
‘If there were neither ETA nor any such armed organization. . .’
(metaphoric) (Joxe Azurmendi, Euskal Herria krisian, 1999)
7 Another possible though not exclusive interpretation of negative sentences with the partitive
is the following: ‘You will not find any[body called] Peter there’ (impersonal), as in the question
Perurik bada hor? ‘Is any[body called] Peter there?’ [lit. ‘Is there any Peter?’] vs. Peru hor da? ‘Is
Peter there?’ (cf. Laka 1996: §2.2).
b. ETA ez ba-lego. . .
ETA NEG if-be:IRR .3 SG
‘If there were not ETA. . .’
Similarly, in non-finite phrases like “no war” or “no more violence against
women” the partitive is also needed:
(10) Gerra-rik ez
war-PAR NEG
‘No war!’
(11) Emakume-en aurkako indarkeria gehiago-rik ez

women-PL .GEN against violence more-PAR NEG
‘No more violence against women!’
On the other hand, although in some cases the Basque partitive may seem to
be the equivalent of ‘any’, it is not always so, because the partitive appears to be
compulsory. Moreover, an equivalent in Basque of “there is not any water left”
would need an additional element to emphasize the negation and give a more
accurate translation of the English sentence, as shown in (10b):8
(10) a. Ez dago ur-ik

NEG it.remains water-PAR
‘There is no water left.’
b. Ez dago batere ur-ik

NEG it.remains any water-PAR
‘There is not any water left.’
8 I mean to say that, although the English translation of (10b) and (11b) would be grammatical
and acceptable for (10a) and (11a) too, the semantic nuance of the English any-sentences,
can only be expressed by the addition of an emphasizing element, and not by removing the
partitive suffix from the neutral-meaning sentences (10a)–(11a).
However, it is not the case that it has always been like that. As Oihenart (1638: 58) said:
“The case of negation or doubt [i.e. the partitive] is which occurs along with verbs and particles
of negation, doubt, or question, and it always bears explicitly or implicitly the specific noun
any. Albeit it [the partitive] has the function of the nominative, it is different (. . .) Being implicit
in all of them [the examples], as I said, the specific noun batere, that is, ‘any’ [Casus negandi
seu dubitandi est qui cum verbis & particulis negandi, dubitandi seu interrogandi concurrit,
habétque semper aut expressum aut subintellectum nomen particulare aliquis. Hic quamquam
sustineat munus nominatiui, est tamen ab eo diuersus (. . .) Subintellecto, ut dixi, in his
omnibus nomine particulari batere id est aliquis].
(11) a. Ez dute jostailu-rik

NEG they.have toy-PAR
‘They haven’t got toys.’
b. Ez dute jostailu-rik batere

NEG they.have toy-PAR any
‘They haven’t got any toy.’
2.2 Conditional mood

Partitive-marked arguments can also appear in Basque in the protasis of condi-
tional sentences, as shown in (12–14).
(12) Inon zorte txarr-ik izan baldin ba-du, Interlagos zirkuitu-an

anywhere luck bad-PAR have COND if-AUX Interlagos track-LOC
izan du
have AUX
‘If he has been unlucky somewhere, it has been in the Interlagos track.’
(Berria, 24-10-2004)
(13) Osasun arazo-rik ba-lute

health problem-PAR if-have:IRR .3 PL>3 SG
‘If they had any health problem. . .’
(Berria, 25-01-2005)
(14) Horrelako-rik behar izan-ez gero

such-PAR need have-INS after
‘If [you] need such a thing. . .’
(Jon Alonso, Euskal karma, 2001)
2.3 Questions
Interrogative sentences also allow for the partitive suffix to be used, both in
partial (mostly rhetorical) (15) and in yes-no questions (16):
(15) Nor-k behar du benetan egia-rik?

who-erg need (s)he.have really truth-PAR
‘Who really needs the truth?’
(L. R. Aurrekoetxea’s blog, 28-01-2010)
(16) Horrek axola-rik ba al du?

that:ERG care-PAR AFF Q it.has
‘Does it matter at all?’ (lit. ‘Does it have any importance?’)
(Denis Guedj / Jon Muñoz, Loroaren teorema, 2006)
In fact, the partitive is so characteristic of this type of questions that in

colloquial speech, to ask someone for something, the partitive suffix alone is
attached to what is asked, usually with the addition of bai ‘yes’ at the end of
the question (17a), or nahi ‘want, desire’ in order to offer something (17b):
(17) a. Su-rik (bai)?

fire- PAR (yes)
‘Got a light?’
b. Garagardo-rik (nahi)?
beer-PAR (want)
‘Want a beer?’
2.4 Some quantifiers

In addition to what I have already said, Basque also displays a quite archaic
construction with indefinite quantifiers modifying a noun phrase in the partitive
case, which retains its original sense of partition: asko / anitz / franko ‘many, a
lot’, gutxi ‘few’, aski ‘enough’ (now the only common use is eskerr-ik asko / anitz
‘a lot of thanks’, being the others much rarer, at least in the extra-literary use).
Moreover, since the first Basque writings, affirmative statements with expres-
sive or emphatic meaning are often combined with an adjectival element marked
for partitive, supposedly with the elision of asko ‘a lot’:
(18) Hetan izan baita eta baita zientzia guzi-etan

among.them be[PTCP ] AUX and be.3SG science all-LOC . PL
letratu handi-rik
erudite great-PAR
‘Since there have been and there are among them [a lot of] great erudites
in all sciences. . .’
(Etxepare, 1545)
(19) Ba=da pasadizo polit-ik ere

AFF =it.is happening nice-PAR also
‘There are also [a lot of] nice happenings.’
(Berria, 27-03-2004)
However, it is becoming more and more frequent to find sentences in which

the partitive adds no special nuance, or indeed it carries a strictly partitival
meaning, i.e., one of partiality, more or less equivalent to the meaning of the
English some:
(20) UGT- k eta CCOO-ek egin dituzte proposamen-ak.

UGT- ERG and CCOO-ERG make[PTCP ] AUX proposals-ABS . PL
‘(The unions) UGT and CCOO have made the proposals.
Dena den ELA-k ere egin du proposamen-ik
anyway ELA-ERG too make[PTCP ] AUX proposal-PAR
Anyway, ELA has made some proposals too.’
(Euskaltzaindia 1995: 167)
In fact, example (19) could also be interpreted as ‘there are some nice
happenings too’.
2.5 Fixed spatio-temporal uses (or the ablative-partitive)

Finally, I also wish to account for another fossilized use in expressions of the
[NOUN - PAR NOUN ] type. In (21) such a construction can be said to have an ablatival
sense, of spatial or temporal separation, or to denote some kind of union, either
serial or reciprocal (de Rijk 1996: 146–147):
(21) alde-rik alde ‘from one side to the other, right through’
esku-rik esku ‘from hand to hand, hand in hand’
herri-rik herri ‘from town to town’
egun-ik egun ‘day by day, from day to day’
And, to a more restricted extent, in temporal expressions like ordu-rik

‘since then’, betidan-ik ‘since always’, or har-ik eta. . . arte ‘until. . .’, lit. [that-PAR
and. . . until] ‘since then and until. . .’.
3 Other archaic uses from the historical data

Up to this point, I have analyzed for the most part the modern meanings of the
partitive in Basque. However, when turning to older texts a more widespread use
of the partitive appears. On the one hand, we can see that the use of the partitive
with some indefinite quantifiers (cf. subsection 2.4. above) and expressions like
atzo-rik ‘since yesterday’ (cf. subsection 2.5.) are far more frequent in such texts.
On the other hand, modern restrictions for partitive-marked constituents do not
seem to apply in Old Basque: for example, in Lazarraga’s manuscript (ca. 1602)
we find partitives also in affirmative sentences, even without an implicit quanti-
fier. This happens mostly with subjunctive or completive verb forms, and with
cardinal numerals:9
(22) Artalasto-a agin egizu ni-k ahal dagidan gauza-rik

corn.silk-ABS . SG order AUX I- ERG can do:SBJV thing-PAR
‘Order corn silk [to prepare the bed] so I can do something [sexually].’
(23) Dama-rik dakusen-ean. . . (corrected from dama bat ‘a lady’, with bat ‘one, a’)
lady-PAR see:COMP- LOC
‘When he sees a lady. . .’
(24) Gure ortu-an madari, gurazau-rik hamabi

our orchard-LOC pear[ABS ] peach-PAR twelve
‘[There are] pears in our orchard, [and] twelve peaches.’
Nowadays, in the first two examples we would say gauza-ren bat [thing-GEN
one] or the plural gauzak (22), and dama(-ren) bat [lady(-GEN ) one] (23).10 The
third one (24) is quite special, since it is a poem fragment, but it would be
gurazau-etarik [peach-ABL . PL] or simply gurazau-ak [peach-ABS . PL], depending
on the meaning we want to give it.
4 The superlative-partitive
This is the last case I will comment in this paper, since the next section explains
a “morpho-syntactically unmotivated” usage of the partitive (that is, not triggered
by any other element in the clause).
9 For a more accurate glossing and better understanding, I have modernized the spelling and
put the sentences more or less according to the modern model of standard Basque.
10 Ricardo Gómez informed me about an old text (nearly contemporary with the Lazarraga’s
manuscript) which shows that the construction ‘-GEN one’ (as it is synchronically analyzed)
is actually formed with the other old ablative -rean (see fn. 14), thus equivalent to partitive:
oraziorean bazuk ‘some prayers’, puntu batreanbat ‘some point or other’, tempora señaladurean
baten ‘in a special date’ (Urquizu 2009). For a more detailed account of this and other issues of
the noun-morphology of Basque, see Santazilia (2013).
(25) Txori-rik eder-en-a

bird-PAR beautiful-SUP-ABS . SG
‘The most beautiful bird.’
(26) Txori-etan/-etarik eder-en-a

bird- LOC . PL /-ABL . PL beautiful- SUP-ABS . SG
‘The most beautiful among / of birds.’
In my opinion, the meaning of (25) is clearly related to the ablatival origin of

the partitive, although obscured in Modern Basque by the emergence of an ellip-
tical construction (similar to that of [26], and quite common cross-linguistically)
which is presumably due to high frequency of usage (cf. [18–19]). Therefore,
while it is true that today it has only the interpretation of (25) (‘the most ADJ x’),
it is reasonable to think that originally the adjective did not refer to the head of
the overt noun phrase, but to an elided one, as if in: the most ADJ (x) of X. For
example, the literal translation of (25) would be originally “the most beautiful
of birds”, from which the current one is inferred through reanalysis of the old
ablative -rik (due to the changes outlined in Fig. 4 (section 6). The modern equiva-
lent of that construction is in (26), where the locative plural (-etan) or the ablative
plural (-etarik) are used instead.11
Thus, this elliptical construction (cf. -rik asko in subsection 2.4.) leads not to
a semantic extension of the suffix -rik but rather to a reinterpretation of its func-
tion, through the elision of the elements that motivated its presence. Nowadays,
the partitive is licensed by the morphosyntax of the superlative phrase.12 We can
draw the change as follows:
11 In such occurrences partitive-marked nouns do not make up a very close constituent with
the adjective, and can be separated from the superlative, unlike partitive-marked ones
(txorietan/txorietarik/**txoririk hau da ederrena ‘among birds, this is the most beautiful one’).
For a more detailed discussion about this issue, see Hualde & Ortiz de Urbina (2003: 835–837).
It has been noted by an anonymous reviewer that this alternation only occurs in the superlative
construction, as the following sentences show:
(i) Hemen ez da txoririk
there NEG is bird-PAR
‘There are not birds in here’
(ii) **Hemen ez da txorietarik/txorietan
Yet, despite the core meaning of (25) and (26) is the same, I think that they have an obvious
semantic nuance regarding the manner in which the same idea is expressed, and that they
should be accounted for individually, especially in a diachronic explanation.
12 As an anonymous reviewer reminds me, it is also possible to use a bare noun, that is, in the
absolutive case (txori-ø ederrena); in fact, I think that there is a diachronic correlation between
different constructions (-ø vs. -rik > -ø/-rik vs. -etarik), and that it would be interesting to
explore the use percentages of each one in a historical database. However, I will leave this
question for future research.
the most ADJ x from X I the most ADJ x of X > the most ADJ x[-PAR ]
Figure 1: Grammaticalization of the modern Basque superlative-partitive
Therefore, continuing with the same example (25), the complete semantic
evolution of the superlative-partitive would be as follows: ‘the most beautiful
bird from (all / these) birds’ → ‘the most beautiful of birds’ → ‘the most beautiful
bird[-PAR ]’.
5 The adverbial partitive

To finish with my account of the different uses of the partitive in Basque, I
should mention the adverbial value it acquires with perfective participles13 and
adjectives.14
On the one hand, from the standpoint of the agent or at least the logical
agent of the verb to which the partitive binds, constructions with the participle
can have a temporal / aspectual meaning, as a near past or a sort of Ablative
Absolute in Latin:
(27) a. Zeru-etako mandatu-a izan-ik

heaven-LOC . PL mandate-ABS . SG COP [ PTCP ]- PAR
‘Having been the mandate of God.’
(The dirge of Milia Lasturko, 15th–16th)
b. zein-ak hartu-rik biguela bat hasi zan

who-ERG . SG take:PTCP- PAR viola one begin:PTCP AUX . PST
kantaetan manera honetan

singing manner this:LOC
‘Who holding (i.e., having taken) a vihuela began to sing as follows.’
(Lazarraga’s manuscript, ca. 1602)
13 For an extensive account of the use of this partitive-marked participles in resultative con-
structions (such as nekaturik nago ‘I am tired’) and its diachronic evolution, see Krajewska
(2012, 2013).
14 Rudolf P. G. de Rijk (1972: 161 and fn. 12) suggested distinguishing between the partitive
case / determiner and what he called “stative -rik” (see also de Rijk 2008: §13.1 and §25.5.2),
here referred to as adverbial partitive (cf. Hualde & Ortiz de Urbina 2003: 204). I think that the
syntactic and semantic divergence between them is just the result of diverse grammaticaliza-
tions of the same morpheme, not an instance of two “entirely different morpheme[s]” (de Rijk
1996: 145).
Moreover, if you refer to an object (28a) or an intransitive subject (28b), the

partitive can be used to indicate the manner in which the action occurs or is
performed, or the state in which the object / subject has resulted.
(28) a. Juanikote-gaz lagundu-rik

Juanikote-COM accompain:PTCP- PAR
‘Accompained by Juanikote.’
(The Song of the Burning of Arrasate, 15th–16th)
b. Hemen natza ehortzi-rik (. . .) ustel eta kirastu-rik

here I.lie bury:PTCP- PAR rotten and reek:PTCP- PAR
‘Here I lie buried (. . .) rotten and smelly.’
(The dirge of Amendux, 1564)
When the partitive relates to adjectives other than deverbal, the resulting
meaning is the same, although they might be frequently considered as predi-
cative adjectives, rather than adverbials (Hualde & Ortiz de Urbina 2003: 194):
bakar ‘single, only one’ > bakarrik ‘alone’, triste ‘sad’ > tristerik ‘sad(ly)’, eder
‘beautiful’ > ederrik ‘beautiful(ly)’, hotz ‘cold’ > hotzik ‘cold(ly), dispassion-
ately’, poz ‘happiness’ > pozik ‘happ(il)y’ etc.
(29) a. Ezin geldi-rik egon

cannot still( ADJ )- PAR stand
‘Can’t stand still.’
(Bilingual song of the 15th–16th c.)
b. Gaur bixi-rik noa-n

today live( ADJ )- PAR I.go-SBJV
‘In order to (me) get out alive today.’
(The dirge of Errodrigo Zarateko, 15th c.)
Assuming that the temporal / aspectual construction is the original, at least

two possible scenarios of grammaticalization can be proposed.
Step 1. ikus-i-rik [see-P TCP -ABL]

→ *‘from / since seen it’ > *‘after seeing it, have just seen it’ > having seen it
Figure 2a: Grammaticalization of the Basque adverbial partitive (1st scenario, step 1)
According to the first scenario, the ablatival meaning of (temporal) separa-

tion may have developed into a sort of (near) past, and then into an adverbial-
shaded expression like the Ablative Absolute construction in Latin, the latter
meaning being the only possibility in Modern Basque. At the same time, as can
be seen in (30), the first step (‘ablative’ > ‘near past’) is attested in many lan-
guages and appears to be quite common in the grammaticalization of a ‘come
(from)’ verb (see section 6). Thus, one understands in (27b) that the musician
has already picked up the viola when he begins to sing, so that the action of tak-
ing it is over, although the expression could be translated into English by means
of the -ing form of the verb, rather than a perfect form.
Pitta-Pitta:
(30) a. Tatyi-ka-inya nganyanta
eat-NOM -ABL I
‘I’ve just eaten.’
(Heine & Kuteva 2002: 34)
Haitian:
b. l-fèk sòt rive kéyi gnou kòk vin bâ mwê
3:SG -TAM come.from arrive gather a nut come give 1:SG
‘He has just gathered a nut for me.’
Jiddu:
c. y-aam-ooku
3: M -eat-come
‘He has just eaten.’
Klao:
d. ɔ dɛ dɛ di
he come thing eat
‘He just ate.’
The second part of this path of grammaticalization starts from the circum-
stantial reading of this construction. Then, a (re)interpretation of such partitive-
marked participles as denoting a current state would arise, bringing on an
overtly marked adjectival use of those non-finite verb forms. The thin line
between adjectives and adverbs does the rest (compare it with English still,
quiet, angry, and Spanish tranquilo, enfadado, etc.).
Step 2. haserre-tu-rik [get.angry-P TCP -rik]

→ ‘having got angry’ > ‘angry’ (adj. > adv.) > ‘angrily’
Figure 2b: Grammaticalization of the Basque adverbial partitive (1st scenario, step 2)
Hence, from adjective-like participles, this suffix was spread to non-deverbal

adjectives and settled as “adverbial partitive”.
Nevertheless, there may be an alternative explanation, based on the partitive
meaning of the suffix -rik, rather than on the older ablative sense, and taking
into account the fact that Old Basque does not distinguish clearly between the
categories of noun, adjective, adverb and even verb (e.g. argi ‘light’, ‘clear/
bright’, ‘clearly’ and ‘to shine’; poz ‘happiness’, ‘happy’, ‘happily’ and ‘to get
happy’). Thus, the following pattern of grammaticalization can be set forth.
1. izutu-rik da [frightened- PAR COP ]

→ ‘(s)he is [one] of the frightened’ > ‘(s)he is one of the frightened’
→ izutu-rik ‘frightened’, lotsatu-rik ‘shamefaced(ly)’, etc.
2. poz-ik da [happy-PAR COP ]
→ ‘(s)he is [one] of the happy’ > ‘(s)he is one of the happy’
→ poz-ik ‘happy(ly)’, zabal-ik ‘open(-ed/-ly)’, etc.
Figure 3: Grammaticalization of the Basque adverbial partitive (2nd scenario)
Arguably, the strong tendency for elliptic constructions to occur in Basque,

and the increasing preference for an overt marking of the different lexical cate-
gories, have also contributed to the reinterpretation of the partitive as an adver-
bial suffix in such cases.
6 On the origin of the Basque ablative-partitive

suffix
The relationship between the ablative and the partitive in many languages is
well known (as in Finnish, French, German, Bulgarian, etc.; see Heine & Kuteva
2002: 32–33). Many years ago, scholars realized that the partitive suffix of
Basque was just an old allomorph of the ablative case (Humboldt 1811–1814
[apud Gómez 1996: 616], Mitxelena 1977: 236). Table 3 shows the history of the
two cases in Basque over the last 500 years:15
15 This table offers a somewhat simplified account of the facts. Actually, in the western dialect
there were two ablative suffixes in complementary distribution in the 16th century: -(r)ik (for
animate nouns, place names and temporal expressions), and -rean (for the rest). However,
this distinction was gradually dropped out, and together with the pleonastic -reanik, the suffix
-(r)ik gained ground, and in turn mixed with the originally prosecutive -ti, both in form and in
meaning (Lakarra 1996: 158–161). On the other hand, in the east we find the variant -(r)ik also
in singular with animates (-gan-ik, modern west. -gan-dik), and in some fixed spatio-temporal
expressions and place names.
Table 3: Diachronic and diatopic variation of the ablative-partitive suffix(es)
Ablative16
Partitive
Singular Plural Indefinite
16th c. west. -(r)ik -eta-rik -ta-rik
east. -tik -eta-rik -ta-rik
-(r)ik
Modern west. -tik -eta-tik -ta-tik
east. -tik -eta-rik -ta-rik
Thus, the ablative was reanalyzed as a partitive presumably just before

the emergence and development of the features of number and definiteness in
spatial cases; hence, the bare mark -(r)ik, originally being an indefinite or
number-unmarked suffix from which developed the partitive acquired the definite
singular meaning as ablative.
Figure 4: Prehistoric evolution of the ablative-partitive and its number distinction
Joseba Lakarra (2008: 482) has proposed to relate this suffix (specifically its
vowel -i-) with the proto-Basque verb-root *nin ‘give’, which he reconstructs in
order to explain (among other things) the origin and grammaticalization of the
Basque dative -(r)i. However, I find it difficult to establish a genetic relationship
between the dative and the ablative, as they are indubitably opposed, both in
meaning and in the path of grammaticalization that they tend to follow.
Indeed, especially in languages which have Serial Verb Constructions
(SVC)17 the notion of source or a “hither direction” is usually expressed by (or
16 The part -ti- of the modern ablative (-tik) was originally the so-called prosecutive, that is,
the case that expresses the notion of moving ‘through’, or ‘along’ the referent of the noun
that is marked with, as in this old Basque proverb: ze eikek maurtu-ti oanean, eder eztanik
kalean ‘when going through the desert, don’t do anything which isn’t fine in the street’ (RS:
36, translation from Hualde & Ortiz de Urbina 2003: 282); such meaning has converged with
the ablatival one (once -rik) on a unique suffix in Modern Basque, cf. zelai-tik dator ‘(s)he
comes from the meadow’ / kale-tik dabil ‘(s)he is walking through/down the street’.
17 For further information on such constructions, see Aikhenvald & Dixon (2006).
grammaticalized from) verbs roughly meaning ‘come (from)’ (31). In this sense, I
assume here that Proto-Basque should have made use of such constructions (see
Lakarra 2008 passim).
(31) Dumo:
a. beh [wa Opi luh]
3sgfpro 3sgfSU.go PSN 3sgfSU. come
‘She came from Dali.’
(Ingram 2006: 210)
Haitian: sòt(i) ‘(out) from’ < French sortir ‘come out’

b. yo pòté bagay sa yo sòt nâ-mòn
they bring thing DEM PL from LOC -hill
‘They bring these things from the hills.’
(Heine & Kuteva 2002: 71–72)
Olutec: -mi:nʔ ‘come’ > directional ‘hither’

c. tzüm-mi:nʔ
carry.on.the.back/shoulders-come
‘Carry something towards ego’
(Zavala 2006: 291)
Tetun Dili: mai ‘come’ > ‘from’ / ‘hither’

d. tuda bola mai
throw ball come
‘Throw the ball over here.’
(Hayek 2006: 243, 251)
Ewe, Swahili, Lingala:

e. tsó ‘come from’ > preposition ‘from’
kutoka ‘to come from’ > ‘from’
-úta ‘come from’ > útá, út’ó ‘since’, ‘from’
Thus, bearing in mind that Lakarra (2006: 585–586) has reconstructed a

proto-Basque verb-root ‘come’ as *din (which through *e-din developed into
the modern jin ‘come’), and making use of the cross-linguistic comparison and
the internal reconstruction, it appears safe to conclude that *din is the perfect
candidate for a possible source of the Old Basque ablative -rik and, therefore,
for the partitive (and its other uses).
*din ‘come (from)’ I ‘from’ (in SVCs) I *#lin18 I *-lin I *-Vri(n)19

+ -ka I -rika I -(r)ik
Figure 5: Grammaticalization of the Old Basque ablative
18 The change of #d- into #l- was a regular phonological change in Proto-Basque. It has been
somewhat productive, at least in some dialects, up to recent times (Mitxelena 1977: 242, 257–
258). Cf., for example, loanwords like Lat. theca ‘a cover, sheath’ > *deka > leka ‘pod’,
‘sheath’, Fr. dangereux ‘dangerous’ > lanjeros ‘id.’, etc., as well as the native betagin >
*detagin > letagin ‘eyetooth’ (from < *beg(i)-hagin, lit. ‘eye-tooth’).
19 This change (VlV > VrV ) is also well attested, or at least proven and accepted (Mitxelena
1977: 311–314, 327); cf. Lat. cōlum ‘colander’ > goru ‘distaff’, Lat. cēlum ‘sky’ > Rom.
*tselu > zeru ‘id.’, or the ancient toponym Ilumberri > modern Irunberri, etc.
I know, however, that there is a matter that I should explain, that is: why the -r- of the suffix
drops after a consonant (gizon-ik) instead of provoking epenthesis (**gizon-e-rik, cf. all. -ra:
behe-ra ‘to the bottom’ vs. gain-e-ra ‘to the top’), and why it behaves as if that consonant
was not its own but a kind of synchronic excrescence between vowels (cf. gen. -(r)en: hainbat
gizon-en ‘of many men’ vs. hainbat emakume-r-en ‘of many women’, from a reanalysis of the
final *-r of the definite article: *seme-(h)ar > seme-a ‘the son’ but *seme-(h)ar-en → seme-a-r-
en ‘of the son’). A fact that goes against an interpretation like that of the genitive is that, as far
as we know, neither the partitive nor the ablative have been ever used along with the definite
article (see Tab. 1–2), so the -r- of -rik could not have been originated in it (but cf. the “archaic”
allative -a, employed with consonant-ending place names as Zarautz-a ‘to Zarautz’, and with
animates as seme-a-gan-a ‘to the son’). On the other hand, the change -l- > -r- took place
only between vowels, hence there would have been an alternation as etorri-rik : *jin-lik. Such
groups of consonants like *-nl-, *-rl- and *-ll- could have been reduced in favour of the stem’s
one (e.g. > jinik), but those with sibilants might have been developed in a different way:
-S + l- > -Sl- or -S-e-l- (e.g. bihotz ‘heart’ → *bihozlik, or *bihotzelik > *bihotzerik (cf. actual
bihotzik). Thus, we can only assume that certain analogical forces regularized the variation on
the ablative according to the pattern of the genitive, rather than of the allative (cf. the proximity
of the ablative and the genitive, not only in Old Western Basque -re-an [GEN - LOC ] = [ABL ], but
also cross-linguistically, as it shows the fact that the genitive is used also in many languages
to render the partitive meaning. However, there may be another explanation: Trask (1995: 231,
1997: 221) suggested that the variation on the onset of some suffixes resulted by generalization
of a previously conditioned allomorphic variability. Thus, the pluralizers of the verb agreement
of the 3rd person (-de, -te, -e) would come from a unique *-de, whose consonant was devoiced
after a sibilant, and it was lost after a vowel. The same would have happened to the dative flag
of verbs *-gi, which developed to the forms -gi, -ki, -i in same contexts. In that way, Professor
Lakarra has found (p. c.) some other plausible instances of such alternation: -tegi, -degi, -egi
(lexicalized as hegi); and -toki, -doki, -oki (lex. toki, hoki). Similarly (and I must thank him
again for his remarks), I think now that it is possible to derive the suffix of the old ablative
(>partitive) -(r)i-k, not from the *lin (+ -ka) above, but directly from *din (+ -ka), as a ø- onset
variant. Its counterparts would be the -ti suffix of the old prosecutive (>ablative), which seems
to be related to the adjective-maker -ti (Lafon 1948), and the older variant -di of unknown
meaning (cf. ardi ‘sheep’, zaldi ‘horse’, ahardi ‘sow’, idi ‘ox’). The latter seems to have been
used also as an adjective-maker (lodi ‘fat’, hordi ‘drunk’ < hor ‘dog’ + -di?). Indeed, the
adjective-making use could be understood through the secondary but well-attested meaning
‘become’ of the proto-verb-root *din ‘come’.
The morpheme -ka20 might seem to be a dark point in the explanation, but it
is clear (Mitxelena himself was convinced of that already in 1977: 236–237) that it
is the same iterative suffix of adverbial expressions like harri-ka [stone-ka
‘throwing stones’] garrasi-ka [scream-ka ‘screaming’], albo-ka [side-ka ‘(go) through
sideways, (walk) staggering’], and also of spatial cases, as shown by the modern
ablative suffix (from the old prosecutive -ti + -ka) and the contemporary non-
standard -raka ‘toward’, formed on the allative -ra ‘to’.
7 Conclusions
In this paper I tried to provide a more detailed explanation of the ablative-to-
partitive grammaticalization in Basque, which, being almost a commonplace,
had yet to be studied in depth. I have also explained the origin of some of the
uses of the partitive, such as the adverbial and the superlative, as being the
result of various processes of reanalysis and semantic change. I have come to
the conclusion that they may have not developed directly from the partitive in
all cases (and even, perhaps, in none of them), but rather from the original abla-
tive meaning, in accordance with cross-linguistic data about the possible gram-
maticalization patterns of the ablatives or ‘come (from)’ verbs.
Along with this, I have brought the analysis to a greater diachronic depth,
locating the origin of the Old Basque ablative morpheme in the grammaticaliza-
tion of the Proto-Basque ‘come’ verb (*din), thus modifying a previous proposal.
This finding may constitute a further example to be added both to the typology
of grammaticalization in general, and to the set of probable SVC-sourced Basque
affixes in particular.
References
Aikhenvald, Alexandra Y. & R. M. W. Dixon (eds.). 2006. Serial Verb Constructions: A Cross-
linguistic Typology. Oxford: Oxford University Press.
Bilbao, Gidor, Ricardo Gómez, Joseba Andoni Lakarra, Julen Manterola, Céline Mounole & Blanca
Urgell. 2012. Lazarraga eskuizkribua: Edizioa eta azterketa [Lazarraga’s manuscript: Edition
and study]. UPV/EHU. http://www.ehu.es/monumenta/lazarraga/ (30 August, 2010).
20 Cf. B. Etxepare’s (1545) yxilica ‘silent(ly)’, cerutica ‘from the heaven’ (>modern isil-ik,
zeru-tik) and Lazarraga’s (ca. 1602) Gaztelatica ‘from Castilla’ (>mod. Gaztelatik), in both
geographic extremes of the Basque language. The change -a# → -ø# affected several mor-
phemes and words both historically and pre-historically (bart < barda ‘last night’, -t [1SG ] <
-da, naiz ‘I am’ < *naiza, etc., Mitxelena 1977: 235–236).
Etxeberria, Urtzi. 2010. -a, -ak eta -(r)ik euskal hizkeretan zehar [-a, -ak, and -(r)ik across
Basque speeches]. In Beatriz Fernández, Pablo Albizu & Ricardo Etxepare (eds), Euskara
eta euskarak: Aldakortasun sintaktikoa aztergai [The Basque language and the Basque
speeches: Syntactic variability under investigation] (Appendix of Anuario del Seminario
de filología vasca Julio Urquijo 52), 65–84. Donostia & Bilbao: University of the Basque
Country (UPV/EHU).
Etxepare, Ricardo. 2003. Negation. In José Ignacio Hualde & Jon Ortiz de Urbina (eds.),
A Grammar of Basque, 516–564. Berlin: Mouton de Gruyter.
Euskaltzaindia. 1995. Partitiboa baiezko perpausetan noiz eta nola [When and how to use
the partitive in affirmative sentences]. http://www.euskaltzaindia.net/dok/arauak/Araua_
0030.pdf (29 July, 2013).
Gómez, Ricardo. 1996. La aportación de W. von Humboldt a la gramática vasca. Revista
Internacional de Estudios Vascos 41(2). 607–622.
Hayek, John. 2006. Serial Verbs in Tetun Dili. In Alexandra Y. Aikhenvald & R. M. W. Dixon
(eds.), Serial Verb Constructions: A Cross-linguistic Typology, 239–253. New York: Oxford
University Press.
Heine, Bernd & Tania Kuteva. 2002. World Lexicon of Grammaticalization. Cambridge: Cam-
bridge University Press.
Hoeksema, Jacob. 1996. Introduction. In Jacob Hoeksema (ed.), Partitives: Studies on the Syn-
tax and Semantics of Partitive and Related Constructions, 1–24. Berlin: Mouton de Gruyter.
Hualde, José Ignacio. 2003. Case and number inflection of noun phrases. In José Ignacio
Hualde & Jon Ortiz de Urbina (eds.), A Grammar of Basque, 171–186. Berlin: Mouton de
Gruyter.
Hualde, José Ignacio & Jon Ortiz de Urbina (eds.). 2003. A Grammar of Basque. Berlin & New
York: Mouton de Gruyter.
Ingram, Andrew. 2006. Serial Verb Constructions in Dumo. In Alexandra Y. Aikhenvald & R. M.
W. Dixon (eds.), Serial Verb Constructions: A Cross-linguistic Typology, 202–222. New
York: Oxford University Press.
Koptjevskaja-Tamm, Maria. 2001. “A piece of the cake” and “a cup of tea”: Partitive and
Maria Koptjevskaja-Tamm (eds.), Circum-Baltic Languages: Grammar and Typology, 523–
568. Amsterdam: Benjamins.
Koptjevskaja-Tamm, Maria & Bernhard Wälchli. 2001. The Circum-Baltic languages: An areal-
typological approach. In Östen Dahl & Maria Koptjevskaja-Tamm (eds), Circum-Baltic Lan-
guages: Grammar and Typology, 615–750. Amsterdam: Benjamins.
Krajewska, Dorota. 2012. The diachrony of resultative constructions in Basque. Vitoria-Gasteiz:
University of the Basque Country (UPV/EHU) MA thesis.
Krajewska, Dorota. 2013. Resultatives in Basque: A Diachronic Study. Lingua Posnaniensis 54
(2). 55–67.
Laka, Itziar. 1996. A brief grammar of euskara, the Basque language. Bilbao: University
of the Basque Country (UPV/EHU). http://www.ei.ehu.es/p289-content/eu/contenidos/
informacion/grammar_euskara/en_doc/index.html (6 May, 2011).
Lakarra, Joseba A. 1996. Refranes y Sentencias: ikerketak eta edizioa (= RS) [Refranes y
Sentencias: research and edition]. Bilbao: Euskaltzaindia & Bizkaiko Foru Aldundia.
Lakarra, Joseba A. 2006. Jaun eta jabe, jaio eta herio, jin eta joan . . . : etimologiaz eta aditz
morfologia zaharraz (Hitz hasierez II) [Jaun ‘lord’ and jabe ‘owner’, jaio ‘to be born’ and
herio ‘death’, jin ‘to come’ and joan ‘to go’. . . : about etymology and old verbal morphology
(Word-initials II)]. In Beatriz Fernández & Itziar Laka (eds.), Andolin gogoan. Essays in
honour of Professor Eguzkitza, 576–611. Bilbao: University of the Basque Country (UPV/
EHU).
Lakarra, Joseba A. 2008. Aitzineuskararen gramatikarantz (malkar eta osinetan zehar) [Towards
a grammar of Proto-Basque (across slopes and pools)]. In Xabier Artiagoitia & Joseba A.
Lakarra, Gramatika jaietan: Patxi Goenagaren omenez [Grammar holidays: In honour of
Professor Patxi Goenaga] (Appendix of Anuario del Seminario de filología vasca Julio
Urquijo 51), 451–490. Bilbao: Gipuzkoako Foru Aldundia & University of the Basque Country
(UPV/EHU).
Lafon, René. 1999 [1948]. Sur les suffixes casuels -ti et -tik. In Vasconiana (IKER 11), 199–207.
Bilbao: Euskaltzaindia.
Mitxelena, Koldo. 1977 [1961]. Fonética Histórica Vasca, 2nd edn. Donostia: Gipuzkoako Foru
Aldundia.
Mitxelena, Koldo & Ibon Sarasola. 1987–2005. Diccionario General Vasco: Orotariko Euskal Hiz-
tegia. Bilbao: Euskaltzaindia, Desclée de Brouwer, Mensajero. http://www.euskaltzaindia.
net/oeh (26 August, 2010).
Oihenart, Arnaud. 1638. Notitia utriusque Vasconiae, tum Ibericae, tum Aquitanicae. Paris:
S. Cramoisy. http://books.google.es/books?id=38hUlGr28x0C&printsec=frontcover&hl=ca#v=
onepage&q&f=false (29 July, 2013).
Rijk, Rudolf P. G. de. 1972. Partitive assignment in Basque. Anuario del Seminario de filología
vasca Julio Urquijo 6(1). 130–173.
Rijk, Rudolf P. G. de. 1996. On the origin of the partitive determiner. Anuario del Seminario de
filología vasca Julio Urquijo 30. 145–158.
Rijk, Rudolf P. G. de. 2008. Standard Basque. A Progressive Grammar. Volume 1. Cambridge,
MA: MIT Press.
Santazilia, Ekaitz. 2013. Noun Morphology. In M. Martínez-Areta (ed.), Basque and Proto-Basque.
Language-Internal and Typological Approaches to Linguistic Reconstruction (Mikroglottika
5), 223–281. Frankfurt am Main, Berlin, Bern, Bruxelles, New York, Oxford, Wien: Peter Lang.
Trask, Robert Lawrence 1995. On the History of the Non-Finite Verb Forms in Basque. In José
Ignacio Hualde, Joseba A. Lakarra & R. L. Trask (eds.), Towards a History of the Basque
Language, 207–234. Amsterdam: Benjamins.
Trask, Robert Lawrence 1997. The History of Basque. London: Routledge.
Trask, Robert Lawrence 2003. The Noun Phrase: nouns, determiners and modifiers; pronouns
and names. José Ignacio Hualde & Jon Ortiz de Urbina (eds.), A Grammar of Basque, 113–
170. Berlin: Mouton de Gruyter.
Urquizu, Patricio. 2009. Un texto olvidado del siglo XVII: SUMARIO BREBEA. . . (Azkoitia, 1614).
Revista de lenguas y literaturas catalana, gallega y vasca (RLLCGV) XIV. 207–217.
Zavala, Roberto. 2006. Serial Verbs in Olutec (Mixean). In Alexandra Y. Aikhenvald & R. M. W.
Dixon (eds.), Serial Verb Constructions: A Cross-linguistic Typology, 273–300. Oxford:
Corpora:
Ereduzko Prosa Gaur [Modern Exemplary Prose]. http://www.ehu.es/euskara-orria/euskara/
ereduzkoa (26 August, 2010).
Klasikoen Gordailua [Depository of Classic Works]. http://klasikoak.armiarma.com (27 August,
2010).
IV Slavic languages
Michael Daniel
9 The second genitive in Russian
This paper is an overview of the so-called second genitive in Russian, a nominal

form available for a minority of Russian nouns but widely used with these
nouns in certain contexts. In many ways, the second genitive is a secondary
case. Thus, it may always be substituted with a regular genitive form, while the
opposite is not true. A major subset of the contexts where the second genitive
may be used fits into what is known as a functional category of partitive, so
this form is sometimes called Russian partitive. To a certain extent, indeed, the
second genitive is the form with which the regular genitive may be substituted
in partitive contexts. The analysis of the distribution of the second genitive
shows, however, that the partitive meaning is not the only function of this
form. Not less if not more widespread are uses in combinations with preposi-
tions. These and other types of contexts should be taken into account to build
a comprehensive picture of the category’s distribution and functional load.1
Keywords: Russian, partitive, genitive, secondary cases, paradigm structure
1 Introduction
Russian has a nominal inflection system typical of those Indo-European lan-
guages that are rich in cases, including e.g. Latin or Ancient Greek. The core
case inventory includes Nominative, Genitive, Dative, Accusative, Instrumental
and Prepositional (the former locative case presently used only in combination
with prepositions – hence the term, traditional in Russian linguistics). Typically
for Indo-European languages with nominal declension, the expression of number
(singular vs. plural) and case is cumulative and there are several (three in
Russian) major declension classes. Ascription of an individual lexical item to a
1 Support from the Basic Research Program of the National Research University Higher School
of Economics and from the Russian Academy of Sciences research program ‘Corpus Linguistics’
(project Rusgram, http://rusgram.ru) is gratefully acknowledged. I would also like to thank
Alexandre Arkhipov, Greville Corbett, Sergei Kniazev, Filipp Minlos, Anna Polivanova, Anna
Ptentsova, Leonid Stunzhas and Ilya Yakubovich as well as all members of the Rusgram
research group for helping me to achieve a better understanding of the range of problems
related to the Russian second genitive, and to Tommaso Claudi for his help with preparation
of the final draft of this paper.
348 Michael Daniel
specific declensional class is on the whole formally and semantically unpredict-

able, though the declension correlates with gender. Most nouns in the first
declension (determined as having ‑a as the nominative singular inflection) and
in the third declension (zero nominative singular and the dative singular -i) are
feminine, while the second declension (the dative singular -u) includes nouns
that are either masculine (zero nominative singular) or neuter (nominative
singular in -o).
In addition to these well-established case forms, however, there are several
nominal forms that are paradigmatically peripheral in one or other way. As dis-
cussed in e.g. (Spencer & Otoguro 2005 and Corbett 2008), these peripheral,
less ‘casey’ forms include the new vocative (zero or, under a different interpreta-
tion, a truncated nominative form, e.g. mam ‘mom (Voc)’), the second preposi-
tional alias locative (e.g. v snegu ‘in the snow’), and the second genitive alias
partitive. Jakobson (1936), on the other hand, without going into discussion
claims that Russian has eight cases, considering both the second genitive and
the second prepositional together with the other, core case categories. Zalizniak
(1967) does the same after a thorough analysis of the data, but seems much
more wary about this claim.
The present paper focuses on the second genitive. It is often used in contexts
that contain some sort of quantification and/or negation. These uses ideally fit
into the domain of the partitive category, and the form is often called the partitive
genitive case (RG80) or a quantitative-separative case (količestvenno-otdelitel’nyj
padež; Zaliznjak 1967). Defining the exact lexical and functional scope of the
second genitive in modern language is complicated by the fact that many actual
examples found in the texts sound archaic or are imitations of dialectal speech,
where the form has a wider use, at least in terms of lexical distribution (Bromley
& Bulatova 1972). The form also seems to be more frequent in spoken language
(cf. RG80) or in its imitations in written texts than in formal style.
Below, we suggest a paradigmatic and distributional analysis of the second
genitive. All examples come from the Russian National Corpus (www.ruscorpora.ru).
Section 2 discusses the paradigmatic status of the second genitive and explains
in what sense it is not a member of the core case inventory. Section 3 gives a
very brief overview of the external (i.e. not observable by means of RNC) history
of the form. Section 4 explains corpus preferences of the second genitive forma-
tion, showing which lexical items allow or disallow, and whether (and to what
extent) they tend to prefer it to the main genitive form. The analysis of the
functional distribution in section 5 shows that the functional scope may only
be described as a combination of interacting factors and certainly can not be
limited to the partitive function. Section 6 is a summary of the discussion.
The second genitive in Russian 349
2 Paradigmatic status
The Russian second genitive (henceforth Gen2) is a special nominal form available
for some inanimate masculine nouns of the second declension. It is formally
identical to the dative singular of such nouns. Yet it cannot be considered to be
a special function of the dative case, or to belong to the dative functional
domain in any way. Gen2 clearly alternates with the genitive, with which it is
almost always freely interchangeable in terms of grammatical acceptability (1).
Nouns that lack Gen2 normally display the genitive in contexts where Gen2 is
used (2).
(1) Николай Василич и виду (+вида) не подал, что он Николай Василич.

Nikolaj Vasilič i vid-u (+vid-a)
Nikolaj.NOM . SG Vasilič.NOM . SG and view-GEN 2 (+view-GEN . SG )
nе poda-l, čto on Nikolaj Vasilič.
not give-PST. M that 3 SG . M . NOM Nikolaj.NOM . SG Vasilič.NOM . SG
‘Nikolai Vasilič didn’t let on (lit. did not give sign, did not show) that he was,
in fact, Nikolai Vasilič.’ (Jurij Koval’. Rusačok-travnik (2000))
(2) В вагоне почти не было народу (cf. людей), все на танцах, на гулянье.
V vagon-е počti nе by-l-o narod-u (cf. ljud-ej),
in car-PREP. SG almost not be-PST- N . SG people-GEN 2 (cf. person-GEN . PL)
vs-е na tanc-ach, na guljan’-е.
all-NOM . PL on dance-PL . PREP on walk-PREP. SG
‘There was almost no one in the coach; everyone was away, dancing or
out for a walk.’ (Viktor Astaf’ev. Proletnyj gus’ (2000))
Moreover, Gen2 is completely absent from the declension of adjectives and pos-
sessive pronouns. Paul Garde’s claim (Garde 1998: 168) that Gen2 is impossible
for a noun with an adjectival attribute is certainly not true; on the other hand, it
is true that, in a complex NP, a Gen2-marked head governs regular genitive
agreement on all agreeing dependents (3). Thus, this optional marking on the
head appears to be ‘invisible’ to agreement: whether the head noun is inflected
for genitive or Gen2, the dependents invariably inflect for regular genitive.
350 Michael Daniel
(3) Захвати, Володя, моего любимого ликеру.

Zachvat-i, Volod-ja, mo-еgo ljubim-ogo likеr-u.
bring-IMP Volodya-NOM . SG POSS .. SG - GEN . M . SG favorite-GEN . M . SG liquor-GEN 2
‘And, Volodya, bring my favorite liquor.’
(Aleksej Novikov-Priboj. Kapitan pervogo ranga (1936–1944))
All this seems to corroborate the analysis of Gen2 as an allomorph of the regular
singular genitive ending -a, similar to variants of e.g. Instrumental plural ending
for some nouns of the third declension (dver’-mi ~ dver-jami, lošad-’mi ~ lošad-
jami, dočer’mi ~ dočer-jami). However, the regular genitive may only be sub-
stituted in certain contexts (cf. section 5), and lexical groups for which Gen2 is
available are at least partly semantically motivated (cf. section 4), which contra-
dicts the idea of free or formally motivated distribution of the two markers. In
the case of the Instrumental plural allomorphs (-mi ~ -jami), on the other hand,
there seem to be no (known) semantic constraints on contexts or lexical items.
Above I have insisted that, in all contexts, Gen2 can alternate with regular
genitives without affecting grammaticality. Zaliznjak (1967) discusses certain
examples that, to his eyes, rule out such substitution, as the form čajku (Gen2
of the diminutive noun čajok derived from čaj ‘tea’):
(4) В столовой большой стол, за которым каждый вечер усаживались

гости попить чайку.
V stolov-oj bol’š-oj stol, za
in dining.room-PREP. SG big-NOM . SG . M table.NOM . SG behind
kotor-ym každ-yj vеčеr usaživa-l-i-s’
which-INS . M . SG every-ACC . SG evening.ACC . SG seat-PST- PL- M / P
gost-i popi-t’ čaj-k-u.
tea-DIMIN - GEN 2 guest-NOM . PL drink.some-INF
‘There was a large table in the dining room where the guests sat down
every evening to have tea.’ (Daniil Granin. Zubr (1987))
At first glance, this suggests that Gen2 ending is not a variant of the regular
genitive. But in fact all contexts where the form čajku may be used are those that
allow Gen2. It can not be used in contexts that disallow it: ?vkus čajka and ?vkus
čajku are both problematic (see Section 5 on contextual preferences and con-
straints). It seems that čajok has a gap in the paradigm (absence of the regular
genitive); what we see here is not a case of contextual impossibility to alternate
with the regular genitive but the absence of the latter. In contexts such as (4),
Gen2 is simply the only option available. In addition, a look at the actual usage
indicates that, even in this case, the competition between Gen2 and regular
genitive is a matter of a very strong tendency rather than grammaticality, as at
least one example of čajka (regular genitive of the diminutive of čaj ‘tea’) is
found in the modern section of the Russian National Corpus (5), and more are
found in the Internet (6):
(5) Это все равно, что где-то пригласили тебя к костру чайка попить,
побеседовать, а ты потом хочешь за чай деньги заплатить.
Èt-o vs-е ravn-o, čto gdе-to
this-NOM . N . SG all-NOM . N . SG equal-NOM . N . SG that where-INDF
priglasi-l-i tеbja k kostr-u čaj-k-a popi-t’,
invite-PST- PL 2 SG . ACC to fire-DAT. SG tea-DIMIN - GEN . SG drink.some-INF
pobеsеdova-t’, a ty potom chočе-š’ za čaj
chat-INF and 2 SG . NOM then want-PRS .2 SG behind tea.ACC . SG
dеn’g-i zaplati-t’.
money-ACC . PL pay-INF
‘It is as if you were invited to chat and have some tea by a camping fire,
and then you want to pay for the tea.’
(Tat’jana Gordeeva. Sportivnyj avtostop (2003))
(6) Решили попить чайка, да и закончить дело . . .

Rеši-l-i popi-t’ čaj-k-a,
decide-PST- PL drink.some-INF tea-DIMIN - GEN . SG
da i zakonči-t’ dеl-o . . .
then and finish-INF business-ACC . SG
‘We’ve decided to have some tea and then to finish it off.’
(http://www.syntone.ru/library/stories/content/5820.html)
However, examples such as (5) and (6) certainly sound less (or even much less)
natural than (4), both in terms of textual frequency and speakers’ evaluation.
There is a strong preference for the second genitive form for certain lexical
items, including -Vk- diminutives and designations of potable liquids (both pref-
erences work for čaëk ‘tea (diminutive)’) in certain contexts, including e.g.
partially affected direct objects (as in 4, 5 and 6). However, these are matters of
stronger or weaker preferences and do not provide unambiguous solution for
the issue of the paradigmatic status of Gen2. I support the gradual approach
to case-hood as described in (Spencer & Otoguro 2005) and (Corbett 2008)
which, in a way, continues hesitations in (Zalizniak 1967). A convenient solution
that reflects the intermediate nature of Gen2 was suggested by Paul Garde who,
352 Michael Daniel
without lengthy theoretical discussion and reasoning, labels Gen2 and the
second prepositional case secondary cases (cas seconds; Garde 1998: 165).
3 The early and external history of Gen2

The paper focuses on the modern history of the Gen2 forms in standard Russian
as represented in the texts of the Russian National Corpus. This section is a very
brief account of where the forms are thought to come from and of their expan-
sion in Russian dialects. The second genitive of modern Russian continues the
genitive of the Proto-Slavic declension in -u. Shakhmatov (1957: 82ff.) suggested
that, at the time of the first written sources, this declension was already repre-
sented as remnants in some lexical items, so that its original lexical scope may
only be tentatively reconstructed. He cites the following words as members of the
-u-declension for the Old Church Slavonic: domъ ‘house, home’, vrьchъ (vьrchъ)
‘top (n)’, synъ ‘son’, polъ ‘half, side’, nizъ ‘down (n)’, volъ ‘ox’, dolъ ‘dale’,
vъnъ ‘outside (n)’, kratъ ‘time, occurrence’, medъ ‘honey’, mirъ ‘peace’, olъ ‘beer’.
The data from birch bark manuscripts, unavailable to Shakhmatov and cited in
(Zaliznjak 1995: 95) further include bebrъ or bobrъ ‘beaver’, borъ ‘forest’, darъ
‘gift’, dъlgъ ‘debt’, rędъ ‘row’, solodъ ‘malt’, stanъ ‘site, camp’ and tъrgъ ‘trade’.
Zaliznjak also notes that some of these nouns in fact alternate between -u- and
-о- (the future second) declensions.
Many of these have Gen2 in modern language; cf. iz domu ‘out of the house,
away from home’, do (samogo) verchu / nizu ‘up to the very top / down to the
very bottom’, bočka medu ‘a barrel of honey’. Other Gen2 forms sound obsolete
but are attested, in adverbs and idioms (as srjadu ‘in a sequence, lit. from the
row’, iz rjadu von vychodjaščij ‘extraordinary, lit. going out of the row’, bez torgu
‘without bargaining’, s boru po sosenke ‘unsystematically, lit. from (each) forest
one pine tree’), being probably more typical of older texts (7, 8) or proverbs (9):
(7) К северу обставлена она хребтом Далеху, на который взбирается

с долу дорожка до Кале-и-Зенджир <. . .>
K sеvеr-u obstavl-еn-a ona chrеbt-om
to north-DAT. SG surround-PTCP. PASS - F 3. SG . F. NOM range-SG . INS
Dalеchu, na kotor-yj vzbira-еt-sja s
Dalekhu.NOM . SG on which-ACC . SG . M climb-PRS .3 SG - M / P from
dol-u dorožk-a do Kalе-i-Zеndžir.
valley-GEN 2 track-NOM . SG towards Kale-i-Zendzhir
‘To its north stands the Dalekhu range, onto which climbs (lit. from the
bottom, from the valley) the track leading to (the fortress of)
Kale-I-Zendzhir.’ (E. Čirikov. Putevoj žurnal (1849–1852))
(8) Саженях в семидесяти от монастырской стены, подле берёзовой

рощицы, среди зелёного луга, стоит пустая хижина, без дверей, без
окончин, без полу; кровля давно сгнила и обвалилась.
Sažеn-jach v sеmidеsjat-i ot monastyr-sk-oj stеn-y,
sazhen-PREP. PL in seventy-PREP from cloister-ADJV- GEN . F. SG wall(F )- GEN . SG
podlе bеrёz-ov-oj roščic-y, srеdi zеlёn-ogo
near birch-ADJV- GEN . F. SG grove-SG .GEN amidst green-GEN . M . SG
lug-a, sto-it pust-aja chižin-a, bеz
meadow-GEN . SG stand-PRS .3 SG empty-NOM . F. SG cabin-SG . NOM without
dvеr-еj, bеz okončin, bеz pol-u;
door-GEN . PL without window.frame-PL .GEN without floor-GEN 2
krovl-ja davno sgni-l-a i obvali-l-a-s’.
roof-NOM . SG long.ago rot-PST- F and drop-PST- F- M / P
‘About five hundred feet away from the monastery wall, near the birch
grove, amidst a green meadow stands an empty hut, no door, no window
frames, no floor; the roof has long since decayed and fallen in.’
(Nikolaj Karamzin. Bednaja Liza (1792))
(9) А что ж я, по-вашему – за сто верст киселю хлебать приехал?

A čto ž ja, po-vashemu ― za sto
and what PTC 1 SG . NOM in.your.opinion behind hundred
verst kisel-ju chleba-t’ priecha-l
versta.GEN . PL jelly.drink-GEN 2 drink-INF arrive-PST. M
‘And what do you think I came for – travelled a hundred verstas to drink
some kissel.’ (Sergej Cvetkov. Ivan Poddubnyj (2007); obtained through
Google.Books search)
(to travel a seven / hundred verstas to drink some kissel – a proverbial
expression meaning to travel a long distance for little gain; versta – an old
unit of length, approx. one km; kissel – a traditional dessert, a thickened
berry juice)
(10) У семи нянек дитя без глазу.

U sеm-i njan-еk dit-ja bеz glaz-u.
at seven-GEN nanny-GEN . PL child-NOM . SG without eye-GEN 2
‘Seven nannies leave the child unwatched.’ (a proverb, lit. no eye after).
(The proverb means that the more people are responsible for something,
the less safe it is)
354 Michael Daniel
(11) Нашего полку прибыло.

Naš-еgo polk-u priby-l-o.
regiment-GEN 2 come.more-PST- N
POSS .1 PL - GEN . M . SG
‘There is more of us now.’
(An idiomatic expression, lit.: ‘Our regiment grew larger.’)
Some words listed by Shakhmatov as -u-stems are either absent from Modern
Russian (as olъ ‘beer’) or ceased to decline (as vъnъ ‘outside’, only used as an
adverb von and preposition vne, or kratъ ‘occurrence’, only used in compound
adverbs and adjectives as dvukratnyj ‘double, one that occurred two times’).
The three animate nouns on the list, syn ‘son’, vol ‘ox’ and bober ‘beaver’, exist
in modern language but are not attested in Gen2 in the Russian National Corpus.
This probably means that, while there was a functional expansion of Gen2 to
various lexical classes, animate nouns were first to lose it. As was discussed
above, in Modern Standard Russian, only inanimate nouns may have Gen2.
To sum up, for all ex‑u-nouns, the -a genitive marker has become a default,
but the -u marker (Gen2) was preserved by some as a variant (sometimes a pre-
ferred variant); it has also spread to some other nouns of the second declension.
As we will see further, this process is at least partly functionally motivated: Gen2
has become typical for certain contexts (including, among other, partitive con-
texts), spreading primarily to those nouns that typically occur in such contexts.
In Russian dialects, Gen2 may have a wider distribution than in the standard
language (Bromley & Bulatova 1972: 66ff.). As in Standard Russian, it is only
available for inanimate nouns of the second declension. Unlike standard Russian,
however, it is well attested for neuter nouns (rather than only in the masculine):
bez mjasu ‘without meat’, iz stadu ‘from the herd’. I am aware of no data that
would allow a careful comparison of the lexical and functional distribution of
Gen2 in standard language vs. dialects. However, it seems feasible to suggest that
the domain of Gen2 is a matter of negotiating the borderline between Standard
Russian, its dialects and substandard2 language, which is constantly fed by dia-
lectal and regional variation (cf. RG80: 497; Garde 1998: 169). To a certain extent,
this complicates working with the RNC data. To a standard ear, many examples
2 By substandard language I mean specifically the so-called prostorečie literally ‘simple talk’,
a language variant used by people of lower educational level and with a stronger dialectal
background. Today, the sociolinguistic definition of prostorečie, inherited from traditional Rus-
sian studies of the 1950s through 1970s, certainly needs to be revisited. Although the term
‘substandard’ should be understood in a much broader sense, for the sake of brevity I will
below refer to prostorečie simply as substandard Russian.
After the paper was submitted to the publisher, my attention was brought to (Seržant,
to appear), whose aim is exactly to analyse the evidence of the use of the second genitive in
Russian dialects.
of Gen2 in literary texts of the Russian National Corpus sound archaic or prover-
bial (see 8-11 above) or dialectal.
4 Lexical distribution
According to RG80, Gen2 formation has several absolute lexical constraints. It is
only formed from some uncountable or abstract inanimate masculine nouns of
the second declension (p. 492). I will start by considering these constraints, and
then will proceed to subtler semantic preferences.
The reasons why RG80 mentions uncountability are clear (cf. section 5);
but, if understood straightforwardly, uncountability/abstractness is hardly a valid
constraint. Some Gen2‑forming nouns, such as dom ‘house’, šag ‘step’, glaz ‘eye’,
čas ‘hour’, god ‘year’, voz ‘cart’, rjad ‘rank’, designate quantifiable non-abstract
entities. Incidentally, however, some of the respective Gen2 forms are used in
contexts that are incompatible with the idea of quantification and individuation.
Thus, in the proverb in (10) above, glaz is not used in its main sense ‘eye’ but in
the (obsolete) sense of ‘control, process of looking after someone’, both abstract
and unquantifiable. In (12), too, the noun does not refer to a specific building
but to a domicile of the person spoken about (cf. English home) which is unique
and thus unquantifiable – and more abstract than the meaning ‘building’.
(12) В первое время он, как огонь, печёт, терзает, и за кишки, и за душу
рвёт, ― человек и бежит из дому.
V pеrv-oе vrеm-ja on, kak ogon’, pеč-ёt,
in first-ACC . N . SG time-ACC . SG 3 SG . M . NOM as fire.NOM . SG burn-PRS .3 SG
tеrza-еt, i za kišk-i, i za duš-u rv-ёt,
torment-PRS .3 SG and from gut-ACC . PL and from soul-ACC . SG tear-PRS .3 SG
― čelovek i bež-it iz dom-u.
person.NOM . SG and run-PRS .3 SG from home-GEN 2
‘At first it (the hunger) burns, and torments like fire, and tears your guts
and your soul – and people run away, away from home.’
(Vasilij Grossman. Vse tečet (1955–1963))
Further, non-animacy should be understood here as a grammatical rather

than semantic category. In Russian, the animacy is grammaticalized as differ-
ence in declension. Animate masculine nouns of the second declension (as well
356 Michael Daniel
all animate nouns in the plural) have the accusative identical to the genitive.
Inanimate nouns have the accusative identical to the nominative:
Table 1: Animate and Inanimate inflection
Nominative Accusative Genitive

8
<‘person’ čelovek čeloveka čeloveka
animate ‘horse’ kon’ konja konja
:
‘toy soldier’ soldatik soldatika soldatika
inanimate ‘bust’ bjust bjust bjusta
Second declension nouns that follow the animate pattern do not form Gen2.
However, Gen2 may occur in the paradigms of such nouns as polk ‘regiment’ (11)
and narod ‘people’ (2), which are collective nouns designating animate entities
(groups). The form narodu is extremely frequent and even preferred (Mustajoki
2006). Cf. also iz tabunu ‘out from the herd’, a group of non-human animates
(which, however, does not occur in RNC at all and only once in the whole
runet):
(13) В итоге, её выгнали из табуну, ну она и сама хотела его покинуть,

так что, этот случай ей помог. . .
V itog-е, eё vygna-l-i iz tabun-u, nu
in result-PREP. SG 3 SG . F. ACC chase-PST- PL from herd-GEN 2 well
ona i sam-a chotе-l-a еgo pokinu-t’,
3 SG . F. NOM and self-NOM . F. SG . want-PST- F 3 SG . M . ACC leave-INF
tak čto, ètot slučaj ej pomog. . .
so that this.NOM . SG event.NOM . SG 3 SG . F. DAT help.PST. M
‘As a result she’s been chased from the herd – well, she wanted to leave
it anyway, so this event only helped her.’
(http://elmundomilagros.ucoz.ru/forum/34-103-1)
Finally, Russian National Corpus provides some rare occurrences of the

second genitive of neuter nouns. Some of them are apparently dialectal, as 14
(wider context contains lots of dialectal items; cf. section 3 for expansion of
Gen2 to neuter nouns in dialects); other sound obsolete, as 15.
(14) А теперь из этого из земству больно разговорчивый начальник пошел.

A tеpеr’ iz èto-go iz zеmstv-u
and now from this-GEN . SG from local.government-GEN 2
bol’no razgovorčiv-yj načal’nik pošе-l.
too.much talkative-NOM . M . SG boss.NOM come-PST. M
‘These days the new bosses, those from zemstvo are rather talkative.’
(Lidija Sejfullina. Virineja (1924))
(zemstvo – local government bodies instituted by the liberal reforms of
the second half of the 19th century.)
(15) Я пошел прогуляться ― более из любопытству, чем для моциону . . .

Ja pošе-l progulja-t’-sja ― bolее iz ljubopytstv-u,
1 SG . NOM go-PST. M walk-INF- M / P more from curiosity-GEN 2
čеm dlja mocion-u. . .
than for excersize-GEN 2
‘I went out for a walk – more out of curiosity than to stretch my limbs. . .’
(Nikolaj Simanovskij. Dnevnik (1837))
In modern texts, there are several occurrences of načalu, mainly in prepositional

constructions, first of all dlja načalu, but also do načalu and ot načalu. But all
these examples lie on the borderline of substandard language:
(16) Для началу надо, как считает премьер, разделить по принципу

отделы в министерствах.
Dlja načal-u nado, kak sčita-еt prеm’еr,
for beginning-GEN 2 need as believe-PRS .3 SG prime.minister.NOM . SG
razdеli-t’ po princip-u otdеl-y v ministеrstv-ach.
divide-INF following principle-DAT. SG department-NOM . PL in ministry-PREP. PL
‘According to the prime minister, one should start with (lit. for the
beginning) systematically separating the ministry departments.’
(unauthored text from the newspaper subcorpus of the RNC, 2003)
The last constraint, which only allows for nouns of the second declension to
have Gen2 marking, is absolute. Apparently, the reason for this is historical: it
was the second declension with which the -u-declension, the assumed source of
Gen2 forms, has merged and where the ending ‘shuffling’ could have happened.
Note, however, that although the u-declension was limited to masculine nouns,
in dialects Gen2 has expanded to neuters.
358 Michael Daniel
On the whole, except for the constraint on uncountability / abstractness, all

main constraints of RG80 (the declension type, the gender and animacy) are
well observed in the Corpus. On the other hand, only a few nouns that meet
these constraints may form the second genitive; or, at least, are attested as
Gen2 in the Corpus. Compare ‘house, home (Gen2)’, which is very frequent in
the corpus, with dvoru ‘(back)yard (Gen2)’, which does not have a single corpus
occurrence (though it does occur in the internet). RG80 (p. 492) suggests that the
following classes of nouns may form Gen2:
– mass nouns (e.g., vozduch ‘air’, beton ‘concrete’);
– physical states (e.g., vizg ‘shrieks’, chochot ‘roars of laughter’);
– human activities or states (gonor ‘pride, foolishly proud behavior’);
– sets of non-individuated elements (ljud ‘people’).
Garde (1998: 167) completes this list with nouns designating meteorological
phenomena (cholod ‘cold’, moroz ‘frost’, veter ‘wind’). Although they do sound
natural as Gen2 to my ear, they are certainly not among nouns most frequently
occurring in Gen2 even though they are quite frequent lexical items3; RG80
classifies them together with physical states rather than as a separate class.
This classification is not straightforward (as dolg ‘debt’ is put together with sets
of non-individuated elements, and the definitions of types of states are probably
too broad.
Even more importantly, Gen2 is so often on the borderline of the standard
language that Gen2-forming nouns simply cannot be listed in an exhaustive,
prescriptive way. RG80 is aware of this and accompanies the list of the words
that it provides with a reserve that it only includes the most common nouns.
But (Garde 1998: 168), when commenting on the classes, simply says that some
of the nouns in these categories do not form Gen2. His examples are chlebu
‘bread (Gen2)’, ovsu ‘oats (Gen2)’, doždju (‘rain (Gen2)’). Unfortunately, two of
the three forms he indicates as non-existent do not sound unnatural (at least to
me) and occur in the Internet: bez chlebu ‘without bread (Gen2)’, zadat’ ovsu
‘provide oats (Gen2)’ (especially to horses), and the latter form even has a
couple of occurrences in RNC. Corpus statistics is a more adequate tool here,
one that allows to describe rather than prescribe the lexical scope of the form
as well as to try to account for comparative frequencies of Gen2 formation for
different nouns.
3 Thus, within the disambiguated subcorpus of RNC, cholod ‘cold’ has 320 occurrences, 97 of
which are regular genitive forms and only one is the second genitive; moroz ‘frost’ has 461 /
109 / 2; and veter ‘wind’ 1020 / 178 / 5, respectively. Cf. this with čaj ‘tea’ that occurs 909
times, 103 times as a regular genitive and 171 as Gen2.
I have limited the search to the disambiguated subcorpus (to exclude iden-
tical dative forms) and to the texts created after 1960 (to exclude older contexts),
which gives a sample of about 1,500 occurrences of the second genitive in a
corpus of about 4,5 million tokens. Below, this sample is called the probe sample.
The resulting lexical distribution is shown in the Appendix 1. This distribu-
tion shows several tendencies; RG80 mentions most of them. The following
generalizations are validated by the table in Appendix 2, which shows that the
rate of occurrences of such nouns among Gen2 forms is significantly different
from the rate of their occurrence in a random corpus sample:
– there is a large amount of idiomatic contexts (cf. RG80: 493–494); some-
times the noun exists in this idiom only: e.g. sbit’ s pantalyku ‘to confuse’;
– many nouns are mass nouns (cf. RG80: 492), primarily potable liquids;
– many nouns have stems in velar -k, -g, -ch;
– the sample shows an unusually high rate of diminutives (cf. RG80: 493); as
the diminutive suffix is -Vk, these nouns also fall within the previous class.
On the other hand,

– there are very few occurrences of the second genitive of palatalized stems.
This shows a curious two-fold, formal and semantic nature of the lexical
distribution of Gen2. RG80 at great length investigates the semantic factors, pro-
viding a detailed semantic classification of the Gen2-forming nouns, but seems
to be unaware of formal factors. Thus, it starts the whole discussion by a reference
to the “genitive in -u and -ju”, where -ju is the orthographic variant of Gen2 with
palatalized stems. Among the Gen2-forming nouns the following palatalized
stems are mentioned: kafel’ ‘ceramic tile(s)’, reven’ ‘rhubarb’, jačmen’ ‘barley’,
imbir’ ‘ginger’, kisel’ ‘jelly drink’, mindal’ ‘almond’, mitkal’ ‘calico’, našatyr’
‘sal ammoniac’, chmel’ ‘hops’, ščeben’ ‘crushed stone’, jantar’ ‘amber’. Although
some of these nouns are found as Gen2 in the Corpus, they tend to occur in
older, 19th century texts, while forms like spirtu ‘alcohol (Gen2)’ or sacharu
‘sugar (Gen2)’ are widely used in modern texts. With two exceptions, gogel’-
mogelju ‘egg-based sweet drink (Gen2)’ and ovošču ‘vegetable (Gen2)’, no pala-
talized stem made its way into the limited sample described above, and in each
case there is an obvious imitation of substandard speech or a quotation from an
older text.
360 Michael Daniel
(17) <. . .> так имей в виду, если вздумаешь отходить, я тебе дам
гогель-могелю, не хуже немецкого молока.
<. . .> tak imе-j v vid-u, еsli vzdumaj-еš’ otchodi-t’,
so have-IMP in view-LOC . SG if decide-PRS .2 SG go.back-INF
ja tеbе da-m gogеl’-mogеl-ju, nе chužе
1 SG . NOM 2SG . DAT give-PRS .1 SG gogle.mogle-GEN 2 not worse
nеmеck-ogo molok-a.
German-GEN . SG milk-GEN . SG
‘. . . so keep in mind that if you try to retreat, I’ll feed you some
gogle-mogle, even better than the German milk!’ (Vasilij Grossman. Žizn’
i sud’ba (1960))
(18) В старинных книгах по огородничеству писалось: “ <. . .> ты только

почни в этот день, непременно огород будет добр и овощу
будет много. . .
V starinn-ych knig-ach po ogorodničеstv-u pisa-l-o-s’:
in ancient-PREP. PL book-PREP. PL following gardening-DAT. SG write-PST- N - M / P
<. . .> ty tol’ko počn-i v ètot dеn’, nеprеmеnno
2SG . NOM only start-IMP in this.ACC . SG day.ACC . SG certainly
ogorod bud-еt dobr
garden.NOM . SG be.FUT-3 SG kind.NOM . SG
i ovošč-u bud-еt mnogo. . .
and vegetable-GEN 2 be.FUT-3 SG a.lot
‘Old gardening books said: “. . . start working on this day, and the garden
will certainly be fertile and there will be plenty of vegetables”.’
(Vera Krippa. Bogatyrskaja skazka (2002); newspaper corpus)
Note that u-declension nouns could not have palatalized stems, either.
The tendency of palatalized stems not to form Gen2 in modern language
should however be further investigated. The stems in palatal -j easily form Gen2
(čaj ‘tea’ and kraj ‘side, edge’ are on the top of the table in appendix 14). All
words with palatalized stem listed by RG80 are textually less frequent than the
words in Appendix 1. On the other hand, the opposite tendency for the stems in
velars ‑k, ‑g, -ch is beyond doubt. As shown in Appendix 2, 46 out of 140 differ-
ent lexemes that occur in the probe sample as Gen2 have stems in velars; while
4 Though, incidentally, these are palatal but not palatalized stems as, unlike the cases dis-
cussed so far, the Russian [j] has no non-palatalized counterpart.
in a random sample of contexts with inanimate masculine genitive forms of the

second declension out of 181 lexemes only 27 have stems in velar; this gives
about 34 per cent on 17 per cent. The higher rate of second genitive among velar
stems can not be explained etymologically: there is only one velar stem among
the words cited by Shakhmatov (1957: 82ff.), vьrchъ (vьrchъ) ‘top’, while all the
rest except domъ ‘house’ have stems ending in a coronal consonant. In some
way or another, the rate of velar stems grew over time.
Such a low start for velar stems, in my eyes, makes improbable or at least
highly speculative the idea that the expansion of velar Gen2 is a matter of formal
analogy with vьrchъ or other Gen2 that emerged later. At any rate, this hypothe-
sis is not more convincing than the suggestion that the observed distribution is a
result of historical coincidence. Alternatively, one could think of the correlation
between labialization and velars, discussed in Croft (2003: 200–201), who, with a
reference to previous research, indicates a higher cross-linguistic frequency of
the presence of labialization on velars and uvulars (as well as labials) than on
coronal and palatal consonants; more recently, the same correlation on a much
vaster database is shown in Maddieson et al. (2014) (also p.c.). In principle, such
correlation may explain the formal factors underlying the lexical distribution
of Gen2.
5 Functional Scope
In his discussion of Gen2, Paul Garde notes (once again without discussion) that
the main genitive inflection and Gen2 are not interchangeable; rather, each of
them is specialized in its own functional domain (Garde 1998: 166). On the one
hand, in many contexts, substituting the regular genitive with Gen2 is indeed
ungrammatical. On the other, however, it is in terms of preference (sometimes,
a very strong preference) that Garde’s claim about the specialization of Gen2
within its domain should be understood. As discussed above in section 2, even
in those cases where the use of the regular genitive seems to be on the border of
admissibility, the use of Gen2 is better described as a very strong preference.
This applies not only to lexical constraints, such as the apparent absence of
the regular genitive ??čajka ‘tea (diminutive)’, but also to contextual preferences.
Thus, ni razu ‘not a single time (Gen2)’ has 8,762 occurrences in RNC, and
??ni raza (same with the regular genitive) would probably be judged impossible
by many speakers. Yet, the latter has 12 occurrences in RNC (let alone in the
Internet), including some classical texts:
362 Michael Daniel
(19) О кокетстве она и не думала ни раза; ей не приходилось даже

воздерживаться.
O kokеtstvе ona i nе duma-l-a ni
about coquetry-PREP. SG 3 SG . F. NOM and not think-PST- F none
raza; ej nе prichodi-l-o-s’ dažе vozdеrživa-t’-sja.
time-GEN 3 SG . F. DAT not need-PST- N - M / P even refrain-INF- M / P
‘She never (lit. not a single time) thought of flirting; she didn’t even
have to restrain herself.’ (Leo Tolstoj. Vojna i mir (1867–1869))
Thus, while substituting regular genitive with Gen2 may yield ungrammatical
sentences, the reverse situation should be thought of in terms of improbability
rather than strict ungrammaticality. The present section will describe this asym-
metrical division of labor between the two forms.
Garde (Garde 1998: 168–169) divides all uses of Gen2 into two major classes,
free (“vivants”) and idiomatic (“figés”). A slightly reworked5 version of this classi-
fication is as follows.
Free uses:
– under measure nouns or quantifiers (cf. 22, 25);
– in quantifying impersonal constructions (cf. 24);
– as transitive objects where the action affects only some quantity of the sub-
stance (cf. 23);
– governed by some prepositions, including bez ‘without’ and iz and ot in the
sense of ‘out of, caused by’ (cf.).
Idiomatic uses:
– negative idioms: ni šagu ‘not a single step’, cf. also (20);
– governed by spatial prepositions: iz lesu ‘from the forest’, s kraju ‘on/from
the edge’, cf. also (21);
– governed by dlja ‘for’: dlja strachu.
Classification by AG80 is similar to that of Garde, though less clearly arti-

culated. In addition to the primary division between idiomatic vs. free uses it
5 Garde’s classification is more detailed. He considers measure nouns and quantifiers sepa-
rately; and puts transitive objects of the verbs chotet’ ‘want’ and prosit’ ‘ask for’ in a separate
class.
classifies all contexts into quantificational and spatial prepositional uses with
e.g. s ‘from upon’, ot ‘from the vicinity of’, iz ‘from inside; from’ and do ‘to the
vicinity of, as far as’, also in their secondary abstract meanings (mainly cause).
There are several problems with both approaches to the systematization of Gen2
uses.
The first problem is the interpretation of the label ‘idiomatic’ or ‘figés’. The
nominal expressions in question are not, strictly speaking, idioms. They are, in
their vast majority, semantically compositional and allow for various modifica-
tions. Thus, (20) uses a diminutive of the head noun, producing ni raziku ‘never
ever (lit. not a small time)’, a variation on ni razu ‘not once’:
(20) ― Меня сам ни разику не бил, ― рассказывала она о Ефиме. ― По

улице на мужиков с кулаками бегал, а меня ― ни разику. . .
Mеnja sam ni raz-ik-u nе bi-l, ―
1 SG . ACC self.NOM . SG none time-DIMIN - GEN 2 not beat-PST. M
rasskazyva-l-a ona o Efim-е. po ulic-е
tell-PST- F 3 SG . F. NOM about Efim-PREP. SG through street-PREP. SG
na mužik-ov s kulak-ami bеga-l, a mеnja ― ni
on man-ACC . PL with fist-INS . PL run-PST. M but 1 SG . ACC none
raz-ik-u. . .
time-DIMIN - GEN 2
‘He never ever beat me, she was saying about Efim. He would run down
the street after men with his fists clenched, but not once did he beat me.’
(Aleksandr Solženicyn. Matrenin dvor (1960))
Expressions like ni šagu ‘not a (single) step’, iz lesu ‘from the forest’ or dlja
strachu ‘for (to induce) fear’, s kraju ‘on/from the edge’ may easily incorporate
adjectives: ni odnogo / edinogo šagu ‘not one / single step’, iz temnogo lesu
‘from the deep (lit. dark) forest’, dlja puščego strachu ‘for (to induce) stronger
fear’, s samogo / dal’nego / samogo dal’nego kraju ‘from/on the very / farther /
most distant edge’. In many cases modifying adjectives may be considered as
intensifiers, so that such extended prepositional NPs may be viewed as lexical
elaborations on the idioms. In these context, adjectives do not bring new referen-
tial meaning but reinforce the meaning of the original simple expression. How-
ever, adjectives may also sometimes be specifying:
364 Michael Daniel
(21) Крэк, как корова, обежал Чарли, сделав полукруг, и зашел к нему
со звучащего боку.
Krèk, kak korov-a, obеža-l Čarli, sdеla-v polukrug,
Crack as cow-NOM . SG run.about-PST Charlie do-CVB half.circle.NOM . SG
i zašе-l k nеmu so zvučašč-еgo bok-u.
and come.into-PST. M to 3 SG . M . DAT from sound.P TCP - GEN . SG side-GEN 2
‘Moving like a cow, Crack (proper name) made a half-loop around Charlie
and approached his other side (lit. approached him from the side),
the side that was emitting sounds.’
(Jurij Mamleev. Amerikanskije rasskazy / Charlie (1975–1999))
As discussed in section 4, there are some lexical preferences of Gen2 forma-

tion for certain nouns. Whether the same noun is more typically inflected for
Gen2 with one preposition that requires a genitive than with another however
remains to be seen. In any case, as the examples above show, these are not
idioms in the common sense of the term. Rather, they are more or less close-
knit constructions allowing to modify its components or to add new lexical
material.
On the other hand, the use of Gen2 probably correlates with the degree
of how tight is the connection between the preposition and the noun or with
its resemblance to the prototypical construction (which I assume to be a plain
collocation of the preposition and the noun inflected for Gen2). Thus, in a prepo-
sitional phrase that allows Gen2, the frequency of the latter seems to be lower
when it includes an adjectival modifier (and probably lower for specifying modi-
fiers than for intensifying modifiers). On the opposite end of the same continuum,
as RG80 notes, are some nominal stems that only occur under a specific pre-
position and strongly prefer Gen2 over the regular genitive inflection. Cf. RNC
statistics for the forms of the noun *prosyp (a derivation from prosypat’sja
‘wake up’) that only occurs under bez: bez prosypu (Gen2) 87 vs. bez prosypa
(genitive) 22. True idioms are then an extreme case of how tight the relation
between a preposition and a noun can be.
The second problem posed by the two classifications is the special focus
they make on the use of Gen2 in spatial prepositional phrases. Garde puts such
uses apart not only from abstract prepositions, as bez ‘without’, but also from
spatial prepositions in their abstract meanings, such as s ‘from upon’ in contexts
like s golodu ‘because of hunger’ or so strachu ‘out of fear’. With some reserves,
RG80 classifies all uses into quantificational and spatial (the latter, however,
include abstract uses of spatial prepositions), also underlining the presence of
the elative component of meaning in most spatial contexts. However, out of the
seven prepositions that require genitive (bez ‘without’, dlja ‘for’, do ‘up to, as far
as’, iz ‘from (inside)’, ot ‘from (the vicinity of)’, s ‘from upon’, u ‘at’) five have
primarily spatial and three more specifically elative semantics6. The statistics of
the choice between the genitive and Gen2 in modern texts of the disambiguated
subcorpus of the RNC, as represented in table 5 in the appendix, does not show
specific semantic association of Gen2 with spatiality and especially with elative
meaning. The preposition that has a highest percentage of Gen2 is indeed s
‘from upon’, but next – with a much lower rate – comes bez ‘without’, next
again iz ‘from inside’; but then again do ‘up to, as far as’. This, to my eyes, is a
kind of evidence that should be interpreted as lexical preferences of individual
lexical items rather than an indication of any semantic correlations.
I suggest to treat all prepositional uses of Gen2 together, including them into
what I will call phrasal uses of Gen2. The label refers to the fact that phrasal
Gen2, as discussed above, relies on lexical relations between the function word
and the head; individual preposition selects Gen2 inflection for individual noun.
Phrasal Gen2 includes primarily prepositional phrases, but the presence of some
highly frequent phrases with ni (ni razu ‘not a single time’, ni šagu ‘not a single
step’) makes its syntactic scope broader (and ultimately also includes some verb
plus noun combinations discussed below).
Phrasal uses are opposed to the first three groups of Garde’s free (“vivants”)
uses: Gen2 in quantifying expressions (22) and in argument position, including
that of the direct object (23) and subject (24) position.
(22) Пойти билет купить на поезд на родину да две бутылочки

лимонаду в дорогу.
Poj-ti bilеt kupi-t’ na poеzd na rodin-u
go-INF ticket.NOM . SG buy-INF on train.ACC . SG on fatherland-ACC . SG
da dv-е butyločk-i limonad-u v dorog-u.
and two-F little.bottle-PL . NOM lemonade-GEN 2 in road-ACC . SG
‘Should I go and buy a train ticket back home and two bottles of soda
water for the journey?’ (Ludmila Petruševskaja. Malenkaja volšebnica (1996))
(23) Да, совсем забыл, я же хотел выпить соку.

Da, sovsеm zaby-l, ja žе chotе-l vypi-t’ sok-u.
and completely forget-PST. M 1 SG . NOM well want-PST. M drink-INF juice-GEN 2
‘Oh, I completely forgot – I wanted to have some juice.’ (Sergej Jurskij.
Bumažnik Chofmanna (1993))
6 The Russian genitive is thought to go back to the Indo-European ablative.

366 Michael Daniel
(24) Как это, чтобы в доме не было шоколаду?! ― возмутилась Мур.

Kak èt-o, čtoby v dom-е nе by-l-o šokolad-u?!
how this-SG . N that in house-PREP. SG not be-P ST- N chocolate-GEN 2
vozmuti-l-a-s’ Mur.
outrage-PST- F- M / P Mur
‘What? No chocolate in the house? – Mur was outraged.’
(Ludmila Ulickaja. Pikovaja dama (1995–2000))
These types of uses have some common properties. Most importantly, there
is no closed set of heads that typically combine with nouns inflected for Gen2
(hence the opposition between this group and phrasal uses headed by a closed
set of prepositions). Naturally, many examples of Gen2 in quantifying expressions
occur with semantically broadest measure adverbs, such as mnogo ‘a lot of’ and
nemnogo and malo ‘some’ (25). But their synonyms (26) as well as any noun that
is contextually understood as a measure unit (27) has the same effect.
(25) <. . .> когда он обнимал её прохладное тело в лёгком крепдешиновом

платье, вся она казалась ему сладкой, словно в эту белизну и впрямь
добавили немного сахару.
Kogda on obnima-l eё prochladn-oе tеl-o
when 3 SG . M . NOM hug-PST POSS .3 SG . F fresh-ACC . N . SG body-ACC . SG
v lёgk-om krеpdеšinov-om plat’-е, vs-ja
in light-PREP. F. SG crepe.de.chine-PREP. F. SG dress-PREP. SG all-F. SG
ona kaza-l-a-s’ еmu sladk-oj, slovno v èt-u
3 SG . F. NOM seem-PST- F- M / P 3 SG . M . DAT sweet-INS . F. SG as.if in this-ACC . F. SG
bеlizn-u i vprjam’ dobavi-l-i nеmnogo sachar-u.
whiteness-ACC . SG and really add-PST- PL some.GEN sugar-GEN 2
‘When he was embracing her fresh body clad in a light crepe-de-Chine
dress, she seemed so sweet, as if someone had indeed added some sugar
to this whiteness.’ (Irina Murav’ёva. Meščanin vo dvorjanstve (1994))
(26) “Полно песку в волосах”, ― пришло в голову Елене. . .

“Polno pеsk-u v volos-ach”, priš-l-o v golov-u Elеn-е. . .
full sand-GEN 2 in hair-PREP. PL come-PST- N in head-ACC . SG Elena-DAT. SG
‘Lots of sand in my hair, thought Elena.’
(Ludmila Ulickaja. Kazus Kukockogo (2000))
(27) Пару полная квартира, а ему хоть бы хны.

Par-u poln-aja kvartir-a, a еmu chot’ by chny.
steam-GEN 2 full-NOM . F. SG flat-NOM . SG and 3 SG . M . DAT even CONJ no.heed
‘The flat is full of steam (lit. a flat-ful of steam) – but he doesn’t care.’
(Andrej Volos. Nedvižimost’ (2000))
Similarly, any verb may introduce Gen2 in an object position, provided that it
allows some kind of partial affectedness7. Again, there is a very typical situa-
tional frame, one of drinking a potable liquid, as in (23). But any similar frame
will do, as eating in (28) or pouring tobacco in (29):
(28) Я выпил водки, съел чесноку.

Ja vypi-l vodk-i, s″е-l čеsnok-u.
1 SG . NOM drink-PST. M vodka-GEN . SG eat-PST. M garlic-GEN 2
‘I drink some vodka and ate some garlic.’
(Evgenij Vesnik. Darju, čto pomnju (1997))
(29) Он вздохнул, вынул папиросную бумагу, насыпал табаку и стал

лепить папироску.
On vzdochnu-l, vynu-l papiros-n-uju bumag-u,
3 SG . M . NOM sigh-PST. M take.out-PST. M cigarette-ADJV-ACC . M . SG paper-ACC . SG
nasypa-l tabak-u i sta-l lеpi-t’ papirosk-u.
pour-PST. M tobacco-GEN 2 and begin-PST. M shape-INF little.cigarette-ACC . SG
‘He sighed, took a piece of cigarette paper, poured some tobacco and
started making himself a cigarette.’
(Jurij Dombrovskij. Chranitel’ drevnostej (1964))
Finally, in the negated subject position, again, there are no lexical con-
straints or preferences on the context; whenever the regular genitive may be
used and the noun may inflect for Gen2, Gen2 may be used, too.
7 Although this does not specifically relate to Gen2, and is in no way a special property of
Russian ‘partitives’, either, a note should be made on how the distinction between complete
and partial affectedness is realized in the alternation of accusative and genitive marking on
DO, respectively. When using a mass noun, speakers of course may refer to all the substance
in the world – but this happens relatively rarely. As a result, the distinction between genitive and
accusative DO of mass nouns in contexts where both are available is similar to non-referential
vs. referential.
368 Michael Daniel
In Garde (1998) or RG80, no attempt is made as to group these three func-

tions on a functional semantic basis. Garde puts them together simply as free (as
opposed to more idiomatic) uses; abstract prepositional uses also belong here.
Unlike Gen2, genitive noun phrase in subject (under negation) and direct object
positions is a very frequent type of context and subject to extensive research in
Russian studies (e.g. by Paducheva; cf. Paducheva 2006 on genitive DO under
negation with further bibliography; as well as Paducheva 2012 and Letuchiy
2012). From the language internal perspective of the study of the use of the
regular genitive, the commonalities between partially affected DO and measure
noun phrases and negative polarity may indeed be weak. However, if we con-
sider the same issue from the viewpoint of the typology of partitives (e.g. Kittilä
& Luraghi, this volume; Miestamo, this volume), it becomes obvious that the
three uses that Gen2 chooses from all genitive contexts are typologically parti-
tive. I will thus call all three uses partitive Gen2. As has been indicated in each
of the three cases, however, even for those nouns that form Gen2, partitive
contexts do not require Gen2 inflection; rather, they allow for the genitive to
alternate with Gen2. It is only in this sense that Gen2 is associated with partitive
functions.
The last use of Gen2 is a rare type of contexts not mentioned in (Garde 1998)
or in RG80. It does not fit under the definitions of phrasal or partitive Gen2 given
above and is of clearly constructional nature. As in many other languages,
Russian, properties of an object may be described by the genitive of a property
noun modified by an adjective: ogromnoj dliny chvost ‘an extremely long tail’,
komsomolka nebol’šogo uma i skromnych sposobnostej ‘a komsomol girl of
limited wits and modest talents’. The genitive NP is typically dependent on the
object NP whose property it describes but may become a free attribute and shift
to the predicative position (cf. 34). In such contexts, for those nouns that inflect
for Gen2, Gen2 may be used; I will call this use feature Gen2. The noun rost
‘height (of a human)’ is the most frequent property noun used in feature Gen2
(30), but adhoc property nouns are also possible, as (31). Incidentally, the vast
majority of the contexts are human features; (32) is clearly a metaphorical exten-
sion, and examples like (33) are rare.
(30) Это оказалась старуха такого большого росту, что редко

встретишь, да и по глазам приметная.
Èt-o okaza-l-a-s’ staruch-a
this-NOM . N . SG turn.out-PST- F- M / P old.woman-NOM . SG
tak-ogo bol’š-ogo rost-u, čto rеdko vstrеti-š’,
thus-GEN . M . SG big-GEN . M . SG height-GEN 2 that rarely meet-2SG
da i po glaz-am primеtn-aja.
and and through eye-DAT. PL remarkable-NOM . F. SG
‘The old woman turned out to be unusually tall, and she had striking
eyes.’ (Pavel Bažov. Dorogoj zemli vitok (1948))
(31) Петляя по земле после фронта и госпиталей, вращаясь, так сказать,

в массах, немало повидал я красивых женщин, хотя бы на той же
Кубани любимая моя медсестра была не последнего ряду, одно
время даже самой красивой на всём белом свете казалась.
Pеtlja-ja po zеml-е poslе front-a i
loop-CVB through earth-DAT. SG after front-GEN . SG and
gospital-еj, vrašča-ja-s’, tak skaza-t’, v mass-ach,
military.hospital-GEN . PL turn-CVB - M / P so say-INF in mass-PREP. PL
nеmalo povida-l ja krasiv-ych žеnščin, chotja
not.few see-PST. M 1 SG . NOM beautiful-ACC . PL woman.ACC . PL though
by na toj žе Kuban-i ljubi-m-aja
CONJ on that.PREP. F. SG even Kuban-PREP. SG love-PTCP. PASS - NOM . F. SG
moj-a mеdsеstr-a by-l-a nе poslеdn-еgo rjad-u,
POSS .1 SG - NOM . F. SG nurse-NOM . SG be-PST- F not last-GEN . M . SG order-GEN 2
odn-o vrеm-ja dažе sam-oj krasiv-oj na

one-ACC . N . SG time-ACC . SG even most-INS . F. SG beautiful-INS . F. SG on
vs-ёm bеl-om svеt-е kaza-l-a-s’.
all-PREP. M . SG white-PREP. M . SG world-PREP. SG seem-PST- F- M / P
‘Traveling in the country after the war and all the hospitals, meeting
simple people so to speak, I’ve seen a lot of beautiful women. For instance,
when I was in Kuban region, the nurse I fell in love with was quite pretty
(lit. not of the last row), I even used to think she was the most beautiful
woman in the world.’ (Viktor Astaf’ev. Oberton (1995–1996))
(32) Да ведь комар-то не обыкновенный, а гвардейского росту!

Da vеd’ komar=to nе obyknovеnn-yj,
and even mosquito.NOM . SG = PTC not usual-NOM . M . SG
a gvard-еjsk-ogo rost-u!
but guard-ADJV- GEN . M . SG height-GEN 2
‘But the mosquito was not an ordinary one, it was enormous!’
(lit. of a guard’s height) (Michail Šolochov. Podnjataja celina (1960))
370 Michael Daniel
(33) Дело у вас серьёзного размеру, а оптиков при деле нету.

Dеl-o u vas sеr’ёzn-ogo razmеr-u,
business-NOM . SG at 2 PL .GEN serious-GEN . M . SG size-GEN 2
a optik-ov pri dеl-е nеtu.
but optician-GEN . PL by business-PREP. SG there.is.no
‘You’ve got serious business (lit. a business of a serious size), but there
are no opticians available.’ (Ivan Kataev. Serdce (1928))
(34) ― Не будешь ты большого росту, не будешь! ― сердито закричал

Кузька, вскочив с места.
Nе budе-š’ ty bol’š-ogo rost-u, nе budе-š’!
not be.FUT-2 SG 2 SG . NOM big-GEN . M . SG height-GEN 2 not be.FUT-2 SG
sеrdit-o zakriča-l Kuz’k-a, vskoči-v s mеst-a.
angry-NOM . N . SG yell-PST. M Kuz’ka-NOM . SG jump-CVB from place-GEN
‘Kuzka jumped up and yelled angrily, you won’t grow tall (lit. you won’t
be of big height), no you won’t!’ (Mark Sergeev. Volšebnaja galoša (1971))
Above, I have described three types of contexts where the genitive may alternate
with Gen2. One of them is directly related to the partitive function; it has been
labeled partitive Gen2. The core of the other, phrasal Gen2, is constituted by
prepositional phrases. Some of these prepositions are primarily elative in their
meaning: s ‘from upon’, оt ‘from near’, iz ‘from within’. Through the concept of
removal, elative uses of Gen2 may be distantly connected to the partitive func-
tion. However, this connection is not specifically associated with Gen2. The
same connection may be observed for the regular genitive which is also used in
all these contexts; and in fact, if the two functions are indeed related, they are
so through the formal and functional merger of the Indo-European genitive and
ablative in the modern Russian genitive (Vaillant 1958: 128). Moreover, synchron-
ically, all prepositions that allow genitive also allow Gen2, and not all of them
are ablative or even spatial (e.g. dlja ‘for’). For the third type of Gen2, feature
Gen2, I am aware of no diachronic or semantic paths that could relate it to the
partitive functions.
What is then the Russian Gen2? (Zalizniak 1967) uses the term quantitative-
separative (količestvenno-otdelitel’nyj), and RG80 starts the discussion of the
form with a reference to the theoretical literature on partitives. However, there
are two other types, phrasal and feature Gen2, that are unrelated to the partitive.
To what extent it may be considered a representative of the partitive category in
Russian, keeping in mind that it is only optional marking, a morphological alter-
native available to the genitive in partitive contexts? There are many different
ways of answering this question, depending on what one considers a language-

specific representative of a cross-linguistic category. If one puts cross-linguistic
and language-specific categories totally apart, into different and independent
perspectives of linguistic research (Haspelmath 2010), then Gen2 as a ‘descrip-
tive category’ is certainly a realization of the ‘comparative concept’ of the parti-
tive. One way of answering the question of how central the partitive function of
Gen2 is to conduct a further corpus research8.
6 Conclusions
The present paper is an overview of the so-called second genitive in modern
Russian. Formally, Gen2 is an alternative genitive inflection available for some
inanimate masculine nouns of the second declension. Its paradigmatic status,
discussed in section 2, is contradictory: in many ways, it behaves like a free
allomorph of the regular genitive ending -a with a rather limited formal distri-
bution; agreement is non-sensitive to the choice of Gen2 vs. regular genitive
ending. On the other hand, there are some semantic functional constraints
on where the genitive alternates with Gen2. In some contexts, as on genitive
dependents other than in measure noun phrases, the choice of Gen2 is ungram-
matical; while in some contexts, such as with mass nouns in direct object posi-
tion in situations of partial affectedness, Gen2 may even be preferred. However,
when the noun does not form Gen2, in such contexts it always uses the regular
genitive ending. I follow (Garde 1998) in adopting the term secondary case that
also includes second prepositional case and the new vocative form (Daniel
2009), which is in line with the theoretical discussion of ‘peripheral’ or ‘non-
canonical’ cases in (Spencer & Otoguro 2005) and (Corbett 2008).
Most probably, this ‘mild’ functional opposition is a result of a relatively
recent diachronic development (section 4). The common opinion on the origins
of Gen2 is that this comes back to the genitive of the u-declension that was
nearly extinct already at early stages of the written history of Russian and later
merged completely with the much more powerful o-declension. The genitive in
-u, however, survived as a variant and changed in two ways. A few former u-
nouns have lost it (all of them are animate). Many new nouns started to combine
with it, and among them a large group of nouns designating substance and
especially liquids, primarily potable liquids, including recent loanwords such
as kon’jak ‘cognac’. It is tempting to describe this evolution as functionally
8 Preliminary counts indicate that the functions labeled as partitive constitute not more than
half of all occurrences of Gen2.
372 Michael Daniel
driven by the compatibility of the noun with typically partitive contexts that
constitute one of the three typical uses of Gen2 – hence both sacking animate
nouns and expansion to mass nouns.
Indeed, mass nouns form a substantial part of Gen2 forming nouns, and
there is an absolute constraint on inflecting animate nouns for Gen2 in modern
Russian. But, however tempting this reconstruction is, it is not unproblematic.
The lexical distribution of Gen2 is at least two-fold (section 4), and in addition
to the semantic motivation there also seem to be a formal correlation with the
type of the auslaut: nominal stems with velar auslaut seem to be more prone to
inflect for Gen2 than other stems. This may be somehow related to the observa-
tion in Croft (2003: 200–201) and Maddieson et al. (2014) that labialization on
velars are cross-linguistically much more widespread than elsewhere; but the
exact articulatory nature of this correlation in Russian, where the velars are
not labialized but simply more easily combine with -u of Gen2, is unclear. In
anyway, the expansion of Gen2 can not be explained in purely functional terms.
Moreover, an inspection of the types of contexts where Gen2 is used also
shows a strong heterogeneity (section 5). Along with the typically partitive contexts
(as a dependent in a noun phrase headed by a measure or quantity word, as a
direct object in the situation of partial affectedness, and in the subject position
under negation, all labeled partitive Gen2), the form is also widely used under
prepositions that govern genitive (prepositional phrasal Gen2). It also has a
peripheral but a very distinctive use on property nouns ( feature Gen2) – this
use occurs rarely but is important because it is the only adnominal use available
to Gen2 except in measure expressions. A certain diachronic link to the partitive
function may be established for at least some prepositional uses, as some prepo-
sitions combining with the genitive have primarily ablative meaning (‘from
inside’, ‘from near’. ‘from upon’), and one other is a caritive preposition ‘with-
out’. The further evolution of phrasal Gen2 may have been governed by lexical
relations between the preposition and the noun; some sources even put most
prepositional uses into the class of idioms (which they are not, strictly speak-
ing). Again, feature Gen2 also adds an important element to the general picture
because, for property expressions, there is no visible link with the partitive func-
tion, and the category remains essentially heterogeneous.
I hope to have shown that whether Russian Gen2 should be considered as a
true partitive is a matter of discussion. It strongly depends on our view on the
relation between typology and language description. It is in any case obvious
that many contexts where the regular genitive optionally alternates with Gen2
are fairly well attested in the typology of the partitives (see the present volume).
However, preliminary counts of the relative frequencies of the three functions in
the corpus (given here only as a note 8) suggest that partitive uses do not exceed
half of the occurrences in terms of token frequency.
Appendix
The subcorpus only includes grammatically disambiguated documents dated
after 1960. Instead of the forms that occur only as part of idiomatic expression,
the expressions themselves are given.
Table 2: Lexical occurences of G2 on nominal lexemes
ni razu ‘not a single time’ (also sinlge 143 duch ‘spirit, soul’ 8
occurences of do razu i оt razu‘at once’)
tolk ‘sense’ 75 pol ‘floor’ 8
čaj ‘tea’ 72 prok ‘use’ 8
narod ‘people’ 66 s razbegu ‘with a run’ 8
(as in a running jump or dive’)
dom ‘house, home’ 61 sneg ‘snow’ 8
vid ‘aspect’, ‘appearance’ 57 tabak ‘tobacco’ 8
chod ‘move’ 44 sluch ‘rumour’ or ‘hearing’ 7
šag ‘step’ 22 supčik ‘soup’ (dimin) 7
kon’jak ‘cognac’ 19 s perepugu ‘out of fear’ 6
čaëk ‘tea’ (dimin) 19 spirt ‘alcohol’ 6
smeх ‘laughter’ 18 bez umolku ‘talking no stop’ 6
kraj ‘edge’ 14 god ‘year’ 5
sachar ‘sugar’ 13 žar ‘heat’ 5
s (odnogo) boku ‘from one side’ 12 rost ‘hight’ (animate, primarily human) 5
les ‘forest’ 12 svet ‘light’ 5
s glazu na glaz ‘personally’ 11 spasu net / ne bylo ‘there is/was no 5
(lit. from an eye to an eye) escape from something’
strach ‘fear’ 11 do zarezu ‘very much’ 4
golod ‘hunger’ 10 zvon ‘peal of bells’ 4
kipjatok ‘boiling water’ 10 s pantalyku as in sbit’ s pantalyku 4
‘confuse someone’
rod ‘descent’ (including nine ot rodu as 10 sok ‘juice’ 4
in pjat’ let ot rodu ‘five years old’)
vozduch ‘air’ 9 šum ‘noise’ 4
čas ‘hour’ 9 vek ‘century, lifelong period’ 4
Other lexemes only occurred less than four times.

Three occurences for: s boru po sosenkе ‘one small pint from every forest’
(proverb), gaz ‘gas’, dym ‘smoke’, žir ‘fat’, kvas ‘kvass’, krik ‘yell’, mach ‘swing’,
374 Michael Daniel
mjod ‘honey’ odekolon ‘cologne’, s perepoju ‘on the next day after strong drink-
ing’, ženskogo polu ‘of the female sex’, tabačok ‘tobacco’ (dimin), do upadu
‘(to dance, laugh etc.) to one’s full, lit. so that one starts falling’, charakter
‘temperament’, jad ‘poison’. Two occurences: voz ‘cart’, vychod ‘exit’, kagor
‘Cahors’, kipjatoček ‘boiling water’ (dimin), limonad ‘soda water’, luk ‘onion’, niz
‘bottom’, bez obmanu ‘no cheating’ , do otkazu ‘crammed full’, bez peredychu
‘without any stop for a rest’, pokoj ‘peace, rest’, prochod ‘passage’, s pylu s žaru
‘right from the stove’, razgovor ‘talk’, samogon ‘home-made spirit’, spirtik ‘home-
made spirit’ (dimin), sporu net ‘certainly; no objection’, sup ‘soup’, chvorost
‘brushwood’, chochot ‘roars of laughter’. One occurence: bez brjochu ‘no lies’
(twice but in the same context), s boju ‘in a fight’, vzdor ‘nonsense’, vizg ‘squeal’,
gipnoz ‘hypnosis’, gogel’-mogel’ ‘gogle-mogle’ (a very sweet liquid dessert made
of yolks and sugar), golos ‘voice’, zagad ‘guess’, zad ‘back’, irjan ‘a kind of
drink’, kajf ‘keef, euphoria’, kvasok ‘kvass’ (dimin), kerosin ‘kerosene’, klej
‘glue’, kompot ‘a drink made of cooked fruits’, kon’jačok ‘cognac’ (dimin), kofeëk
‘coffee’ (dimin), kofej ‘coffee’, kraešek ‘edge’ (dimin), kuraž ‘thoughtless bravery’,
lad ‘harmony’, ledochod ‘drifting ice’, ljot ‘the process of flying’, likjor ‘liquor’,
lišku ‘extra’, medok ‘honey’ (dimin), mir ‘peace’, nos ‘nose’, ovošč ‘vegetable’,
ogonëk ‘fire’ (dimin), odekolončik ‘cologne’ (dimin), par ‘steam’, (dali) percu ‘give
someone a punishmet’, pesok ‘sand’, pomin ‘memory of’, poroch ‘(gun)powder’,
poryv ‘impulse’, procent ‘percent’, razgon ‘run’, s razmachu ‘with a swing’,
rozdych ‘break’, rjad ‘row’, samogončik ‘home-made spirit’ (dimin), sаchаrок ‘sugar’
(dimin), svist ‘whistle’, smrad ‘odour’, syr ‘cheese’, talant ‘gift’, t’vorožok ‘soft
cheese’ (dimin), temp ‘tempo’, tyl ‘home front’, uderžu ‘(no) restraint’, dlja forsu
‘in order to show-off’, chrap ‘snore’, chrеn ‘horse raddish’, cvet ‘color’, čifirëk
‘very strong tea’ (dimin), šik ‘chic’, ves ‘weight’, bul’on ‘broth’.
Table 3: Formal distribution of Gen2
Random sample Second genitive sample

of the of the of the of the lexical items
contexts lexical items contexts lexical items used only once
velar auslaut 31 (13%) 27 (17%) 309 (31%) 47 (34%) 19 (31%)

mass nouns 6 (3%) 7 (4%) 236 (24%) 49 (35%) 23 (38%)
including liquids 1 (0,4%) 1 (0,6%) 170 (17,1%) 30 (21,4%) 15 (25%)
diminutives 1 (0,4%) 1 (0,6%) 43 (43%) 15 (11%) 10 (16%)
palatalized auslaut 31 (13%) 16 (10%) 2 (0,2%) 2 (1,4%) 2 (3%)
-j auslaut 5 (2%) 4 (2%) 92 (9,3%) 6 (4,3%) 3 (5%)
Total: contexts: lexemes: contexts: lexemes: lexical items used

230 161 992 140 only once: 61
Columns 2 and 3 represent the control sample. The second column gives the
occurrences of all singular genitive forms of the lexemes of the type indicated in
the first column in a random sample of singular genitive forms of inanimate
masculine nouns of the second declension. The third columns gives the number
of different lexemes in the same sample. The fourth column gives the occurrences
of the second genitive (disambiguated texts created after 1960) for each class.
The fifth column gives the number of different such lexemes. The last sixth
column gives the occurrences of the lexemes that only occurred once in the
second genitive.
It is important to note the following. Not only the first four lines show
higher frequencies in the second genitive than in the control sample, while in
the fifth line (palatalized stems) it is lower. Importantly, the frequency of the
second genitive is also significantly higher in the last column. This means that
the difference between the control sample and the second genitive sample is
not caused simply by several very frequent lexical items of this type. For
instance, tolku is as high as 8% of all second genitive occurrences which could
be a lexical factor in increasing the frequency of contexts where the velar is
used – simply because this lexeme is so frequent in this form. But looking at
the one before the last and the last column, we discover that the percent is just
as high; which means that having a velar auslaut is something peculiar of those
nouns that form the second genitive. The stems in -j are different: the token
frequency is again considerably higher than in the control sample, but the type
frequency is so close to that of the control sample that the high token frequency
may be due to the very high frequency of čaju (‘tea’ gen2 – 7 percent).
Table 4: Comparative frequencies of prepositional genitive and Gen2 for dom ‘house’ and
strach ‘fear’
dom ‘house’ strach ‘fear’

gen gen2 gen gen2
iz ‘from’ 67,3% (2841) 32,7% (1383) 100% (196) 0,0%
do ‘up to’ 64,3% (616) 35,7% (342) – –
ot ‘from’ 98,7% (924) 1,3% (12) 98,5% (1806) 1,5% (28)
dlja 100 (223) 0% 50,0% (9) 50,0% (9)
bez 100 (30) 0% 99,6% (279) 0,4% (1)
s(о) 96,7% (29) 3,3% (1) 16,6% (38) 83,4% (191)
u 100 (411) 0% 100% (26) 0%
These ‘blind’ (i.e. without looking into the context) calculations have been made
on the subcorpus of the RNC documents written after 1960.
376 Michael Daniel
Some examples of nouns in Gen2 that only exist in the idiom.
(sbit’) s pantalyku ~ pantalyka ‘confuse, lit. knock down from ?’

dat’ deru ~ dera
bez prosypu ~ bez prosypa
do otvalu ~ do otvala
do upadu ~ do upada
Table 5: Occurrences of various prepositions with the genitive

and Gen2 in the disambiguated subcorpus of modern texts
Gen Gen2 Total:

s ‘with’ 89,3% 10,7% 2108
1882 226
bez ‘without’ 95,9% 4,1% 1014
972 42
iz ‘from inside’ 96,7% 3,3% 3868
3742 126
do ‘up to’ 97,3% 2,7% 1603
1559 44
ot ‘from’ 98,6% 1,4% 2223
2192 31
dlja ‘for’ 98,8% 1,2% 1042
1030 12
u ‘at, near’ 100,0% 0% 810
810
Note that, in column 2, this table puts together genitives of those nouns that
may and may not have Gen2, so it should not be interpreted as a direct reflection
of prepositional preferences of Gen2; actual rates in the column Gen2 should be
higher. However, I assume that it does reflect, even if in a rough way, preferences
of the prepositions.
References
Bromley, Sofja V. & Lidia N. Bulatova. 1972. Ocherki morfologii russkikh govorov. [A survey of
Russian dialectal morphology]. Moscow: Nauka.
Corbett, Greville. 2008. Determining morphosyntactic feature values: the case of case. In
Greville G. Corbett & Michael Noonan (eds.), Case and grammatical relations: papers in
honour of Bernard Comrie, 1–34. Amsterdam: Benjamins.
Croft, William. 2003. Typology and universals. Cambridge: Cambridge University Press.
Daniel, Mikhail. 2009. “Novyj” russkij vokativ: istorija formy usechennogo obrashshenija skvozj
prizmu korpusa pis’mennych tekstov. [New Russian vocative. The history of the truncated
address in the written corpus]. In Xenia Kisseleva, Ekaterina Rakhilina, Vladimir Plungian
& Sergei Tatevosov (eds.), Korpusnyje issledovanija po russkoj grammatike, 224–244.
Moscow: Probel.
Garde, Paul. 1998. Grammaire russe. Paris: Institut d’études slaves.
Haspelmath, Martin. 2010. Comparative concepts and descriptive categories in cross-linguistic
studies. Language 86(3). 663–687.
Jakobson, Roman. 1971 [1936]. Beitrag zur allgemeinen Kasuslehre. In Roman Jakobson, Selected
writings II, 23–71. The Hague – Paris: Mouton.
Letuchij, Alexander. Perexodnostj [Transitivity]. www.rusgram.ru
Maddieson I., Flavier S., Marsico E., Pellegrino F., 2014. LAPSyD: Lyon-Albuquerque Phonological
Systems Database. http://www.lapsyd.ddl.ish-lyon.cnrs.fr/
Mustajoki, Arto. 2006. Pochemu narodu mnogo, ili novyje nabljudenija nad upotrebleniem
vtorogo roditelnogo padezha v sovremennom russkom jazyke [Why narodu mnogo, or
some new observations on the use of the second genitive case in Modern Russian]. In
Arto Mustajoki & Olga Pussinen (eds.), Integrum: tochnye metody i gumanitarnye nauki,
216–269. Moscow: Letnij sad.
Paducheva, Elena V. 2006. Genitiv dopolnenija v otricatelnom predlozhenii. Vorposy jazykoznanija
6. 21–43.
Paducheva, Elena V. 2011. Otricanie [Negation]. www.rusgram.ru
Russian National Corpus. 2012. www.ruscorpora.ru
Seržant, Ilia. Independent partitive genitive in North Russian and Russian. To appear in Ilia
Seržant and Björn Wiemer (eds.), Contemporary approaches to dialectology: the area of
North, Northwest Russian and Bielorusian vernaculars. Slavica Bergensia, 13.
Shvedova, Natalia Yu. (ed.). 1980. Russkaja grammatika. Moscow: Nauka.
Shakhmatov, Alexei A. 1957. Istoricheskaja morfologija russkogo jazyka. Moscow: AN SSSR.
Spencer, Andrew & R. Otoguro. 2005. Limits to case: a critical survey of the notion. In Mengistu
Amberber and Helen de Hoop (eds.), Competition and variation in natural languages: the
case for case, 119–145. Amsterdam: Elsevier.
Vaillant, André. 1958. Grammaire comparé des langues slaves. Tome II: Morphologie. Première
partie: flexion nominale. Lyon: Editions IAC.
Zalizniak, Andrej A. 1967. Russkoje imennoje slovoizmenenie. Moscow: Nauka.
Zalizniak, Andrej A. 1995. Drevnenovgorodskij dialekt. Moscow: Jazyki russkoj kultury.
Katia Paykin1
10 The Russian Partitive and Verbal Aspect
The present article provides some new ways of solving a well-known mystery
of the aversion of the Russian partitive genitive for the imperfective aspect (cf.
Jakobson 1936, Padučeva 1998). According to our analysis, the functioning of
the Russian partitive genitive should not be considered comparable to that of
the Finnish or French partitive as far as verbal aspect is concerned, as the aspec-
tual marking in Russian is explicitly present on the verb and is independent
from the tense marking. Therefore, the verb (im)perfectivity is always expressed
independently from the case of the object (or the subject). We argue that the
partitive marked argument denotes a certain indefinite quantity presented as
an indivisible whole, the delimitation or bounding of the referent is thus manda-
tory. It can come from the process, which explains the predilection of the partitive
for the perfective aspect, or from other elements in the context, thus making the
imperfective compatible with the partitive, including in its progressive reading.
Keywords: partitive, genitive, verbal aspect, (in)definiteness, (im)perfectivity,

bounding
1 Introduction
The correlation between the use of the partitive and verbal aspect in Russian has
been regarded as an intriguing topic since Jakobson’s (1936) study, although few
authors have ventured to examine the subject in detail. The goal of the present
paper is to revisit the most recent analyses of this question, providing several
critiques and alternative solutions. We will start by clarifying what exactly the
term “partitive” means when considered in the context of verbal aspect (cf.
section 2). Having established that the reality covered by the label in this par-
ticular context corresponds to the partitive genitive, we will briefly discuss its
functioning as opposed to that of the nominative and accusative cases with
1 This article owes much to Marleen Van Peteghem, with whom I discovered the tricky question
presented here while working on the genitive in Russian. I would also like to thank Cédric
Patin, Sarah Growas and Ilja Seržant for their comments and our fruitful discussions that
helped clarify several of the points under scrutiny. Needless to say, all errors and inadequacies
that remain are the sole responsibility of the author.
380 Katia Paykin
which it can compete (cf. section 3). However, our main emphasis will be on the
relationship between the partitive genitive and the aspect of the verb assigning
it. We will, therefore, summarize the existing points of view on the subject (cf.
section 4) before presenting our own analysis (cf. section 5).
2 One label, two different notions

When applied to Russian, the label “partitive” can refer to two different linguistic
phenomena: a separate case or a particular use of the genitive. It can thus
designate a second genitive form available for some nouns, distinguishing a
supplementary case next to the genitive, or apply to the usage of one and the
same genitive case available in certain contexts.
Indeed, it has been argued (cf. Jakobson 1936, Neidle 1988, Paus 1994,
Franks 1995) that Russian has a separate case called “partitive”, considering
that a restricted set of masculine nouns in the singular possess a second genitive
form, morphologically identical to the dative case. As seen from example (1),
such masculine nouns as čaj ‘tea’ or sachar ‘sugar’ possess two genitive forms:
a regular form ending in –a/–ja and another form ending in –u/–ju. Feminine
and neuter nouns in the singular as well as all nouns in the plural have only
one form.2 It is the existence of this irregular u-form that eventually justifies the
introduction of a separate partitive case.
(1) a. kusok sachara / sacharu3

piece:NOM sugar:GEN / sugar:GEN 2
‘piece of sugar’
b. stakan čaja / čaju

glass:NOM tea:GEN / tea:GEN 2
‘a glass of tea’
The so-called u-form appears mostly on mass nouns (sachar ‘sugar’, čaj
‘tea’, mëd ‘honey’, ris ‘rice’, etc.) and is generally used in semantically partitive
2 Most feminine nouns in the singular take the ending –y, while most neuter nouns take the
genitive ending –a. In the plural, the most frequent ending for masculine nouns is –ov,
whereas most feminine and neuter nouns take a zero ending. For more details, see Van
Peteghem & Paykin (2013).
3 The majority of our examples come from the Russian National Corpus, texts after 1960 (RNC –
http://www.ruscorpora.ru) and Internet (Google). We have also used data from various linguistic
studies.
The Russian Partitive and Verbal Aspect 381
contexts. Yet whenever the u-form is possible, it is always a variant of an a-form

(cf. Brown & Franks 1995)4.
The plausibility of distinguishing a separate partitive case on the basis of
the u-variant of the genitive can thus be questioned. Even exceptional cases
known from textbooks to give rise to different interpretations, as in ((2)a), no
longer seem to provide two distinct readings, as can be seen from examples in
((2)b), taken from Russian National Corpus.
(2) a. Lesa net. vs. a0 . Lesu net.

forest:GEN NEG forest:GEN 2 NEG
‘There is no forest.’ vs. ‘There is no wood.’
b. Da vojna. Net stekla, net lesa, net kirpiča. (RNC)

yes war NEG glass NEG forest:GEN NEG brick
‘Yes, there is war. There is no glass, no wood, no brick.’
b0 . Pokatosti gor obrosli lesom; mesta, gde net

slopes mountains got.covered forest places where NEG
lesu, imejut vid želto-krasnyj. (RNC)
forest:GEN 2 have appearance yellow-rouge
‘The mountain slopes are covered with forest; places without forest
are of yellowish-reddish color.’
Besides, the u-morphology is possible with count nouns, such as čas ‘hour’, dom
‘house’, etc., as in (3).
(3) Progulivalsja ja bol’se času po svoej komnate. (RNC)

wandered 1 SG . NOM more hour:GEN 2 on own:DAT room:DAT
‘I wandered for more than an hour around my room.’
Moreover, the u-form can be used without any partitive meaning, namely when
governed by prepositions, as in (4).
(4) Zajac vyskočil iz lesu i pobezal polem. (RNC)

hare:NOM rushed.out from forest:GEN 2 and ran field:INS
‘A/The hare rushed out of the forest and ran through the field.’
4 It can be argued that certain –u forms do not have an –a form equivalent, like the diminutive
čajku ‘tea’ in čajk-u by sejčas litt. tea-GEN 2 COND now ‘I wish I could have some tea now’, which
does not have the corresponding –a form. However, this phenomenon is extremely rare and we
will not take it into account in the present paper.
382 Katia Paykin
Therefore, in the present paper, we will consider the u-form as a mere

variant of the a-form and reserve the label “partitive” for a particular use of the
genitive, i.e. the genitive that appears on internal verb arguments containing
mass nouns and bare plurals, as in (5).
(5) Ja prinesla sachara / vody / drov.

1 SG . NOM brought sugar(M ): GEN . SG / water(F ): GEN . SG / firewood(M ): GEN . PL
‘I brought (some) sugar / water / firewood.’
3 Partitive genitive vs. other cases

Normally, when a verb allows various case assignments (cf. Jakobson 1936,
Babby 1980, Neidle 1988, among many others), the case choice is associated
with different semantic interpretations. Most frequently, the partitive genitive
competes with the accusative, as in (6), where it marks the surface object.
(6) a. Ja kupila vody / jablok.

1 SG . NOM bought water:GEN / apples:GEN
‘I bought (some) water / apples.’
b. Ja kupila vodu / jabloki.

1 SG . NOM bought water:ACC / apples:ACC
‘I bought the water / the apples.’ or ‘I bought water / apples.’
However, it can also compete with the nominative case when it appears on
subjects of unaccusative verbs used in their impersonal form, as in (7).5
(7) a. Nočju snega navalilo! (Google)

night:INS snow(M ): GEN . SG piled.up:N
‘During the night, there fell piles of snow!’
b. Nočju sneg navalil – skol’zko, grjazno. (Google)

night:INS snow(M ): NOM . SG piled.up:M . SG – slippery dirty
‘During the night snow has piled up – it got slippery and dirty.’
5 It has to be noted, however, that partitive genitive subjects are quite rare, unlike negative
genitive subjects, as in (i):
(i) Zdes’ ne rastët gribov. (Babby 1980)
here NEG grows:3SG mushroom:GEN . PL
No mushrooms grow here.’
For a general overview of the question, see Van Peteghem & Paykin (2013).
When the partitive genitive alternates with other cases, the genitive typically
gives rise to an indefinite reading, whereas the accusative or the nominative
can be interpreted either as definite or indefinite.
3.1 Partitive genitive vs. accusative

When competing with the accusative, the partitive genitive marking of the direct
object yields an indefinite interpretation (cf. (6)), similar to the absence of a
determiner in English and to the partitive article du / de la or the indefinite
plural des in French (cf. the French translation of ((6)a), ‘J’ai acheté de l’eau /
des pommes’). However, the use of the accusative-marked direct object does
not impose a definite interpretation, as it may give rise to either a definite or an
indefinite interpretation, as illustrated in ((6)b). Therefore, both the accusative
and the partitive genitive can be used with an indefinite interpretation. As has
been shown in Paykin & Van Peteghem (2002), the indefiniteness expressed by
both the genitive and the accusative is semantically different, the genitive NP
emphasizing quantity, and the accusative NP denoting a class.
3.2 Partitive genitive vs. nominative

As we have noted earlier, the partitive genitive appears not only on objects, but
also on subjects of unaccusative verbs used in their impersonal form. The use of
the nominative case on the subject entails the agreement with the predicate, as
in ((8)a) and ((9)a), justifying the label “personal construction”. The genitive
marking of the subject, in its turn, entails the use of the impersonal construc-
tion, as the predicate does not show agreement with the subject but appears in
the neuter/third person singular form, as in ((8)b) and ((9)b).
(8) a. Starik poskol’znulsja v luže [. . .], ljudi

old.man slipped in puddle people:NOM
nabežali na nego. (RNC)
came.running:PL on 3 SG . ACC
‘The old man slipped in the puddle [. . .], people came running onto
and over him.’
b. Srazu ljudej nabežalo! (Google)

immediately people:GEN came.running:N . SG
‘Immediately there came people running!’
384 Katia Paykin
(9) a. On postavil na pol portfel’, i iz nego

3 SG . NOM put on floor briefcase and from 3 SG .GEN
natekla voda. (RNC)
leaked:F. SG water(F ):NOM . SG
‘He put his briefcase onto the floor and (some) water leaked out of it.’
b. Ot ego zipuna na lavku nateklo vody. (RNC)

from 3 SG .GEN coat on bench leaked:N water(F ):GEN
‘There leaked water on the bench from his coat.’
Although quite rare, the partitive genitive subject appears mostly in exclamatory
contexts putting the emphasis on the implicitly large quantity of the noun
referent.6
Similarly to what happens when the genitive competes with the accusative,
both the genitive and the nominative can give rise to indefinite readings, the
genitive marked subject insisting on quantity, and the nominative subject denot-
ing a class. The fact that the genitive subject is used with an impersonal form of
the verb accounts for yet another difference between the two case assignments:
the impersonal sentence, analyzed as thetic and all rhematic, presents the com-
ing into existence of an event, while the personal one needs more anchoring in
terms of the information structure, being potentially ambiguous between thetic
and categorical, in Kuroda’s (1979) terms.7
As shown in Van Peteghem & Paykin (2013), the partitive genitive cannot be
analyzed as a lexical case, although it is semantically conditioned and obliga-
tory with certain verbs. It marks internal arguments of verbs denoting a process
affecting a certain quantity, but it is not imposed by the thematic structure of the
verb. Following Neidle (1988), Brown & Franks (1995), the authors argue that the
partitive genitive is assigned by a null quantifier, inherent to certain verbs and
triggered with others in certain contexts, suggesting that one of these triggers
could possibly be the perfective aspect.
6 The large quantity reading comes from the verb where it is signaled by the presence of such
prefixes, as na- or po-.
7 The distinction between categorical and thetic judgments has been initially proposed by
Brentano, then adopted by Marty (1918) and finally reintroduced by Kuroda (1979). The cate-
gorical judgment contains two separate acts, the recognition of the subject, on the one hand,
and the assertion or the negation expressed by the predicate about this subject, on the other.
The thetic judgment, in its turn, solely represents the recognition or the rejection of the judg-
ment matter.
4 Partitive genitive and verbal aspect:

previous studies
Indeed, in Russian, the partitive genitive is much more frequent with perfective
than with imperfective verbs. According to Kiparsky (1998), this phenomenon
can be viewed as an “unexpected wrinkle”, since many studies assimilate the
partitive case in general to the imperfective aspect. Therefore, it could have
been logical to suppose a correlation between the use of the Russian partitive
genitive and the use of the imperfective aspect. Yet as early as 1936, Jakobson
argued that the use of the partitive genitive is not compatible with the use of
the imperfective aspect, as can be seen from his examples, given in (10).
(10) a. poel chleba (Jakobson 1936)

ate:PFV bread:GEN
‘(he) ate (some) bread’
b. *el chleba
ate:IPFV bread:GEN
c. el chleb
ate:IPFV bread:ACC
‘(he) ate (some) bread’
According to Padučeva (1998), however, Jakobson’s claim has to be simultane-

ously narrowed and broadened: on the one hand, only the use of the imperfec-
tive in the progressive meaning should be excluded; on the other hand, the
restriction is not limited to the partitive case but concerns all quantifier phrases.
In what follows, we will first present Kiparsky’s (1998) point of view on the
partitive in terms of aspect correlation; we will then examine Padučeva’s (1998)
treatment of the question before proposing some revisions to the existent analyses
of the seeming incompatibility between the genitive partitive and the imperfective
aspect.
4.1 Aspect – partitive correlation according to Kiparsky (1998)

Kiparsky (1998) claims that the use of the partitive can have an NP-related func-
tion as well as an aspectual function: in the former case, it is assigned “to quan-
titatively indeterminate NPs, even if the verb denotes a bounded event”, and in
the latter, “to the objects of verbs which denote an unbounded event” (Kiparsky
386 Katia Paykin
1998: 1). Thus, under Kiparsky’s analysis, the alternation between the accusative
and partitive object, for example, can entail an aspectual difference. Indeed,
Finnish data (Kiparsky 1998: 13) illustrate perfectly how the difference in the case
assignment can have implications for the aspectual interpretation of the verb, as
the presence of the partitive marked object in ((11)a) entails the unbounded
reading of the process, authorizing the use of the duration complement ‘for an
hour’. When the object appears in the accusative case, we obtain the bounded
reading of the process, compatible with the complement ‘in an hour’ (cf. Vendler
1967).
(11) a. Matti osti maitoa (tunnin) (Kiparsky 1998: 13)

Matti:NOM . SG buy:PST.3 SG milk:PAR . SG (hour:ACC )
‘Matty bought milk (for an hour).’
b. Matti osti maidon (tunnissa)

Matti:NOM . SG buy:PST.3 SG milk:ACC . SG (hour:INE )
‘Matty bought the milk (in an hour).’
We observe the same phenomenon in French, where the use of a partitive

article or an indefinite plural confers a non-bounded reading to the predicate
(cf. Bosveld 2000). In examples (12) and (13), taken from Bosveld (2000: 53),
the use of the partitive article with the subject and the use of the indefinite plural
object respectively make the bounded reading impossible, thus excluding the
presence of the French equivalent of an ‘in’-complement.
(12) a. Du gaz s’est échappé du tuyau pendant

PAR gas escaped from.the pipe during
des heures.
ART. INDF. PL hours
‘Gas leaked from the pipe for hours.’
b. *Du gaz s’est échappé du tuyau en une heure.

PAR gas escaped from.the pipe in one hour
(13) a. Marie a cueilli des fraises pendant des heures.

Mary picked ART. INDF. PL strawberries during ART. INDF. PL hours
‘Mary picked strawberries for hours.’
b. *Marie a cueilli des fraises en une heure.

Mary picked ART. INDF. PL strawberries in one hour
The use of the partitive case can therefore “imperfectivize” the process ex-
pressed by the verb that stays unchanged. As the partitive case in languages
like Finnish and the imperfective aspect in Russian both express the unbounded-
ness of the VP, the predilection of the partitive for the perfective aspect in Russian
is viewed by Kiparsky (1998) as rather unexpected. The partitive/accusative dis-
tinction is thus not “fully exploited [in Russian] to yield a four-way paradigm
where unbounded reference in the nominal and verbal domain are separately
marked” (Kiparsky 1998: 26).
However, the aspectual marking in Russian is explicitly present on the verb,
independent of the tense marking, and the verb (im)perfectivity is always ex-
pressed independently of the case of the object. The switch between the accusa-
tive and the partitive therefore has no impact on the aspectual interpretation of
the process. Indeed, the change of the verbal aspect does not necessarily entail
the case change of the object, as can be seen from example (14), just as the use
of the accusative and of the partitive cases on the object are both compatible
with the perfective aspect, as in example (15). The compatibility of the adverbial
time complements thus depends solely on verbal aspect.
(14) a. Petja čital knigu v tečenii časa / *za čas.

Peter:NOM read:IPFV book:ACC in duration hour within hour
‘Peter was reading (not to the end) a / the book for an hour.’
b. Petja pročital knigu *v tečenii časa / za čas.

Peter:NOM read:PFV book:ACC in duration hour / within hour
‘Peter read (completely) a / the book in an hour.’
(15) a. Petja kupil sliv *pjat’ minut / za pjat’ minut.

Peter:NOM bought:PFV plums:GEN five minutes / in five minutes
‘Peter bought (some) plums in five minutes.’
b. Petja kupil slivy *pjat’ minut / za pjat’ minut.

Peter:NOM bought:PFV plums:ACC five minutes / in five minutes
‘Peter bought (the) plums in five minutes.’
Moreover, as mentioned above, the use of the accusative case for objects
and of the nominative case for subjects can both yield an indefinite interpreta-
tion, which is thus not exclusively linked to the use of the Russian partitive.
Therefore, its functioning should not in fact be considered comparable to that
of the Finnish or French partitive as far as verbal aspect is concerned.
388 Katia Paykin
4.2 “Semantic obstacle” of indefiniteness according to

Padučeva (1998)
Let us now turn to the actual correlation between the partitive genitive and verbal
aspect in Russian. Jakobson’s (1936) original restriction of the partitive genitive
to verbs in the perfective aspect has been revisited by Padučeva (1998) who
claims that not only arguments marked by the partitive genitive but all quanti-
fier phrases (QP) seem to be averse to the imperfective aspect, a phenomenon
not specific to Russian but also characteristic of other Slavic languages (cf.
Wierzbicka 1967 for Polish).
(16) a. Ja pročël tri stranicy. (Padučeva 1998: 73)

1 SG . NOM read:PFV three pages:GEN
‘I read three pages.’
b. ?Ja čitaju tri stranicy.

1 SG . NOM read:IPFV three pages:GEN
Yet according to Padučeva (1998), the constraint concerns exclusively the

progressive use of the imperfective, since the non-progressive imperfective, i.e.
the one expressing potentiality, habituality, praesens historicum, etc., is perfectly
compatible with quantifier phrases. Thus, examples ((17)a) and ((17)b), both
taken from Padučeva (1998: 75), illustrate the compatibility between the QP and
the imperfective aspect in praesens historicum and the so-called general-factual
readings respectively.
(17) a. On delaet pjat’ šagov k stene.

3 SG . NOM makes:IPFV five steps:GEN to wall
‘He takes five steps toward the wall.’
b. On sažal vse derev’ja v étom

3 SG . NOM planted:IPFV all:ACC trees:ACC in this:PREP
sadu. (from Glovinskaja 1982)
garden:PREP
‘He planted all the trees in this garden.’
Padučeva (1998) notes, however, that there are contexts containing verbs in
the progressive imperfective that do authorize the use of QPs. Indeed, example
((18)a) contains such a verb but, since there is no “accumulation of result”
(Padučeva 1998: 80), the use of a numerical QP is acceptable. The use of the
partitive in the same exact context seems impossible, however. This can be seen
from example ((18)b).
(18) a. On nesët dva arbuza. (Padučeva 1998: 80)

3 SG . NOM carries:IPFV two watermelon:GEN
‘He is carrying two watermelons.’
b. *On nesët drov.

3 SG . NOM carries:IPFV firewood:GEN
According to Padučeva (1998: 80), the impossibility of using the partitive case
in the contexts of imperfective progressive compatible with the use of the QP
may lie in the “inherent indefiniteness of the Russian Partitive”, or even its
non-referentiality. The incompatibility between the partitive and the progressive
imperfective is thus semantically motivated. In the rare cases where one does
find the partitive genitive with the imperfective progressive aspect, it is “the
context that cancels the non-refentiality of the partitive” (Padučeva 1998: 81),
as in example (19).
(19) Nesu žene jablok.

carry:1 SG wife:DAT apples:GEN
‘I am carrying apples for my wife.’
The possible objection to the semantic motivation of the incompatibility

between the partitive and the imperfective (i.e. the fact that French and English
equivalents of the ungrammatical Russian sentences are perfectly well-formed
and contain respectively a partitive article and the quantifier some) is answered
by Padučeva in the following way: there is no semantic synonymity between
French and English examples, on the one hand, and Russian examples, on the
other hand. Russian sentences containing the partitive genitive emphasize quan-
tity, which French and English equivalents do not do.
(20) a. *Ja p’ju vody. (Padučeva 1998: 81)

1 SG . NOM drink:IPFV water:GEN
vs. b. He is drinking some water.
c. Je bois de l’ eau. (French)

I drink ART. PAR water
‘I am drinking water.’
390 Katia Paykin
5 Critiques and some alternative answers

5.1 From Padučeva back to Jakobson
If we examine Padučeva’s (1998) main claims closely, it turns out that the major-
ity of cases analyzed in the article as examples of the use of the non-progressive
imperfective aspect authorizing quantifier phrases display strong incompatibility
with the partitive genitive.
(21) a. *On delaet šagov k stene.

3 SG . NOM makes:IPFV steps:GEN to wall:DAT
b. *On sažal derev’jev v étom sadu.

3 SG . NOM planted:IPFV trees:GEN in this:PREP garden:PREP
Therefore, it does seem that the partitive genitive NP does not have the exact
same behavior as ordinary QPs and that it is not always compatible with the
non-progressive imperfective aspect.
Moreover, some contexts authorizing QPs are not compatible with the parti-
tive genitive at all, including those cases where the verb is used in the perfective
aspect, as in (22).
(22) *On sdelal šagov k stene.

3 SG . NOM made:PFV steps:GEN to wall:DAT
Furthermore, there are contexts that do authorize the use of the partitive
genitive with the progressive imperfective aspect, as in example (23), where the
question ‘What are you doing?’ forces the progressive ‘here-now’ reading.
(23) Čto éto ty delaeš’?

what this you do:IPFV
‘What are you doing?’
a. V bajan kipjačënoj vody nalivaju. (Google)

into accordion boiled:GEN water:GEN pour:IPFV
‘I am pouring boiled water into the accordion.’
b. Moloka nalivaju v bljudce. (Google)

milk:GEN pour:IPFV into saucer
‘I am pouring milk into the saucer.’
Yet Padučeva’s explanation in terms of referentiality seems rather odd. Indeed, it

appears strange to consider that examples under (23) and examples under (24)
differ according to the referentiality status of the partitive.
(24) a. *Ja varju boršča. (Padučeva 1998: 80)

1 SG . NOM cook:IPFV borsch:GEN
b. *On nesët drov.

3 SG . NOM carries:IPFV firewood:GEN
If the non-referentiality implies a generic use, where the noun denotes a class
or a notion, and does not have a particular referent in the extra-linguistic reality,
it should rather apply to the use of the accusative. In other words, to refer to a
situation that can be roughly described as ‘I am engaged in the activity “cooking
borsch”’, Russian would rather resort to the accusative case, capable of having
an indefinite reading emphasizing a class, as in (25).
(25) Ja varju boršč.

3 SG . NOM cook:IPFV borsch:ACC
‘I am cooking (the) borsch.’
As we have mentioned earlier, the use of the accusative case is in fact ambiguous
between two readings: one referring to a class of dishes and another one to a
particular substance in a specific pot. The impossibility of using the partitive
genitive in (25), therefore, comes neither from its “inherent indefiniteness” nor
its non-referentiality.
To summarize, Jakobson’s (1936) analysis should be nuanced but not exactly
as Padučeva (1998) suggests. Russian partitive genitive NPs do not always com-
pare to QPs, they are not perfectly compatible with the non-progressive readings
of the imperfective aspect and they are not entirely averse to the progressive
imperfective aspect. Not all verbs take the partitive genitive to begin with,
and those that do can authorize it in their imperfective version under certain
circumstances.
5.2 Partitive taking verbs

According to Klenin (1978), verbs that govern the partitive genitive are in general
“semantically restricted to those that affect or effect a change in the status of
their objects, rather than expressing the attitude of the speaker”. Without pro-
viding any detailed explanations, the author also counts the verb xotet’ ‘to
392 Katia Paykin
want’ and its synonyms among the verbs assigning the partitive, although it
does not entail any change in its object.
In our view, the definition in terms of “affecting or effecting” does not seem
operational, as certain verbs compatible with QPs and affecting the referent of
their object exclude the use of the partitive, as in (26).8
(26) a. Maša otkryla pjat’ banok.

Masha:NOM opened:PFV five cans:GEN
‘Masha opened five cans.’
a0 . *Maša otkryla banok.

Masha:NOM opened:PFV cans:GEN
b. Maša pokrasila pjat’ sten.

Masha:NOM painted:PFV five walls:GEN
‘Masha painted five walls.’
b0 . *#Maša pokrasila sten.9

Masha:NOM painted:PFV walls:GEN
As far as the verb xotet’ ‘to want’ and its synonyms are concerned, we con-
sider that they should rather be analyzed as assigning the intensional genitive,
which has its own functioning10, and not the partitive genitive.
We argue that, in order to be compatible with the partitive genitive, a verb
has to be able to envisage the referent of its internal argument as a delimited
whole emphasizing the indefinite character of the quantity. In the case of what
can be called a bare plural, for example, the process denoted by the verb cannot
apply distributively to each element of the set. Therefore, the combination of the
verb otkryt’ ‘to open’ with the partitive is impossible, as can be seen from ((26)
8 In order to define which verbs do take the partitive genitive, we will use exclusively the
perfective aspect.
9 The prefixed verb po-krasit’ can have two different readings depending on the interpretation
of the prefix po-:
(i) it can have a perfective reading, insisting on the limit reached, equivalent to ‘I was painting
a / the wall’ – ‘I have painted a / the wall completely’;
(ii) it can have a perfective reading resulting from the bounding of an activity ‘to paint’ and
insisting on the fact that the activity in question lasted some time and stopped, the prefix
po- functioning like the prefix in the case of such verbs, as po-kurit’ ‘to smoke a bit’,
po-guljat’ ‘to walk a bit’, etc.
It is only the second meaning that is compatible with the partitive genitive.
10 For more details on the question, see Van Peteghem & Paykin (2013).
a0 ), because the action of opening necessarily applies to each can separately.

The verb naotkryvat’ ‘to open a large quantity’, however, is perfectly compatible
with the partitive genitive, as in example (27), since the denoted action concerns
an indefinitely large quantity, a set of cans, viewed as a whole.
(27) Maša naotkryvala banok.

Masha:NOM opened.a.lot.of:PFV cans:GEN
‘Masha opened a large quantity of cans.’
In fact, the majority of verbs compatible with the partitive genitive are pre-
fixed and hence perfective. Some of them, explicitly insisting on quantity, such as
na-pit’sja11 ‘to drink, to quench one’s thirst’ or na-rvat’ ‘pick (a great quantity)’,
take exclusively the partitive, rejecting the accusative.
(28) a. Ona napilas’ vody / *vodu.

3 SG . F. NOM drank:PFV water:GEN / *water:ACC
‘She quenched her thirst with water.’
b. On narval cvetov / *cvety.

3 SG . M . NOM picked.a.lot.of:PFV flowers:GEN / *flowers:ACC
‘He picked a great quantity of flowers.’
On the contrary, some prefixed verbs, such as do-est’ ‘to eat up’, for example,
reject the genitive due to the fact that the prefix used (do-) yields a holistic
11 Verbs with the affix –sja (cf. naest’sja ‘eat one’s fill’, napit’sja ‘quench one’s thirst’, etc.),
always assign the genitive case and can never combine with the accusative. When combined
with NPs containing an explicit quantifier, as in (i), these verbs impose the use of the instru-
mental case. Thus, these verbs function as intransitive verbs of saturation and the NP in the
instrumental simply denotes the quantity necessary to achieve the final saturation state.
(i) Ona naelas’ pjat’ju grušami / *pjat’ gruš.
3 SG . F. NOM ate.her.fill five:INS pears:INS / *five:ACC pears:GEN
‘She ate her fill [by eating] five pears.’
The use of the instrumental is also possible when the explicit quantifier is absent, as in (ii).
(ii) Ona naelas’ grušami.
3 SG . F. NOM ate.her.fill pears:INS
‘She ate her fill [by eating] pears.’
We can conclude, therefore, that the object marked by the partitive genitive insists on an in-
definite quantity absorbed viewed as a closed whole. However, the final saturation state is
not necessarily achieved. In other words, someone can eat sufficient amount of pears without
being able to continue consuming any more of this fruit, and yet, not be full or saturated in
general, result exclusively available through the instrumental case.
394 Katia Paykin
meaning, compatible with definite reference only, completely excluded for the
partitive genitive.
(29) Ona doela pirog / *piroga.

3 SG . F. NOM ate.up:PFV pie:ACC / pie:GEN
‘She ate up the pie.’
5.3 Different ways of delimitation

As noted above, the partitive genitive arguments denote a delimited whole
emphasizing the indefinite character of the quantity. Moreover, this delimited
whole gives no access to its components. Therefore, it cannot refer to intrinsi-
cally delimited entities. The delimitation has to be provided by the context.
Indeed, unlike the accusative or nominative marked arguments, the partitive
marked arguments cannot be immediately anaphorized by a personal subject
pronoun.
(30) a. ?*Maša kupila chlebai . Oni byl sliškom čerstvyj.12

Masha:NOM bought:PFV bread:GEN 3 SG . NOM was too stale
b. Maša kupila chlebi . Oni byl sliškom čerstvyj.

Masha:NOM bought:PFV bread:ACC 3 SG . NOM was too stale
‘Masha bought bread. It was too stale.’
(31) a. ?*Nočju navalilo snegai . Oni bystro rasstajal.

night fell:N . SG . PFV snow(M ): GEN 3 SG . NOM quickly melted
b. Nočju navalil snegi . Oni bystro rasstajal.

night fell:M . SG . PFV snow(M ): NOM 3 SG . NOM quickly melted
‘During the night, snow piled up. It melted quickly.’
The necessary delimitation or bounding comes most naturally from the pro-
cess, which explains the predilection of the partitive for the perfective aspect.
However, the delimitation can come from other elements of the context, such
as prepositional locative phrases v bajan ‘into the accordion’ or v bljudce ‘into
the saucer’, as found in (23), or a beneficiary dative, as mentioned in Padučeva
(1998: 81) and given here under (32).
12 The (non)acceptability of these examples can depend on the idiolect of the speaker.
(32) Nesu žene jablok.

carry:IPFV wife:DAT apples:GEN
‘I am carrying apples for my wife.’
It can even be provided by a special set-up of a spatio-temporal situation, suffi-

cient for such a delimitation, as in (33).
(33) Prichožu v magazin, pokupaju chleba i vdrug menja

come:IPFV in store buy:IPFV bread:GEN and suddenly 1 SG . ACC
kto-to za plečo trogaet. (Google).
somebody on shoulder touches:IPFV
‘I come to the store, buy some bread, and then all of a sudden somebody
touches me on the shoulder.’
Thus, the compatibility of the partitive genitive with the imperfective aspect
does not seem to come, as Padučeva (1998) claims, from the fact that the context
cancels the non-referentiality of the partitive, but from the possibility to obtain
the necessary delimitation imposed on the referent, which cannot be provided
by the unbounded imperfective aspect alone. We can therefore nuance Klenin’s
(1978) remark that the use of the partitive requires a double existential com-
mitment. According to Klenin (1978), in sentences, such as in (34), using the
genitive means not only that there was some tea that Sasha drank. “In addition,
there must also have been a larger supply of tea out of which Sasha’s tea was
taken, and the remainder of which he is assumed not to have drunk”.
(34) Saša vypil čaju.

Sasha:NOM drank:PFV tea:GEN
‘Sasha drank (some) tea.’
In our view, it seems extremely difficult to postulate this double existence com-
mitment for all uses of the partitive unless we postulate the double existence in
terms of the species’ existence and the existence of a certain quantity taken out
of it, delimited either by the aspect of the verb or by some other elements in the
context.
6 Conclusions
Let us now summarize our main conclusions:
396 Katia Paykin
– When the partitive genitive alternates with other cases, the genitive typically
gives rise to an indefinite reading, whereas the accusative or the nominative
can be interpreted either as definite or indefinite. The indefiniteness expressed
by the genitive, on the one hand, and the accusative or the nominative, on
the other hand, is semantically different, the genitive NP emphasizing quan-
tity, the accusative or the nominative NP denoting a class.
– Aspectual marking in Russian is explicitly present on the verb, independent
from the tense marking, and the verb (im)perfectivity is always expressed
independently from the case of the object (or the subject). Therefore, the
switch between the partitive and the accusative or the nominative has no
impact on the aspectual interpretation of the process. Moreover, the use of
the accusative and the nominative can both yield an indefinite interpre-
tation, thus refusing the exclusivity of the indefiniteness to the use of the
Russian partitive.
– Considering that the partitive marked argument denotes a certain indefinite
quantity presented as an indivisible whole, the delimitation or bounding of
the referent is mandatory. It can come from the process, which explains the
predilection of the partitive for the perfective aspect, or from other elements
in the context, thus making the imperfective compatible with the partitive,
including in its progressive reading.
References
Babby, Leonard H. 1980. Existential sentences and negation in Russian. Ann Arbor: Karoma
Publishers.
Bosveld-de Smet, Léonie. 2000. Les syntagmes en des et du: Un couple curieux parmi les
indéfinis. In Léonie Bosveld-de Smet, Marleen Van Peteghem & Danièle Van de Velde
(eds), De l’indétermination à la qualification: Les indéfinis, 17–116. Artois Presses Univer-
sité.
Brown, Sue & Steven Franks. 1995. Asymmetries in the scope of Russian negation. Journal of
Slavic Linguistics 3(2). 239–287.
Franks, Steven. 1995. Parameters of Slavic morphosyntax. New York: Oxford University Press.
Glovinskaja, Marina Ja. 1982. Semantičeskije tipy vidovych protivopostavlenij russkogo glagola
[Semantic types of Russian verbal aspect oppositions]. Moskva: Nauka.
Jakobson, Roman. 1962 [1936]. Beitrag zur allgemeinen Kasuslehre: Gesamtbedeutungen der
Russischen Kasus. In Selected Writings. Vol. 2, 23–71. The Hague: Mouton.
Kiparsky, Paul. 1998. Partitive case and aspect. In Miriam Butt & Wilhelm Geuder (eds.), The
Projection of Arguments: Lexical and Compositional Factors, 265–307. Stanford: CSLI
Publications.
Klenin, Emily. 1978. Quantification, partitivity, and the genitive of negation in Russian. In Bernard
Comrie (ed.), Classification of Grammatical categories, 163–182. Urbana & Edmonton:
Linguistic Research Inc.
Kuroda, Sige-Yuki. 1979. The ’w’hole of the doughnut: Syntax and its boundaries. Ghent:
E. Story-Scientia.
Marty, Anton. 1918. Gesammelte schriften. Halle: Abteilung.
Neidle, Carol. 1988. The role of case in Russian syntax. Dordrecht: Kluwer.
Padučeva, Elena. 1998. On non-compatibility of partitive and imperfective in Russian. Theoretical
Linguistics 24(1). 73–82.
Paus, Charles. 1994. Social and pragmatic conditioning in the demise of the Russian partitive
Case. Russian Linguistics 18(2). 249–266.
Paykin, Katia & Marleen Van Peteghem. 2002. Definiteness in a language without articles: A
case-study of Russian. Recherches linguistiques de Vincennes 31. 97–112.
Van Peteghem, Marleen & Katia Paykin. 2013. Genitive in Russian. In Anne Carlier & Jean-
Christophe Verstraete (eds.), Case and Grammatical Relations Across Languages: The
Genitive, 55–104. Amsterdam: Benjamins.
Vendler, Zeno. 1967. Linguistics in philosophy. Ithaca, NY: Cornell University Press.
Wierzbicka, Anna. 1967. On the semantics of the verbal aspect in Polish. In To honor Roman
Jakobson: Essays on the Occasion of His Seventieth Birthday, 2231–2249. The Hague:
Mouton.
Elżbieta Tabakowska
11 Double government in Polish: a case study
The paper deals with possible semantic and pragmatic motivation for case selec-
tion in alternate constructions with the accusative and the genitive in Polish. As
illustration, the Catholic priest’s invocation following the intercessions during
the Holy Mass is used, which the corpus shows to alternate between the two
cases. Corroborated for further Polish data, a more general claim is made that
the opposition between structures with transitive verbs followed by the direct
object in either the genitive or the accusative case reveals a significant difference
in meaning, which results from an intricate interplay of lexical semantics, aspec-
tual meaning of verbs, the meaning of verbal prefixes, pragmatic factors and
overall discourse structure.
Keywords: double government, case selection, motivation, aspectual meaning,

verbal prefixes
1 Preliminaries
The inspiration for this paper came from personal experience, shared by some
native speakers of Polish, as certified by the following extract from a letter ad-
dressed to an Internet counseling service on “correct Polish” (http://poradnia.
pwn.pl): “In church during Holy Mass I often hear the phrase wysłuchaj nasze
prośby [lit. hear our-ACC requests-ACC: ‘Hear our prayers’]. It jars, as it seems to
me that the correct expression calls for the genitive: wysłuchaj naszych próśb [lit.
hear our-GEN requests-GEN: ‘Hear our prayers’]”1. The linguistically naïve client
obviously gave no justification of their judgement. Looking for other similar con-
texts, I soon discovered that sometimes the accusative “jars” the users of the
service, while at other occasions it does not. The following discussion aims at
finding a possible explanation of those feelings. Consequently, I intend to focus
on a specific case of what is formally called dwojaki rząd (‘double government’),
i.e. alternative constructions with transitive verbs followed by direct objects
in either the genitive or the accusative case. An analysis of just two alternate
constructions employing a single verb does not allow generalizations. However,
it might provide a theoretically interesting illustration of a more general phe-
nomenon: semantic and pragmatic motivation for the use of Polish genetivus
partitivus.
1 All Polish sources are quoted in my translation – E.T.

400 Elżbieta Tabakowska
2 Facts
In contemporary Polish direct objects, used in structures with transitive verbs,
are typically coded as the accusative or the genitive form of the noun (or the
nominal), with the former construction being what Polish linguists traditionally
describe as “basic” (Jodłowski 1976: 98, Saloni, Świdziński 1985: 145). The use of
the genitive with negation is a rule. Double government – alternate construc-
tions with the accusative or the genitive – is considered as “semantically justi-
fied” in typical uses of genetivus partitivus, that is, when the nominals refer to
“objects that can be divided, most often with verbs that justify or make probable
such division”. Semantic justification results in semantic differentiation between
genetivus partitivus on one hand and other uses of the case on the other, the
opposition which is “well known to traditional syntax” (Saloni, Świdziński
1985: 146). However, it is possible to find such occurrences of double govern-
ment in which the two structures seem to be used indiscriminately, that is, to
appear as free variants. Simple Internet searching machines easily supply the
interested searcher with ad hoc corpora. But in grammatical descriptions, this
type of opposition is usually either ignored or taken for granted. The construction
presently under discussion, wysłuchać+ nominal-ACC/GEN, is a case in point.
Interestingly, it is not discussed in the earliest account of Polish verbs prefixed
by wy- (‘out-’), written almost forty years ago in the cognitivist vein (Rudzka-
Ostyn 1984). And it is precisely the framework of cognitive linguistics that seems
best suited to a description of structures whose assumed status of semantic free
variants may be questioned on pragmatic grounds.
Polish does not use either a separate case or a specialized case marking for
partitives; within the rich case system of contemporary Polish the “partitive
meaning” is considered as an extension of the prototypical meaning of the geni-
tive. In traditional grammar books, both descriptive and prescriptive – notably
textbooks for native and foreign speakers (cf. e.g. Szober 1962: 346, Kaleta
1995: 96) – the difference between the two structures in question is described
in terms of the “holistic” meaning of nouns in the accusative case as opposed
to the “part-of-the-whole” meaning of genetivus partitivus:
(1) a. dopić herbaty / herbatę

drink-up tea.GEN tea.ACC
‘to drink up one’s tea’
b. ustąpić miejsca / miejsce

give-up seat.GEN seat.ACC
‘to give up one’s seat’ (after Kaleta 1995: 96)
Double government in Polish: a case study 401
The motivation behind the genitive-accusative opposition in expressions like

(1b) above is unclear for linguists and native speakers alike (cf. the Internet
query quoted in Section 1. above). Explanations given both by naïve language
users and by language experts seem impressionistic, vague, and non-systematic.
Attempts at diachronic analyses are often plainly contradictory. For instance,
while the Internet counseling service which was quoted above claims that in
contemporary Polish one can observe a tendency towards generalization “with
the acusative taking over as the case of direct object, and the genitive being
gradually eliminated” (http://poradnia.pwn.pl; cf. also Kempf 2007: 98), at
some other place the very same service states that “today collocations with
the accusative seem obsolete” (http://poradnia.pwn.pl). An analysis of actual
data clearly points to the need of a more subtle analysis and a more detailed
description.
3 Hypothesis
In the present paper I wish to substantiate the claim that the opposition between
structures with transitive verbs followed by the direct object in either the genitive
or the accusative case reveals a significant difference in meaning, which results
from an intricate interplay of lexical semantics, aspectual meaning of verbs, the
meaning of verbal prefixes, pragmatic factors and discourse structure. An attempt
at such an integrated approach was actually made over forty years ago – in
what would be today legitimately called the cognitivist vein – by a Polish lin-
guist Zdzisław Kempf (2007: 97 ff.). Kempf stated that “varieties of the genitive
caused either by the scope of use or by the dependence of certain parts of
speech do not change the fact that, from logical and semantic points of view,
what lies at the core is always the same relation between part and whole”
(2007: 96.). He saw the crucial property of divisibility of referents as pertaining
to either mass (i.e. substance) or group (i.e. countable objects), but also to
actions which “cover only part of an object, without involving the remainder”
(2007: 97). Predicting subsequent claims of cognitive linguists, who assumed a
quarter of century later that in the process of human cognition spatial concepts
become extended to cover conceptualizations pertaining to time, he placed in
the latter category “actions that last only for a fragment of time” (2007: 98).
Finally, he realized that the overall meaning of structures with genetivus partitivus
depends on lexical meaning of verbs and their aspectual variants, notably the
meaning of perfectivising verbal prefixes.
In the taxonomy that Kempf proposed, the semantic area of Polish genitivus
partitivus is divided into six subcategories (2007: 98–105). The two classes that
seem most relevant for the present discussion are “the partitivus of a weakened
activity” (Kempf’s Type V) and “ the temporal partitivus” (Type VI), which con-
vey, respectively, the meaning of a low extent of involvement of the object in
the event described (cf. Moravcsik 1978) and the meaning of the event’s short
duration.
4 A case study: analysis

The following analysis will focus upon a single case study: the Catholic priest’s
invocation following the intercessions during the Holy Mass, which, as was said
before, is found to alternate between:
(2) Panie, wysłuchaj nasze modlitwy.

Lord, out-hear our.ACC prayers.ACC
and:
(3) Panie, wysłuchaj naszych modlitw.

Lord, out-hear our.GEN prayers.GEN
The ad hoc corpus of data was arrived at by means of the Google searching
machine (accessed on several occasions), and thus the analysis that was sub-
sequently carried out must be defined as corpus-illustrated rather than corpus-
based (for principled distinction between these two types of analyses, see
Geeraerts 2006). It reveals that the difference in meaning between (2) and (3)
results from an interplay of the following factors:
– lexical semantics of the verb stem, semantics of preverbal prefix wy- and
lexical semantics of the resulting prefixed verb
– semantics of case
– discourse pragmatics.
4.1 Semantics of the verb wysłuchać

A standard dictionary of contemporary Polish gives the following basic meaning
of the verb słuchać: “consciously receive auditory signals by means of one’s
sense of hearing” (SWJP), which is an equivalent of the basic (context-free)

meaning of the English verb listen.2 In Old Polish, the verb słuchać appeared in
contexts that suggest a more specialized basic meaning: “to listen attentively to
somebody or something, to hear what someone is saying” (Janowska 2007: 100;
emphasis mine). The diachronic development of its lexical semantics shows
further specification, among other forms, the combination of the verb stem with
the prefix wy-.
The first comprehensive semantic account of contemporary meanings of the
Polish verbal prefix wy- was made in 1984 by the prominent Polish linguist of
cognitivist persuasion, Brygida Rudzka-Ostyn. Rudzka-Ostyn proposed two basic
semantic schemas (1984: 228):
1. “to (re)move something out of a container” (in older, traditional accounts
this was defined as “taking out, an outward motion”, Janowska & Pastuchowa
2005: 148), as in (4):
(4) wypić wody/ę

out-drink water.ACC/GEN PAR.
‘drink [out of a container] the water/some [of the] water’
2. “to expand so as to completely fill a container”, as in (5):
(5) a. wypełnić pustkę

out-fill void.ACC
‘to fill up the void’
b. wymalować mieszkanie
out-paint flat.ACC
‘to paint [the entire] flat’
It is the first of these two schemas that has become “so salient and so
well established in the system of Polish that the meaning of derivatives is easily
recognized even in those derived forms whose stems had already disappeared”
(Janowska & Pastuchowa 2005: 150). In Janowska & Pastuchowa (2005) the
above claim is put forward in a comprehensive diachronic account of the prefix.
Written in the traditional descriptive style, the account makes a distinction
between aspectual (i.e. “purely aspectual” or “semantically empty”) and creative
2 The reason why English translation equivalents of expressions like (2) and (3) above require
the use of the lexeme hear (and not listen) in itself requires a discussion, which goes well
beyond the limits of this paper.
derivation, without, however, setting the demarcation line between the two
categories (2005: 150). Traditionally, Polish grammarians postulate the existence
of what they call pary aspektowe (‘aspectual pairs’), claiming that two verbs
making a pair are “semantically identical but different in aspect” (Kurzowa
1997:16, emphasis mine), as in (6):
(6) iść.IMPFV – wyjść.PFV

‘to go’ ‘to go out’ (after Kurzowa 1997: 16)
Further in this paper two claims will be made: first, that “purely aspectual”
or “semantically empty” derivations do not exist in language use and, second,
that even if the opposition between aspectual and creative derivations were to
be postulated, drawing a line that would separate the two would be impossible
by definition. Although in their analysis of the diachronic semantic development
of the prefixed verb in question Janowska & Pastuchowa admit the existence of
aspectual pairs (in their corpus they classify c. 90 lexical items as belonging
to this category; 2005: 150), they fail to account for, or to analyze, metaphorical
extensions of formations with wy-, which clearly indicate their semanto-pragmatic
differentiation, as in (7):
(7) minąć.IMPFV – wyminąć.PFV

‘to pass’ ‘to pass, to go by’
The authors classify wyminąć (in its Old Polish core meaning) as “tautological”
in respect of minąć. But they provide a context that points to a metaphorical
extension, whereby the “three years” are conceptualized as moving (metaphori-
cally) out of “time”, which is conceptualized – in agreement with the general
cognitive model – as an abstract schematic container:
(8) pierwej niźli trzy lata wyminą. . .

earlier than three years out-pass.
‘before three years go by. . .’ (2005: 150)
In contemporary Polish analogous examples can be easily found:
(9) a. wyjść z pokoju

out-go from room
‘to go out of the room’
b. wyjść z kłopotów
out-go from trouble
‘to free oneself from difficulties’
In (9) above wyjść means “to go out of a container”, literal in (9a) (the room)
and metaphorical in (9b) (difficulties).
In Old Polish, the prefixed verb wysłuchać had two basic meanings: 1. listen
attentively and, while listening, take note of what has been heard, and – inter-
estingly – 2. take note of something and grant somebody’s request. In the latter
sense, the agent was “God, Holy Virgin, or, exceptionally, a ruler” (Janowska
& Pastuchowa 2005: 162). Analogous contexts were attested for the iterative
form wysłuchiwać (the “secondary imperfective” – cf. Grzegorczykowa 1997, used
exclusively with reference to God) and for the (now obsolete) perfective non-
iterative forms wysłuchnąć, wysłuszać and wysłyszeć (= wysłuchać; Janowska
& Pastuchowa 2005: 162). Relevant specifications of the two meanings – given
in Janowska & Pastuchowa for Old Polish – are also attested in contemporary
language in the Google corpus:
(10) a. wysłuchać = słuchać czegoś do końca/w całości

hear something.GEN till end/ in entirety
‘to hear something till the end’
b. wysłuchać = spełnić czyjeś (usilne) prośby/błagania

fulfill someone’s (insistent) requests/begging.ACC
‘to answer someone’s (insistent) prayers’
As exemplified below, these two extensions are directly related to contextual

distribution of the perfective wysłuchać in contemporary Polish.
Defined within the framework of cognitive grammar, the structure wysłuchać
+ nominal.ACC (as in (10b) above) might be described in terms of Langacker’s
notion of boundedness (cf. Langacker 1991a, 1991b, passim), whereby the object
of a predication is conceptualized as a bounded region in physical or abstract
space. The definition is compatible with traditional descriptions, which present
the object profiled by the noun in the accusative case as being (or being con-
ceived of as) taken in its totality (i.e. as a “single indivisible whole”; cf. also the
notion of “totality” in Dickey 2000: 16). The structure conforms to Rudzka-Ostyn’s
second schema: the action of hearing expands till it completely fills the meta-
phorical container designated by the expressions requests or prayers. Thus it
is necessarily conceived of as non-partitive. The interpretation is corroborated
by the fact that core meaning of the perfective verb spełnić, used as standard
intralingual translation of wysłuchać followed by the nominal in the accusative

(cf. [10b]), is to ‘fill up/in’.
On the other hand, the construction wysłuchać + nominal.GEN (as in [10a])
exemplifies the semantic feature described as “temporal definiteness”: it expresses
“a totality which is uniquely located relative to contiguous, qualitatively different
states of affairs” (Dickey 2000: 20; cf. example [12] below). However, the descrip-
tion alone does not explain why the verb should be followed by an object in
the genitive case.
4.2 The semantics of case

The following constructions with the verb wysłuchać were found in the Google
corpus for contexts other than that of the Holy Mass liturgy:
(11) wysłuchać + nominal 1.ACC + nominal 2.GEN

Dane nam było wysłuchać część kompozycji.
was to us given out-hear part.ACC composition.GEN
‘We had a chance to listen to a part of the composition.’
(12) wysłuchać + nominal.GEN

Jan wysłuchał koncertu i poszedł do domu.
Jan out-heard concert.GEN and went to home
‘Jan listened to the concert and went home.’ ”
(13) wysłuchać + nominal.ACC

Bóg wysłuchał nasze modlitwy i dał Janowi zdrowie.
God out-heard our.ACC prayers.ACC and gave Jan health
‘God heard our prayers and gave John health.’
In (11) the totality of the direct object (i.e. the entire part [of the composition])
is – predictably - referred to by the noun in the accusative. The grammatical
case of the second nominal is the so-called “selective genitive” (przypadek
selektywny, Saloni, Świdziński 1985: 167): it selects “composition” as complement
of “part”. In cognitive grammar, this particular use of the genitive is explained
as an instance of the reference point construction, whereby language users
“invoke the conception of one entity in order to establish ‘mental contact’ with
another” (Langacker 2009: 83). In other words, “part” is conceptualized via
reference to “composition”.
By contrast, the genitive in (12) can be interpreted an instance of genetivus

partitivus. Although it does not seem to translate readily into any of the two
relevant varieties of genitivus partitivus, that is, “the partitivus of a weakened
activity” or “the temporal partitivus”, an analysis reveals its close affinity to the
former category. Kempf defines this use of the genitive case as the construction
in which the object of an action is undivided (i.e. conceived of in its totality as a
“single indivisible whole”), but the action itself is weakened (Kempf 2007: 103).
According to Kempf, such an action can either involve only a part of an object or
embrace the whole object while at the same time implying a limited involvement
on the part of the agent. Such is indeed the case in (12): it contains the notion of
a completed action (the hearing “fills up” the concert), but fails to imply its
intensity. There is no indication that the agent “listened attentively and, while
listening, took note of what has been heard”. Full involvement would require
strengthening the overall meaning by explicit reference to the totality of the
object. The point is illustrated by the contrast between (14) and (15):
(14) posłuchać płyty

[around-hear CD.GEN]
‘to listen to a CD for some time’
(15) przesłuchać płytę

through-hear CD.ACC
‘listen attentively to a CD’.
In (14) the meaning of the prefix po- (“around a limited area, for a limited
time,”) precludes the agents’ full involvement and/or long duration of the
action, and thus – predictably – requires an object in the genitive (genetivus
partitivus of a weakened activity). By contrast, the prefix prze- (‘through’) in (15)
implies thoroughness, thus requiring an object in the accusative. Predictably
again, neither of the two verbs admits double government.
Finally, (13) exemplifies full involvement of the agent, whose action (God’s
having listened to prayers) led to a tangible final result (Jan’s health).
In this connection, several questions arise, which have to be answered
before any conclusions from the analysis can be drawn. First, as the material in
the corpus reveals, double government occurs with the verb wysłuchać, but not
with other verbs of sensuous perception, even though their stems can also be
preceded by the prefix wy-:
(16) a. wypatrzyć piękną broszkę na prezent

out-look lovely brooch.ACC/*GEN as a present
‘to find out a lovely brooch as a present’
d. wymacać zgrubienie
out-feel swelling.ACC/*GEN
‘to feel a swelling’
e. wyczuć pieprz
out-sniff pepper.ACC/*GEN
‘to sniff pepper’
f. wyczuć niechęć
out-feel reluctance.ACC/*GEN
‘to feel reluctance’, etc.
Why is double government possible with wysłuchać, but not with other verbs
of sensuous perception, which can only take objects in the accusative? One
plausible answer seems to be that, unlike the action of hearing, other acts of
sensuous perception tend to lead to tangible results, following from the agent’s
involvement combined with the “holistic” character of the objects: the brooch
will most probably be bought and given as present; the swelling will be probably
reported to the doctor; the peppery dish will be appreciated or rejected; the reluc-
tance will influence the agent’s behaviour, etc. To refer once again to Rudzka-
Ostyn’s schemas, objects will be taken, literally or metaphorically, out of their
present containers, literal or metaphorical, as the result of particular actions.
An analysis of further examples found in the corpus reveals yet another
interesting regularity. Namely, while in “church” contexts the constructions
with the accusative prevail, “lay” contexts show the tendency for the genitive.
This can be illustrated by the following data:
(17) wysłuchać.GEN:
argumentów [‘arguments”.GEN], głosu swojego dziecka [‘the voice of your
child. GEN’], reportażu [‘report.GEN’], reklamy [‘advertisement.GEN’],
raportu [‘journalist report.GEN’], mej pieśni [‘my.GEN song.GEN’], swego
serca [‘one’s.GEN heart.GEN’], albumu Imagine [‘the album by Imagine
group.GEN’], Krzysztofa Ibisza [Krzysztof.GEN Ibisz.GEN], prezentacji
[‘presentation.GEN’], informacji [‘information.GEN’], etc.
as opposed to:
(18) wysłuchać.ACC:
a. Pani, wysłuchaj modlitwy nasze.
Lady, out-hear prayers.ACC our.ACC
‘O, Lady, hear our prayers.’ (from Canonical hours)
b. Wysłuchaj pokorne modlitwy swojego Kościoła.

out-hear humble prayers.ACC your.GEN Church.GEN
‘Hear the humble prayers of your Church.’ (Litany to All Saints)
c. Wysłuchaj, wierny Boże prośby moje. . . Wczas

out-hear, faithful God, prayers.ACC mine.ACC In due time
mię. wysłuchaj
me.ACC out-hear
‘Hear, o faithful God, my prayers. . . In due time, listen to me.’
(church hymn)
d. Wysłuchaj Panie wszechmocny naszą wieczorną modlitwę.

out-hear Lord almighty, our.ACC evening.ACC prayer.ACC
‘Listen, O Lord almighty, to our evening prayers.’ (liturgy of hours)
e. Odkupicielu prawdziwy/wysłuchaj nasz głos płaczliwy.

Saviour true wy-hear our.ACC voice.ACC tearful.ACC
‘O, Saviour true, hear our tearful voice.’ (church hymn)
f. Racz wysłuchać me westchnienie.

Deign to wy-hear my.ACC sigh.ACC
‘Deign to hear my sigh.’ (church hymn), etc.,
The counseling service informs its clients that structures with the accusative,
although in general discourse felt to be obsolete, are rather frequent in religious
contexts, because “the language used in the Church may preserve old syntactic
schemata longer than other language varieties” (http://poradnia.pwn.pl). The
explanation reports a statistical fact, but it does not explain much. It is then
advisable to look for more facts.
First, besides structures exemplified by (18a)–(18f) above, constructions with
the genitive are also attested, albeit less frequently, in ‘religious’ contexts:
(19) wysłuchać.GEN:
a. Wysłuchaj prośby naszej.
Out-hear prayer.GEN our.GEN
‘Hear our prayer.’
b. Dzieciątko Boże, wysłuchaj naszych próśb.

Child God’s out-hear our.GEN requests.GEN]
‘O, Holy Child, hear our requests.’
On the other hand, one can find “lay” contexts with the direct object in the
accusative:
(20) wysłuchać.ACC
a. Wysłuchaj hejnał miasta Chełm.
out-hear bugle-call.ACC town.GEN Chełm
‘Listen to the bugle-call of the town of Chełm.’
b. [Dziecko] potrafi uważnie wysłuchać tekst.

[The child] is able attentively out-hear text.ACC
‘The child is able to listen to the text attentively.’
c. wysłuchać drugą stronę

out-hear second side.ACC
‘to hear the other party’
d. Sąd może wysłuchać organ założycielski.

Court can out-hear committee.ACC founding.ACC
‘The court can hear the founding Committee.’3
In this connection, two observations can be made.

First, (20a) and (20b) sound unnatural, with the former being more unnatural
than the latter. In (20a) there seems to be no reason for emphasizing the agent’s
involvement in a fairly standard routine activity. Due to its overall meaning, in
(20b) such involvement is more justified.
Second, examples (20c) and (20d) refer to legal activities, and as such
they exemplify contexts traditionally attested for the Old Polish wysłuchać.
As testified by the sources, the second of the two core meanings postulated
3 Note the occurrence of listen rather than hear in the English translations of (20) a.–b. The
difference – although the reasons seem rather obvious in view of the analysis presented in
the body of the present text – cannot be discussed at length at this point – cf. footnote 1
above.
for Old Polish, “take note of something and grant somebody’s request”, was
further specified as “to hear and take note of a court testimony” (Janowska &
Pastuchowa 2005: 162), predictably calling for the accusative government. The
tendency for conservatism in “preserving old syntactic schemata”, pointed out
by the Internet counseling service, can thus be found not only in the language
of the Church but also in that of the Court. It might then seem that in “lay” and
“non-legal” contexts the construction with the accusative is less firmly entrenched
– or less readily chosen. General, non-specified contexts will obviously be more
numerous and therefore more strongly conventionalized. With the (statistical)
tendency to use constructions with the genitive, more specified, and less en-
trenched, “church” contexts may also be found “jarring”, as was the case with
the internet interpellation quoted in section 1. above. Yet, as was said before,
providing statistical data is not tantamount to providing explanations. As will be
argued in the next section, usage phenomena seem more convincingly explained
as being motivated by pragmatic factors.
4.3 Discourse pragmatics

It was seen that while the wysłuchać + Nominal-GEN construction implies parti-
tive interpretation in terms of low involvement of the agent in an event, the
wysłuchać + N-ACC evokes the “holistic” interpretation, involving full engage-
ment in an activity or/and its all-embracing character. It might then be legiti-
mately claimed that the choice of the direct object in the accusative, as in (21):
(21) wysłuchać szmery w płucach

out-hear murmurs.ACC in lungs
‘to hear pulmonic murmurs’
is justified by the expected involvement of the medical doctor, presupposed by

the activity and crucial in view of possible grave effects. For the same reason
(22a) sounds, so to say, more optimistic than (22a):
(22) a. Panie, wysłuchaj nasze modlitwy.

Lord, out-hear our.ACC prayers.ACC
‘O, Lord, hear our prayers.’
b. Panie, wysłuchaj naszych modlitw.

Lord, out-hear our-ACC prayers-ACC
‘O, Lord, hear our prayers.’
The speaker assumes that the prayers will be heard in their entirety, with
full engagement on the part of the one who is asked to listen. Moreover, the
“expansion to fill in a container” schema is a semantic prerequisite to the actual
effect, that is, the answering of the prayers. This, in turn, requires an assumption
of adequate prerogatives or competences thanks to which the intended effect
may be actually achieved. Naturally, such prerogatives belong to agencies that
have the power to grant people’s requests or to pass judgments – naturally
then, the agents are divine or legal, authorized to answer people’s prayers or/
and to decide about their fate.
The “schema of agent involvement” can be extended to general contexts.
The accusative is justified in expressions like (20b) or (21) by pragmatic salience
of the effects of actions taken up by the attentive child or the responsible medical
doctor.
5 Conclusions
An analysis of expressions like those listed under (17)–(20) reveals the existence
of some clear-cut pragmatic factors that restrict the use of either of the two types
of genetivus partitivus, thus pointing out discursive motivation for its usage in
opposition to structures with the accusative. As revealed by the case study dis-
cussed above, the distributional opposition between the accusative and the
genitive cannot be explained unless we consider – apart from lexical semantics
of the verb and the object – the meanings of the particular verbal prefix and of
the particular case (cf. also Tabakowska 2001: 218). It is the “conspiracy” between
lexical and grammatical meanings that ultimately accounts for the “linguistic”
meaning of an expression, which is then fitted into a pragmatic context as the
language user thinks fit.
In particular, the analysis confirms Rudzka-Ostyn’s postulate of the “bipartite”
schema for the verbal prefix wy- as well as Kempf’s reflection on the semantics
of Polish genetivus partitivus. The former is a reflection of early cognitivist
thought, the latter is its prediction. Both bring in, albeit indirectly, pragmatic
factors. The analysis given in this paper might perhaps contribute to their way
of thinking about language structure and language use.
As a case study on grammatical case, it points out the need to investigate
the integral role played by semantics, morphology and pragmatics of discourse
structure in the historical development of case systems. Discussions of mecha-
nisms where change can be observed would include verb morphology, semanti-
cally and aspectually motivated synchronic case variations, discourse motivated
object marking, extensions of case marker distribution, and subjectification (cf.

Barðdal & Chelliah 2009).
Possible implications for contrastive linguistics and language pedagogy
deserve separate studies. At this point, what must suffice is a current remark
on translation. The type of double government discussed above finds no direct
counterpart in languages with less developed case systems, for instance, in
English. The fact that English translation equivalents of most examples given
above, in which the accusative – genitive opposition brings in subtle but signif-
icant differences in meaning, sound alike heralds translational problems. How-
ever, their more detailed discussion has to be left for another occasion.
References
Barðdal, Jóhanna & Shobhana Chelliah (eds.). 2009. The Role of Semantic, Pragmatic and Dis-
course Factors in the Development of Case. Studies in Language Companion Series 108.
Dickey, Stephen M. 2000. Parameters of Slavic Aspect. A Cognitive Approach. Stanford: CSLI
Publications.
Geeraerts, Dirk. 2006. Methodology in Cognitive Linguistics. In Gitte Kristiansen, Michel Achard,
Rene Dirven & Jose Ruiz de Mendoza Ibanez (eds.), Cognitive Linguistics: Current Applica-
tions and Future perspectives, 21–49. Berlin: Mouton de Gryuter.
Grzegorczykowa, Renata. 1997. Nowe spojrzenie na kategorię aspektu w perspektywie semantyki
kognitywnej [A new view upon the category of aspect from the perspective of cognitive
semantics]. In Renata Grzegorczykowa & Zofia Zaron, Semantyczna struktura słownictwa i
wypowiedzi [Semantic structure of lexicon and utterance], 25–38. Warszawa: Wydawnictwo
Uniwersytetu Warszawskiego.
Janowska, Aleksandra. 2007. Polisemia staropolskich czasowników. Źródła, swoistość, konsek-
wencje [The polysemy of Old Polish verbs. Sources, peculiarities, consequences]. Katowice:
Wydawnictwo Uniwersytetu Śląskiego.
Janowska, Aleksandra & Magdalena Pastuchowa. 2005. Słowotwórstwo czasowników staro-
polskich. Stan i tendencje rozwojowe [Word formation of Old Polish verbs. Status quo
and the directions of development]. Kraków: Universitas.
Jodłowski, Stanisław. 1976. Podstawy polskiej składni [Foundations of Polish syntax]. Warszawa:
Polskie Wydawnictwo Naukowe.
Kaleta, Zofia. 1995. Gramatyka języka polskiego dla cudzoziemców [Polish grammar for foreign
learners]. Kraków: Wydawnictwa Uniwersytetu Jagiellońskiego.
Kempf, Zdzisław. [1970] 2007. Próba teorii przypadków. Część II. [An attempt at the theory of
case]. Opole: Wydawnictwo Uniwersytetu Opolskiego.
Kurzowa, Zofia. 1997. Tackling Polish Verbs. Kraków: Baran & Suszczyński.
Langacker, Ronald W. 1991a. Foundations of Cognitive Grammar. Volume II: Descriptive Applica-
tion. Stanford: Stanford University Press.
Langacker, Ronald W. 1991b. Concept. Image and Symbol. Berlin: Mouton de Gruyter.
Langacker, Ronald. W. 2009. Cognitive Grammar. A Basic Introduction. Oxford: Oxford Univer-
sity Press.
Moravcsik, Edith. 1978. On the case marking of objects. In: Joseph Greenberg et al. (eds.),
Universals in human language. Vol. IV. Syntax, 249–290. Stanford: Stanford University
Press.
Rudzka-Ostyn, Brygida. 1984. Cognitive Grammar and word formation. The case of Dutch uit
and Polish wy. In Irena Kałuża, Irena Przemecka, Marta Gibińska & Elżbieta Tabakowska
(eds.). Litterae et lingua. In honorem Premislavi Mroczkowski, 227–240. Wrocław etc.
Ossolineum.
Saloni, Zygmunt & Marek Świdziński. 1985. Składnia współczesnego języka polskiego [The syn-
tax of modern Polish]. Warszawa: Polskie Wydawnictwo Naukowe.
Szober, Stanisław. 1962. Gramatyka języka polskiego [A grammar of Polish]. Warszawa: Polskie
Wydawnictwo Naukowe.
SWJP: Słownik Współczesnego Języka Polskiego [Dictionary of contemporary Polish].
Tabakowska, Elzbieta. 2001b. O motywacji związku rządu derywatów prefiksalnych polskich
czasowników z dopełnieniem. [On the motivation of government of prefixed derivatives of
Polish verbs over their complements] In Wojciech Kubiński & Danuta Stanulewicz (eds.)
Językoznawstwo kognitywne II. Zjawiska pragmatyczne [Cognitive Linguistics II. Pragmatic
phenomena], 212–224. Gdańsk: Wydawnictwo Uniwersytetu Gdańskiego.
V Historical perspectives on Indo-European
languages
Eystein Dahl
12 Partitive Subjects and Objects in
Indo-Iranian and beyond*
The genitive is occasionally used to express subjects and objects with finite
verbs in the Indo-Iranian languages, a phenomenon which is well known from
other Indo-European languages as well. In this paper I explore the semantic
domain of these constructions with particular regard to the lexical semantic
properties of the predicates selecting genitive-marked core arguments, showing
that while partitive subject constructions are relatively rare and appear to be
subject to strong lexical semantic restrictions in Indo-Iranian, partitive object
constructions are found with a broad range of predicate types, including ingestion
verbs, perception verbs, authority/possession verbs and change of state verbs.
Keywords: partitive genitive, subject, object, lexical restrictions, Indo-Iranian,

Vedic, Avestan, Indo-European
1 Introduction
This paper takes a fresh look at some interrelated readings associated with the
genitive in Indo-Iranian. Among the various functions traditionally ascribed to
this case category, its so-called possessive and partitive readings appear to be
most salient (cf. e.g. Delbrück 1888: 151–155, Reichelt 1909: 256–258). Both of
* This paper represents a revised version of my talk ‘Partitive Subjects and Objects in Indo-
Iranian’ presented at the Workshop on Partitives, SLE 43rd Annual Meeting, Vilnius University
Sunday 5 September 2010. I am grateful to the audience for inspiring and thought-provoking
comments after my presentation, in particular Silvia Luraghi and Michela Cennamo. Moreover,
I would like to thank Silvia Luraghi, Maria Carmela Benvenuto and an anonymous reviewer for
commenting upon one or more previous draft versions of the paper, thereby having contributed
significantly to its final form. Evidently, I alone assume the responsibility for all remaining
weaknesses. During the time this paper was written, I was a guest researcher in Paris from
October to December 2011 where the Fondation Colette Caillat of the Institut de France generously
provided me with free housing and at the Dipartimento di Scienze Documentarie, Linguistico-
Filologiche e Geografiche, Università della Sapienza, Roma from January to June 2012. I hereby
wish to express my sincere gratitude to the Fondation Colette Caillat as well as to Professor
Georges-Jean Pinault, EPHE and Professor Paolo di Giovine, La Sapienza for their kind hospitality
during my stays in Paris and Rome.
418 Eystein Dahl
these relations can be argued to have a “nominal” character in the sense that
they appear most clearly when a noun qualifies another noun. However, the
genitive is also frequently used as an object marker and, somewhat less fre-
quently, as a subject marker in the Indo-Iranian languages. In such cases the
question arises whether these latter uses of the genitive can be derived from its
possessive reading, from its partitive reading, from neither or from both. Here I
argue that the use of the genitive as an argument marker in Indo-Iranian is
closely connected with its partitive reading. The present paper has a diachronic-
comparative scope, seeking to establish a basic range of argument realization
constructions which may be reconstructed for the Indo-Iranian partitive genitive,
partly by examining the attested constructions in Vedic and Avestan and partly
by a comparison with Homeric Greek. The paper is organized as follows. Section
2 discusses a few basic philological and theoretical issues. Section 3 discusses
the use of the Indo-Iranian genitive in argument realization, in Section 3.1 I
examine its use as an object marker and Section 3.2 deals with the use of the
genitive in subject position. Section 4 contains a brief summary and conclusion.
2 Philological and theoretical preliminaries

Before proceeding to the main topic of this paper, a clarification of some basic
philological issues and theoretical assumptions is in order. In this paper, “Vedic”
means the language of the Vedas, the sacred texts of Hinduism and the oldest
attested stage of Old Indo-Aryan, “Avestan” means the language of the Avesta,
the oldest attested Iranian collection of texts and “Old Persian” means the lan-
guage of the inscriptions of the Achaemenids. “Homeric Greek” is the language
of the Homeric Epics. “Indo-Iranian” or “Proto-Indo-Iranian” is understood as
a hypothetical language reconstructed on the basis of Vedic, Avestan and Old
Persian, whereas “Proto-Indo-European” refers to the hypothetical ancestral
proto-language common to Indo-Iranian, Homeric Greek and other related lan-
guage families such as Italic, Germanic, Baltic, Slavonic, Celtic etc.
This paper is based on the assumption that the various readings associated
with a morphological category as a rule can be accounted for as lexically and
contextually determined variants of a basic general meaning. Under this approach,
the traditionally acknowledged distinction between possessive and partitive
genitive may ultimately be regarded as two interrelated types of readings arising
from a more general meaning associated with the Indo-Iranian genitive. At
present, I shall adopt a version of what may be labeled a “localistic” or perhaps
a “concretistic” approach to case semantics, assuming that, as far as Indo-
Iranian goes, all of the case categories except the nominative, vocative and
Partitive Subjects and Objects in Indo-Iranian and beyond 419
possibly the accusative may be ascribed a local or concrete meaning which

underlies the various readings associated with each of them. This generalization
is based on the observation that several case categories, e.g. the ablative, loca-
tive or instrumental are associated with a concrete, localistic meaning which is
characteristically apparent when they are used as adverbial modifiers and which
is transparent in their other uses, for instance when they appear in argument
realization (cf. e.g. Dahl 2009a: 36).1
On the basis of the Vedic and Avestan evidence we may reconstruct the
inventory of eight morphological case categories for Indo-Iranian. Old Persian
has lost the dative at the beginning of its tradition. This is an early manifest
result of a more general drift towards the substitution of the dative by the geni-
tive in the Indo-Iranian daughter languages. Note that the dative is also on the
verge of disappearing in Young Avestan and in later stages of Old Indo-Aryan
and is almost completely obsolete in Middle Indo-Aryan. Table 1 summarizes
the most important readings associated with the Indo-Iranian case categories.
Note that this table does not aim at a comprehensive account of the functions
of the various case categories.
Table 1: The Indo-Iranian case categories and their canonical functions
Case Categories Canonical Functions

Nominative Grammatical Subject
Vocative Addressee
Accusative Grammatical Object, Patient, Goal
Dative Benefactive, Recipient, Goal, Internal/External Possessor
Genitive2 Partitive, Possessor
Instrumental Instrument, Comitative, Cause, Perlative
Locative3 Location, Goal
Ablative Cause, Origin
1 A reviewer points out to me that a similar claim could be made with regard to the Indo-
Iranian accusative, in particular the (adverbial) accusative of direction and the temporal accu-
sative vs. the accusative as an object marker. Case categories like the nominative and vocative,
on the other hand, are characteristically used to express prominent discourse participants and
are not associated with any concrete or local meaning. Hence, a localistic approach to case
meaning does not have universal validity in Indo-Iranian.
2 A reviewer points out that the Indo-Iranian genitive is used as a ‘structural case’ with nominal
derivations and hence may encode a rather broad variety of thematic roles. However, at present
I prefer to regard such readings of the genitive as contextually determined variants of its
semantically more general possessive or partitive readings.
3 A reviewer notes that the Indo-Iranian locative occasionally has a partitive-like reading. How-
ever, most examples of this reading in Vedic appear to represent special cases of the basic
function of the locative to express location (among).
420 Eystein Dahl
Another point which is important in the following discussion concerns the

factors determining the range of argument realization options for a given verb
in a given language. The notion of inherent transitivity plays an important role
in this respect. There is a growing consensus that transitivity represents a scalar
notion and that verbal predicates may be classified hierarchically according to
their relative transitivity on the basis of a restricted number of semantic proper-
ties or lexical entailments (cf. e.g. Hopper & Thompson 1980, Tsunoda 1985,
Levin & Rappaport Hovav 1995 and elsewhere, Grimm 2005, Beavers 2006,
Dahl 2009a). Simplifying matters somewhat, I shall assume in the following
that at least three lexical semantic properties are relevant for the classification
of verbs and verb classes, namely whether the verb denotes a (total) change of
state or not [+BECOME], whether it entails that one or more of the core argu-
ments are animate [+SENTIENCE] and, if so, whether it entails that the animate
argument(s) control(s) the outcome of the situation [+VOLITION]. I take these
properties to be privative, so that a verb is either specified with regard to or
underspecified and thus in principle compatible with a given feature. Verbs that
are associated with all of these entailments leave less room for morphosyntactic
variation than verbs lacking some or all of them and tend to be associated with
canonical argument realization patterns across languages (cf. also Levin 1999
for discussion).
The notion of change of state needs some further refinement. In a recent
study, Beavers (2011) develops a relational model of change involving two partici-
pants, a theme participant undergoing a change of state and a scale participant
measuring the change. Furthermore, he distinguishes four predicate types which
form the basis of a principled model of relative affectedness or affectedness
scale. In Beavers’ hierarchy, the lexical entailment that the theme argument
undergoes a quantized change (e.g. accomplishments/achievements such as
break, shatter, destroy or devour) constitutes the highest degree of affectedness,
followed by the weaker entailment that the theme undergoes a non-quantized
change (e.g. degree achievements like widen or cool or verbs of cutting like cut
or slice), which in turn is followed by the entailment that the theme has poten-
tial for change (e.g. verbs of surface contact/impact like wipe, scrub, rub, punch,
hit, kick, slap), while the lowest degree of affectedness is constituted by the lack
of an entailment of change (e.g. activities/states like see, laugh at, smell, follow,
ponder or ogle) (cf. Beavers 2011: 358–359). In Beavers’ words, ‘the more specific
a predicate is about the theme’s progress on a scale, the higher the degree of
affectedness’ (2011: 357). Under this approach, the notion of telicity may be
elegantly defined as ‘a property a predicate has when all of the theme crosses
all of the path’ (Beavers 2011: 356). In the following pages, the distribution of
the partitive genitive over different types of predicate classes in Indo-Iranian
will be scrutinized.
3 The use of the Indo-Iranian genitive in

argument realization
The genitive is frequently used as an object marker with finite forms of two-
place verbs and considerably less commonly used as a subject marker with
finite forms of one-place verbs in Vedic as well as Avestan. In the following,
I review the semantic domain of these constructions with particular regard to
the lexical semantic properties of the predicates selecting genitive-marked core
arguments.
It should be noted from the outset that most of the verbs that allow genitive
case marking of their core arguments are also attested with canonical case mark-
ing patterns, i.e. nominative subject case marking or accusative object case
marking. Consider the following examples from Vedic and Avestan.
(1) a. pácanti te vṛṣabhā́ m̐ átsi téṣām

cook:PRS .3 PL 2SG . DAT bulls:ACC eat:PRS .3 SG 3PL . M .GEN
‘They cook bulls for you, you eat (some) of them.’
(Vedic, Rigveda X 28.3)
b. svadhā́ m pīpāya subhú ánnam atti /

wantonly swell:PRF.3 SG strong:ACC food:ACC eat:PRS .3 SG
‘He has swollen wantonly. He eats excellent food.’
(Vedic, Rigveda II 35.7)
c. ho pərəQβe aiŋ́hā̊ zəmo

3SG . M . NOM surface:LOC 3 SG . F.GEN earth:GEN
upa.Qβaršti uruuaranąm
cut.off:PRS .3 SG plants:GEN
‘He cuts down plants on the surface of the earth.’
(Avestan, Vendīdād 9.2)
d. daṣ̌inəm he gaoṣ̌əm upa.Qβərəsaiiən

right:ACC 3SG . M .GEN ear:GEN cut.off:PRS .OPT.3 PL
‘They shall cut of his right ear.’ (Avestan, Vendīdād 13.32)
e. ákāri vām ándhaso

make:AOR . PASS .3 SG 2 D. DAT soma.juice:GEN
‘Some soma-juice has been prepared for you two.’
(Vedic, Rigveda VI 63.3)
422 Eystein Dahl
f. eṣá vām̐ stómo aśvināv akāri

3SG . M . NOM 2 D. DAT hymn:NOM Aśvins:VOC make:AOR . PASS .3 SG
‘This hymn has been made for you two, O Aśvins.’
(Vedic, Rigveda I 184.5a)
g. uruuaranąm zairi.gaonanąm zaramaēm paiti zəmāδa

uzuxšiieiti
(Avestan, Yašt 7.4)
h. zəmāδa uzuxṣ̌iieiṇti uruuarā̊

earth:ABL grow:PRS .3 SG plants:NOM
‘Across the earth plants are growing.’ (Avestan, Yašt 13.14)
These examples illustrate that the core arguments of certain verbs may be alter-
nately expressed by the genitive or by the (canonical) accusative or nominative
in Vedic and Avestan. Note that minimal pairs like that cited in (1a) through (1d)
are by far more common than the type illustrated in (1e) through (1h). This is due
to the fact that the use of the genitive as a subject marker is extremely rare
in Vedic and Avestan. In the following, we shall take a closer look at the use
of the genitive in argument realization in Indo-Iranian. Section 3.1 deals with
genitive-marked object arguments, whereas Section 3.2 focuses on genitive-
marked subject arguments.
3.1 The genitive as an object marker in Indo-Iranian

This section examines a number of verb types that select an object argument
in the genitive, in order to establish whether they have any semantic properties
in common that may contribute to explaining the distribution of the genitive in
this position. As we shall see in the following, the Indo-Iranian genitive is
attested as an object marker with ingestion verbs, perception/comprehension
verbs, authority/possession verbs, one verb of praising as well as certain change
of state verbs.
Ingestion verbs constitute the perhaps most well-known group of verbs
occasionally taking a genitive object in the Indo-Iranian languages. The following
examples illustrate that verbs like Vedic pā- ‘drink’ and Avestan fra-xvar- ‘con-
sume’ allow their object argument to be alternately expressed by the accusative
or the genitive (cf. also Vedic ad- ‘eat’ in (1a) and (1b) above).
(2) a. ápāḥ sómam. ástam indra prá yāhi

drink:AOR .2 SG soma:ACC home:ACC Indra:VOC PREV go:IMP.2 SG
‘You have drunk the soma. Go home, Indra!’ (Vedic, Rigveda III 53.6)
b. arvā́ ṅ éhi sómakāmaṃ tvā āhur ayáṃ

hither come:IMP soma.lover:ACC 2SG . ACC say:PRF.3 PL 3 SG . M . NOM
sutás tásya pibā mádāya
juice:NOM 3SG . M .GEN drink:IMP.2 SG inebriation:DAT
‘Come hither! They say you are a lover of soma. This is the juice.
Drink of it to inebriation!’ (Vedic, Rigveda I 104.9)
c. nā aṣ̌auua yaoždātąm zaoQrąm fraŋuharāt ̰

man:NOM having.truth:NOM purified:ACC libation:ACC consume:SBJV.3 SG
‘Truth-owning man shall drink a purified libation!’ (Avestan,
Yašt 10.120 after Gershevitch 1959: 133)
d. mā.ciš mē ā̊ŋhąm zaoQranąm fraŋuharāt ̰

no.one:NOM 1SG .GEN 3 PL . N .GEN libations:GEN consume:PRS . SBJV.3 SG
‘Let no one drink of these libations of mine!’ (Avestan, Yašt 10.122, cf.
also Gershevitch 1959: 135)
Somewhat surprisingly, there appear to be no clear examples of cognate verbs

attested in Vedic as well as Avestan within this particular semantic field. Cheung
(2007: 148) suggests that the Iranian verb *hṷar1- ‘eat’ originally was identical
to the root *hṷar3- ‘take’ and that it replaced the inherited Indo-Iranian verb
*Had- ‘eat’, reflected in Vedic ad- ‘eat’. Significantly, a similar point could be
made with regard to the inherited verb *paH(i ̯)- ‘drink’, reflected in Vedic pā-
‘drink’, which is only attested in nominal forms in Avestan. The fact that we
find exactly the same alternation pattern with ingestion verbs in Homeric Greek
(cf. also Conti & Luraghi this volume), as illustrated by the examples in (3),
suggests that this alternating construction is inherited from Proto-Indo-European
and that it was secondarily assigned to the Iranian verb fra-x var- ‘consume’.
(3) a. haímatos ófra píō kaí toi

blood:GEN so.that drink:AOR . SBJV.1 SG and 2 SG . DAT
nēmertéa eípō.’ (. . .)
truth:ACC say:AOR . SBJV.1 SG
‘So that I may drink blood and speak the truth to you.’ (Homeric
Greek, Hom. Od. 11.95–96 after Napoli 2006: 113)
424 Eystein Dahl
b. ho d’ epeì píen haîma kelainón

3 SG . M . NOM PTC when drink:AOR .3 SG blood:ACC dark:ACC
‘And when he had drunk the dark blood.’ (Homeric Greek, Hom.
Od. 11.98 after Napoli 2006: 113)
The use of the genitive as an object marker with ingestion verbs has traditionally
been interpreted as directly reflecting its partitive meaning, implying that only
an unspecified part of the object argument is consumed. The accusative, on the
other hand, is typically interpreted as implying that the object argument is com-
pletely consumed. Drawing on the discussion in Beavers (2011), I take ingestion
verbs to entail that the object argument undergoes a quantized change of state
and, on this assumption, the alternation between the accusative and the parti-
tive genitive with ingestion verbs in Indo-Iranian is exploited to specify whether
the object argument is totally consumed or only in part, in more technical terms,
whether all of the theme crosses the entire scale of being consumed. From this
perspective it is interesting to note that Hettrich (forthcoming: 6) observes that
the genitive is far less frequently attested with ingestion verbs than the accusa-
tive in Vedic. The accusative/genitive alternation is found with nine ingestion
verbs which select the accusative in approximately 445 cases and the genitive
in 132 cases.4 Hettrich also notes that the genitive is exclusively used when
only a part of the direct object is consumed, whereas the accusative may either
be used to express that the direct object is completely consumed or to express
that it is only partially consumed. He concludes that the opposition between
the accusative and the genitive is privative and that the genitive represents the
marked member of the opposition, as schematically represented in Table 2.
Table 2: The accusative/genitive alternation

as a privative opposition with ingestion verbs
Accusative Genitive
Ø [+ partitive]
With ingestion verbs, then, the use of the partitive genitive to express the object
appears to be licensed by the fact that such verbs entail that the object argu-
ment undergoes a quantized change of state but leave the exact measure open.
Significantly, the fact that ingestion verbs strongly tend to select an accusative-
marked object in Indo-Iranian may be interpreted as an indication that verbs of
4 The verbs are ad- ‘eat’ (28/1), aś- ‘consume’ (7/6), gar i- ‘devour’ (6/1), ghas- ‘consume’ (12/1),
joṣ- ‘enjoy’ (ca. 180/2), pā- ‘drink’ (163/106), bhaj- ‘enjoy as a share’ (28/13), bhas- ‘devour’
(5/1) and reh- ‘lick’ (26/1).
this type are prototypically used with the implicature that the object argument is
fully consumed. Under this analysis, the genitive represents a marked morpho-
syntactic option expressing that the object argument is not totally consumed, a
construction which, in turn, tends to yield an atelic meaning (cf. Dahl 2009a,
Dahl & Napoli 2008 for discussion). The question remains whether these proper-
ties are of a more general nature or whether they reflect idiosyncratic peculiari-
ties of ingestion verbs.
Other verb classes selecting the genitive as an object case marker include
perception/comprehension verbs like *ćrau̯h- ‘hear, listen’.5 Consider the examples
in (4).
(4) a. śyāvā́ śuvasya sunvatás táthā śṛṇu yáthā́

Śyāvāśuva:GEN extracting:GEN thus hear:PRS . IMP.2 SG like
áśṛṇor átreḥ kármāṇi kṛṇvatáḥ
listen:IPF.2 SG Atri:GEN sacred.deeds:ACC performing:GEN
‘Listen thus to Śyāvāśuva, who is extracting (soma), as you listened to
Atri, who was performing sacred deeds!’ (Vedic, Rigveda VIII 36.7)
b. surunaoiti zaota upa.sraotaranąm

listen:PRS .3 SG high.priest:NOM assistant.priests:GEN
nōit ̰ upa.sraotarō zaotarō
not assistant.priests:NOM high.priest:GEN
‘The high priest listens to the assistant priests, not the assistant priests
to the high priest.’ (Avestan, Nīrangastān 10.32)
Interestingly, perception/comprehension verbs like *ćrau̯h- ‘hear, listen’ are

sometimes also attested with accusative-marked object arguments in Vedic as
well as Avestan, as illustrated by the examples in (5).
(5) a. víśve devāḥ śṛṇutá imáṃ hávam me

all.gods:VOC hear:PRS . IMP.2 PL 3SG . M . ACC invocation:ACC 1sg.gen
‘O all-gods, hear this invocation of mine.’ (Vedic, Rigveda VI 52.13)
b. surunuiā̊ nō yasnəm ahurāne ahurahe

listen:PRS .OPT.2 SG 1 PL .GEN prayer:ACC Ahurānī:VOC Ahura:GEN
‘May you listen to our prayer, o Ahurānī (daughter) of Ahura!’
(Avestan, Yasna 68.9)
5 Other examples include Iir. *man- ‘remember, imagine’ (Ved. man-, Av. man-), *u̯ai ̯d- ‘find
out, recognize, know’ (Ved. ved-, Av. vaēd-), Ved. cet- ‘perceive, take care of’, bodh- ‘be aware’.
426 Eystein Dahl
We may note that about 9 verbs of this class are attested with an object argument
alternately expressed by the accusative or by the genitive in Vedic. Moreover, the
relative frequency of the accusative and genitive with perception/comprehension
verbs is somewhat less strikingly different than with ingestion verbs, the accusa-
tive being attested approximately 344 times and the genitive being attested
about 152 times (cf. also Hettrich forthcoming 9–10).6 Moreover, the examples
cited in (4) and (5) illustrate that the distribution of the accusative and genitive
with perception/comprehension verbs like *ćravh- ‘hear, listen’ is subject to
significant lexical semantic constraints. Specifically, the accusative is generally
restricted to cases where the object argument is an abstract noun denoting a
speech act type, whereas the genitive is restricted to cases where the object
argument denotes an animate referent. Significantly, these two case categories
show an analogous distribution in Homeric Greek, as illustrated by the examples
in (6a) and (6b). However, the situation in Homeric Greek is somewhat more
complex, as the genitive and accusative in some cases can be coordinated in
cases where the stimulus object denotes an inanimate referent, as illustrated by
example (6c).7 Hence, the ability of perception/comprehension verbs to alternately
select an object argument in the accusative or genitive is common to Indo-Iranian
and Greek and may be reconstructed for Proto-Indo-European, but as the exact
distribution of the two case categories with this class of verbs diverges slightly in
these two language families, their original distribution remains unclear.8
(6) a. hṑs éphath’, hoì d᾽ ára toû mála

Thus speak:AOR .3 SG 3 PL . M . NOM PTC PTC 3 SG . M .GEN very
mèn klúon ēdè píthonto:
indeed hear:AOR .3 PL and obey:AOR .3 PL
‘Thus he spoke and, indeed, they listened to him and obeyed him.’
(Homeric Greek, Hom Il. 15.300)
6 These verbs include ádhi-ay-/gam-/gā- ‘consider, regard’ (5/7), cet- ‘perceive’ (50/16), bodh-
‘be aware’, man- ‘remember, think’ (31/5), ved- ‘find out, realize’ (ca. 130/ca. 60) and śrav-
‘hear, listen’ (ca. 120/ca. 40).
7 I am grateful to Silvia Luraghi for having drawn my attention to this fact.
8 There appears to be little or no difference in meaning between the construction with the
genitive and the one with the accusative with inanimate stimuli in Homeric Greek, as illustrated
by the example in (6c). Moreover, verbs like akoúō ‘hear, listen’ exclusively select the genitive
when the meaning is intentional (‘listen to’) or when it is factual (‘hear about’), as pointed out
to me by Silvia Luraghi (p.c.). These facts readily explain the tendency of associating the genitive
with animate referents in Indo-Iranian and may be taken as an indication that the Homeric
Greek situation is more archaic and the Indo-Iranian situation is slightly more innovative, as
the distribution patterns of the two case categories appear to be more restrictive in Indo-
Iranian and hence may be regarded as more grammaticalized than in Homeric Greek.
b. ēé tin᾽ aggelíēn stratoû ékluen erkhoménoio

or some:ACC message:ACC army:GEN hear:AOR .3 SG coming:GEN
‘Or has he heard a message that an army is coming.’ (Homeric Greek,
Hom. Od. 2.30)
c. mukēthmoû t’ ḗkousa boôn aulizomenáōn

lowing:GEN PTC hear:AOR .1 SG cows:GEN lodge:PTCP. PRS .GEN
oiôn te blēkhḗn
sheep:GEN PTC bleating:ACC
‘I heard the lowing of the cattle lying (in the courtyard) and the
bleating of the sheep.’ (Homeric Greek, Hom. Od. 12.265–266 after
Luraghi et al. 2012)
At this point we need to consider two slightly distinct sets of problems. The first
concerns the relationship between the accusative and the genitive with verbs of
this type. The second concerns the semantic relationship between the allegedly
partitive genitive and its use as an object marker with perception/comprehension
verbs.
From the examples discussed so far, the somewhat incoherent impression
arises that the accusative is the preferred object realization option with percep-
tion/comprehension verbs and that the distribution of the accusative and geni-
tive at the same time is determined by the distinction between abstract and
animate nouns. This situation is very different from the one we encountered
earlier with ingestion verbs where it is less immediately obvious that there are
any similar lexical semantic motivation for the distribution of the two object
realization options. An important problem concerns the fact that abstract nouns
appear to be characterized by fundamentally different set of lexical entailments
than animate nouns or proper names. If the distribution of the accusative and
genitive is motivated by this distinction, it remains unclear to what extent it
may be directly compared with the alternation pattern shown by ingestion pre-
dicates, where the distribution of these two case categories is motivated by a
relatively clear-cut privative grammatical opposition.
However, a closer look at the data reveals that the accusative in certain
cases may be used when the object argument is animate. Most notably, this is
the case when verbs like *ćrav h- ‘hear, listen’ are used with an object predica-
tive, as illustrated by the following examples from Vedic and Avestan:
(7) a. bhiṣáktamaṃ tvā bhiṣájāṃ śṛṇomi

most.healing:ACC 2 SG . ACC healing:GEN hear:PRS .1. SG
‘I hear that you (are) the most healing of the healing ones.’
(Vedic, Rigveda II 33.4 after Dahl 2009a: 42)
428 Eystein Dahl
b. vīspanąmca aēṣ̌ąm zaraQuštrəm paoirīm vahištəm

all:GEN 3PL . M .GEN Zarathustra:ACC first:ACC best:ACC
āhūirīm sūsrūma t ̰kaēṣ̌əm
belonging.to.Ahura.Mazdā:ACC hear:PRF.1 PL teacher:ACC
‘We have heard that Zarathustra (is) the first and best teacher of
Ahura Mazdā’s doctrine among them all.’ (Avestan, Yašt 13.148)
The available data thus suggest that the accusative may be used with abstract as
well as animate nouns whereas the genitive is restricted to animate nouns with
perception/comprehension predicates. Hence, it is reasonable to conclude that
the accusative/genitive alternation may be analyzed as a privative grammatical
opposition, with the genitive as a marked member, as schematically represented
in Table 3.

as a privative opposition with Perception/
Comprehension verbs (preliminary)
Accusative Genitive
Ø [+ animate]
Although an analysis along these lines may account for the distribution of
the accusative and genitive as object markers with perception/comprehension
verbs, it begs the question as to whether the restriction of the genitive to animate
nouns may be derived from the other uses associated with this case category or
whether it represents a reading which is less obviously synchronically related
with the core readings of the genitive. It was noted previously that the genitive
has two principal readings in Indo-Iranian, a partitive and a possessive reading,
but it is not immediately clear that the above restriction can be plausibly taken
to represent a special case of either of these readings. Note, however, that the
accusative noun phrases imáṃ hávam me ‘this invocation of mine’ and nō
yasnəm ‘our prayer’ in (5) provide an unambiguous endpoint for the situation
in an intuitively clear way, the resultant verb phrase yielding an unmistakable
telic reading. In contrast, the referents of the genitive noun phrases śyāvā́ śuvasya
sunvatás ‘extracting Śyāvāśuva’, átreḥ kármāṇi kṛṇvatáḥ ‘Atri making sacred
deeds’ and upa.sraotaranąm ‘assistant priests’ and zaotarō ‘high priest’ in (4) do
not introduce an endpoint for the situation denoted by the verb *ćrau̯h- ‘hear,
listen’. Significantly, the accusative noun phrases bhiṣáktamam tvā ‘you the most
healing’ and zaraQuštrəm paoirīm vahištəm āhūirīm t ̰kaēṣ̌əm ‘Zarathustra the first
and best teacher of Ahura Mazdā’s doctrine among them all’ also do not seem to
imply that the situation has a natural endpoint. These considerations suggest that
the genitive is exclusively associated with an atelic reading whereas the accusa-
tive is compatible with a telic as well as an atelic reading, as schematically repre-
sented in Table 4.

as a privative opposition with Perception/
Comprehension verbs
Accusative Genitive
Ø [+ atelic]
From a general typological perspective, the use of the genitive as marker of

atelic verb phrases may be taken to directly reflect its partitive meaning as this
is an aspectual reading commonly associated with partitive categories across
languages. Under this account, the restriction of the genitive to animate noun
phrases may be regarded as a consequence of the fact that noun phrases of this
type characteristically give rise to an atelic meaning with perception verbs. Note
that perception/comprehension verbs are underspecified with regard to the
change of state feature and, consequently, belong at the bottom end of the
affectedness scale. Hence, the data we have reviewed so far allow us to conclude
that the distribution of the genitive object construction does not seem to be
determined by the relative affectedness of the object argument but that it rather
tends to be employed as a kind of oppositional grammatical marker of atelicity.
Authority/possession verbs like *aić- ‘be lord, control, possess’9 constitute
another group of verbs that take an object argument with genitive case marking
in Indo-Iranian, as illustrated by the examples in (8).
(8) a. ayám agníḥ suvīŕ yasya

3SG . M . NOM Agni:NOM rich.in.sons:GEN
īś́ e maháḥ saúbhagasya
control:PRS .3 SG great:GEN happiness:GEN
‘This Agni here controls the great welfare of having many sons.’
(Vedic, Rigveda III 16.1)
b. kat ̰ mōi uruuā isē cahiiā auuaŋhō

PTC 1 SG .GEN soul:NOM command:PRS .3 SG any:GEN help:GEN
‘Does my soul command any (kind) of help?’
(Avestan, Yasna 50.1 after Humbach et al. 1991: 183)
9 Other examples include Iir. *ćsā- ‘rule, control’ (Ved. kṣā- , Av. xšā-) and Ved. irajya- ‘rule
over, straighten’ and rāj- ‘rule over’.
430 Eystein Dahl
As many verbs that select genitive objects in Indo-Iranian also allow their object
argument to be expressed by the canonical accusative, this is what we would
expect with authority/possession verbs as well. This expectation seems to be
borne out, as illustrated by the following data from Vedic:
(9) a. yád indra yā́ vatas tuvám etā́ vad ahám

if Indra:VOC as.much:GEN 2 SG . NOM so.much:ACC 1 SG . NOM
īś́ īya stotā́ ram íd didhiṣeya
possess: PRS .OPT.1 SG singer:ACC indeed want.to.win:PRS .OPT.1 SG
‘If I got to possess so much as you (are possessing) Indra, then I
would indeed want to win the singer for me.’ (Vedic, Rigveda VII
32.18 after Dahl 2009a: 46)
b. ā́ satvanaír ájati hánti

to warriors:INS assail:PRS .3 SG defeat:PRS .3 SG
vṛtráṃ kṣéti kṣitīḥ́
enemy:ACC control:PRS .3 SG races.of.men:ACC
‘With his warriors he attacks and defeats his enemy with his warriors,
he gains control over the races of men.’
(Rigveda V 37.4 after Dahl 2009a: 46)
Somewhat surprisingly, we do not seem to find parallel examples in Old Iranian.

This might, of course, reflect an accidental gap in the corpus, especially given
the fact that the Old Iranian sources are not very extensive. Interestingly, how-
ever, locative-marked object arguments of authority/possession verbs are mar-
ginally attested in Vedic and Avestan, as illustrated by the following examples:
(10) a. índur devéṣu patyate

drop:NOM gods:LOC be.lord:PRS .3 SG
‘The drop (Soma) is lord of the gods.’
(Rigveda IX 45.4 after Dahl 2009: 52)
b. naēciš xšaiiāt ̰ (. . .) aŋrō mainiiuš

not be.lord:PRS . SBJV.3 SG Aŋru:NOM Mainiius:NOM
xvaēšu dāmōhu
POSS .3 SG . LOC . PL . M beings:LOC
‘Aŋru Mainiius shall not be lord of his (i.e. Ahura Mazdā’s) beings.’
(Fragmentum Westergaard 4.2)
Although this object realization pattern appears to be restricted to the verb

patya- ‘be king’ in Vedic and to the one isolated example found in Avestan, a
comparative look at data from Homeric Greek suggests that the Indo-Iranian
construction is not isolated in an Indo-European perspective. Consider, for
instance, the following examples:
(11) a. Tenédoió te îphi anásseis

Tenedos:GEN and with.might be.king:PRS .2 SG
‘And you rule Tenedos with might.’ (Homeric Greek, Hom. Il. 1.38)
b. Trṓessin anáxein hippodámoisi

Trojans:DAT be.king:INF horse.taming:DAT
‘(. . .) to rule the horse-taming Trojans.’
(Homeric Greek, Hom. Il. 20.180)
These examples illustrate that authority/possession verbs alternately select the

genitive or the dative in Homeric Greek. What is remarkable from a comparative
perspective is that the Homeric dative has subsumed the function of the Indo-
European locative. As a consequence, Homeric examples like (11b) may be
directly compared with Vedic examples like the one cited in (10a) and with the
Avestan example in (10b). If this comparison is correct, the Indo-Iranian and
Indo-European genitive did not compete with the accusative as an object marker
of authority/possession verbs but rather with the locative. It is therefore tempt-
ing to regard Vedic examples like those cited in (9) as secondary, being due to a
generalization of canonical argument realization patterns in Vedic. However,
more detailed research is needed in order to substantiate this claim.
Before concluding the discussion of this class of verbs, we may note that
most of the verbs involved are synchronically transparent denominatives. Hence,
it is reasonable to understand the genitive as partitive in these cases. I also
believe that it is uncontroversial that authority/possession verbs do not strictly
entail that the object argument undergoes a change of state and verbs belonging
to this class appear to belong at the low end of the affectedness hierarchy.
The preceding discussion indicates that, as a rule, the genitive appears to
have been one among two (or more) object realization patterns associated with
certain classes of verbs in Indo-Iranian as well as Indo-European. Interestingly,
however, a few verbs are only attested with genitive-marked objects, e.g. *karh-
‘praise, speak highly of’. Consider, for instance, the following examples:
432 Eystein Dahl
(12) a. maháś carkarmi árvataḥ kratuprā́

great:GEN praise:PRS .1 SG steed:GEN being.inspired:NOM
dadhikrā́ vṇaḥ puruvā́ rasya vṛ́ṣṇaḥ
Dadhikravan:GEN rich.in.gifts:GEN mighty:GEN
‘Being inspired, I praise Dadhikravan, the mighty steed which is rich
in gifts.’ (Vedic, Rigveda IV 39.2)
b. fšūmā̊ astī ašavā vәrәQrajā

cattle.holder:NOM be:PRS .3 SG righteous:NOM victorious:NOM
vahištō fšūšə̄ carәkәrәmahī
most.successful:NOM possessing.cattle:GEN praise:PRS .1 PL
‘He who has cattle is righteous, victorious and most successful.
We praise him who is possessing cattle.’ (Avestan, Yasna 58.4)
Apart from the fact that the verb *kar h- ‘praise, speak highly of’ is exclusively
attested with an object argument in the genitive, it is exceptional also because
it seemingly represents the only verb of praising that selects a genitive-marked
object argument in Indo-Iranian.10 Other verbs with a similar meaning either
select the accusative or the dative in Indo-Iranian, as illustrated by the following
examples:
(13) a. índram̐ stavā nṛ́tamaṃ yásya

Indra:ACC praise:PRS . SBJV.1 SG most.manly:ACC who:GEN
mahnā́ vibabādhé rocanā́ ví
greatness:INS drive.asunder:PRF.3 SG lights.of.heaven:ACC asunder
jmó ántān /
earth:GEN ends:ACC
‘I will praise the most courageous Indra who by his might drove
asunder the lights of heaven, the ends of the earth.’ (Vedic,
Rigveda X 89.1ab)
b. yauuat ̰ā Qβā mazdā staomī11 ufiiācā

how.much you:ACC Mazdā:VOC praise:PRS .1 SG extol PRS .1 SG =and
‘According to how much I praise and extol thee.’ (Avestan, Yasna 43.8
after Humbach et. al. 1991: 153)
10 This construction may safely be assumed to be restricted to Indo-Iranian, as the etymon of

this verb only appears as a noun outside of Indo-Iranian, cf. Mayrhofer (1992: 310–311) and Rix
et al. (2001: 353).
11 Note that Geldner’s edition reads staumī.
c. yó mitrā́ ya váruṇāya ~ ávidhaj jáno

who:NOM Mitra:DAT Varuṇa:DAT honor:AOR .3 SG man:NOM
‘The man who honors Mitra and Varuṇa.’ (Vedic, Rigveda I 136.5a)
d. tə̄ m zī və̄ spərədā nivarānī

3. SG . M . ACC for 2. PL .GEN enthusiasm:INS convert:AOR . SBJV.1 SG
yā fəδrōi vīdāt ̰
who:INS father:DAT honor:AOR . SBJV.3 SG
‘For with enthusiasm I shall convert that one amongst you by whom
one will honor the father.’ (Avestan, Yasna 53.4)
Examples like these are quite representative for verbs denoting an act of prais-
ing or homage and the verb *karh- ‘praise, speak highly of’ appears to be the
only verb within this semantic field that selects an object argument in the geni-
tive. Note that Hettrich (forthcoming: 9) classifies the Vedic verb kari- ‘praise,
speak highly of’ as a ‘verb of mental activity’ which corresponds to the percep-
tion/comprehension verbs in the present terminology. Although a classification
along these lines has the immediate advantage that the verb under discussion
would thereby fall under a more general pattern, it is not clear to me that it is
justified on semantic grounds. Specifically, unlike perception/comprehension
verbs like *ćrau̯h- ‘hear, listen’ or *u̯ai ̯d- ‘find out, recognize, know’ which appar-
ently only presuppose that the subject is animate, the verb *karh- ‘praise, speak
highly of’ entails a volitional subject argument. Finally, I would like to draw
attention to the fact that verbs of praising like *kar h- ‘praise, speak highly of’
are activity predicates, i.e. they characteristically denote a temporally unbounded
or atelic process. From this perspective, the use of the genitive with verbs of this
type may be regarded as directly related to its partitive meaning.
Most of the examples so far appear to corroborate the conclusion that the
genitive object construction was primarily used in cases where a more or less
markedly atelic reading was intended. In this connection, the data cited in (14)
are of some interest.
(14) a. ho pərəQβe aiŋ́hā̊ zəmo

3SG . M . NOM surface:LOC 3SG . F.GEN earth:GEN
upa.Qβaršti uruuaranąm
cut.off:PRS .3 SG plants:GEN
‘He cuts down plants on the surface of the earth.’
434 Eystein Dahl
b. daṣ̌inəm he gaoṣ̌əm upa.Qβərəsaiiən

right:ACC 3SG . M .GEN ear:GEN cut.off:OPT.3 PL
‘They shall cut off his right ear.’ (Avestan, Vendīdād 13.32)
c. janaiti vīspaēšąm aŋrō mainiiuš

smite:PRS . SBJV.3 SG everyone:GEN Aŋru:NOM Mainiius:NOM
‘Aŋru Mainiius will smite everybody.’ (Avestan, Yašt 3.5)
d. vidanvān vai bhārgava indrasya pratyaham̐ s

Vidanvat:NOM indeed Son.of.Bhṛgu:NOM Indra:GEN smite:IPF.3 SG
‘Vidanvat, the son of Bhṛgu (once upon a time) smote Indra.’
(Vedic, Pañcaviṃśabrāhmaṇa XIII 11.10)
Example (14a) shows that the genitive object construction is compatible with verbs
like upa.Qβars- ‘cut down’ in Avestan, that is, verbs entailing a non-quantized
change of state. As shown in (14b), this verb is also attested with an object argu-
ment in the accusative and it is therefore reasonable to regard the use of the
genitive object construction as an instantiation of its more general preference
for atelic contexts. Similar observations apply to the examples in (14c) and (14d)
where the genitive object construction occurs with the verbs Av. gan- ‘smite’ and
Ved. prati-han- ‘smite’ which clearly entail that the theme undergoes a quantized
change of state. In these and similar cases, the genitive appears to induce a
markedly atelic reading, yielding a conative-like meaning and indicating that
the subject argument makes an unsuccessful attempt to perform the situation
denoted by the predicate. These considerations suggest that the Indo-Iranian
genitive in some cases could be used with verbs denoting a quantized change
of state to indicate that an atelic or perhaps even an imperfective meaning is
intended, a discourse function which is fairly frequently found with partitives
cross-linguistically.
From this perspective it is remarkable that there appear to be few aspectual
constraints on the use of the genitive as an object marker. Dahl (2011a, 2011b)
argues that the comparative evidence from Vedic, Avestan and Ancient Greek
indicates that aspect played a central role in the Indo-Iranian verbal system.
Specifically, there was a distinction between the neutral-imperfective present
stem, the perfective aorist stem and the anterior/perfect perfect stem (for a defi-
nition of these aspectual relations cf. also Dahl 2009b, 2010). At first glance, the
following examples appear to show that the genitive could be used as an object
marker with forms of the present, aorist as well as the perfect stem in Early Vedic.
(15) a. simá ukṣṇó avasṛṣṭā́ m̐ adanti

themselves:NOM oxen:ACC released:ACC eat:PRS .3 SG
‘They eat/are eating the released oxen themselves.’
(Vedic, Rigveda X 28.11 after Dahl 2009a: 37)
b. kuvít sómasya ápām

whether soma:GEN drink:AOR .1 SG
‘Have I drunk/been drinking soma?’
(Vedic, Rigveda X 119.1 after Dahl 2009a: 39)
c. índro no asyá pūrvyáḥ papīyād

Indra:NOM 1 PL .GEN 3SG . M .GEN first:NOM drink:PRF.OPT.3 SG
dyukṣó mádasya somyásya rā́ jā //
heavenly:NOM intoxicating.drink:GEN soma:GEN king:NOM
‘Indra, the heavenly king, shall have drunk/been drinking the
inebriating soma as the first among us.’ (Vedic, Rigveda VI 37.2cd)
These examples illustrate that genitive objects are attested with forms of the
present (adanti ‘they eat’), aorist (ápām ‘I have drunk’) and perfect (papīyāt ‘he
shall have drunk’). Significantly, however, whereas examples like (15a) or (15b)
are fairly frequently met with, examples of the type illustrated by (15c) are
exceedingly rare in Vedic. Comparative data from Avestan show a similar distri-
bution pattern. This fact suggests that the genitive object construction was com-
patible with the neutral/imperfective aspect as well as the perfective aspect in
Vedic but that it tended not to be combined with the anterior aspect. Under the
assumption that the genitive object construction was associated with a markedly
atelic or perhaps even progressive-like interpretation arising from its partitive
reading, this restriction indicates that some languages, like Vedic and probably
Indo-Iranian, tend to avoid the combination of progressive and anterior markers
whereas others, like English, are more permissive in this respect. However, a full
evaluation of this possibility will have to be pursued elsewhere.
At this point, an interim summary is in order. We have seen that the data
reviewed in the previous paragraphs suggest that the use of the genitive as an
object marker is closely related with the basic partitive meaning of this case
category and that verbs selecting the genitive as an object marker also tend to
allow other object case marking patterns. In such cases, the genitive is often
associated with a markedly atelic interpretation, something which is directly
derivable from its basic partitive meaning. The genitive object construction is
frequently attested with forms of the present and aorist stems but very rarely
occurs with forms of the perfect stem, a fact which could indicate that Indo-
Iranian tended not to combine partitive and anterior morphology.
436 Eystein Dahl
3.2 The genitive as a subject marker in Indo-Iranian

The genitive is far less commonly used as a subject marker than as an object
marker in Indo-Iranian. Interestingly, however, it is one of only two relatively
clear-cut non-canonical subject constructions which are attested in Indo-Aryan
as well as Iranian, the other selecting the accusative. Consider, for instance, the
following examples:
(16) a. xvarəQanąm he barәtąm

food:GEN 3SG . M . DAT be.carried:PRS . IMP.3 SG
zaremayehe raokhnahe
springtime:GEN butter:GEN
‘As food shall springtime butter be brought to him!’ (Avestan,
Haδoxt Nask 2.18)
b. yat ̰ hē stārąm baχō.dātanąm

so.that 3SG . M . DAT stars:GEN set.up.by.the.gods:GEN
aiβi raocaiiā̊ṇte
round shine:PRS .3 PL
‘So that the stars which are set up by the gods shine around for
him.’ (Avestan, Vendīdād 19.23)
c. ápāyi asya ándhaso mádāya

drink:AOR . PASS .3 SG 3 SG . N .GEN soma.juice:GEN inebriation:DAT
‘(Some of) this soma-juice has been drunk for inebriation (i.e. in order
to get inebriated).’ (Vedic, Rigveda II 19.1)
d. ákāri vām ándhaso

make:AOR . PASS .3 SG 2 D. DAT soma.juice:GEN
‘Some soma-juice has been prepared for you two.’
(Vedic, Rigveda VI 63.3)
These examples illustrate that the genitive subject construction is compatible

with non-active voice morphology in Vedic as well as Avestan. The Avestan
forms barәtąm ‘shall be carried’ and raocaiiāṇ̊ te ‘shine’ in (16a) and (16b) are in
the middle voice, whereas the Vedic construction illustrated by (16b) and (16c) is
in the passive voice.12 Some readers might object that the two construction types
illustrated in (16) are not directly comparable, as the so-called passive Aorist of
12 Note that the term ‘passive’ may not be entirely accurate, as this construction is found with
two-place as well as one-place predicates.
the type ápāyi ‘is drunk, has been drunk’ and ákāri ‘is made, has been made’ is
only found in Vedic (cf. Insler 1968) and that the construction instantiated in
(16b) and (16c) is obviously secondary. Still, I have chosen to compare these ex-
amples because they neatly illustrate that the use of the genitive as a subject
marker with non-active voice forms of the verb reflects its partitive meaning in
a fairly direct manner.
On way of interpreting the non-active voice morphology in (16) is that it has
substantial impact on the lexical semantics as well as the argument structure
associated with the verbal root. In their canonical use, the Indo-Iranian middle
and passive constructions tend to reduce a given verb’s inherent semantic tran-
sitivity, in some cases by suppressing agentive subjects and promoting patientive
objects to subject position. In other cases, non-active voice morphology may be
taken to indicate that the privileged argument of the predicate lacks any control
over the outcome of the situation denoted by the predicate. From this perspec-
tive, cases like those cited in (16c) and (16d) may be regarded as the non-active
counterparts of constructions like those cited in (1a) where an active verb has
a partitive object. Note, however, that we do find a few examples of genitive
subjects with active voice morphology in Vedic as well as Avestan, as illustrated
by the examples in (17), a fact which indicates that the genitive subject con-
struction was compatible with active as well as non-active voice morphology in
Indo-Iranian.
(17) a. uruuaranąm zairi.gaonanąm zaramaēm paiti zəmāδa

uzuxšiieiti
(Avestan, Yašt 7.4)
b. yat bā paiti fraēshtəm bavain̩ti

where indeed again to.greatest.extent be:PRS .3 PL
angrō.mainiiauuanąm gərədhąm
belonging.to.evil.spirits:GEN caves:GEN
‘Where indeed there are most caves housing evil spirits.’
c. yád vái púruṣasya ~ āmáyaty

if PTC man:GEN be.ill:PRS .3SG
‘If a man is ill.’ (MS I 8.9)
438 Eystein Dahl
Note that all of the predicates cited in (17) are unaccusative, but that the Avestan
examples both have an inanimate subject whereas the Vedic example has an
animate (experiencer) subject. These examples differ in significant respects and
it is not clear how the apparently partitive use of the genitive in (17ab) is related
to the use of the genitive to express the experiencer in (17c). Significantly, we
do not find any examples of experiencer predicates with a genitive subject in
Avestan where this construction type seems to be restricted to inanimate sub-
jects and to have a relatively clear-cut partitive meaning, characteristically denot-
ing an indefinite quantity. In Vedic, on the other hand, this construction is partly
found when non-active verbal forms have an inanimate subject characteristi-
cally showing a properly partitive meaning and partly when active or non-active
verbal forms have an animate subject with a less obviously partitive meaning.
These data can be interpreted in various ways. One could, for instance, take
the Vedic data at face value and assume that the Indo-Iranian genitive subject
construction was simply incompatible with the [+VOLITION] feature and that
cases with animate subjects are simply unattested in Avestan. On the other
hand, one might also take the Avestan situation to be more archaic and assume
that this construction was incompatible with the [+SENTIENCE] feature. On the
latter assumption, the use of the genitive subject construction with experiential
predicates in Vedic might be taken to represent an innovation. Interestingly, we
find some sporadic examples in Homeric Greek where the subject is expressed
by the genitive, as illustrated by the following examples (cf. Conti 2008: 101,
Conti & Luraghi this volume):
(18) a. áll’ oú pēi khroòs eísato pâs

but not PTC skin:GEN be.seen:AOR .3 SG all:NOM
d’ára khalkôi smerdaléōi kekáluphth’
PTC armour:DAT terrible:DAT be.hidden:PPF.3 SG
‘But no part of the skin was seen, for it had all been hidden in the
terrible armor.’ (Homeric Greek, Hom. Il. 13.191–192)
b. hṑs aikhmês apélamp’ euḗkeos. . .

thus spear:GEN be.resplendent:IPF.3 SG sharp:GEN
‘Thus (the point of) the sharp spear was resplendent.’
(Homeric Greek, Hom. Il. 22.319)
These passages appear to be the only clear-cut cases where the genitive is used
as a subject marker in Homeric Greek. From the perspective of the preceding dis-
cussion it is significant that the subject argument is inanimate in both of these
cases, something which provides prima facie evidence in favor of the assump-
tion that Indo-Iranian had inherited a genitive subject construction with a parti-
tive meaning from Proto-Indo-European which was restricted to unaccusative
one-place predicates with an inanimate subject argument and that the use of
the genitive with experiencer predicates is a Vedic innovation. This hypothesis
is further strengthened by the fact that Vedic examples like the one cited in
(17c) do not appear in the oldest stage of the language whereas examples like
those cited in (16cd) do.
Before concluding the discussion of the use of the genitive as a subject
marker in Indo-Iranian, one might be tempted to ask whether it is possible to
reconcile this pattern of use with its use as an object marker, i.e. to what extent
the genitive subject construction has an atelic or progressive-like meaning.
Although the available evidence is somewhat meager, there are some potentially
significant distributional similarities between these two constructions. First,
both are primarily found with forms of the present or aorist. Second, they seem
to be restricted to verbs that have a relatively low degree of transitivity. Third,
the examples cited in (16) through (18) are perfectly compatible with an atelic
reading. Thus one may stipulate that the Indo-Iranian genitive subject construc-
tion was also associated with an atelic meaning, although the arguments in
favor of this hypothesis are not of the strongest possible kind. The fact that there
are comparably few instances of predicates with a genitive-marked subject in
the Indo-Iranian languages reflects a more general pattern, namely that there
are comparably few cases of non-canonically case-marked subjects in these
languages. In contrast, both Vedic and Avestan have a relatively rich inventory
of non-canonical object constructions. Taken together, these facts suggest that
the Indo-Iranian languages differ in typologically significant respects from
certain other Indo-European languages, like for instance Latin, which have a
relatively rich set of non-canonical subject-like constructions but employ non-
canonical object constructions to a somewhat less pervasive extent.
4 Summary and conclusion

This paper has discussed some salient interrelated readings associated with the
Indo-Iranian genitive. It was argued that this case category has a basic partitive
meaning which underlies its use in argument realization. Although the Indo-
Iranian genitive is attested both as a subject marker and as an object marker,
it is far less frequently used to express the subject argument than to express
the object argument. The former construction type appears to be restricted to
440 Eystein Dahl
unaccusative predicates with inanimate subjects, i.e. it seems to be incompatible

with verbs entailing that the subject argument is animate. The use of the genitive
as an object marker, on the other hand, is found with a broad range of lexical
classes of verbs and that the genitive object construction was strongly associated
with an atelic meaning. We have also seen that the evidence from Homeric Greek
indicates that many, if not all of the various readings discussed here can be
ascribed to the Proto-Indo-European genitive but more detailed research is
needed in order to establish to what extent the Indo-Iranian genitive represents
an archaic case category or whether it also must be ascribed certain innovative
readings. In any case, the Indo-Iranian genitive appears to show certain basic
readings characteristic of partitive categories cross-linguistically.
References
Beavers, John T. 2006. Argument/oblique alternations and the structure of lexical meaning.
Stanford, CA: Stanford University dissertation.
Beavers, John T. 2011. On affectedness. Natural Language and Linguistic Theory 2011(29). 335–
370.
Cheung, Johnny. 2007. Etymological dictionary of the Iranian verb. Leiden: Brill.
Conti, Luz. 2008. Synchronie und diachronie des altgriechischen genitivs als semisubjekt.
Historische Sprachforschung 2008(121). 94–113.
Dahl, Eystein. 2009a. Some semantic and pragmatic aspects of object alternation in Early Vedic.
In Jóhanna Barðdal & Shobhana Chelliah (eds.), The Role of Semantics and Pragmatics in
the Development of Case. 23–55. Amsterdam: Benjamins.
Dahl, Eystein. 2009b. Reconstructing inflectional semantics: The case of the Proto-indo-
European imperfect. In Stephanie W. Jamison, H. Craig Melchert & Brent Vine (eds),
Proceedings of the 20th Annual UCLA Indo-European Conference, 37–52. Bremen: Ute
Hempen Verlag.
Dahl, Eystein. 2010. Time, tense and aspect in Early Vedic grammar: Exploring inflectional
semantics in the Rigveda. Leiden: Brill.
Dahl, Eystein. 2011a. Tense and aspect in Indo-Iranian Part 1: The present and aorist. Language
and Linguistics Compass 5(5). 265–281.
Dahl, Eystein. 2011b. Tense and aspect in Indo-Iranian Part 2: The Perfect, futurate, participial
and periphrastic categories. Language and Linguistics Compass 5(5). 282–296.
Dahl, Eystein & Maria Napoli 2008. Case as an aspect marker? The partitive case in Indo-
European. Paper read at the XIII. Fachtagung der Indogermanischen Gesellschaft, Salzburg.
22–27 September, 2008.
Delbrück, Berthold. 1888. Altindische Syntax. Halle a. S.: Verlag der Buchhandlung des
Waisenhauses.
Gershevich, Ilya. 1959. The Avestan Hymn to Mithra. Cambridge: Cambridge University Press.
Grimm, Scott. 2005. The lattice of case and agentivity. Amsterdam: University of Amsterdam
dissertation.
Hettrich, Heinrich. Forthcoming. Some remarks on the adverbal genitive in Rigvedic Sanskrit. To
appear in Jared Klein & Elizabeth Tucker (eds.), Vedic and Sanskrit Historical Linguistics:
Papers from the 13th World Sanskrit Conference. New Dehli: Motilal Banarshidas.
Hock, Hans Heinrich. 1991. Oblique subjects in Sanskrit? In Manindra Verma & K. P. Mohanan
(eds.), Experiencer Subjects in South Asian Languages, 119–139. Stanford, CA: CSLI.
Hopper, Paul & Sandra Thompson. 1980. Transitivity in grammar and discourse. Language 56.
251–299.
Humbach, Helmut, Josef Elfenbein & Prods O. Skjærvø. 1991. The Gāthās of Zarathushtra and
the other Old Avestan Texts. Part I. Introduction – Text and Translation: Heidelberg: Carl
Winter.
Insler, Stanley. 1968. The origin of the Sanskrit passive aorist. Indogermanische Forschungen 73.
312–346.
Levin, Beth. 1999. Objecthood: An event structure perspective. Chicago Linguistic Society 35(1).
223–247.
Levin, Beth & Malka Rappaport Hovav. 1995. Unaccusativity: At the syntax-semantics interface.
Cambridge, MA: MIT Press.
Luraghi, Silvia, Barbara McGillivray & Eleonora Sausa. 2012. Verbal valency and classes of
verbs in Ancient Greek: Bivalent verbs in Homer. Paper read at the workshop Contrastive
studies of verbal valency in European languages at the 45th Annual Meeting of the
Societas Linguistica Europaea, Stockholm 29 August – 1 September, 2012.
Mayrhofer, Manfred. 1992. Etymologisches Wörterbuch des Altindoarischen. I. Band. Heidelberg:
Carl Winter.
Napoli, Maria. 2006. Aspect and actionality in Homeric Greek: A contrastive analysis. Milano:
Franco Angeli.
Reichelt, Hans. 1909. Awestisches Elementarbuch. Heidelberg: Carl Winter.
Rix, Helmut, Martin Kümmel, Thomas Zehnder, Reiner Lipp & Brigitte Schirmer. 2001. Lexikon
der indogermanischen Verben. Wiesbaden: Reichert.
Tsunoda, Tasaku. 1985. Remarks on transitivity. Journal of Linguistics 21. 385–396.
Luz Conti and Silvia Luraghi
13 The Ancient Greek partitive genitive in
typological perspective
The paper describes the usage of the partitive genitive in Ancient Greek. We
show that the partitive genitive can take a variety of syntactic functions, includ-
ing subject, direct object, complement of other bivalent and trivalent verbs,
adverbial, and complement of preposition. Hence, the partitive genitive is not
connected with a specific grammatical relation: rather, it indicates partial involve-
ment of a referent in an event, which is usually reflected in a low degree of tran-
sitivity. In addition, the partitive genitive may indicate indefiniteness. From the
point of view of discourse organization, partitive subjects, which often occur
in presentative clauses, convey new information and are never topical. Beside
partitive subjects in personal constructions, partitive arguments also occur in
impersonal constructions, in which they can display some behavioral properties
of subjects. Partitive adverbials occur in space and time expressions, mostly
limited to Homeric Greek. Especially when occurring with prepositions, the par-
titive genitive indicates that a portion of space is conceived as constituted of
detachable units, as opposed to the accusative. This has various consequences
on possible landmarks and on the structure of trajectories when the same prepo-
sition occurs with either case.
Keywords: partitive genitive, Ancient Greek, indefiniteness, degrees of

affectedness, low transitivity, low involvement.
1 Introduction
In this paper, we give a description of the partitive genitive in Ancient Greek
(henceforth AG). We show that the partitive genitive can occur in a wide variety
of syntactic constructions, partly parallel to the genitive in other Indo-European
(henceforth IE) languages, or to other partitive markers discussed in this book.
Significantly, both direct objects and subjects can be encoded through the partitive
genitive, which does not, thus, indicate a specific grammatical relation. As we
show in the paper, the partitive genitive may be obligatory in a certain construc-
tion, or it may alternate with another case. We argue that, in the latter type of
occurrences, the original motivation for the use of the partitive genitive is the
encoding of partial/low involvement of a referent in an event.
444 Luz Conti and Silvia Luraghi
The paper is organized based on the syntactic functions of the partitive geni-
tive, and is divided into four main sections, devoted to the genitive used as a
second argument (sec. 2), as a subject (sec. 3), as a third argument (sec. 4), and
as an adverbial (sec. 5). Section 6 summarizes the findings. In the remainder of
this section, we give a preliminary survey of syntactic, semantic and denota-
tional features of the AG partitive genitive.
The AG genitive results from the syncretism of the IE genitive and ablative,
as shown in Fig. 1:
Figure 1: Case syncretism and the Ancient Greek genitive
The Proto-Indo-European (henceforth PIE) ablative essentially expressed

origin, separation or distance with respect to a landmark, while the genitive
had a partitive value, which constituted the basis for this syncretism.1 From a
synchronic point of view, distinguishing usages of the AG genitive which must
be accounted for as based on the partitive meaning from those that result from
syncretism with the ablative is not always straightforward. Comparison with the
other IE languages allows us to reconstruct verb specific constructions which
took a (partitive) genitive second argument, such as those occurring with some
perception verbs.2 However, this task may become more complicated when the
comparative method fails to yield conclusive evidence, and the genitive depends
on verbs which, given their meaning, could in principle take either a genitive or
an ablative complement.
In its partitive usage, grammars traditionally describe the AG genitive as de-
noting an entity that is only partially involved in the verbal action (cf. Schwyzer
1 The functions of the PIE genitive are traditionally divided into adnominal and adverbal (see
Brugmann & Delbrück 1911: 565–567). In its adnominal function, the genitive indicated the
relation between a head noun and a modifier noun, as it does in all IE languages that preserved
and still preserve case marking. The partitive meaning that we discuss in this paper was typical
of the adverbal genitive (Brugmann & Delbrück 1911: 567).
2 In such cases, the verb can be said to indicate an event unfolding away rather than toward
an entity: consider examples such as ‘to hear something or someone’, i.e. to perceive a hint
from something or someone making a noise. This means that the semantic distance of the
genitive from the ablative was relatively small. Along with other factors, this overlap contributed
to the syncretism of the IE genitive and ablative in AG.
The Ancient Greek partitive genitive in typological perspective 445
& Debrunner 1950: 101 and Chantraine 1953: 50–51 among others), being a
whole that can be divided into detachable parts, as shown in Fig. 2:
Figure 2: Partitive meaning of the genitive
In fact, such description of the partitive meaning does not make any distinc-
tion between cases in which the partitive genitive must be understood as a part
of a partitive construction, a pseudo-partitive, (see Koptjevskaja-Tamm 2001), or
as an indefinite. We come back to this issue later on in this section.
From a syntactic point of view, the AG partitive genitive has several uses: it
can function as a second argument, frequently with direct object properties (this
is its most often described usage), as a subject, as a third argument, as an adver-
bial,3 or as the complement of a preposition. In practice, it can take any syntac-
tic function, with the exception of indirect object of verbs of giving or saying: as
we show in sec. 4, genitive third arguments are not indirect objects. Thus, the
partitive genitive does not encode a specific grammatical relation, and does
not fulfill the function which is typical of case, that is, to indicate the relation
between and NP and the predicate or the sentence in which it occurs (Blake
1994: 1; see further Luraghi 2003: 61–62 and 2009: 243–249). Notably, in many
of these functions the partitive genitive is a possible alternative to another
case, which does indicate a grammatical relation (e.g. the nominative for the
subject, the accusative for the direct object, etc.).
The genitive also functions as a predicate in copular and pure nominal
sentences as in (1). This example contains a so-called ‘genitive of material’,
discussed below, sec. 4.
(1) hē krēpìs mén esti líthōn megálōn

ART. NOM . SG basement:NOM PTC be:PRS .3 SG stone:GEN . PL big:GEN . PL
‘The basement is (made) of big stones.’ (Hdt. 1.93.6).
3 After Homer, use of the partitive genitive without a preposition became mostly restricted to
arguments. When functioning as an adverbial the genitive normally occurs in prepositional
phrases, except when it expresses time with a temporal referent (cf. nuktós ‘at night’).
Remarkably, in each of the uses we analyze in this paper, the partitive genitive is
generally (much) less frequent in comparison with other expressions.
As noted above, traditional descriptions of the meaning of the AG partitive
genitive do not capture completely all its semantic facets. Let us consider examples
(2)–(5):
(2) tòn pûr kêai ánôige. . . optêsaí

DEM . ACC fire:ACC light:INF. AOR order:IMPF.3 SG roast: INF. AOR
te kreôn
PTC meat:GEN . PL
‘He ordered him to light the fire and roast some meat.’ (Hom. Od. 15.97–98);
(3) tôn kēríōn hósoi éphagon tôn

ART.GEN . PL honeycomb: GEN . PL REL . NOM . PL eat: AOR .3 PL ART.GEN . PL
stratiōtôn pántes áphronés te egígnonto
soldier: GEN . PL all: NOM . PL crazed: NOM . PL PTC become: IMPF.3 PL
‘The soldiers who ate of the honeycomb (= specific) all went off their heads.’
(X. An. 4.8.20);
(4) polláki moi katédeusas. . . khitôna oínou apoblúzōn

often 1SG . DAT stain: AOR .2 SG tunic:ACC wine:GEN spit.out: PTCP. PRS . NOM
‘You often stained my tunic when you spat out wine.’ (Hom. Il. 9.490–491);
(5) en d’ Apóllōnos pólei nemomisménon

in PTC Apollo:GEN town:DAT establish:PTCP. PRF.NOM
estì krokodeílou phageîn pántōs hékaston
be:PRS .3 SG crocodile:GEN eat:INF. AOR in.all.cases each:ACC
‘In the town of Apollonopolis it is an established custom for every person
without exception to eat crocodile.’ (Plu. Mor. 371.D.5–6).
In (2) and (3) the NPs kreôn and tôn kēríōn refer to parts of specific entities: they
can be translated as ‘a part of that meat’, ‘some of that honeycomb’. In (4) and
(5), on the other hand, the NPs oínou ‘some wine’ and krokodeílou ‘some cro-
codile (meat)’ have generic reference: the partitive genitive does not denote a
part of a specific entity; rather, it functions as a marker of indefiniteness, and
indicates an indefinite quantity of an unbounded, non-specific entity.
Note that in examples (2) and (4) from Homer it is only the context that can
disambiguate between partitive construction and indefinite partitive: in the
Homeric epic the definite article is not yet fully grammaticalized. Consequently,
it is not always easy to determine if the referent of a noun is definite or indefinite.
Examples (3) and (5), on the other hand, show a different distribution of the
definite article with respect to the two types of partitive genitive in Classical
Greek, after full grammaticalization of the definite article. In (3), in which we
find a partitive construction, the article also occurs, while in (5), in which
the genitive functions as a marker of indefiniteness, we do not find the article.
According to Napoli (2010), who has analyzed the use of the partitive genitive
limited to consumption verbs, occurrence vs. non-occurrence of the definite article
indicates a distinction between partitive vs. pseudo-partitive construction (as
in ‘a cup of that coffee’ vs. ‘a cup of coffee’, in which the NP ‘of coffee’ is non-
referential; cf. Koptjevskaja-Tamm 2001, 2006). However, the above examples do
not straightforwardly support this hypothesis, as the usage of the partitive without
an article does not seem to be limited to pseudo-partitive constructions: indeed,
examples (4) and (5) do not contain pseudo-partitives. Rather, the genitive in-
dicates indefiniteness and non-specificity. In addition, the definite article with
partitive subjects seems to work in the contrary way, as it occurs with indefinite
partitives (see below, sec. 3).
2 Genitive second arguments

When the genitive is used as the second argument of a verb, one finds two dif-
ferent types of construction. In the first type, the genitive, though originally a
partitive, does not indicate that only a part of a referent is affected. Rather, the
occurrence of the genitive is conditioned by the verb. In Homer and, to a lesser
extent, in later authors, the genitive in such constructions usually alternates
with the dative or with a prepositional phrase. In addition, some verbs (typically
verbs of perception or mental activity) may occur with the genitive or with the
accusative, with no clear semantic difference that can be connected with the
partitive value of the genitive. All verbs in this group are low transitivity predi-
cates, which do not indicate a change of state. Low involvement is reflected by
low transitivity, and in particular by the absence of a change of state undergone
by a Patient: genitive second arguments are semantically Themes, rather than
Patients.
In the second type of construction, which is less frequent, the genitive can
occur with transitive verbs, some of which indicate change of state, and alternate
with the accusative. This alternation is semantically relevant: the genitive in such
occurrences indicates either low or partial affectedness, while the accusative in-
dicates high or total affectedness. This type of usage is more frequent in Homeric
Greek, and becomes less frequent in later authors. In principle, it is possible
with any transitive verb; in practice, however, it is most often attested with spe-
cific groups of verbs, such as verbs of consumption and verbs that mean ‘touch’/
‘reach’, such as lambánō. Genitive objects in this second type of construction
can be real Patients and undergo a change of state. Low involvement is substan-
tiated by the fact that only a part of their referent is affected.
2.1 Non-motivated alternation between the genitive and

other cases
As shown in examples (6)–(8), with certain verbs the dative or a prepositional
phrase with the dative can refer to the same extra-linguistic reality as the geni-
tive. Therefore, it is difficult to determine whether the meaning of the genitive
differs from that of other types of expression:
(6) hòs mega pántōn Argeíōn ḗnasse

REL . NOM mightily all: GEN . PL Argives: GEN . PL rule: IMPF.3 SG
‘Who ruled mightily over all the Argives.’ (Hom. Il. 10.32–33);
(7) . . . pâsin nekúessi kataphthiménoisin anássein

all: DAT. PL corpse: DAT. PL perish: PTCP. AOR . MID. DAT. PL rule: INF. PRS
‘. . . to be the lord over all the dead that have perished.’
(Hom. Od. 11.491);
(8) andrássin en polloîsi kaì iphthímoisin anássōn

man:DAT. PL in many: DAT. PL PTC brave:DAT. PL rule:PTCP. PRS . NOM . SG
‘Ruling over many brave heroes.’ (Hom. Od. 19.110).
Remarkably, the verb ánassein ‘reign’, ‘govern’ belongs to a group of verbs

that can take the locative in IE languages (see Brugmann & Delbrück 1911: 513
and Dahl, this volume with examples from Indo-Iranian). Such verbs often have
an animate second argument which does not undergo a change of state. Simi-
larly, verbs that take the dative often have animate second arguments and do
not indicate change of state, as the verb amúnō ‘help’ in example (9):
(9) ḕ Hellēnída eînai ḕ automoléein ek

or Greek:ACC . SG be:INF. PRS or flee: INF. PRS from
tôn barbárōn kaì autoîsi amúnein
ART.GEN . PL barbarian: GEN . PL and 3PL .DAT. PL help: INF. PRS
‘[The ship] was either Greek, or a fugitive from the barbarians helping them.’
(Hdt. 8.87.4).
In general, all verbs with a dative/genitive alternation are low transitivity predi-
cates, which indicate a low degree of affectedness, as the patient does not undergo
a change of state.
Examples (10) and (11) feature another kind of dative/genitive alternation,
based on the instrumental meaning of the dative:
(10) epeì k’ olooîo tetarpṓmestha góoio

When PTC dire:GEN enjoy:PRF. M / P.1 PL groan:GEN
‘When we have taken our fill of dire lamenting.’ (Hom. Il. 23.10);
(11) philótēti trapeíomen

love:DAT enjoy:AOR . MID.1 PL
‘We take our joy in love.’ (Hom. Il. 3.441).
The AG dative is a syncretic case, which resulted from the merger of the IE
dative, locative and instrumental. Alternation of the instrumental and the geni-
tive with the class of verbs discussed here is also quite typical of IE languages
(see Brugmann & Delbrück 1911: 583–584). Examples (10) and (11) show that
in AG the verb térpesthai ‘enjoy’ can take either an instrumental dative or a
partitive genitive (cf. English ‘satiate with/of’). Such alternation also occurs
with certain types of third arguments and adverbials (see section 4 below).
As in the case of the alternations with the dative, the alternation of the
genitive with the accusative does not seem to give rise to semantic differences
with certain types of verb, such as verbs of perception and mental activity:4
(12) tákha dè mnēsthai émellon Argeîoi . . .

soon PTC remember:INF. AOR be.about.to:IMPF.3 PL Argives:NOM . PL
Philokḗtao ánaktos
Philoctetes:GEN king:GEN
‘But soon the Argives would remember their king, Philoctetes.’
(Hom. Il. 2.724–725) ;
(13) Tudéa d’ ou mémnēmai

Tydeus:ACC PTC NEG remember:INF. PRF
‘But I don’t remember Tydeus.’ (Hom. Il. 6.222).
4 On genitive-accusative alternation with perception verbs, see further Luraghi and Sausa
(2012).
Occasionally, the genitive may indicate an entity which is fully affected by

the event denoted by the verb, as in (14). In such cases, the genitive alternates
with the accusative, which occurs in (15):
(14) laboménē tôn gounátōn

clasp:PTCP. AOR . NOM . SG . F ART.GEN . PL knees:GEN . PL
‘And she, clasping his knees. . .’ (Hdt. 9.76.11);
(15) . . . lábe dískon

clasp:AOR .3 SG disc:ACC
‘He clasped a disc.’ (Hom. Od. 8.186).
Even though synchronically there is no semantic difference connected with case

alternation in (14) and (15), a semantic motivation can be detected for the origin
of the usage of the partitive genitive here: partitivity in such occurrences is
connected with the fact that the entity is conceived as a part of a whole.
2.2 Semantically motivated alternation between the genitive

and the accusative
In some cases, mostly with verbs of consumption, the genitive/accusative alterna-
tion reflects an obvious difference in meaning, based on partial vs. total affected-
ness. This alternation can be understood in two different ways:
a) A referent is affected to a low extent, i.e. it does not undergo a change of
state. The verb can have different degrees of transitivity, which are specified
by the case: again, as with the verbs discussed in sec. 2.1, the genitive is
connected with low transitivity, while the accusative indicates high transi-
tivity and change of state.
b) Only a part of a referent is affected by an event denoted by a high transi-
tivity predicate, and undergoes a change of state, while the rest of it remains
unaffected. In this latter case, the difference between the genitive and the
accusative does not lie in different degrees of transitivity, but rather in the
extent to which a referent undergoes a change of state.
Partial affectedness as in (a) is shown in examples (16) and (17). Whereas

in (16) the accusative object indicates that a referent is actually reached, that
is, that the action denoted by the verb is carried out, and achieves its intended
effects, in (17) the genitive indicates the opposite: in fact, Hector does not reach
his boy, who gets scared at his father’s helmet plume (cf. Hom. Il. 6.467–470, and
see the discussion in Luraghi (2011: 337) regarding case alternation with this verb):
(16) hoppóterós ke phthêisin orexámenos

INDF. NOM . SG PTC overtake: SBJV. AOR .3 SG reach: PTCP. AOR . NOM
khróa kalón
flesh: ACC fair: ACC
‘W hich of the two will first reach the other’s fair flesh.’ (Hom. Il. 23.805);
(17) hṑs eipṑn hou paidòs oréxato

so say:AOR .3 SG 3 SG .GEN child:GEN reach: AOR .3 SG
phaídimos Héktōr
glorious: NOM Hector: NOM
‘So saying, glorious Hector tried to reach his boy.’ (Hom. Il. 6.466).
Examples (18)–(21) demonstrate partial affectedness and low involvement as

understood in (b): the referent of the direct object undergoes a change of state in
all occurrences, but (19) and (21) refer to a whole, while (18) and (20) only refer
to an indefinite quantity of the referent involved.
(18) óphra píoi oínoio

for drink:OPT.3 SG wine: GEN
‘I n order to drink some wine.’ (Hom. Od. 22.11);
(19) pine te oînon

drink:IMP.2 SG PTC wine:ACC
‘Drink the wine!’ (Hom. Od. 15.391);
(20) pleúsantes es Leukáda tḕn Korinthíōn

sail:PTCP. PRS . NOM . PL to Leucas: ACC ART. ACC Corinthian: GEN . PL
apoikían tês gês étemon
colony: ACC ART.GEN land: GEN ravage: AOR .3 PL
‘Sailing to Leucas, the colony of the Corinthians, they ravaged part of
the country.’ (Thuc. 1.30.2);
(21) hoi Ēleîoi . . . Lepreatōn tḕn gên étemon

ART. NOM . PL
Elean: NOM . PL Lepreate: GEN . PL ART. ACC land: ACC ravage: AOR .3 PL
‘The Eleans ravaged the land of the Lepreates.’ (Thuc. 5.31.3.)
In certain occurrences, with plural count nouns, the partitive genitive does
not indicate that an indefinite quantity of the entities that constitute the whole
is affected. Rather, it refers to a single, unspecified individual, as in (22):
(22) Adrḗstoio d’ égēme thugatrôn

Adrastos:GEN PTC marry: AOR .3 SG daughter: GEN . PL
‘He married (one) of Adrastos’ daughters.’ (Hom. Il. 14.121).
In such cases, too, the partitive genitive can alternate with the accusative, as
shown in (23). Note that gunaîka ‘a woman’ in (23) is also indefinite, and it does
not refer to a specific woman already known form the previous context:
(23) éntha d’ égēme gunaîka

there PTC marry:AOR .3 SG woman:ACC . SG
‘There he married a woman.’ (Hom. Od. 15.241).
Remarkably, reference to a single, indefinite entity belonging to a given set

is also possible with verbs that normally admit alternation of the genitive and
the accusative with no change of meaning, and do not indicate change of state,
such as perception verbs. Consider example (24):
(24) hōs ep’ exódōi klúō tôn éndothen khōroûntos

as at exit:DAT hear: PRS .1 SG ART.GEN . PL inside run: PTCP. PRS .GEN . SG
‘I hear as though [one] of them inside was running out at the exit.’
(S. El. 1322; from Nachmanson 1942: 17).
In (24) the fact that the genitive NP tôn éndothen ‘those inside (pl.)’ functions as
a partitive is shown by the occurrence of khōroûntos ‘running (sg.)’, which limits
the reference of the genitive NP to one.
As the examples show, the genitive typically occurs in the place of a direct
object: as already remarked, it indicates that a part of a referent undergoes the
effects of the event, possibly a change of state, indicated by the verb. As often
noted, the partitive indicates partial affectedness; crucially, however, in the
occurrences discussed in this section partial affectedness does not always coin-
cide with a low degree of affectedness. To the contrary, the part of the referent
which undergoes the effects of the state of affairs may be affected to any degree,
including high, as it can undergo a change of state: typically, partitive expres-
sions occur with verbs of ingestion, which imply that the referent of the direct
object is consumed.
To sum up, in AG the partitive genitive can indicate indefiniteness of direct
objects, but just sporadically (cf. examples (4) and (5)). This limitation partly
follows from the fact that, as we have seen above, the genitive also enters parti-
tive constructions; in addition, genitive direct objects are obligatory with numerous
verbs. Such constituents are real direct objects, as shown by possible passivization
(see Conti 1998, Luraghi 2010); typically, they occur with low transitivity verbs,
which do not indicate a change of state, such as verbs of perception, mental

activity, or verbs that mean ‘touch’ or ‘reach’. Some verbs allow for accusative/
genitive variation; in some cases, such variation triggers different meanings
of the verb, but in most cases factors that affect it are harder to indicate. In
any case, when different meanings are available in such contexts, the genitive
typically indicates lower transitivity and a lesser degree of affectedness. An
example is the verb orégein ‘reach (out)’ in Homeric Greek, which indicates that
the referent of the direct object is actually reached only with the accusative,
while the genitive indicates failure to accomplish the action (see the discussion
of examples (16) and (17) above). Indeed, partial involvement, or low partici-
pation, of the Patient, which is characteristic of the genitive, is generally con-
sidered a clear indicator of a low degree of transitivity.5 Other factors that also
determine transitivity, such as the degree of control by the Agent or Experiencer
over the event, the absence or presence of a negation in the sentence, and the
telic or atelic nature of the state of affairs described in the sentence, also seem
to have some influence over use of the AG genitive. The specific importance of
each of these factors and their possible interconnection, however, remain to be
determined.6
Both the verbs that only take the genitive and the low transitivity verbs that
allow for alternation with the accusative, here described in (a), are in sharp con-
trast with the verbs in (b), with which the genitive has a clear partitive inter-
pretation, and can indicate indefiniteness, as seen in sec. 3.1. Indeed, when the
genitive functions as a partitive or as a partitive construction it does not indicate
low transitivity or a low degree of affectedness. Rather, low participation of the
referent in the event is borne out by the fact that only a part of a certain referent
is affected by the event itself. The degree of affectedness can be high, and it can
imply a change of state, as with verbs of consumption. In the case of genitive/
accusative alternation, as shown in (18)–(21), there is no difference triggered by
degrees of affectedness: rather, the difference lies in the fact that the partitive
genitive signals that a referent is affected to the same degree as signaled by the
accusative, but that affectedness, albeit high, only concerns a part of it.
3 Genitive subjects
In AG the partitive genitive is used very sporadically as a (non-canonical) subject.
This use is attested after Homer; one possible occurrence is also attested in the
5 On the parameters that determine transitivity in a sentence, see Hopper & Thompson (1980).
6 For partial analyses of some of these factors in AG, see Conti (2002), (2010a), Napoli (2010)
and Riaño (2005).
Odyssey, depending on the interpretation. Undoubtedly, the phenomenon is an

IE legacy.7 Before describing this usage, we briefly summarize the properties of
canonical subjects in AG.
Prototypical subjects, canonically encoded in the nominative, display a
number of pragmatic, semantic and morphosyntactic properties that reflect both
their relevance in comparison to other arguments and their close ties to the verbal
predicate. This is shown by the tendency for subjects to be chosen as topic, and
take semantic roles such as Agent, Experiencer or Force.8
Syntactic properties of subjects are more difficult to pinpoint. In Greek,
control of reflexive pronouns, possible elision, and control of elliptical (null)
arguments, often taken as behavioral properties of subjects in other languages,
are not solely characteristic of subjects.9
In contrast, coding properties, in particular agreement in number and person
with the verbal predicate, do seem to differentiate the subject from other argu-
ments. Typically, nominative subjects agree with the verb. Although some sporadic
instances are found involving lack of agreement between a nominative subject
and the verb, as we will see in sec. 3.1, no agreement is documented between
the predicate and other arguments.10 Thus, nominative marking is characteristic
of canonical subjects in AG.
Constituents which, although not encoded in the nominative case, display
some subject properties, whether these are semantic, pragmatic or morpho-
syntactic, are considered non-canonical subjects. Remarkably, the genitive can
function as a non-canonical subject both in personal and in impersonal con-
structions in AG. The following sections illustrate its features.
7 In recent years, many works have been devoted to the analysis of non canonical subjects
in both Indo-European and non Indo-European languages. On this topic see, among others,
Aikhenvald & Dixon (2001), Baños (2003), Bhaskararao & Subbarao (2004), Barðdal (2006)
Barðdal & Eythórsson (2003, 2009) Eythórsson & Barðdal (2005), and Seefranz & Montag
(1983, 1984). Dahl (this volume) discusses partitive genitive subjects in Indo-Iranian.
8 On the functions of prototypical subjects cross-linguistically, see Keenan (1976), Keenan &
Comrie (1977) and Dik (1997: 357–358).
9 The syntactic properties of subjects as defined by Cole, Harbert, Hermon & Sridhar (1980) in
a typological paper remain a point of reference in the analysis of Indo-European languages. On
the peculiarities of subjects in AG, cf. Conti (2010b).
10 Note that, in argument clauses with the infinitive, the subject is encoded by the accusative
case. All other arguments remain unchanged: ho nomothétēs títhēsi toùs nómous ‘The policy-
maker (nom) establishes the laws (acc)’ vs. oîmai tòn nomothétēn toùs nómous tithénai ‘I
believe that the policymaker (acc) establishes the laws (acc)’. In the event of coreference with
any of the arguments of the main clause, the subject of the infinitive clause may be omitted, or
it can remain in the nominative, see Luraghi (1999).
3.1 Non-canonical subject in personal constructions
3.1.1 Syntactic properties
In contrast to other ancient and modern IE languages, in which partitive sub-

jects often do not agree in number and person with the predicate, in AG agree-
ment between genitive subjects and the predicate is quite systematic.11 Homer
gives us the first possible example of agreement between a genitive, in the plural,
and the verb, also in the plural. Leaving out other possible interpretations (on
which see Conti, 2010d), this passage might allows us to analyze the genitive as a
non-canonical subject. Due to the fact that in Homer the forms of autós are still
employed -leaving aside a few exceptions- with an emphatic value, and not as
forms of the third person anaphoric pronoun, autôn may be interpreted as
expressing the focus (Conti 2010d: 9–10):
(25) autôn gàr sphetárēisin atasthalíēisin ólonto

DEM .GEN . PL PTC own: DAT. PL wickedness: DAT. PL perish: AOR . MID.3 PL
‘They (viz. Odysseus’ comrades) perished, indeed, through their own blind
folly’ (Hom. Od. 1.7–9).
Thus far we have not been able to find any example of lack of agreement
between the genitive and the verb,12 except for some occurrences involving
forms of existential verbs such as e.g. eimí ‘to be, exist’. However, it must be
noted that for the verb eimí lack of agreement is not limited to genitive subjects:
rather, it can also be found with nominative (i.e. with canonically marked) sub-
jects. Compare example (26) with (27) and (28):
11 Comparison with other IE languages suggests that lack of agreement between the genitive
and the predicate precedes agreement between the two: non-canonical subjects acquire agree-
ment with the predicate as they gradually move closer to prototypical subjects. With regard to
this phenomenon, see among others Sasse (1982) and Seefranz & Montag (1984). On lack of
agreement between partitive subjects and verbal predicates in Latin, Germanic Slavic, see
Conti (2010a: 101–102 and fn. 18 and 19). He development outlined above is in line with the
‘behavior before coding’ principle (Haspelmath 2010), which predicts that behavioral properties
precede coding properties (e.g. agreement) when a new construction arises, as in this case the
partitive subject construction.
12 For the analysis of the genitive in subject function we selected works by Homer, Aeschylus,
Herodotus, Sophocles, Euripides, Thucydides, Aristophanes, Plato, Xenophon, Aristotle and
Plutarch. We also reviewed the contents of the Hippocratic Corpus.
(26) tôn dè polemíōn ên hoùs

ART.GEN . PL PTC enemy:GEN . PL be:IMPF.3 SG REL . ACC . PL
hupospóndous apédosan
under.truce:ACC . PL return: AOR .3 PL
‘And there were some of the enemies, who they returned under a truce.’
(Xen. Hell. 7.5.17);
(27) kaì éstin hoì etúnkhanon

and be:PRS .3 SG REL . NOM . PL reach:IMPF.3 PL
kaì thōrákōn kaì gérrōn
and breastplate:GEN . PL and shield:GEN . PL
‘And there were some that made it to breastplates and shields.’
(Xen. Cyr. 2.3.18);
(28) ésti dè heptà stádioi ex Abúdou es

be:PRS .3 SG PTC seven stadia:NOM . PL from Abydos:GEN to
tḕn apantíon
ART.ACC opposite.shore:ACC
‘There are seven stadia between Abydos and the opposite shore.’
(Hdt. 7.34.4).
Thus, rather than being an archaic feature of partitive subjects, this usage reflects
grammaticalization of the genitive in its function as non-canonical subject.13
In addition to agreement with the predicate in number and person, the
genitive also displays another interesting morphosyntactic property: it can head
attributive or predicative constituents in the nominative. This is the case with the
nominative participle thēreuthéntes in (29), which has a predicative function with
respect to the genitive autôn (see further the nominative mégiston in example (33),
which functions as an attribute of the genitive subject tês diaitētikês):
(29) eisì gàr autôn kaì parà basiléi tôn

be:PRS .3 PL PTC they:GEN . PL and by king:DAT ART.GEN . PL
Perséōn entheûten thēreuthéntes
Persian:GEN . PL there capture:PTCP. AOR . PASS . NOM . PL
‘There are (some) of these (sc. ants), captured there, even by the king of
the Persians.’ (Hdt. 3.102.2).
13 Lack of agreement between the subject and the predicate is not limited to eimí, but also
occurs with forms of gígnomai ‘become, exist, come into being’. An in-depth analysis of the
semantic and pragmatic peculiarities of both verbs in these contexts could turn up interesting
results.
Genitive subjects can also occasionally occur with the infinitive in comple-
ment clauses, in which one normally finds the accusative (cf. fn. 10), as shown
in the following example, in which the genitive pronouns sphôn and ekeínōn are
subjects of the infinitive epimignúnai:
(30) éphasan . . . epimignúnai sphôn te pros ekeínous

say:AOR .3 PL intermingle:INF. AOR 3PL .GEN PTC toward DEM . ACC . PL
kaì ekeínōn pros heautoús
and DEM .GEN . PL toward REFL . ACC . PL
‘They said that some of them (sc., the Carduchians) intermingled with
those people, and some of those people with them.’ (X. An. 3.5.16).14
Apparently, there is a difference in the use of the definite article with partitive
subjects and partitive objects. As remarked in sec. 1, when the partitive genitive
indicates indefiniteness it occurs without the definite article (cf. example (5)
above). On the other hand, the definite article tends to occur in partitive con-
structions. In the case of genitive subjects, apparently, indefiniteness can also
be indicated by genitive NPs with the definite article, as shown in example (31):
(31) en hósoisi toû liparoû enên

in INDF. DAT. PL ART.GEN fat.GEN be: IMPF.3 SG
‘There was fat in them (sc. the bones).’ (Hp. Carn. 4.6).
3.1.2 Semantic and denotational properties
The denotational properties of genitive subjects display no peculiarities in

comparison with genitive second arguments: case marking indicates that one
or several entities are partially involved, to an indeterminate extent, in the event
expressed by the verb.
From a semantic perspective, the genitive is generally associated with the
expression of semantic roles far removed from the Agent. Genitive subjects are
normally Patients or Themes, as in the examples already discussed, and in (32)
and (33):
14 The predicate allows for a reciprocal reading. It seems, then, that the degree of agency of
the subjects is not high.
(32) hṓste ouk apéthanon autôn

so.that NEG die: AOR .3 PL DEM .GEN . PL
‘So none of them (viz. of the Athenians) died.’ (X. Hell. 4.2.21);
(33) tês diaitētikês esti mégiston

ART.GEN diet:GEN be:PRS .3 SG great:SUP. NOM
paratēréein kaì phulássein
observe:PRS . INF and supervise:PRS . INF
‘Some aspects related to the diet are the most necessary things to observe
and supervise.’ (Hp. Acut. 22.1).
The genitive is also used as a subject of predicates expressing culmination,

such as those meaning ‘arrive’, ‘appear’, ‘disappear’, ‘fall asleep’, and so on,
which denote achievements. Such verbs are cross-linguistically most often un-
accusative, and their subjects are syntactically treated as patients (see Levin &
Rappaport Hovav 1995). An example is (34):
(34) en khṓrai épipton hekatérōn

in place:DAT fall:IMPF.3 PL both:GEN . PL
‘Men of both sides fell in their position.’ (X. Hell. 4.2.20).
Partitive subjects of unaccusative verbs also occur in Slavic and in Basque

(cf. Miklosich 1883: 357 on Slavic languages in general and Timberlake 2004 on
Russian; on Basque, see Etxeberria this volume and Ariztimuño this volume). In
other languages, both IE and non-IE, like Lithuanian and Finnish, use of
partitive expressions also extends to unergative verbs, but remains much more
frequent with unaccusatives (see Senn 1966: 392–395, Huumo 2003 respectively,
and Luraghi & Kittilä this volume).
3.1.3 Pragmatic aspects
From a pragmatic point of view, genitive subjects normally have a non-topical

function, as they refer to participants that are newly introduced into discourse.
Indeed, they often appear in presentative clauses, in accordance with the fact
that the partitive genitive is a marker of indeterminacy/indefiniteness. When the
genitive denotes several human entities, both the identity and the number of
individuals involved in the event are also indefinite. This feature contrasts with
typical characteristics of topical elements, which must be easily identifiable both
for the speaker and for the hearer.
The non-topical nature of genitive subjects is in accordance with their fre-

quent post-verbal position in the sentence, as seen in examples (29), (32) and
(34). As remarked above, items chosen as topics are typically definite, as they
have been introduced earlier in discourse, while partitive subjects are indefinite.
Indeed, post-verbal position is a frequent feature of indefinite subjects across
languages. For example, Lyons (1999: 88) remarks that in Chinese “[t]he verbs
that allow a post-verbal subject are those of appearance or location, some verbs
of motion and few other”: in other words, the same verbs that typically allow for
partitive subjects in AG. Lyons further observes that these verbs are unaccusa-
tive, and concludes that unaccusative verbs allow for post-verbal subjects in
many languages including English, under the constraint that the subject NP is
indefinite. In the Romance languages, in which the partitive article functions as
an indefinite, such indefinite and non-topical subjects are also typically post-
verbal, as shown in the following Italian examples (see further Carlier & Lamiroy,
this volume):
(35) Si è sparso del sangue.

REFL be:PRS .3 SG shed:PTCP. SG . M of.DEF. ART. SG . M blood( M ): SG
‘(Some) blood was shed.’
(36) Arrivano dei soldati.

arrive:PRS .3 PL of.DEF. ART. PL . M soldier( M ): PL
‘Some soldiers are coming.’
The genitive systematically denotes entities that do not participate in the

speech act: this is the pragmatic reflex of low involvement, or reduced participa-
tion, the main feature of partitive genitive arguments. As a consequence, parti-
tive genitive subjects are typically encoded in the third person, either singular or
plural. The non-topical nature of genitive subjects is also a reason for this: as is
widely known, first and second person pronouns are closely linked to expres-
sion of the topic. Indeed, speech act participants are often chosen as topics in
discourse.15
15 Note further that first and second person pronouns are located at the top of the animacy
hierarchy. The partitive genitive, in its turn, rather denotes entities that do not occupy the highest
position in The animacy hierarchy, but are located lower down on the scale, such as human
entities that are not speech-act participants, non-human animate entities and inanimate entities.
On animacy and topic > comment hierarchies, see among others Dik (1997: 357–358).
3.2 Negation
Handbooks of AG and other ancient IE languages highlight the frequent occurrence
of the partitive genitive in negative sentences. In fact, in Lithuanian and Slavic,
second arguments of verbs in negative sentences are, with few exceptions, encoded
through the partitive genitive (cf. Miklosich 1883: 498–499, 1966: 392–419 and
Timberlake 2004). Furthermore, in Slavic languages genitive subjects mostly
occur in sentences with negative polarity (cf. Miklosich 1883: 357–358 and Večerca
1993: 75). As shown in Miestamo (this volume), the occurrence of partitive cases
with negation is not limited to IE languages. However, it seems to be a phenom-
enon best detectable in (some) languages of Europe, as “[t]he requirement that a
case with a partitive function be used on NPs under the scope of negation is not
found outside the familiar European languages.” (See Etxeberria this volume,
and Aritmuño, this volume on Basque, and Miestamo this volume for examples
from Balto-Finnic languages.)
In AG occurence of the genitive under the scope of negation is sporadic, and
is apparently limited to genitive subjects.16 Example (32) and the passage cited
below in (37) provide examples of genitive subjects in negative sentences:
(37) pánta péphraktai kouk éstin opês

all:NOM . PL . N fence.in: PRF. M / P.3 SG and+not be: PRS .3 SG hole: GEN
oud’ ei séphrōi diadûnai
not if mosquito: DAT creep: INF. AOR
‘Everything is squeezed together and there is no room even for a mosquito
to go through.’ (Ar. Vesp. 352).
The tendency observed in languages of different genetic affiliation to employ

the partitive genitive, or the partitive case if available, as the subject of intransi-
tive verbs, and more specifically, unaccusative verbs (cf. Miestamo this volume)
under the scope of negation is explained in Conti (2010a) as follows. Unlike
canonical subjects, partitive subjects are especially characterized by a low
degree of agentivity. In intransitive clauses, sentence negation is a factor that
diminishes the subject’s agentivity. Indeed, sentence negation indicates that the
state of affairs described in the sentence cannot be brought about, and, conse-
quently, that the agent/subject cannot perform or carry out its potential agentivity.
Remarkably, the semantic effect achieved by the occurrence of the partitive
16 We did not find evidence of a similar tendency for the use of the genitive as second argument.
However, as far as we know an exhaustive analysis of all available data is needed to enable
definite conclusions to be made on this point.
genitive under the scope of negation is specular to the effect achieved by the
occurrence of partitive genitive objects in affirmative clauses. In the latter, the
partitive genitive indicates low affectedness of the Patient, whereas in the former
it indicates low agentivity. Thus, it is hardly surprising that negation and the
partitive genitive or the partitive case can occur in intransitive sentences featur-
ing a predicate that takes a non-agentive subject, and in transitive sentences in
which the verb does not indicate a change of state.
3.3 Impersonal constructions

The partitive genitive is also used in a small set of impersonal constructions.
In such occurrences, genitive NPs co-occur with NPs in the dative or the accusa-
tive, which denote animate, generally human beings. Genitive NPs in their turn
may indicate either animate or inanimate entities (see Conti 2010b, c). Typically,
dative or accusative NPs are Experiencers, while genitive NPs are Stimuli.
In recent studies on non-canonical, or non-nominative subjects, the construc-
tion described above has been extensively studied in numerous IE and non IE
languages (see for example the papers in Bashkararao & Subbarao 2004 and,
on IE languages, Barðdal & Eythórsson 2009 among others). Cross-linguistically,
it occurs with verbs of emotion, perception, mental activity, and physical sensa-
tion, that is, experiential predicates.17 The Experiencer NP is normally under-
stood as the non-canonical subject of these verbs, based on the fact that it tends
to show behavioral properties of subjects, though not being coded in the canon-
ical subject case (i.e. the nominative). However, genitive Stimuli also exhibit
some subject properties, as we will show below.
In AG, constructions of this type are restricted to verbs of feeling and verbs
of lacking or needing: mélei tiní (dat) tinos (gen) ‘there is care for something to
somebody’, ‘somebody cares for something’, metamélei tiní (dat) tinos (gen)
‘somebody repents for something’, deî tiní (dat) / tiná (acc) tinos (gen), ellépei
tiní (dat) tinos (gen) and khrḗ tiná (acc) tinos (gen) ‘there is need of something
for somebody’:
(38) Zēnì tôn sôn mélei pónōn

Zeus:DAT ART:GEN . PL POSS .2 SG .GEN . PL care:PRS .3 SG sorrow:GEN . PL
‘Your sorrows interest Zeus.’ (Eur. Heracl. 717);
17 See further Moreno (1990a) and (1990b), Bossong (1998), Bauer (2000) Haspelmath (2001),
Cuzzolin & Napoli (2008), and Barðdal & Eythórsson (2009). On Germanic languages, see
Barðdal (2006).
(39) autòn gár se deî promēthéōs

DEM . ACC . SG PTC 2 SG . ACC need:PRS .3 SG provident:GEN
‘You yourself need someone who thinks ahead.’ (A. Pr. 86).
Contrary to some other IE languages, notably Latin (Cuzzolin & Napoli 2008)
and Germanic (Barðdal 2006), in AG we are dealing with constructions that are
only documented sporadically. In Homer and Herodotus, these constructions are
even less frequent than in later authors. This fact is difficult to explain if we
accept, as comparison with other IE languages suggests, that verbs of emotion
already featured non-canonical subjects in PIE (see Wackernagel 1920: 117,
Brugmann 1925: 24–26, Hermann 1926: 290–291 and more recently Barðdal &
Eythórsson 2009). In addition, it must be borne in mind that in AG constructions
featuring non-nominative subjects are gradually replaced by canonical construc-
tions, with nominative subjects. We can therefore conclude that the constructions
that we are now analyzing are residual, although their evolution features some
rather disharmonious stages.18
Leaving aside the diachronic interpretation of these constructions in AG, the
characteristics of the genitive in the selected documents are described below.19
From a semantic point of view, the genitive denotes the triggering factor behind
the state of affairs described in the sentence, i.e. the Stimulus (see, among others,
Verhoeven, 2007: 54–55 on this semantic role). A loose relationship between the
notion of Stimulus and those of Agent and Force, all triggering factors of a
certain state of affairs, explains the existence of personal constructions such as
those shown in examples (40)–(41). In these passages, the Stimulus is coded in
the nominative; it corresponds to the genitive of the impersonal constructions
discussed in this section (see examples (42)–(44)):20
18 The prevailing world view in works by a given author and the stylistic resources that
characterize each literary genre also unquestionably act as conditioning factors, determining
the extent to which impersonal constructions are used. Therefore, it is not surprising that
impersonal constructions are much more frequent in tragedy, where reality is governed by
uncontrollable, distant forces, than in the Homeric poems, whose heroes and gods fight for
control over the events they experience.
19 Works analyzed include those of Homer, Aeschylus, Sophocles, Euripides and Aristophanes.
Among prose writers, works have been selected from Herodotus, Thucydides, Plato, Xenophon,
Demosthenes and Plutarch (Conti 2010b).
20 It can be argued that the dative Experiencer has subject properties in these occurences, as
in those in which the Stimulus is in the genitive, see below, the discussion of examples (43)
and (44). We do not pursue this issue further here, as it lies beyind the scope of the present
paper.
(40) eím’ Oduseùs Laertiádēs, hòs . . .

be:PRS .1 SG Odysseus:NOM son.of.Laertes:NOM REL . NOM
ánthrṓpoisi mélō . . .
man:DAT. PL care:PRS .1 SG
‘I am Odysseus, son of Laertes, of great interest to men.’ (Hom. Od. 9.19–20);
(41) . . . hōs autoîsi metamélēi pónos

that DEM . DAT. PL repent:SBJV. PRS .3 SG penalty:NOM
‘. . . that the penalty will make them repent.’ (Aesch. Eum. 771).
In so-called impersonal constructions, the genitive features one behavioral

property of subjects: it can be used in coordination with clauses that contain in-
finitives and accusative subjects. It is widely accepted that coordination between
two elements is only possible if their syntactic, semantic and pragmatic func-
tions are equivalent (cf. Dik, 1997: 189–190):
(42) ou nautikês kaì phaúlou stratiâs mónon

NEG fleet:GEN and insignificant:GEN army:GEN only
deî, allà kaì pezòn polùn xumpleîn
need:PRS .3 SG but also infantry:ACC large:ACC sail.with:INF. PRS
‘Not only are a fleet and an insignificant army needed, but also a large
infantry force to sail with them.’ (Th. 6.21.1).
In complement clauses, genitives in impersonal constructions are not elided

in the case that they are coreferential with the subject of the governing clause.
Canonical subjects are either elided, or they may surface in the nominative or in
the accusative (see Luraghi (1999) and above, fn. 9):
(43) póthen? Tí d’ autêi soû mélein

how why PTC 3SG .DAT. F 2SG .GEN matter:INF. PRS
dokeîs, téknon?
think:PRS .2 SG son:VOC
‘How is that? Why do you think, son, you matter to her?’ (Eur. El. 657).
In (43), the genitive does not function syntactically as a canonical subject.

The passage does not provide syntactic cues as to the status of the dative: note
however that in other occurrences the dative NP does in fact display the syn-
tactic behavior typical of subjects (cf. Conti 2010c: 263–264). Thus, the dative
may qualify as a non-canonical subject as well, as we will show below.
In terms of pragmatic properties, the data show that the genitive is usually
rhematic, and expresses the comment, rather than the topic. Generally speaking,
it is the dative that encodes the topic while the genitive, when placed in initial
position, is often the focus of the clause. In this passage it is the dative constitu-
ent that can be interpreted a non-canonical subject:
(44) pleiónōn dḕ geōrgôn te kaì tôn állōn

more:GEN . PL PTC laborers:GEN . PL PTC and ART.GEN . PL other:GEN . PL
dēmiourgôn deî hēmîn têi pólei
craftsman:GEN . PL need:PRS .3 SG 1 PL . DAT ART. DAT city:DAT
‘More laborers and all the other craftsmen are what our city needs.’
(Pl. Rep. 371.a).
In conclusion, subject properties of genitive NPs in personal and impersonal

constructions are partly different. In personal constructions, such as those dis-
cussed in sec. 3.1.1, the genitive displays two coding properties of prototypical
subjects: it agrees in number and person with the verb, and triggers nominative
agreement with attributive and predicative adjectives. Its semantic properties,
however, clearly differ from the properties that are most typical of subjects, i.e.
to express the Agent.
In impersonal constructions, in contrast, the genitive only displays one
behavioral property which is typical of subjects: use in coordination with control
infinitives. One significant feature is shared by the genitive of personal construc-
tions and the genitive of impersonal constructions, which keeps both of them
apart from prototypical subjects: the fact that they indicate discourse referents
which are not topics. Non-topicality can be seen as a discourse reflex of low
participation, as indicated by the partitive genitive.
4 Genitive third arguments

As already remarked in sec. 1, the partitive genitive can also function as the third
argument of certain three place predicates. Crucially, however, such third argu-
ments are never indirect objects, and do not occur with verbs of giving or saying.
This is an interesting restriction, as the indirect object function is apparently the
only syntactic function which is not accessible to partitive genitives. We cannot
offer an explanation for this restriction. Tentatively, we suggest that it might be
connected with the fact that indirect object is a syntactic function partly different
from others, as it is virtually always taken by NPs which refer to human entities.
Dative coding seems to be strictly associated with human referents that take
semantic roles typical of third arguments, such as Recipient and Addressee, as
well as neighboring roles, such as Beneficiary.
Basically, verbs which can have a genitive third argument belong to two
groups:
(a) Verbs referring to legal actions, like ‘accuse’ and ‘condemn’, as in (45).
Here, the occurrence of the genitive can hardly be explained from a
synchronic point of view:
(45) ḕn Lakedaimoníous tês exapátēs timōrēsṓmetha

if Spartan:ACC . PL ART.GEN trick:GEN punish:SBJV. AOR .1 PL
‘If we punish the Spartans for their trick. . .’ (X. An. 7.1.25.3).
(b) Verbs of filling and commercial transaction that can take either a genitive or
an instrumental dative. This group corresponds to two-place verbs such as
térpomai, described in sec. 2.1; case alternation is also attested in other IE
languages. An example of this alternation is given below:
(46) en d’ ōteilàs plêsan aleíphatos enneṓroio

in PTC wound:ACC . PL fill:AOR .3 PL ointment:GEN nine.years.old:GEN
‘They filled the wounds with an ointment that was nine years old.’
(Hom. Il. 18.351);
(47) kalámēs plḗsantes pân tò ploîon

reed:GEN fill: PTCP. AOR . NOM . PL all: ACC ART. ACC boat:ACC
toûto apieîsi kata tòn potamòn
DEM . ACC send.away: PRS .3 SG along ART. ACC river. ACC
‘They then fill all this boat with reeds and send it floating down the river.’
(Hdt 1.194.2);
(48) dakrúois gàr Hellád’ hápasan éplēsen

tear:DAT. PL PTC Greece: ACC all: ACC fill: AOR .3 SG
‘For she filled all Greece with tears.’ (E. Or.1363).
The IE origin of this usage of the partitive genitive can be demonstrated by

parallels with other languages, such as for example Latin:
(49) plenus timoris

full:NOM fear: GEN
‘full of fear’;
(50) plenus expectatione

full:NOM expectation: ABL
‘full of expectation.’
5 Adverbial
A peculiarity of the AG partitive is its adverbial usage, which also has some
parallel in the other IE languages, though to a limited extent. Adverbial genitives
occur in spatial and temporal expressions, mostly in Homeric Greek. In addition,
in Homeric Greek one can still appreciate the usage of the partitive genitive with
adpositions. This was apparently an innovation in AG, whose semantic impact
lost relevance after Homer.
5.1 Space and time

Below are some local expressions in the genitive:
(51) ê ouk Árgeos êen

PTC NEGA.:GEN be:IMPF.3 SG
‘Was he not in Argos?’ (Hom. Od. 3.251);
(52) hína mḗ . . . ḕ halos ḕ epì gês algḗsete

for NEG or sea: GEN or on land: GEN suffer: FUT.2 PL
pêma pathóntes
misery: ACC . PL endure:PTCP. AOR . NOM . PL
‘In order for you not to undergo trouble either at sea or on land.’
(Hom. Od. 12.26–27).
In such examples, the genitive could alternate with a dative locative, as in:
(53) patḕr d’ emòs Árgeï násthe

father:NOM PTC POSS .1 SG . NOM Argos:DAT abide:AOR .3 SG
‘My father lived in Argos.’ (Hom. Il. 14.119).
Elsewehere, it may alternate with a PP which basically indicates a locative, as

shown in (52), in which halòs is coordinated with an epí phrase.21
Even though such occurrences are relatively few, they are very important
when compared to occurrences of the so-called ablative genitive. Indeed, the
genitive without prepositions could express the meaning of an ablative only
under special contextual conditions, with verbs or adjectives that required such
meaning.22 When no such indication was available from the context, apparently,
the most readily available interpretation of a genitive was the partitive. This
is shown by example (51), in which we find the verb ‘be’. The meaning of a
dedicated ablative case here would be ‘being away from’: however, that the AG
genitive is not a dedicated ablative is shown by the fact that such a meaning can
only be expressed through a PP in Homer, as in (54):
(54) kaì gár tís th’ héna mêna

and PTC INDF. NOM PTC one:ACC month:ACC
ménōn apò hês alókhoio . . .
remain:PTCP. PRS . NOM from POSS .3 SG .GEN wife:GEN
‘For he that abides but one single month far from his wife . . .’
(Hom. Il. 2.292).
Similarly, example (52) does not contain any indication encoded by the verb
regarding the meaning of the spatial expression. The locative meaning of the
NP halós is encoded only by the genitive case.
As compared to other cases (the dative, which could functioned as a loca-
tive, and the accusative, which could appear in allative and perlative expres-
sions), the partitive genitive points toward the conceptualization of a portion of
space as constituted by subparts, separable from one another, among which an
entity can be located. This has consequences on the structure of possible trajec-
tories within genitive landmarks in PPs, as we will see in the next section.
21 The occurrence of the genitive in (52) is also remarkable in this respect. In Homeric Greek,
adpositional phrases were not yet fully grammaticalized as such (see Hewson & Bubenik 2004
for a thorough discussion); especially on the spatial plane, adpositions often specified mean-
ings that could be independently expressed by cases. This is especially clear in (52), where
both halós and gês indicate location, and epí adds the specification of a relation which holds
on the vertical axis. Interestingly, in similar occurrences with epí the dative also seems to
express pretty much the same meaning as the genitive, see Luraghi (2003: 298, 302–303) for
examples and discussion.
22 Remarkably, even with motion verbs the genitive does not per se express Source, and does
not function as an ablative: it does so only when the verb itself requires a Source expression.
Otherwise, the genitive indicates Direction, as does the dative with certain verbs of motion,
and as does the accusative. See Chantraine (1953: 52–53), and Luraghi & Sausa (2012).
After Homer, the partitive genitive in spatial expressions is no longer found.

However, time expressions also occur in Classical Greek, as in (55):23
(55) pínein te kaì eupathéein, oute hḗmerēs oute

drink:INF. PRS PTC and enjoy: INF. PRS NEG day: GEN NEG
nuktós aníenta
night: GEN let.go: PTCP. PRS . ACC
‘and would drink and enjoy himself, not letting up day or night,’
(Hdt. 2.133.4).
5.2 Complement of adposition

The occurrence of the partitive genitive with adpositions is an innovative feature
of Homeric Greek. In other ancient IE languages, the genitive does not normally
occur with adpositions, unless its origin is adnominal: an example is the genitive
with causā and gratiā in Latin. Remarkably, the independent meaning of cases
was still quite strong in Homeric Greek even with adpositions (the adpositional
phrase was still developing, see Hewson & Bubenik 2006, Luraghi 2010). The
result of the extension of the partitive genitive to adpositional phrases had the
effect that cases developed different oppositions between one another.
Inherited from PIE was the threefold opposition:
– dative ! locative
– accusative ! allative
– genitive ! ablative
The inherited spatial meaning of cases had the consequence the AG dative encoded
the locative relation (cf. example (53)). This fact depends on case syncretism, as
the dative was the merger of the PIE dative and locative. However, as we have
seen above, the independent spatial meaning of the partitive genitive is also
locative: it indicates an area in which a certain event takes place. In addition,
even the accusative could indicate a location, rather than a direction. When in-
dicating location, the accusative is generally said to occur in situations in which
a certain extension of space is envisaged, and often adds a perlative meaning,
as in:
23 The use of the partitive genitive in time expressions can be reconstructed for PIE, as it is
also attested in other ancient (and some modern) languages, for example in Germanic, cf.
Gothic nahts, Modern German nachts ‘at night’.
(56) póthen pleîth’ hugrà kéleutha; . . . mapsidíōs

whence sail:PRS .2 PL wet: ACC . PL path: ACC . PL randomly
alálēsthe . . . hupeìr hála . . .?
wander:PRS .2 PL . M / P over sea: ACC
“Whence do you sail over the watery ways? . . . do you wander at random
over the sea?” (Hom. Od. 71–73).
Example (56) is especially interesting when compared to (52): in both examples,

similar spatial relations are expressed twice, the second time with the addition
of an adposition that indicates a relation on the vertical axis. As the occurrence of
the adverb mapsidíōs ‘at random’ shows, the accusative profiles non-directional
movement on a surface: hence the name of ‘accusative of extension’ (see
Chantraine 1953: 45).
Especially when occurring with prepositions, the genitive and the accusative
indicate two different types of extended areas. The opposition between the two
cases can be understood as a spatial correspondence of the opposition between
partial and total affectedness with partitive genitive or accusative direct objects
of verbs of consumption. The partitive indicates a surface that can be divided
into parts; the accusative, on the contrary, indicates an extension which is
conceived as an indivisible whole. This difference affects the type of landmarks
that can occur in the two cases. The difference is neatly examplified by metá,
‘among’. With this preposition, the accusative only occurs with singular collec-
tive nouns or with plurals modified by the adjective pâs ‘all’. An example is:
(57) toîsi dè thumòn enì stḗthessin órine

DEM . DAT. PL PTC soul:ACC in breast:DAT. PL stir:AOR .3 SG
pâsi metà plēthún
all:DAT. PL among crowd:ACC
‘He moved the soul of everyone in the crowd.’ (Hom. Il. 2.142–143).
The genitive, on the other hand, only occurs with plural count nouns, as shown
in:
(58) hoì mèn . . . metà Boiōtôn emákhonto

DEM . NOM . PL PTC among Boeotian:GEN . PL fight:IMPF. M / P.3 PL
‘These were fighting among the Boeotians.’ (Hom. Il. 13.699–700).
In the above examples, landmarks are multiplex following the terminology in

Talmy (2000). However, while genitive landmarks are multiplex and discontinuous,
accusative ones are continuous. This means that genitive landmarks are formed
by sub-units which can be singled out separately, while accusative landmarks

cannot be further analyzed. Such difference is reflected by the occurrence of
plural count nouns (genitive) or collective or plural nouns with pâs (accusative).24
Another way in which the opposition between the genitive and the accusa-
tive can affect the conceptualization of the landmark is shown by the usage of
diá. This preposition has a perlative meaning ‘through’, and indicates that a
trajector is moving along a trajectory inside a landmark (without reference to
the initial or endpoint of the trajectory). With this prepostion, case alternation
alters the structre of the trajectory. The partitive genitive, which indicates a
surface that can be divided into separate units, indicates a unique trajectory,
along which the trajector can be traced down at any point in its movement. The
accusative, instead, indicates that the trajectory is internal to the landmark, but
cannot be traced in a precise manner. In much the same way as the ‘accusative
of extension’ in example (56), the accusative with diá indicates a random, non-
directional motion. Consider the examples:
(59) kephalḕn d’ hapalês apò deirês kópsen

head:ACC PTC tender:GEN from neck:GEN cut:AOR .3 SG
Oïliádēs . . . hêke dé min sphairēdòn
son.of.O.:NOM throw:AOR .3 SG PTC 3 SG . ACC like.a.ball
helixámenos di’ homílou
roll:PTCP. AOR . MID. NOM through crowd:GEN
‘The son of Oïleus cut the head from the tender neck, and with a swing
he sent it rolling through the throng like a ball.’ (Hom. Il. 13.202.204);
(60) helixámenos dià bḗssas

turn:PTCP. AOR . MID. NOM through glen:ACC . PL
‘(A wild boar) turning around through the glens.’ (Hom. Il. 17.283).
24 The dative also occurs with metá in Homeric Greek: indeed, it is the most frequent case.
It occurs 215 times, while the accusative occurs 164 times and the genitive, which was most
likely a recent innovation, only occurs five times, in occurrences similar to the one in (58). In
location expressions, the dative was not constrained by specific types of landmark: in particular,
it could occur both with count and with mass nouns. Thus, contrary to the other two cases, the
dative was underspecified regarding the mass/count distinction in location expressions. Note
that metá could occur with other types of expression, notably it could mean ‘between’ (only
with the dative), or it could indicate motion after a landmark (with the accusative). See Luraghi
(2003: 244–249) and (2005). Remarkably, the dative, which was the most frequent case in
Homeric Greek, is no longer used with metá in Classical Attic-Ionic.
In (59) and (60) the same verb form, helixámenos, indicates two different types
of motion. The head of the champion in (59), cut off from his neck, rolls on itself
along a straight trajectory inside an area defined by the crowd: here, the genitive
landmark is a surface which can be divided into parts. Hence, the trajectory can
be traced down. The wild boar in (60), instead, runs around in different direc-
tions among the glens. The accusative landmark does not allow for precise track-
ing of the trajector, and movement is performed at random. Thus, the difference
between diá with the genitive and diá with the accusative with motion verbs can
be captured in terms of trajectory structure: while diá with the genitive indicates
a single path, diá with the accusative indicates multiple path.25
A similar distinction appears when the two types of prepositional phrase
occur with verbs that denote static situations. Compare the following examples:
(61) polloì dè súes thaléthontes

many:NOM . PL PTC swine:NOM . PL bloom:PTCP. PRS . NOM . PL
aloiphêi heuómenoi tanúonto dià
grease:DAT singe:PTCP. PRS . M / P. NOM . PL stretch:IMPF. M / P.3 PL through
phlogòs Hephaístoio
flame:GEN H.:GEN
‘Many swine, rich with fat, were stretched to singe over the flame of
Hephaestus.’ (Hom. Il. 9.467–468);
(62) autàr ho Kúklōpas megál’ ḗpuen, hoí rhá

then DEM . NOM K.:ACC . PL loudly call:IMPF.3 SG DEM . NOM . PL PTC
min amphìs ṓikeon en spḗessi di’
3 SG . ACC around live:IMPF.3 PL in cave:DAT. PL through
ákrias ēnemoéssas
height:ACC . PL windy:ACC . PL
‘Then he called aloud to the Cyclopes, who dwelt round about him in
caves among the windy heights.’ (Hom. Od. 9.399–400).
In (61), diá with the genitive denotes a situation in which a (number of) straight
trajector(s) is stretched through an area identified by the flame. In (62) instead
diá with the accusative indicates that the units which constitute the trajector
are located randomly within a certain area (more details and other examples
can be found in Luraghi 2003 ch. 3.9).
25 In a limited number of occurrences, diá takes the directional accusative in Homer and indi-
cates movement across a landmark, see Luraghi (2012).
6 Recapitulation
In this paper, we discussed various usages of the AG genitive which may be
regarded as connected with its partitive value. We have shown that, when func-
tioning as a partitive, the genitive can take virtually any syntactic function,
except apparently that of third argument of verbs of giving and communication.
Accordingly, we have analyzed the semantic, syntactic and pragmatic features of
the partitive genitive in specific syntactic functions.
Originally, the partitive genitive indicates that an entity is made up of
divisible parts. With such an entity, it is conceivable that an event does not affect
all its parts. As a consequence, the partitive genitive indicates a low degree
of involvement, or a reduced extent of participation, of a certain referent in a
state of affairs. This general feature has different instantiations, depending on
other variables, as e.g. the fact that the partitive genitive functions as an object,
as a subject, or as an adverbial. In the case of partitive direct objects, low
involvement may mean that a referent does not undergo a change of state, or
that only a part of it does. These features are most clear with verbs that admit
case variation for direct objects, whereby accusative objects indicate high in-
volvement. Especially verbs that do not admit case variation and take partitive
objects are low transitivity predicates, which do not indicate change of state
(this is also true of some verbs that admit case variation).
Genitive subjects, too, typically occur with low transitivity predicates, most
often in presentative constructions or with unaccusative verbs. As they imply
indefiniteness and are only seldom clause initial, they are not topical: rather,
they introduce new participants into the discourse. In so-called impersonal con-
structions, typically with experiential predicates, genitive constituents display
some behavioral properties of subjects. In such constructions, they typically
indicate the Stimulus and co-occur with Experiencer NPs mostly in the dative
(less frequently in the accusative), which can also display some behavioral proper-
ties of subjects. Thus, both genitive and dative NPs with experiential predicates
qualify as non-canonical subjects.
In several languages, partitives shown a strict connection with negation.
This connection is also present in AG, even though to a limited extent: partitive
subjects can occur under the scope of negation, but they are far from being obliga-
tory, as the nominative case can occur as well, and is in fact more frequent.
As already remarked, partitive third arguments are infrequent: in particular,
they never occur in the function of indirect object, that is, with verbs of giving
or saying, with which the indirect object is typically human. Indeed, they are
virtually restricted to verbs of judging or condemning and to verbs of filling
and of economic transaction, with which they alternate with the instrumental
dative. Note that such third arguments are typically inanimate.
Partitive adverbials can indicate a point in time or in space, the latter usage
being virtually limited to Homer. They occur in passages in which the locative
dative would also be possible. In addition, the partitive genitive can also occur
in adpositional phrases, again mostly limited to Homeric Greek. With adposi-
tions, the partitive genitive tends to be interpreted either as a locative, again
competing with the dative, or as a perlative, which indicates unidirectional path.
In the latter case, we find an opposition with the accusative, which denotes multi-
directional path. The partitive nature of the genitive, when referring to space,
envisages a landmark as composed by detachable units. A trajectory moving on
such a landmark is conceived as moving along a clearly individuated trajectory,
hence the implication of unidirectionality, as opposed to the accusative, which
has no similar implications.
In conclusion, the partitive genitive shows features which are also found
in other Indo-European languages, and are typical of partitive cases in some
non-Indo-European ones, thus confirming various cross-linguistic tendencies of
partitives. Among them, the most notable is the fact that the partitive genitive
does not indicate the syntactic function taken by a NP: rather, it can occur in
virtually any syntactic function. This is at odds with what is normally considered
the function of case, and makes the partitive genitive even more remarkable.
References
Aikhenvald, Alexandra, R. M. W. Dixon & Masayuki Onishi (eds.) 2001. Non-canonical marking
of subjects and objects. Amsterdam: Benjamins.
Baños, José Miguel. 2003. Paenitet y los verbos impersonales de sentimiento en latín: Sintaxis y
pragmática del acusativo personal. In José Miguel Baños, Concepción Cabrillana, Esperanza
Torrego & Jesús de la Villa (eds.), Praedicativa, Complementación en griego y latín, 51–77.
Santiago de Compostela: Universidad Santiago de Compostela.
Barðdal, Johanna. 2006. Construction-specific properties of syntactis subjects in Icelandig and
German. Cognitive Linguistis 17(1). 39–106.
Barðdal, Johanna & Thórhallur Eythórsson. 2003. The Change that never happened: The story
of the oblique subjects. Journal of Linguistics 39(3). 439–472.
Barðdal, Johanna & Thórhallur Eythórsson. 2009. The origin of the oblique-subject construc-
tion: An Indo-European comparison. In Vit Bubenik, John Hewson and Sarah Rose (eds.),
Grammatical Change in Indo-European Languages, 179–193. Amsterdam: Benjamins.
Bauer, Brigitte. 2000. Archaic Syntax in Indo-European. Berlin & New York: Mouton de Gruyter.
Benedetti, Marina. 2006. Mehr als Passiv: Über einige Verbalmorpheme in altindogermani-
schen Sprachen. International Journal of Diachronic Linguistics and Linguistic Reconstruc-
tion, 3. 91–110.
Bhaskararao, Peri & Karumuri Venkata Subbarao (eds.). 2004. Non-Nominative Subjects.
Bossong, George. 1998. Le marquage de l’expérient dans les langues d’Europe. In Jack Feuillet
(ed.), Actance et Valence dans les Langues de l’Europe, 259–294. Berlin: Mouton de
Gruyter.
Brugmann, Karl. 1913. Griechische Grammatik II. München: C. H. Beck.
Brugmann, Karl. 1925. Die Syntax des einfachen Satzes im Indogermanischen. Berlin & Leipzig:
Mouton de Gruyter.
Brugmann, Karl & Berthold Delbrück. 1911. Grundriss der vergleichende Grammatik der
indogermanischen Sprachen, 2nd edn, vol. 2: Vergleichende Laut-, Stammbildungs- und
Flexionslehre. Strassburg: Trübner.
Carlier, Anne. 2007. From Preposition to Article: The Grammaticalization of the French Partitive.
Chantraine, Pierre. 1953. Grammaire homérique II: Syntaxe. Paris: Klincksieck.
Cole, Peter, William Harbert, Gabriella Hermon & S.N. Sridhar (eds.). 1980. The acquisition of
subjecthood. Language 56(4). 719–743.
Conti, Luz. 1998. Zum Passiv von griechischen Verben mit Gen. bzw. Dat. als zweitem Komple-
ment. Münchener Studien zur Sprachwissenschaft (MSS), 58. 13–50.
Conti, Luz. 2002. Kasussyntax bei Homer: Überlegungen zum adverbalen Akkusativ. In Heinrich
Hettrich & Jeong-So Kim (eds.), Indogermanische Syntax: Fragen und Perspektiven, 1–19.
Wiesbaden: Reichert Verlag.
Conti, Luz. 2010a. Synchronie und Diachronie des altgriechischen Genitivs als Semisubjekt.
Historische Sprachforschungen 121. 2008 (2010):94–113.
Conti, Luz. 2010b. Weiteres zum Genitiv als Semisubjekt im Altgriechischen: Analyse des Kasus
bei impersonalen Konstruktionen, Historische Sprachforschungen 122. 2009 (2010):182–
207.
Conti, Luz. 2010c. Análisis del dativo en construcciones impersonales: Los conceptos de sujeto
y de semisujeto en griego antiguo. Emerita 78(2). 249–273.
Conti, Luz. 2010d. Nota sobre Odisea 1.7. Exemplaria 14. 3–14.
Crespo, Emilio, Luz Conti & Helena Maquieira. 2003. Sintaxis del griego clásico. Madrid:
Gredos.
Cuzzolin, Pierluigi & Maria Napoli. 2008. An Overview of Impersonal Verbs in Indo-European. In
Rosemarie Lühr & Sabine Ziegler (eds.), Protolanguage and Prehistory: Proceedings of
the Twelfth Congress of the Indogermanische Gesellschaft in Krakow, 75–81. Wiesbaden:
Reichert Verlag.
Dik, Simon C. 1997. The Theory of Functional Grammar, Part 2, Complex and Derived Construc-
tions (ed. by Kees Hengeveld). Berlin & New York: Mouton de Gruyter.
Eckert, Rainer, Emilia Crome & Christian Fleckenstein. 1983. Geschichte der russischen Sprache.
Leipzig: VEB Verlag.
Eythórsson, Thórhallur & Johanna Barðdal. 2005. Oblique subjects: A common Germanic in-
heritance. Language 81(4). 824–881.
Haspelmath, Martin. 2001. Non-canonical marking of core arguments in European languages.
In Alexandra Aikhenvald, R. M. W. Dixon & Masayuki Onishi (eds.), 53–83. Amsterdam:
Benjamins.
Hermann, Eduard. 1926. Die subjektlosen Sätze bei Homer und der Ausdruck der Tätigkeit,
des Vorgangs und des Zustands, Nachrichten von der Gesellschaft der Wissenschaften zu
Göttingen. Philologisch-Historische Klasse, Heft 3.
Hettrich, Heinrich. 1990. Rektionaler und autonomer Kasusgebrauch. In Heiner Eichner &
Helmut Rix (eds.), Sprachwissenchaft und Philologie, Jakob Wackernagel und die Indo-
germanistik heute, 82–98. Wiesbaden: Reichert Verlag.
Hettrich, Hettrich. Some remarks on the adverbal genitive in Rigvedic Sanskrit. In Jared Tucker
& Elizabeth Klein (eds.), Proceedings of the 13th World Sanskrit Conference. Edinburgh,
Delhi: Motilal Banarsidas, in press.
Hewson, John & Vit Bubenik. 2006. From Case to Adposition: The Development of Configura-
tional Syntax in Indo-European Languages. Amsterdam: Benjamins.
Hopper, Paul J. & Sara A. Thompson. 1980. Transitivity in Grammar and Discourse. Language
56. 251–99.
Hualde, José Ignacio & Jon Ortiz de Urbina (eds.) 2003. A grammar of Basque. Berlin: Mouton
de Gruyter.
Keenan, Edward. 1976. Towards a universal definition of ‘subject’. In Charles N. Li (ed.), Subject
and Topic, 303–334. New York: Academic Press.
Keenan, Edward & Bernard Comrie. 1977. Noun phrase accessibility and universal grammar.
Linguistic Inquiry 8. 63–99.
Kiparsky, Paul. 1998. Partitive case and aspect. In Miriam Butt & Wilhelm Geuder (eds.), The
Projection of Arguments, 266–307. Stanford, California: CSLI.
Koptjevskaja-Tamm, Maria. 2001. “A piece of the cake” and “a cup of tea”: Partitive and
Maria Koptjevskaja-Tamm (eds.), The Circum-Baltic Languages: Typology and Contact,
vol. 2., 523–568. Amsterdam: Benjamins.
Kirschbaum, Ernst-Georg & Elizabeth Tauscher Kretschmar. (198013) Grammatik der russischen
Sprache. Düsseldorf: Brücken Verlag.
Kopjevskaja-Tamm Maria. 2006. Partitives. In Keith Brown (ed.), Encyclopedia of Languages
and Linguistics, vol. 9, 2nd edn., 218–221. Amsterdam: Elsevier.
Kühner, Rapahel & Bernhard Gerth. 1898. Ausführliche Grammatik der griechischen Sprache,
II.1. Hannover-Leipzig: Hahn.
Lasso de la Vega, José Luis. 1958. Genitivo partitivo sujeto en griego. Sociedad española de
Estudios Clásicos 2. 462–472.
Lestrade, Sander. 2006. Marked Adpositions. Paper presented at the SIGSEM Workshop on
Prepositions. University of Trento: University of Trento, 3–6 April, 2006.
Levin, Beth & Malka Rappaport Hovav. 1995. Unaccusativity. Cambridge, MA: MIT Press.
Luraghi, Silvia. 1999. The subject of complement clauses with the infinitive. In Bernard Jaquinod
(ed.), Les complétives en grec ancien, 199–213. St. Etienne: Publications de l’Université.
Luraghi, Silvia. 2003. On the Meaning of Prepositions and Cases: A Study of the Expression of
Semantic Roles in Ancient Greek. Amsterdam: Benjamins.
Luraghi, Silvia. 2005. The history of the Greek preposition metá: From polysemy to the creation
of homonyms. Glotta 81. 130–159.
Luraghi, Silvia. 2006. Greek prepositions: Patterns of polysemization and semantic bleaching.
In Emilio Crespo, Jesús de la Villa & Antonio Revuelta (eds.), Word Classes and Related
Topics, 487–499. Louvain-la-Neuve: Peeters.
Luraghi, Silvia 2009. The internal structure of adpositional phrases. In Johannes Helmbrecht,
Mouton de Gruyter.
Luraghi, Silvia 2010. The extension of the passive construction in Ancient Greek. In Acta
Linguistica Hafniensia 42(1): 60–74.
Luraghi, Silvia. 2011. Two theoretical approaches to cases in comparison. In Thomas Krisch
& Thomas Lindner (eds.), Akten der 13. Fachtagung der Indogermanischen Gesellschaft,
331–341. Wiesbaden: Reichert.
Luraghi, Silvia. 2012. The spatial meaning of diá with the accusative in Homeric Greek.
Mnemosyne 65(3). 357–386.
Meillet, Antoine & Joseph Vendryes. 1927. Traité de grammaire comparée. Paris: Champion.
Miklosich, Franz. 1883. Vergleichende Syntax de slavischen Sprachen 4: Syntax. Wien: Brau-
miiller.
Moreno, Juan Carlos. 1990a. Universal and typological aspects of impersonality. In Werner
Bahner, Joachim Schildtand & Dieter Viehweger (eds.), Proceedings of the Fourteenth
International Congress of Linguists, III, 2396–2398. Berlin: Academy Verlag.
Moreno, Juan Carlos. 1990b. Processes and actions: internal agentless impersonals in some
European languages. In Johannes Bechert, Giuliano Bernini & Claude Buridant (eds.)
Toward a Typology of European Languages, 255–272. Berlin: Mouton de Gruyter.
Nachmanson, Ernst. 1942. Partitives Subjekt im Griechischen. Göteborg: Elanders boktryckeri
aktiebolag.
Napoli, Maria. 2010. The case for the partitive case: The contribution of Ancient Greek. Trans-
actions of the Philological Society 108(1). 15–40.
Riaño, Daniel. 2005. La sintaxis de los verbos ‘‘comer’’ y ‘‘beber’’ en griego antiguo: un estudio
sobre el genitivo partitivo. Emerita 73. 263–302.
Sasse, Hans Jürgen. 1982. Subjektprominenz. In Sieglinde Heinz & Ulrich Wandruszka (eds.),
Fakten und Theorien: Beiträge zur romanischen und allgemeinen Sprachwissenschaft.
Festschrift fur Helmut Stimm zum 65. Geburtstag, Tübingen: Narr, 266–86.
Schwyzer, Eduard & Albert Debrunner. 1950. Griechische Grammatik II. München: C.H. Beck.
Seefranz-Montag, Ariane von. 1983. Syntaktische Funktionen und Wortstellungsveränderung.
Die Entwicklung ‘subjektloser’ Konstruktionen in einigen Sprachen. München: Fink.
Seefranz-Montag, Ariane von. 1984. Subjectless Constructions and Syntactic Change. In Jacek
Fisiak (ed.), Historical Syntax, 521–553. Berlin: Mouton de Gruyter.
Senn, Alfred. 1966. Handbuch der litauischen Sprache I. Heidelberg: Carl Winter Verlag.
Večerca, Radoslav. 1993. Altkirchenslavische (altbulgarische) Syntax II. Freiburg: Weiher.
Wackernagel, Jacob. 1920. Vorlesungen über Syntax I. Göttingen: Akademie der Wissenschaften
zu Göttingen.
Wandruszka, Ulrich. 1981. Typen romanischer Subjektinversion. In Horst Geckeler, Brigitte
Schlieben-Lange, Jürgen Trabant & Harald Weydt (eds.), Logos semantikos: Studia linguis-
tica in honorem Eugenio Coseriu (1921–1981) IV, Berlin: Mouton de Gruyter.
14 The grammaticalization of the
prepositional partitive in Romance
The paper is devoted to the adverbal use of the partitive in Romance. Romance
languages developed a partitive marker out of the Latin adposition de (+ NP),
whose original meaning was ‘away from’. This grammaticalization process is
represented in a five-step model, defining the different stages of the shift of
morpho-syntactic categorization, from Late Latin onwards: primitively a preposi-
tion, de (+ NP) turns into an article. We show that, although the source expres-
sion of the partitive is the same in French, Italian and Spanish, the outcome of
the evolution differs significantly according to the language. As to Spanish, we
find in Old Spanish a hybrid use of de, between preposition and determiner,
with a proper partitive meaning, but the grammaticalization process stops at
that stage. In French and Italian, on the contrary, de turned into a full-fledged
indefinite article, thus changing its morpho-syntactic status as well as its mean-
ing. Italian however differs from French in that the process has not fully reached
its endpoint, as the partitive article remains optional (next to zero marking) and
it is more widely spread in the North than in the South. The authors argue that
the different patterns which characterize the emergence of the partitive in
Romance can be linked to global typological properties regarding word order
and information structure.
Keywords: French, Spanish, Italian, Late Latin, grammaticalization,

adpositions, indefinite article
1 Introduction
From a syntagmatic viewpoint, the Romance partitive is used in two configurations:
– in the adnominal use, the partitive is linked to a nominal or pronominal
quantifying expression (e.g. Fr. Un morceau du gâteau ‘a piece of the cake’
cf. Koptjevskaja-Tamm 2001);
– in the adverbal use, the partitive is used for the expression of core arguments,
without any quantifier (e.g. Fr. Je bois du café ‘I drink coffee’).
This paper will be devoted to the latter case.

478 Anne Carlier and Béatrice Lamiroy
1.1 The partitive in Indo-European

The adverbal use of the partitive appears to be an endemic feature of Indo-
European, since it appears in languages or language stages belonging to different
branches of the Indo-European family, such as Sanskrit, Slavonic, Old Germanic,
Gothic, and Ancient Greek. Contrary to languages such as Finnish or Basque,
endowed with a partitive case, Indo-European languages however do not have
a specific partitive marker, but use either the genitive case or – especially in
language stages where nominal declension is weakening or is missing – an
adposition meaning primitively ‘away from’.
Cases, generally speaking, indicate dependency relations, both syntactic
and semantic, with respect to another term. For instance, the genitive typically
marks a dependency relation with respect to a noun. The partitive use of the
genitive, as observed in several Indo-European languages, is atypical, because,
unlike other inflectional cases, it does not create a specific relationship between
the inflected NP and some external element (Carlier 2007). This flexibility explains
why it can be used instead of other inflectional cases (for a detailed discussion,
see Humbert 1960, Luraghi 2009, and Conti & Luraghi, this volume with respect
to Greek, and Serbat 1996 for Latin). The Homeric epics provide some nice illus-
trations of the syntactic flexibility of the partitive genitive: it occurs not only in the
object position of verbs meaning ‘to drink’ or ‘to eat’ (1a), but is also used in
other syntactic functions such as subject, locative (2) or instrumental function.
(1) a. Haímatos óphra píō.

blood:GEN . SG ( N ) in.order.that drink: PRS .1 SG
‘So that I drink of the blood.’ (Hom.Od. 11.96)
b. Epeì píen haíma kelainón.

after drink:INF. AOR blood:ACC . SG ( N ) black:ACC . N . SG
‘After having drunk the dark blood.’ (Hom. Od. 11.98)
(2) Loessámenos potamoĩo. [instead of: en tō(i) potamō(i) [DATIVE ]]

bathe:PTCP. MID. AOR . M . SG river: GEN . SG ( M )
‘After taking a (little) bath in the river.’ (Hom. Il. 21.560)
The partitive genitive is however not a syntactic ‘joker’ (Meillet & Vendryes
1927: §797, Serbat 1996), i.e. its use instead of another inflectional case is not
indifferent: the partitive genitive marks an operation within its constituent,
which consists in isolating an indeterminate quantity from a whole. Humbert
(1960: 269–70) explains the difference between the two examples in (1) along
The grammaticalization of the prepositional partitive in Romance 479
this line: the first example, with the partitive genitive, relates the desire of
Tiresias to drink some of the blood of the victims killed by Odysseus, whereas
the second example, with the accusative, evokes the strength he draws from
drinking the substance of blood.
1.2 The partitive in Latin

In Classical Latin, the partitive genitive is commonly used for adnominal com-
plements. It indicates the definite or indefinite whole from which a fraction is
isolated:
(3) Multum operae.

much:N . SG time:GEN . F. SG
‘A lot of work’ (Cic., Brutus, 89, LXXXIX, 304)
(4) Magna copia frumenti.

great:NOM . F. SG abundance:NOM . SG ( F ) corn:GEN . SG ( N )
‘Great supplies of corn’ (Civ. 1, 52, 4)
Interestingly, the prepositional phrase headed by ex or de ‘from’ already enters

in competition with the genitive case in this context:
(5) Unus ex capitivis.

one:NOM . M . SG of prisoner:ABL . PL
‘One of the prisoners’ (Caes. Gal. 6.35.8)
(6) Nulla de virtutibus tuis plurimis.

none:NOM . F. SG of virtue:ABL . PL ( F ) POSS .2 SG .ABL . PL very.numerous:ABL . PL
‘None of your very numerous qualities.’ (Cic. Pro Q. Ligario, 37)
As to the adverbal partitive genitive, it is sporadic in Latin, to the point that

it would probably have gone unnoticed if it were not rather widespread in other
Indo-European languages such as Homeric Greek, Sanskrit and Slavonic. The
partitive genitive is nevertheless attested in Pre-Classical Latin, mainly in non-
literary, technical texts, such as medical and culinary treatises. Witness the
following example, quoted from Väänänen (1981), where the object of addito
‘add’ is expressed by the partitive genitive, whereas the object of indito ‘put in’
has the canonical form corresponding to its syntactic function, i.e. the accusa-
tive case:
(7) Farinam in mortarium indito;

flower:ACC . SG ( F ) in mortar:ACC . SG ( N ) put.into: IMP. FUT.2 SG
aquae paulatim addito.
water:GEN . SG ( F ) little.by.little add:IMP. FUT.2 SG
‘Put the wheat in the mortar; add (some) water little by little.’
(Cato Agr. 74, 2nd c. BC, quoted from Väänänen 1981)
The tendency to make use of a partitive genitive instead of another case is

repressed in the Classical Latin period, privileging the marking of clear syntactic
relations over the expression of subtle semantic distinctions. But the partitive
construction surfaces again in Late Latin, not only in the form of the genitive
case but also as a prepositional construction with de. Examples are legion in
the popularizing texts of the 4th and 5th century written in Gaul, in particular
by Christian authors.
A remarkable fact is noted by Väänänen (1981) on the basis of the inventory
of all the occurrences of the prepositional partitive construction in the Vulgate
Bible: in a language (or language stage) lacking a grammaticalized article, all
but two occurrences consist of a noun which is either preceded or followed by
a demonstrative or a possessive determiner (8b–8c) or followed by a relative
clause (9) that indicates the spatiotemporal location of the partition set denoted
by the noun. Hence, we find the following distribution between accusative, on
the one hand, and partitive genitive, on the other hand:
– the accusative is normally used when no such partition set is available (8a);
– the partitive construction with the preposition de, ex or ab can be used
when there is a contextually specified partition set (8b–8c–9).
(8) a. Ut comedatis carnem et bibatis

so.that eat:SBJV. PRS .2 PL meat:ACC . SG ( F ) and drink:SBJV. PRS .2 PL
sanguinem.
blood:ACC . SG ( M )
‘So that you eat my flesh and drink my blood.’ (Ezechiel 39,17)
b. Et sic de pane illo edat.

and thus of bread:ABL . SG ( M ) DEM .ABL . M . SG eat:SBJV. PRS .3 SG
‘And so let him eat of that bread.’ (Vulgata, I Corinthians 11, 28)
c. Comede de venatione mea.

eat:IMP. PRS .2 SG of venery:ABL . SG ( F ) POSS .1 SG . ABL . F. SG
‘Eat of my venery.’ (Vulgata, Gen. 27, 19)
(9) Nam et catelli edunt de micis quae

for and puppy:NOM . PL ( M ) eat:PRS .3 PL of crumb:ABL . PL ( F ) REL . NOM . F. PL
cadunt de mensa dominorum suorum.
fall:PRS .3 PL from table:ABL . SG ( F ) masters:GEN . PL ( M ) poss.GEN . M . PL
‘Yet the young dogs eat of the crumbs which fall from their masters’ table.’
(Vulgata, Matthew 15:27)
The expression of the object by means of a PP introduced by de instead of a

prepositionless NP in the accusative, as exemplified in the above examples, is
the common starting point for the evolution of the partitive in Romance.
1.3 The partitive in Romance
1.3.1 The different stages of the grammaticalization process:

from preposition to article
In the Romance area, the preposition de, combined with the definite article (de-
rived from the Latin demonstrative ille), grammaticalized into a partitive article.
The different stages of the grammaticalization process from Latin to French are
represented in Figure 1. The label “partitive” applies from stage III onwards.
Crucially, the emergence of the partitive article has to be conceived as a shift
in morpho-syntactic category: de, a preposition, turns into an article. Therefore,
for each of the stages, we will specify the semantic and syntactic features that
reveal to which paradigm de belongs and to which degree it is integrated in the
paradigm. Importantly, it has to be pointed out that the shift from one stage to
another does not mean that the earlier stage is discarded. According to Hopper
(1993), grammaticalization phenomena often display layering, i.e. the coexis-
tence of more conservative and more innovative evolutionary stages. This means
for example that when de + NP reaches stage III, usages that were typical of
stage I or II still may occur.
S TAGE I
– Syntax: The PP headed by de has the status of an adjunct with respect to
the verb.
– Semantics: The preposition de has originally a spatial meaning: it heads an
NP that introduces a reference point and, in its primitive meaning, marks a
downward movement (10). As a result of desemantization, the notion of
downward movement will progressively fade out, so that Latin de can also
simply express a movement away from a reference point, as in (11).
482
Figure 1: Grammaticalization chain: from preposition to partitive article (Carlier 2007)

(10) Decido de lecto.

fall.from: PRS .1 SG from bed:ABL . SG ( M )
‘I fall from my bed.’ (Latin: Plautus, Casina 931)
(11) De provincia clam abire.

from province:ABL . PL ( F ) secretly go.away:INF
‘Leave in secret the province.’
(Latin: Cic, Verr. II, 2,55)
– Paradigm: the Latin preposition de alternates with zero, on the one hand
(the prepositionless ablative also expresses distancing), and with preposi-
tions such as ex + ablative ‘out of’ and ab + ablative ‘from’, on the other.
In the case of ex, the reference point is conceived of as a containing space,
and the preposition indicates a movement out of it, e.g. profluit ex monte ‘it
flows out of the mountain’. In the case of ab, the starting point of the move-
ment is located outside of the reference point, e.g. aps te abire ‘go away
from you’ (Plautus, Miles gloriosus 4,1), caput a cervice revulsum ‘the head
pulled off from the neck’ (Vergilius, Georgica 4, 523). Contrary to ex and ab,
de does not specify the position of the moved entity with respect to its refer-
ence point, but at least in its primitive meaning, it describes the pathway
as a downward movement. In the evolution from Latin to Romance, de
will replace ex and ab in most Romance languages. In Italian the situation
is more complex, as Italian developed two distinct prepositions, viz. di
(<Lat. de) and da (<Lat. de + ab). However, and interestingly, only di will
evolve into a partitive article1.
S TAGE II
– Syntax: de, still a preposition, introduces an argument of the verb, giving it
the status of an oblique complement.
– Semantics: de gradually loses its spatial meaning. This process is ongoing
in Latin, where de has developed a variety of non-spatial meanings such as
lineage, extraction and partition, and temporal meaning. In a more advanced
stage, the same desemantization process even yields uses where de combines
1 The distinction between the ablative case and the genitive case in Latin is expressed in
Italian by means of the distinction between the preposition da and di. The fact that di gives
rise to the partitive article in Italian shows that the partitive replaces the Latin genitive rather
than the Latin ablative, as was already suggested by Väänänen (1981). As to Spanish and
French, they did not maintain the distinction between the Latin ablative case and genitive
case. On the contrary, this distinction is blurred, because “de + ablative” used for adjuncts in
Latin enters into competition with the genitive and eventually replaces it (Carlier, Goyens &
Lamiroy, 2013).
with verbs expressing a spatial movement towards the reference point, rather
than away from it. Witness the following Old French example:
(12) Il approucha de la dicte

PRO.3 SG . NOM . M approach:PST.3 SG from DEF. ART. F. SG say:PTCP. PST. F. SG
fontaine.
fountain:SG ( F )
‘He approached the fountain.’ (Old French: Jean d’Arras, Mélusine,
end 14th C.)
– Paradigm: there is a reduction of the paradigm, the only other preposition

introducing an oblique complement in Romance languages being a (<Latin
ad / ab, cf. Goyens, Lamiroy & Melis 2002).
S TAGE III
– Syntax: de is no longer a full preposition. As pointed out by Lehmann
(1982/2002: 75) and by Melis (2003), a preposition is a two-sided relator:
f on the one hand, it establishes a relationship between the PP it heads
and some external element, for instance the verb;
f on the other hand, it behaves as a preposition within the PP insofar as it
governs the NP.
In stage III, de evolves from a two-sided towards a one-sided relator:

f De + NP no longer has the status of a PP with respect to some external
element. It can be a direct argument of the verb, for instance a direct
object of a transitive verb (Spanish ex. 13) or a nominal predicate of the
verb be (French ex. 14). It can even combine with a preposition (Italian
ex. 15):
(13) Cogió del agua en él e a

take:PST.3 SG of.DEF. ART. F. SG water:SG ( F ) in PRO.3 SG . M and to
sus primas dio.
POSS .3. PL cousins:PL ( F ) give:PST.
‘He took some water into it [his hat] and gave it to his cousins.’
(Old Spanish, end 12th C. Cid 2801)
(14) Blancandrins fut des plus saives paiens.

B.:NOM ( M ) be:PST.3 SG of.DEF. ART. PL most wise:PL heathen:PL
‘Blancadrin was amongst the wisest heathens.’
(Old French: Roland [1100], v. 24)
(15) Abbiamo affitato una casa con

have:PRS .1 PL rent:PTCP. PST. M . SG one:F. SG house:SG ( F ) with
dei nostri cugini.
of.DEF. ART. M . PL POSS .1 PL . M . PL cousin:PL ( M )
‘We rented a house with our cousins.’
(Italian, quoted from Korzen 1996: 494)
f On the other hand, de still acts as a preposition with respect to the NP it

governs. This appears clearly in examples such as (16), where the NP
corresponds to a pronoun2.
(16) Seignors, du vin de qoi il

Lord:PL of.DEF. ART. M . SG wine:SG ( M ) of REL . N . SG PRO.3 PL . NOM
burent avez oï.
drink:PST.3 PL have:PRS .2 PL hear: PTCP. PST
‘Lords, you heard about the wine of which they drank.’
(Old French, 13th C.: Béroul, Tristan & Iseut, 2133–2135)
– Semantics
The prepositional status of de with respect to the nominal constituent has a
semantic corollary: the NP refers to a contextually (deictically or anaphori-
cally) defined partition set and de isolates an indeterminate part of it. For
instance, with respect to example (13), del agua isolates a portion of a parti-
tion set previously mentioned in the text as la linpia fuont ‘the clear source’.
Hence, de acquires a quantitative meaning which foreshadows its status of
determiner. This specific meaning of stage III explains why the partitive fre-
quently, though not exclusively (cf. examples (14) and (15)), occurs in object
position of verbs meaning ‘to eat’, ‘to drink’, ‘to take’ or ‘to give’. The high
incidence of stage III partitives in culinary and medical treatises containing
recipes therefore comes as no surprise. For the same reason, stage III parti-
tive only occurs in combination with concrete nouns, i.e. mass terms or
countable nouns, and is not attested with abstract nouns.
– Paradigm: as in stage III the partitive has lost its role of relating to an
external element, for instance a verb, it only alternates with zero (and no
longer with any other preposition).
2 Cf. Carlier (2007: 16–22) for a detailed syntactic argumentation.

S TAGE IV
– Syntax:
f The prepositional status of de is completely lost and the shift from prepo-
sition to article is accomplished. Not only does de combined to an NP
no longer confer the status of an indirect or prepositional constituent
to this NP with respect to some external element (typically a verb), but
moreover, within the constituent, de does not behave as a preposition
anymore with respect to the NP.
f As to its syntactic function, de + NP still dominantly occurs in object
position, but spreads to other verbs than those meaning ‘to eat’, ‘to
drink’, ‘to take’ or ‘to give’, e.g. avoir ‘have’ in (17b). If we admit that
the presence of an article is a sign of the independence of the object
with respect to the verb, this extension of the partitive article at the
expense of zero determination can be conceived as a progress in the
marking of transitivity along two axes: affectedness and individuation
(Hopper & Thompson 1980, Lazard 1996). In stage III, only highly individ-
uated objects, definite or singular, were introduced by an article, whereas
less individuated objects were not, unless they were highly affected.
In stage IV, even objects that have a low degree of individuation and
affectedness may be introduced by the partitive:
(17) a. Pren des grains de poyvre.

take:IMP.2SG of.DEF. ART. PL grain:PL ( M ) of pepper:SG (M )
‘Take some peppercorns.’ (Old French translation of Albertus Magnus,
De falconibus, BNF fr. 2003, 15th C).
b. quant le faulcon a des pouez.

when DEF. ART. M . SG falcon:SG ( M ) have.PRS .3 SG of.DEF. ART. PL louse:PL ( M )
‘When the falcon has lice.’ (Old French translation of Albertus
Magnus, De falconibus, BNF ms. fr. 25342, 15th C).
Zero marking is however maintained when the object is an integral part

of the verbal expression (e.g. avoir lieu ‘take place’, avoir mestier ‘need’),
because in this case there is no real transitivity.
– Although the object position remains dominant, de + NP is more fre-
quently attested in other syntactic functions. In example (18a), de + NP
occurs in subject position3. De can even occur after a preposition (18b):
3 As pointed out by Luraghi (Unpublished ms.), the progressive extension of the partitive to other
syntactic functions follows a pathway from object to subject in a certain order: first to subject of
unaccusative predicates which in fact is object (typically existential constructions, as i a ‘there
are’ in (18a)), then to subject of unergative predicates and finally to subject of transitive verbs.
(18) a. Il trova .i. ostel en selve

PRO. NOM . 3SG . M find:PST.3 SG one(M ) dwelling:SG ( M ) in forest:SG ( F )
clere: De sains moines i a
sparse:SG . F of saint:M . PL monk:PL ( M ) there have. PRS .3 SG
de sa contree Qui por l’ amor
of POSS .3 SG . F. SG region:SG ( F ) REL for DEF. ART. M . SG love:SG ( M )
de Dieu bien l’ ostelerent.
of God:SG ( M ) well PRO.3 SG . ACC lodge:PST.3 PL
‘He found a dwelling in a sparse forest: there were some saint
monks from his land that lodged him for God’s sake.’
(Old French: Aïol, Eds J. Normand et G. Raynaud, p. 23)
b. Et le lendemain le fault tresbien

and the following.day PRO.3 SG . ACC . M must:PRS .3 SG very.well
oindre avecques du savon.
rub:INF with of. DEF. ART. M . SG soap:SG ( M )
‘And the following day, you have to rub him very well with soap.’
(Old French translation of Albertus Magnus, De falconibus, BNF
ms. fr. 1304, 16th C.)
– Semantics: The semantic shift of the partitive, from stage III to stage IV, not
only involves semantic loss, but also semantic enrichment, as is shown in
figure 2.
Figure 2: The semantic evolution of the partitive, from STAGE III to STAGE IV (Carlier 2007)
f In stage IV, de no longer acts syntactically as a preposition with respect

to the NP it governs. As a consequence, the notion of partition and
partition set, which is the semantic correlate of the semi-prepositional
status of de, fades away.
f The feature of unspecified quantity remains however intact.
f By a process of pragmatic strengthening, i.e. the integration as a semantic
feature of what is at first no more than a pragmatic inference, the feature
of unspecified quantity gives rise to the feature of indefiniteness (cf.
Luraghi & Kittilä, this volume): the referent is no longer presented as
unambiguously identifiable by the addressee.
– Paradigm: The partitive enters into a new paradigm, that of the article. The
semantic evolution, as described in Figure 2, is conditioned by this integra-
tion into the paradigm, which is structured in terms of two parameters:
singular versus plural or non-singular and definiteness versus indefiniteness.
f The notion of partition set disappears because it does not contribute to
the differentiation of the partitive article with respect to the other articles
already in place.
f The notion of unspecified quantity is maintained, because it allows the
new article to enter in contrast with the indefinite article derived from
the numeral ‘one’.
f The feature of indefiniteness develops in a binary opposition with the
definite article.
S TAGE V
– Syntax: The partitive article becomes more frequent in all syntactic positions.
– Semantics: The use of the partitive article, which in the early stage is sensi-
tive to the opposition between abstract and concrete nouns, is extended to
abstract nouns and this at the expense of zero determination.
– Paradigm: As a consequence of this double evolution, zero determination is
further reduced. Whereas in the earlier stages, there is some freedom to
either specify the category of determination or to let it unmarked, this
option becomes more and more constrained and finally disappears. This
“obligatorification” (Lehmann 2002: 136) in turn contributes to the further
tightening of the paradigm.
1.3.2 The grammaticalization process of the partitive in French,

Italian and Spanish
In Romance, the destiny of the partitive article is very different according to the
language:
– In French, the use of the partitive article is compulsory for indefinite NPs
containing a mass noun or plural count noun and extends even, in most
syntactic positions, to abstract nouns.
– Italian also developed a partitive article, but contrary to its French equiva-
lent, it is almost never obligatory. In other words, it still alternates with
zero and is most often optional. Moreover, it is heavily subject to regional
variation (cf. Luraghi, unpublished ms.), being far more widespread in
Northern Italian than in the Centre and the South, a point to which we will
come back in section 5.
– In Spanish, although attestations have been retrieved in medieval corpora,
the partitive article did not develop into a full-fledged article.
Table 1: The Partitive Article in Romance
FRENCH Pierre mange du pain * Pierre mange pain

SPANISH * Pedro come del pan Pedro come pan
ITALIAN Piero mangia del pane Piero mangia pane
* This sentence is grammatical in Spanish, but with the strictly partitive

meaning of stage III ‘He eats a piece of the bread’
A more thorough investigation of the empirical data will show that, with
respect to the Grammaticalization Chain in Figure 1, French went from stage I
to stage V, Spanish from I to III and although Italian shows a development
from I to V, the process, which shows important regional variation, is not totally
completed.
1.3.3 Research questions
The present paper, which intends to account for the different outcome of the
grammaticalization chain in the three Romance languages, will also tackle the
following research questions:
(i) Are the parameters which were significant in the rise of the partitive in
Middle French (Carlier 2007) equally significant in Modern Italian?
(ii) Is the evolution of the partitive article consistent with other typological
shifts in the area of Romance languages on the one hand and in a broader
(Indo-)European context on the other hand?
In order to answer these questions, we will identify the different stages of the
evolution of the partitive through the history of each of the three Romance lan-
guages and link the language-specific evolution of the partitive to other charac-
teristics that define the place of the language in the typology of Romance. As a
preliminary, we will consider the partitive in Late Latin, the common starting
point of the three Romance languages.
2 Latin
2.1 Stage III: Late Latin
As has been mentioned in section 1.2., the stage III partitive is occasionally attested
as a genitive-inflected NP in Pre-Classical Latin, in particular in technical texts.
It becomes exceptional during the Classical period, but surfaces again in Late
Latin: it occurs in popularizing texts, and takes often the form of a PP. De is the
most common preposition, but ex and ab are also used, without any semantic
difference4, as is shown by the following three translations of the same Bible
verse.
(19) De ligno quod est scientiae

of tree-ABL . SG ( N ) rel.nom.n.sg be:PRS .3 SG knowledge-GEN . SG ( F )
boni et mali non editis.
good-GEN . SG ( N ) and evil-GEN . SG ( N ) NEG eat-SBJV. PRS .2 PL
A(b) ligno sciendi bonum et malum non
from tree-ABL . SG ( N ) know: GERUND good-ACC . SG ( N ) and evil-ACC . SG ( N ) NEG
manducabitis.
eat-FUT.2 PL
Ex arbore diagnoscentiae boni et
Out-of tree-ABL . SG ( F ) knowledge-GEN . SG ( F ) good-ACC . SG ( N ) and
mali ne tangerent.
evil-ACC . SG ( N ) NEG touch-SBJV. IPFV.3 PL
‘From the tree of the knowledge of good and evil you shall not eat.’
(Genesis 2: 17, Vetus Latina, Ed. Bonifatius Fischer (Beuron), Freiburg:
Heider, 1951)
– Syntax: The partitive occurs mainly in object position of transitive verbs

instead of the accusative.
– Meaning: The partitive presupposes a contextually defined partition set and
isolates an unspecified portion of this partition set. This meaning accounts
for some peculiarities of the partitive in Late Latin.
f As has been pointed out by Väänänen (1981: 115), the object of verbs of
this type is marked as a partitive by means of de only when there is a
determiner (20) or a modifier, e.g. a relative clause (21) or a genitive
complement (22), which ensures the contextual anchorage of the parti-
tion set. Moreover the head noun of the object NP always refers to a
concrete referent.
4 Originally, the Latin prepositions de, ex and ab have distinct spatial meanings: ex ‘out of’, ab
‘away from’, expressing increasing distance with respect to a certain point without implying
contact with that point, de ‘coming from (an origin)’ (cf. §1.3.1, stage I). This distinction is no
longer relevant in Late Latin.
(20) Si quis manducav-erit ex hoc pane,

if PRO. INDF.NOM . SG . M eat:PRF.FUT.3 SG out.of DEM .ABL . N . SG bread:ABL . SG ( N )
vivet in aeternum.
live:FUT.3 SG in eternal:ACC . N . SG
‘If anyone shall eat from this bread, he will live forever.’ ( John 6, 52)
(21) Nam et catelli edunt de micis

for and puppy:NOM . PL ( M ) eat:PRS .3 PL of crumb:ABL . PL ( F )
quae cadunt de mensa dominorum suorum
REL . NOM . F. PL fall:PRS .3 PL from table:ABL . SG ( F ) masters:GEN . PL ( M ) POSS .GEN . M . PL
‘Yet the young dogs eat of the crumbs which fall from their masters’ table.’
(Vulgata, Matthew 15:27)
(22) Ampullam, in qua de oleo beati

phial:ACC . SG ( F ) in REL .ABL . F. SG of oil:ABL . SG ( M ) saint:GEN . M . SG
Martini continebatur.
M.: GEN ( M ) contain:PST.3 SG . P
‘The phial, in which (some) oil of St Martin was contained.’ (Vita Aridii)
f The partitive occurs mostly in combination with verbs compatible with

the idea of fragmentation: ‘to eat’, ‘to drink’, ‘to give’, ‘to take’ (cf. ex.
19, 20, 21).
– Paradigm: « De + NP-ABL » enters into competition with the prepositionless
accusative.
3 French
3.1 The Grammaticalization Chain from Old to Modern French
3.1.1 Stage III: Old French
The partitive is rather infrequent in Old French and would perhaps even not
have been noticed if it were not the ancestor of a newly created article in Middle
French. There is considerable controversy about its status: whereas it is considered
as a full-fledged article by Foulet (1916), scholars today are less affirmative and
express their doubt by using quotes (partitive ‘article’) or even accept that de is
still a preposition. In our viewpoint, the Old French partitive is still in the same
stage of evolution as the Late Latin partitive: it has a hybrid status, between
preposition and determiner, which is characteristic of stage III.
– Syntax: del N occurs as a direct argument of the verb, mostly the direct
objet. The following above mentioned example and repeated here, shows
however that it occasionally appears in predicate position from the earliest
texts5 from the earliest texts on.
(14) Blancandrins fut des plus saives paiens.

B.:NOM 6(M ) be:PST.3 SG of.DEF. ART. PL most wise:PL heathen:PL
‘Blancadrin was amongst the wisest heathens.’
(Old French: Roland [1100], v. 24)
Although del N is not a PP with respect to the verb, de still behaves as a prepo-
sition with respect to the NP. Witness the following contrast between (16a) and
(16b):
(16) a. La poison qu’ il but.

DEF. ART. F. SG
poison: SG ( F ) rel pro.nom.3sg.m drink:PST.3 SG
‘The poison which he drank.’
(Old French: Chrétien de Troyes, Cligès, v. 6687–6689)
b. Du vin de qoi il
of.DEF. ART. M . SG wine:SG ( M ) of REL . N . SG PRO.3 PL . NOM
burent avez oï.
drink:PST.3 PL have:PRS .2 PL hear: PTCP. PST
‘Did you hear about the wine of which they drank.’
(Old French: Béroul, Tristran & Iseut, v. 2133–2134)
– Semantics: The semantic operation performed by this preposition upon the

NP is that of isolating an unspecified part from a contextually specified par-
tition set expressed by the NP. Hence, the Old French partitive presupposes
a partition set, which is specified in the context.
5 For early Old French, the Chanson de Roland is one of the very few available documents of
significant length. It is therefore impossible to establish a precise chronology of the rise of
the partitive in this syntactic position.
6 Old French has a two case system, consisting of the nominative case, marking the subject of
the sentence, and the oblique case, for the other arguments and for adjuncts. This system is
defective, however, since only masculine nouns and determiners have case, whereas feminine
nouns and determiners do not.
(23) Le gastel et le vin leur

DEF. ART: OBL . SG . M pastry:SG and DEF. ART.OBL . SG . M wine:SG PRO.3 PL . DAT
baillent. Del vin volentiers bevaient.
bring:PRS .3 PL of.DEF. ART. M . SG wine:SG ( M ) gladly drink:IPFV.3 PL
‘They bring them the pastry and the wine. They drank gladly (some) of
the wine.’ (Old French: Chrétien de Troyes, Erec, 3178 [1170])
This specific meaning of the Old French partitive explains why it occurs fre-
quently as an object of verbs meaning ‘to eat’, ‘to drink’, ‘to take’, ‘to give’, com-
patible with the notion of partition. It also accounts for the fact that the Old
French partitive only occurs in combination with concrete nouns – mass terms
or countable nouns – but not with abstract nouns.
– Paradigm: the partitive alternates with zero determination, e.g. boivre vin
‘drink wine’/ boivre del vin ‘drink some of the wine (from the bottle on the
table)’
From the 13th century on, the first occurrences of stage IV partitive are
attested in theatre texts that reproduce the oral register.
(24) Ha ! biaus dous fiex, seés vous

Ha handsome:M . SG sweet:M . SG son:NOM .SG ( M ) seat:IMP. PRS .2 PL REFL .2 PL
cois, ou vous arés des eviaus.
quiet:PL or you–PL have:FUT.2 PL of. DEF. ART. PL . M hits:PL ( M )
‘Ha ! Dear sweet son, keep quiet, or you will take hits.’
(Adam de la Halle, Jeu de la Feuillée, 396–397 [1276], quoted Foulet (1916)
These occurrences have the status of ‘exploratory expressions’ (Harris & Campbell
1995: 72). Exploratory expressions can be optionally produced by the existing
grammar and are not intended to live a long life. However, when they are gram-
maticalized, they will be retrieved retrospectively as preludes to change.
3.1.2 Stage IV: Middle French
The frequency of the partitive dramatically increases during the 15th Century.
Evidence for this is provided in Figure 3, representing the relative frequency of
the partitive in four translations of the same Latin source text7.
7 The four translations are conserved at the French National Library with the following references:
nouv. acq. fr 18800, fr. 2003, fr. 25342, fr. 1304. Although there is a frequency increase in the
fifteenth century’s translations, there is a frequency fall in the latest text, which is due to a
difference of register: in the spirit of the incipient Renaissance, the sixteenth century’s transla-
tion is closer to the Latin text and written in a more learned, less natural style, which results in
an avoidance of the partitive.
Figure 3: Relative frequency in ‰ of the partitive in four translations of De Falconibus (Albertus

Magnus)
– Syntax:
As to the function with respect to the verb, del N still occurs dominantly in
object position (25), but becomes less exceptional in other functions, for
instance as a subject (26) or a PP (27). These statistical tendencies are repre-
sented in Figure 4.
(25) Prenez du papier fait de

take:IMP.2 PL of.DEF. ART. M . SG paper: SG ( M ) make:PTCP. PST. M . SG of
fil de cotton.
yarn:SG ( M ) of cotton:SG ( M )
‘Take paper made of cotton yarn.’ (Old Fr. translation of Albertus Magnus,
De Falconibus, BNF, ms. fr. 25342, 15th C.)
(26) Soit prins pour icelle guerir
be:SBJV. PRS .3 SG take:PTCP. PST. M . SG for DEM . ACC . F. SG cure:INF
du lart.
of.DEF. ART. M . SG bacon: SG ( M )
‘Bacon should be taken to cure that one.’ (Old French transl. of Albertus
Magnus, De Falconibus, BNF, ms. fr. 1304, 16th C.)
(27) Fault prandre trois morceaux de lart telz
must:PRS .3 SG take:INF three piece:PL ( M ) of bacon:SG ( M ) such:M . PL
que le faulcon les puisse avaller,
that DEF. ART. M . SG falcon PRO.3 PL . ACC ( M ) can:SBJV. PRS .3 SG swallow:INF
et les plonger dedans du miel.
and PRO.3 PL . ACC ( M ) dip:INF into of.DEF. ART. M . SG honey:SG ( M )
‘You have to take three pieces of bacon such that the falcon can swallow
them and dip them in honey.’ (Old French transl. of Albertus Magnus,
Figure 4: Frequency of the partitive and zero determination according to the syntactic function
in the 15th C. translations of De Falconibus (Albertus Magnus)
With respect to the object position, the partitive is no longer restricted to

verbs meaning ‘to eat’, ‘to drink’, etc. but extends to other verbs which do not
necessarily involve the idea of partition or fragmentation:
(28) Faulcon aiant les piez azurins ne

falcon:SG ( M ) have: PTCP. PRS DEF. ART. PL foot:PL blue:PL NEG
assault pas souvent de plus grans oyseaulx
attack:PRS .3 SG NEG often of(DET ) more big:M . PL bird:PL ( M )
que pie ou corneille.
than magpie:SG ( F ) or crow:SG ( F )
‘A blue-footed falcon does not often attack bigger birds than magpies or
crows.’ (Old French transl. of Albertus Magnus, De Falconibus, BNF,
ms. fr. 1304, 16th C.)
(29) Et confians de l’ ayde de leurs compaignons,

And trust- PTCP. PRS . M . PL of DEF. ART. SG help of their-PL companion-PL
font de beaux faiz d’ armes
make:PRS .3 PL of.(DET. M . PL) nice:M . PL fact:PL ( M ) of weapon:PL ( F )
et tressouvens obtiennent de glorieuses victories.
and very.often get:PRS .3 PL of.(DET. M . PL) glorious.F. PL victories.PL ( F )
‘And trusting on the help of their companions, they perform nice feats
and very often get glorious victories.’ (Old French transl. of Albertus
Magnus, De Falconibus, BNF, ms. fr. 1304, 16th C.)
(30) Quant le faulcon a des

when DEF. ART. M . SG falcon:SG ( M ) have.PRS .3 SG of.DEF. ART. PL ( M )
pouez, . . .
louse:PL ( M ) . . .
‘When the falcon has lice, . . .’ (Old French transl. of Albertus Magnus,
As already noticed in §1.3.1. (stage IV), this extension of the partitive article
can be conceived of as a progress in the marking of transitivity along two axes:
affectedness and individuation (Hopper & Thompson 1980, Lazard 1996). The
article, being a sign of the autonomy of the object with respect to the verb,
is no longer restricted to highly individuated objects, definite or singular, but
extends to less individuated objects, even if they are not highly affected by the
verbal process in terms of fragmentation.
Besides the syntactic function, the internal structure of the NP also proves
to be a relevant factor (Carlier 2004). As is shown in Figure 5, in our 15th Century
corpus, the partitive article and zero are equi-probable when there is no modifier,
the partitive article is less likely to occur when there is a prenominal modifier,
and it occurs more readily when there is a post-nominal modifier.
Figure 5: Relative frequency in % of the partitive and zero determination according to the
internal structure of the NP in the 15th century translations of De Falconibus (Albertus Magnus)
The following example, which corresponds to two translations of the same Latin
sentence, clearly illustrates the impact of the internal structure of the NP: the
partitive article is used in the second translation, where the adjective comes after
the noun, but not in the first translation, where the adjective is placed before the
noun.
(31) a. Pren vif argent.

take:IMP. PRS .2 SG quick:M . SG silver:SG ( M )
(transl. Albertus Magnus, De Falconibus, BNF, f. fr. 2003)
b. Prenez de l’ argent vif.

take:IMP. PRS .2 PL of DEF. ART. M . SG silver:SG ( M ) quick:M . SG
(transl. Albertus Magnus, De Falconibus, BNF, ms. fr. 25342)
‘Take some quicksilver.’
– Semantics:
As mentioned above, the stage III partitive isolates an unspecified quantity from
a contextually defined partition set. In stage IV, it still refers to an unspecified
quantity but the partition set is not necessarily explicit in the context. As illus-
trated by example (30) repeated here, the feature of unspecified quantity
evolves, by pragmatic strengthening, into the feature of indefiniteness: the referent
is presented as not univocally identifiable by the addressee:
(30) Quant le faulcon a des

when DEF. ART. M . SG falcon:SG ( M ) have:PRS .3 SG of.DEF. ART. PL ( M )
pouez, . . .
louse:PL ( M ) . . .
‘When the falcon has lice, . . .’ (Old French transl. of Albertus Magnus,
However, at stage IV, the combination of de with abstract nouns remains un-
common.
– Paradigm:
The partitive becomes a full-fledged article. Its meaning shift is conditioned by
its structural place in the paradigm of the articles, defined by two binary oppo-
sitions: by contrast with un ‘a’, it marks the feature of non-singular; by contrast
with the definite article le(s), it marks indefiniteness.
3.1.3 Stage V: Classical and Modern French
– Syntax:
The partitive article becomes more frequent in all positions and, hence, con-
tinues to reduce the primitive domain of zero determination.
– Semantics:
The partitive article is no longer necessarily linked to the notion of unspecified
quantity and becomes common in combination with abstract nouns.
(32) Il éprouve de la haine pour

PRO. NOM .3 SG . M feel:PRS .3 SG of DEF. ART. F. SG hatred:SG (F ) for
son frère.
POSS .3 SG . M . SG brother:SG ( M )
‘He feels hatred for his brother.’
This evolution also leads to the reduction of zero determination. It is however

not fully completed: the very low relative frequency of avec de l’amour, with the
partitive article, in comparison with the pattern without article, in a corpus of
Modern French, for the period of 1700 until the present day, shows that when
the abstract noun is embedded in a PP, the partitive article remains exceptional:
(33) a. avec amour : 632 occurrences in the Frantext corpus for the period
1700–2010 or 98%
with love
b. avec de l’amour: 13 occurrences in the Frantext corpus for the period
1700–2010 or 2%
with of-the love’
– Paradigm:
As a result of the above mentioned evolution, the expression of the article be-
comes increasingly obligatory. The progressive elimination of zero determination
is a sign of the further paradigmatization of the article.
3.2 The French partitive and the typology of Romance
3.2.1 From OV to VO
From Latin to Romance, there is globally a shift in word order (S)OV > (S)VO.
French is the most consistent SVO language among the Romance languages
(Lahousse & Lamiroy 2012). During the period of Middle French, the last resi-
dues linked to the OV paradigm are evacuated at an accelerated rate (Buridant
1987, Combettes 1988 and Marchello-Nizia 1995). Following Greenberg (1963),
Lehmann (1973) and Vennemann (1974), Buridant (1987) argues that the shift
from OV to VO goes along with a shift within the NP from “modifier + N” towards
“N + modifier”.
In Middle French, the situation is as follows: syntactically heavy nominal
modifiers, such as PPs, relative clauses, participles, are located at the right of
the head noun, whereas adjectives can be either at the left or the right, as is
shown in example (34):
(34) a. Samblans au noir faucon. [Adj N]

resemble:PTCP. PRS to-DEF. ART. M . SG black:(M . SG ) falcon:(M . SG )
‘Similar to the black falcon’ (Old French Transl. of Albertus Magnus,
De Falconibus, nouv. acq. fr 18800, end 14th C.)
b. Laquele li emplist et engrosce plus

REL . F. SG PRO.DAT.3 SG fill:PRS .3 SG and enlarge:PRS .3 SG more
les jambes que du faucon

DEF. ART. PL ( F ) leg:PL ( F ) than of-DEF. ART. M . SG falcon( M . SG )
noir. [N Adj]
black( M . SG )
‘That fills and enlarges his legs more than the legs of the black
falcon’ (ibid. nouv. acq. fr 18800)
From the 15th Century on, adjectives tend to be consistently located at the
right of the noun. Only adjectives which have a role of degree marker, bearing
on the quantity (un homme simple ‘a simple minded man’ versus un simple
homme ‘a single man’) or on the categorial content of the noun (un piètre amant
‘a poor/ mediocre lover’, i.e. a lover who does not really deserve the status of
lover) stay in prenominal position. It is in this position that they undergo gram-
maticalization, involving semantic weakening, and acquire a role of nominal
determination.
According to Lehmann (1973) and Geisler (1982), there is a correlation between
the position of the complement or the modifier and the position of the grammatical
morphemes. OV languages tend to express grammar by means of nominal or
verbal suffixes, whereas in VO languages, grammatical markers tend to be
located before the nominal or verbal head:
Lex – V or N – Gramm
Gramm – V or N – Lex
This hypothesis could help us to understand the dramatic increase of the

partitive during the 15th C. and its shift from an intermediate status (stage III)
to that of full-fledged article (stage IV). As a result of phonetic erosion, final –s

is lost. The plural marker –s is however not replaced by some other suffix. We
rather observe an increase of the use of the definite article le(s), of the indefinite
singular article un and the creation of the new indefinite non singular, i.e. parti-
tive article. As a result of the extension of the articles, zero determination becomes
exceptional. In other words, the category of number is systematically expressed
by means of a prenominal article. This further grammaticalization of the articles
is, on the formal side, reflected in the fact that they become clitic.
The correlation between shift in word order and the development of gram-
matical markers before the nominal head is confirmed by the fact that the first
occurrences of the partitive article are more frequently encountered in the case
of “N + modifier” than in the case of “modifier + N”, as is illustrated by the
following contrast between two translations of the same Latin sequence (cf.
example (31) and Figure 5).
Lex – N – Gramm pren vif argent take quick silver

Gramm – N – Lex prenez de l’argent vif take of the silver quick
3.2.2 Information structure, linear order and determination
Leiss (2007: 74) observes that languages with a relative flexible word order have
the possibility to express (in)definiteness by word order. Witness the following
Czech example, quoted from Kramsky (1972: 42):
SV / OV → NP: thematic and definite

(35) a. Kniha je na stole.
book:NOM . SG ( F ) be:PRS .3 SG on table:LOC . SG ( M )
‘The book is on the table.’
VS / OV→ NP: rhematic and indefinite

b. Na stole je kniha.
on table:LOC . SG ( M ) be:PRS .3 SG book:NOM . SG ( F )
‘There is a book on the table.’
In Latin, word order is not strictly constrained by syntactic principles and can
be put to service for the expression of information structure and (in)definiteness.
From Latin to French, word order becomes more and more rigid (Marchello-
Nizia 1992). French represents the most advanced stage in this evolution: word
order is strictly constrained by syntax and hardly available for the marking of
discourse function (Lahousse and Lamiroy 2012).
Hence, when word order can no longer be exploited to express the thematic
or rhematic status of the NP (and the associated opposition between definiteness
and indefiniteness), a welcome bonus of the new article system which has been
enriched with the partitive article is that of allowing systematic marking of (in)
definiteness.
4 SPANISH
4.1 The Grammaticalization Chain
4.1.1 S TAGE III : Old Spanish
– Syntax:
As to its syntactic function, del N appears in Old Spanish texts8 in a way com-
parable to that of Old French, i.e. in object position after verbs with concrete
meaning, as is shown by the verb comer ‘to eat’ in (36a) and in subject position,
typically with unaccusatives (36b):
(36) a. Cogió del agua en él e a

take:PST.3 SG of.DEF. ART. F. SG water:SG ( F ) in PRO.3 SG . M and to
sus primas dio.
POSS .3 cousins:PL ( F ) give:PST.3 SG
‘He took some water into it [his hat] and gave it to his cousins.’
(Old Spanish, end 12th C. Cid 2801)
b. Et salieron a él de los
and come-out:PST.3 PL to PRO.3 SG . M of DEF. ART. M . PL
omnes buenos.
men:PL (M) good:PL (M)
‘And some good men came to him.’ (Alfonso X el Sabio, Primera
Crónica general, 1260–1284, quoted from Luraghi, unpublished ms.)
8 The period covered by Old or Medieval Spanish runs from the 10th to the beginning of the
15th C. Spanish historical linguists (e.g. Lapesa 1983) do not distinguish a period of “Middle
Spanish” (cf. Middle French or Middle English). Old Spanish is followed by “Classical Spanish”
(from the 15th to the 17th C.), which precedes “Modern Spanish” (from the 18th C. on).
– Semantics:
The partitive in Old Spanish is also comparable to that of Old French from a
semantic point of view: it has a real partitive meaning and presupposes extrac-
tion from a contextually (deictically or anaphorically) defined partition set. Thus,
in example (36a), del agua refers to water that has been previously mentioned in
the text as la la linpia fuont ‘the clear source’.
– Paradigm:
The partitive appears in alternation with the zero marker, as exemplified
in (37):
(37) Old Spanish Dexado ha heredades

leave:PTCP. PST. M . SG have:PRS .3SG inheritance:PL (F)
e casas e palaçios.
and house:PL (F) and palace:PL ( M )
‘He has left objects to inherit and houses and palaces.’ (Cid 115)
4.1.2 Stage IV
Surprisingly, the partitive did not pursue its grammaticalization process as it did
in French and Italian: examples of partitive de + NP are sporadic after the 15th
Century. A unique attestation is found in a writing of Santa Teresa de Jesus (17th
C.), viz. dar de la fruta ‘to give some of the fruit’ (Adillo Rufo, unpublished ms).
The stage III partitive can still be found in Modern Spanish; however, it occurs
only in object position with transitive verbs, typically ‘to eat’ (38a), ‘to drink’, ‘to
take’, ‘to give’ and its meaning clearly remains partitive, and not indefinite. With
subject NPs, Spanish del is totally ungrammatical (38b), which proves that the
partitive has not turned into a determiner as it did in French and Italian (Fr.
Des jeunes entraient et sortaient – It. Dei giovani entravano e uscivano):
(38) Spanish a. Comió del pastel y luego

eat:PST.3 SG of.DEF. ART. F. SG cake:SG ( M ) and afterwards
se encontró mal.
REFL .3 SG . M find:PST.3 SG bad
‘He ate a piece of the cake and felt sick afterwards.’
b. *De los jóvenes entraban y

of DEF. ART. F. SG youngsters:PL ( M ) enter:PST.3 PL and
salían.
go-out:PST.3 PL
‘Young people went in and out.’
In Modern Spanish, indefiniteness can be expressed, in combination with

count nouns, by an indefinite article which goes back to Latin unum ‘one’ both
in singular (un, una) and in plural (unos, unas), as shown in (38a). Note that
Modern French uses the partitive in this case (par des bâtons in 38b)9:
(39) a. Una lona sostenida en unos

INDF. ART. F. SG cloth:SG ( F ) hold:PTCP. PST. F. SG in INDF. ART. M . PL
palos servía de toldillo.

stick:PL ( M ) serve:PST.3 SG of canopy
‘A piece of cloth sustained by sticks served as canopy.’
(Poyato, quoted from Herslund 2007)
b. Un tissu soutenu par des

INDF. ART. F. SG cloth:SG ( F ) hold:PTCP. PST. M . SG by of.DEF. ART. M . PL
bâtons servait d’auvent
stick:PL ( M ) serve:PST.3 SG of canopy
‘A piece of cloth sustained by sticks served as canopy.’
Mass nouns however take the zero marker to indicate an unspecified quantity.
Compare example (40) to (38a):
9 Old French used to have the plural indefinite article uns/unes:

(i) Et avoit unes grandes joes et un grandisme nez plat et unes grans narines lees et unes
grosses lèvres.
‘And she had big cheeks and a very big flat nose and big and ugly nostrils and thick lips . . .’
(Aucassin et Nicolette, quoted by Foulet 1916).
The plural form of un never attained high frequency and was mostly used to refer to an entity
composed of identical parts, i.e. an internal plural. For a discussion, see Guillaume (1969),
Carlier (2001) & Herslund (2003). This plural form of un for the expression of the internal plural
disappears during the period of Middle French and is replaced by the newly created article des.
Witness the following two examples, of which the oldest already contains des:
(ii) Qui a ses soulliers percez, il a besoin d’avoir des chausses.
‘The one who has worn-out shoes, he needs to have breeches.’
(Les deux savetiers [end 15th C.], 39, in: Recueil de farces)
(iii) le Duc [estoit] vestu simplement d’unes chausses de veloux noir mouchetté, d’un propoint
de treillis et d’une casaque de mesme.
‘The Duke was dressed in a simple way, with breeches made of black mottled velvet, a
doublet of hemp and a coat of the same fabric.’ (R. de Lucinge, Dialogue du François et du
Savoysien [1593])
(40) Cogió agua y se la dió

take:PST.3 SG water:SG ( F ) and REFL .3 PL pro.ACC . SG . F give:PST.3 SG
a sus primas.
to POSS .3. PL cousins:PL ( F )
‘He took (some) water and gave it to his cousins.’
In other words, in Spanish, the partitive never became an indefinite article,

nicely showing that although certain grammaticalization chains seem to be cross-
linguistically stable as the one which leads from partitive nominal constructions
to indefiniteness marking (cf. Luraghi & Kittilä this volume, for a typological
overview), nothing can prevail them from being cut off at a certain point or
from not reaching the end stage.
The gradual loss of the partitive parallels that of the pronominal adverb
ende ‘from there’ (Lat. inde) which is also found until the 15th C.: the disappear-
ance of the two are most probably linked (Posner 1996: 335), as suggested a con-
trario by the fact that French both developed a partitive article and maintained
the pronominal adverb in parallel (e.g. Il mange du chocolat / Il en mange ‘He
eats chocolate / He eats some’). So did Italian. In the two languages, the parti-
tive became an indefinite article and the adverb turned into a clitic, viz. Fr. en
and It. ne. Spanish maintained neither the partitive nor the pronominal adverb.
Why this happened cannot be fully answered here, but a number of factors
which we summarize hereafter in 4.2. may have played a role.
4.2 The Spanish partitive and the typology of Romance
4.2.1 From OV to VO
In contrast with French, Spanish maintained a strong verbal and nominal mor-
phology based on word endings, inherited from Latin. Thus -s marks plural in
Spanish NPs in an unambiguous way, which could be one of the reasons why
the partitive did not flourish, as there is less need to mark number with free pre-
posed morphemes when it is already marked by bound suffixal endings.
N – Gramm rather than Gramm – N

Worth noting is that Spanish not only differs from French, whose nominal
system hardly has any plural marking at all (in spoken language), but also
from Italian (Stark 2007): whereas one and the same plural morpheme, viz. -s
is used in all occasions in Spanish, Italian has several plural morphemes (e.g.
can-i ‘dogs’, donn-e ‘women’, labr-a ‘lips’), some of which in addition can
also act as singular morpheme (e.g. legg-e ‘law’, ragazz-a ‘girl’). Of the three
Romance languages under study Spanish thus displays the most “efficient” plural
marking system based on suffixation, because it is the most unambiguous: one
form / one function.
This property of Spanish of being a far more synthetic language than French
for example (and hence, closer to Latin in this respect) could account for the
fact that Spanish never fully developed a sophisticated prenominal determiner
system as French did. In fact, Company Company (1991) points out that the
spreading of the article in general, i.e. including that of the definite article, was
also less general in medieval Spanish than in medieval French (e.g. Vassallos
de mio Çid seyense sonrissando ‘the vassals of the Cid sit down and smile’, from
Lapesa 1983: 211), which could be due to the same reason.
Another important factor here may be word order10. As already noted, languages
with a free word order may express (in)definiteness by giving the NP a particular
position in the sentence (Leiss 2007: 74). Compared to French, which during the
course of its history acquired a rigid SVO order (Marchello-Nizia 1992), Spanish
displays a flexible word order: although discursive constraints are obviously at
play, all word orders can still be found in Spanish, viz. SOV, SVO, VSO, VOS, OSV,
OVS (Lahousse and Lamiroy 2012; Zubizarreta 1999). As we will see, Italian also
10 Zamboni (1998) also ascribes the profound morpho-syntactic restructuration which took
place between the Late Latin system and proto-Romance to the change of basic word order
from SOV to SVO and the loss of nominal flexion. Importantly, these changes did not affect
Northern and Southern Romania in the same way: one of the different properties that opposes
the North (Gallo-Romance and Northern Italo-Romance) to the South (Ibero-Romance and
Southern Italo-Romance) according to Zamboni is, for the object, the rise of the partitive in
the North, and the rise of the prepositional accusative in the South, and for the subject, the
rise of personal pronouns in the North vs their absence in the South. The intermediate position
that Italian occupies between French and Spanish regarding the partitive may thus go back to a
major shift that originated in archaic Romance. The typological hypothesis which establishes a
relationship between partitives and differential object marking is however rejected by Luraghi
(unpublished ms.).
has a much more flexible word order than French, but less than Spanish. As
words can be moved around in the Spanish sentence to indicate their thematic
(definite) vs rhematic (indefinite) character, there is less need to mark (in)definite-
ness by an explicit prenominal marker.
5 Italian
5.1 The Grammaticalization Chain
5.1.1 Stage III
– Syntax
In Old Italian11, del N appears in object position after verbs with concrete mean-
ing such as ‘eat’, ‘give’, etc., as the examples in (41a–b) illustrate. The example
(41c) shows that the partitive even occurs in subject position in Old Italian,
typically in existential sentences:
(41) a. Ela mançà del pomo que

PRO.NOM .3 SG . F eat:PST.3 SG of.DEF. ART. M . SG apple:SG ( M ) REL . ACC . M .
li de’ un serpent.
pro.DAT.3 SG . F give:PST.3 SG one:SG ( M ) snake:SG ( M )
‘She ate of the apple that a snake gave her.’
(Uguccione da Lodi, beginning of 13th C., quoted from Luraghi,
unpublished ms.)
b. Se tu ai corno, del
if PRO. NOM .2SG have: PRS .2 SG wheat:SG ( M ) of.DEF. ART. M . SG
vino ti do volentieri.
wine:SG ( M ) PRO. DAT.2SG give:PRS .1SG with pleasure
‘If you have wheat, I will be happy to give you some wine.’
(Novellino, end of 13th C., quoted from Tekavčic 1980: 115)
c. Che del ben non vi sia.

that of.DEF. ART. M . SG good: SG ( M ) NEG there be:SBJV. PRS .3 SG
‘That there is not some good.’
(Ubertino del Bianco d’Arezzo, a. 1269, quoted from Luraghi,
unpublished ms.)
11 Old Italian covers the period that stretches from the 10th till the 16th C., i.e. till 1582, the
founding date of the Accademia della Crusca (Tekavčic 1980).
– Semantics
Not surprisingly, the meaning of del vino here is truly partitive: del indicates
extraction from a deictically defined partition set. In (41b) for example, which
is a dialogue between two speakers, one of them possesses wheat while the
other has wine, the two agree upon sharing their respective goods.
– Paradigm:
From its origin up till now, de + NP enters in a paradigm in which it appears in
alternation with the zero marker:
(42) a. Abbiamo bevuto vino.

have:PRS .1 PL drink:PTCP. PST. M . SG wine:SG ( M )
b. Abbiamo bevuto del vino.12

have:PST.1 PL drink:PTCP. PST. M . SG of.DEF. ART. M . SG wine:SG ( M )
‘We drank wine / we drank some wine.’
5.1.2 Stage IV
– Syntax:
From a syntactic point of view, the partitive gradually extended to other posi-
tions than the (proto)typical direct object. In the same way as in French, it
spread to regular subject position (i.e. not subjects of unaccusatives, which are
in fact objects), e.g. subject of transitive verbs (as in 43 a–b) and to the PP, as
shown by the examples in (44):
SUBJECT
(43) a. Delle feste solenni riuniranno

of.DEF. ART. F. PL party:PL ( F ) solemn:PL ( F ) bring-together:FUT.3 PL
i nuovi figli della Libertá.
DEF. ART. M . PL new:PL ( M ) child:PL ( M ) of.DEF. ART. F. SG freedom
‘Solemn celebrations will bring together the new children of the
Freedom.’ (from the newspaper Monitore Napoletano dated 1799,
quoted by Serianni 1989: 154)
12 A Google hit (April 10, 2012) on this sentence does not show an enormous difference
between the number of speakers who use the partitive (883.000) and those who do not
(913.000).
b. Delle donne sconosciute hanno

of.DEF. ART. F. PL women:PL ( F ) unknown:PL ( F ) have:PRS .3 PL
aperto la porta.
open:PTCP. PST. M . SG DEF. ART. F. SG door: SG ( F )
‘Unknown women opened the door.’ (Native speaker judgment)13
PP (a, con, da, in, per, su, sotto, verso)
(44) a. Ecco una ricetta per dei buonissimi

here one. F. SG recipe:SG ( F ) for of.DEF. ART. M . PL very-good: PL ( M )
biscottini di pasta frolla da
biscuits: PL ( M ) of pastry:SG ( F ) shortcrust: SG ( F ) to
accompagnare con del tè o caffé.
accompany:INF. PRS with of.DEF. ART. M . SG tea:SG ( M ) or coffee:SG ( M )
‘Here comes a recipe for very good biscuits of shortcrust pastry to
have with tea or coffee.’ http://www.ricettaidea.it/cucina-base (29.7.2010)
b. Può essere infatti coperto da del

can:PRS .3 SG be:INF. PRS indeed cover:PTCP. PST. M . SG by of.DEF. ART. M . SG
cioccolato oppure proposto in varie
chocolate:SG ( M ) or propose:PTCP. PST. M . SG in various: PL ( F )
dimensioni e forme in base
dimensions: PL ( F ) and forms: PL ( F ) in basis: SG ( F )
alle proprie esigenze.
to.DEF. ART. F. PL proper:PL ( F ) needs:PL ( F )
‘It can indeed be covered with chocolate or presented in various
dimensions and forms according to one’s own need.’
http://www.ricettaidea.it/cucina-base (26.7.2010)
c. Il viandante che fosse

DEF. ART. M . SG traveler:SG ( M ) REL . ACC . M . SG be:SBJV. PST.3 SG
incontrato da de’ contadini,
find:PTCP. PST by of.DEF. ART. M . PL peasant:PL ( M )
fuor della strada maestro.
outside of.DEF. ART. F. SG street:SG ( F ) main:SG ( F )
‘The passerby who should be found by some peasants outside the
main road.’ (Manzoni, I promessi sposi, chapter XXXII)
13 We wish to thank here the Italian native speakers to whom we submitted our data: Elisabetta
Bonvino, Claudio Iacobini, Stefania Marzo, Sara Scoz and Giuseppe Massangioli.
Examples such as (44b–c) show that some of the constraints still mentioned
in Italian grammars, according to which *da del is ungrammatical or regional
(e.g. Renzi et al. 2001: 318 ff), in fact do not hold, since this combination already
appears in Manzoni’s 19th Century novel, as already remarked in Luraghi
(unpublished ms.).
– Semantics:
With respect to its meaning, the partitive spreads, from the 14th C. on, to con-
texts in which del N is no longer a contextually defined partition set, as shown
by the following examples from Dante quoted from Tekavčic (1980: 115):
(45) a. Chi del fango ingozza.

Who of.DEF. ART. M . SG mud:SG ( M ) swallow_in_great_quantities:PRS .3 SG
‘Who swallows mud in great quantities.’ (Dante, Inferno, VII, 129)
b. Rompendo della scorza, non che dei

break:GER . PRS . of.DEF. ART. F. SG bark:SG ( F ) not that of.DEF. ART. M . PL .
fiori e delle foglie nove.
flower:PL ( M ) and of.DEF. ART. F. PL . leave:PL ( F ) new:PL ( F )
‘Breaking some bark, not only flowers and new leaves’
(Dante, Purgatorio XXXII, 113)
– Paradigm:
Although the partitive up till now still largely alternates with zero marking, its
status has changed over time: it no longer introduces a prepositional comple-
ment, but fully belongs to the paradigm of the nominal determiners where it
functions as an indefinite article (Spore 1987). As pointed out by Renzi (1982),
the “real plural of the article uno” in Modern Italian is the partitive degli/delle.
This is indeed clearly suggested by the fact that they can appear in preverbal
subject position, as in the example (43c). However, although most Italian norma-
tive grammars (e.g. Brunet 1979: 111, Fogarasi 1969: 123, Regula & Jernej (1975:
119) analyze del/degli as indefinite articles, they still often condemn it as
“unpleasant for the ear” (Brunet) or “unelegant” (Fogarasi) and recommend
zero marking especially with a negation and after certain prepositions.
5.1.3 Stage V
– Syntax:
Although the partitive is not excluded in any syntactic position in Modern Italian,
a small corpus study based on journalistic texts (Corriere della Sera and 24 Ore
Il Sole) and recipes on Internet (www.Ricettaidea.it.)14 shows that its favorite

function is still that of determiner of a direct object NP15, as was the case in
Middle French (cf. supra Fig. 3):
Figure 6: Frequency of the partitive according to the function with respect to the verb
Interestingly, the study of the same sample with respect to the internal
structure of the NP reveals that, as in Middle French (cf. supra Fig. 4), postposed
adjectives and participles, relative clauses or prepositional complements that
follow the noun tend to favor the use of the partitive.
Figure 7: Frequency of the partitive according to the internal structure of the NP
The same tendency is further confirmed by the answers of our native speakers: in
(42b) produced by our informants, the NP is introduced by the partitive when
followed by a postposed Adj. / Participle but by the zero marker otherwise, as
in (42a):
14 The data were downloaded from 18.7 till 18.8.2010 with glossanet, a software developed at
the Université Catholique de Louvain (http://glossa.fltr.ucl.ac.be).
15 Interestingly, a corpus study by Luraghi (unpublished ms.) yielded similar results.
(42) a. Abbiamo bevuto vino.

have:PRS .1 PL drink:PTCP. PST. M . SG wine:SG(M)
‘We drank wine.’
b. Abbiamo bevuto del vino

have:PRS .1 PL drink:PTCP. PST of.DEF. ART. M . SG wine:SG(M)
maturato in cantina.
ripen:PTCP. PST. M . SG in cellar:SG(F)
‘We drank wine that riped in the wine cellar.’
The following examples illustrate the combination of the partitive with a relative
clause following the NP. According to Renzi et al. (2001: 378 ff), example (46a) is
a case in which the partitive is obligatory:
(46) a. Ecco del vino che hanno

here of.DEF. ART. M . SG wine:SG ( M ) REL . ACC . M . SG have:PRS .3 PL
portato dall’ Alsazia.
bring:PTCP. PST. M . SG from.DEF. ART. F. SG Alsace
‘Here you have some wine which they brought from Alsace.’
b. Passa del tempo che sembra

pass:PRS .3 SG of.DEF. ART. M . SG time:SG (M ) REL . NOM . M . SG seem:PRS .3 SG
lunghissimo in attesa dei soccorsi.
very-long:M . SG in wait:SG (F ) of.DEF. ART. M . PL help: PL (M )
‘Time seems very long, when you are waiting for help.’
(Corriere della Sera, 26.5.2010)
– Semantics:
With respect to its semantics, the use of the Italian partitive is not limited to any
particular noun type, i.e. it can precede all kinds of nouns, including those with
abstract referents:
(47) a. Bisogna avere della pazienza.

need:PRS .3 SG have:INF. PRS of.DEF. ART. F. SG patience:SG ( F )
‘One needs to have patience.’ (quoted from Tekavčic 1980: 116)
b. Vogliamo soltanto fare del bene

want:PRS .1 PL only do:INF. PRS of.DEF. ART. F. SG good:SG ( M )
e dare una mano, mai saremmo
and give:INF. PRS one: F. SG hand:SG ( F ) never be:COND.1 PL
andati contro la chiesa, davvero

go:PTCP. PST. M . PL against DEF. ART. F. SG church:SG ( F ) really
non capisco le polemiche.
NEG understand:PRS .1 SG DEF. ART. F. PL polemic:PL ( F )
‘We only want to do good and give a hand, we would never have
criticized the Church, I really don’t understand the controversy.’
(Corriere della Sera, 1.6.2010)
However, although for many native speakers there is no meaning difference

at all between zero and the partitive, several linguists (e.g. Brunet 1979 in (48),
but also Korzen 1996) as well as some of our native informants (ex. 49 and 50)
mention that semantic constraints still characterize the partitive. According to
them, the partitive often has some of its residual (original) partitive meaning
and is not totally synonymous with the zero marker:
(48) a. Fare delle smorfie.

make:INF. PRS of.DEF. ART. F. PL face:PL ( F )
‘to pull faces’ (Brunet 1979:109)
b. Fare smorfie.
make:INF. PRS face:PL ( F )
‘to grin’
(49) a. Abbiamo bevuto del vino caldo.

have:PRS .1 PL drink:PTCP. PST. M . SG of.DEF. ART. M . SG wine:SG(M) hot:SG ( M )
‘We drank hot wine.’
b. Abbiamo bevuto vino caldo.

have:PRS .1 PL drink:PTCP. PST. wine:SG(M) hot:SG ( M )
‘We drank mulled wine.’
(50) a. Piero beve dell’ acqua.

Peter drink:PRS .3 SG of.DEF. ART. F. SG water:SG ( F )
‘Peter drinks some water.’
b. Piero beve acqua.

Peter drink:PRS .3 SG water:SG ( F )
‘Peter (usually) drinks water.’
– Paradigm:
Although Modern Italian still displays a large alternation of the partitive with
zero marking, which suggests that the partitive is still in an overlap stage (Heine
& Kuteva 2002) in Modern Italian, many cases are found where the two are used
without any meaning difference. Thus, in the oral example in (51), the postverbal
NP is used once with zero marker and once with the partitive, after the same
verb esistono and by the same speaker:
(51) Esistono per vostra conoscenza queste proprio

exist:PRS .3 PL for POSS .2 PL .F. knowledge:SG ( F ) DEM . F. PL properly
non lo so. Eh ma esistono
NEG pro.ACC .3 SG . M . know:PRS .1 SG eh but exist:PRS .3 PL
dizionari etimologici (. . .) questo per dire
dictionary:PL ( M ) etymological:PL ( M ) DEM . M . SG for say:INF. PRS
che praticamente esistono dei livelli di
that practically exist:PRS .3 PL of.DEF. ART. M . PL level:PL ( M ) of
approssimazione.
approximation:SG ( F )
‘There are, if you want to know, these – actually I do not know eh – but
there are etymological dictionaries (. . .) , that is, there are actually
different levels of approximation’ (quoted from Bonvino 2005)
5.2 The partitive and the typology of Romance
5.2.1 From OV to VO
A very important property of Italian is its regional variation. In many respects,

the Northern variants differ from those of Centre and the South. The partitive
seems to be more frequent in the North (Tekavčic 1980: 116; Luraghi unpublished
ms.), as was also confirmed by our informants, those from Milan using the par-
titive more willingly than those from Naples.
Interestingly, the frequent use of the partitive in the North patterns with
a similar variation in other domains of « neo-Italian » (Sabatini 1985) that are
consistent with OV → VO, in particular the gradual loss of the past tense which
is being replaced by the present perfect, more so in the North than in the South.
South North
V – Gramm Gramm – V
As we have already mentioned, number and gender end marking of N is, in

contrast with the Spanish system, ambiguous in Italian (Stark 2007), which
could be a further reason why Italian introduced pre-posed free morphemes,
like French did. Recall however that Italian occupies an intermediate position,
between Spanish, which expresses plural by means of one unique suffix, and
French, which has virtually no suffixal plural marking on the noun at all (in
spoken language at least).
Table 2: Plural endings in Italian, Spanish and French
Italian
SINGULAR PLURAL
CasA ‘house’ CasE
CanE ‘dog’ CanI
LibrO ‘book’ LibrI
BracciO ‘arm’ BracciA
Spanish
SINGULAR PLURAL
CasA ‘house’ CasaS
MentE ‘mind’ MenteS
LibrO ‘book’ LibroS
French
SINGULAR PLURAL
Rose [Roz] ‘rose’ Roses [Roz]
Livre [livR] ‘book’ Livres [livR]
A similar observation regarding Italian occupying an intermediate position between

Spanish and French holds for word order. On the one hand, it is less free than in
Latin and in Spanish; on the other hand, it is far more free than French, as it
allows almost all types, albeit with particular constraints (Belletti and Shlonsky
1995, Lahousse and Lamiroy 2012, Sornicola 1994). If there is a significant rela-
tionship between free word order and lack of nominal determiners, as we have
suggested above in 4.2.2. for Spanish, it comes as no surprise that the use of the
partitive in Italian is in between the French situation, where word order is rigid
and the partitive obligatory, and the Spanish one, where word order is free and
the partitive inexistent.
6 Concluding remarks
The grammatical category of the articles is typologically rather uncommon.
Moreover, its spread in Western-European languages is a recent phenomenon,
probably due to contact (Heine & Kuteva 2005). Schroeder (2006) has shown
that there is a continuum from West to East:
Western-European languages Central-European languages Eastern-European languages

e.g. English, French, Spanish, e.g. German e.g. Russian, Finnish
Italian
→
No article-use of demonstrative
Definite article ≠ demonstrative Definite article = demonstrative Incipient article-use use of
demonstrative
What we have shown in this paper is that there is also a continuum in the Romance group:
French (North-)Italian Spanish
→
Figure 8: The spread of the articles in Europe
Although the source expression of the partitive is available for all Romance
languages, we have showed here that the outcome of the evolution is quite dif-
ferent according to the language: whereas the partitive does not grammaticalize
into an article in Spanish at all, it becomes an article both in Italian and French.
However, in Italian it remains basically optional (and depends heavily on the
region), i.e. – in Italian the grammaticalization process from preposition to
article has not totally reached its endpoint, as it did in French. Our data also
provide evidence for the hypothesis that there is a significant relation between
the overall typological evolution of the Romance languages regarding other
independent domains and the emergence of the partitive (Lamiroy & De Mulder
2011; Carlier & Combettes, to appear).
On the one hand, there seems to be a significant correlation between the
evolution of word order in Romance and the position of the grammatical mor-
phemes: insofar as there is a shift from OV to VO, the language tends to recruit
existing free morphemes at the left of the verb or the noun for the expression of
grammatical functions, instead of keeping or renewing the existing suffixal
morphology. For instance, in the verbal domain, clitic pronouns take over the
role of personal inflection and the present perfect ‘have/be’ + past participle
(Fr. il est venu) is used instead of the aoristic past tense (“passé simple”, e.g.
Fr. il vint). In the nominal domain, prepositions are grammaticalized in order to
express the syntactic functions of core arguments formerly expressed by case
inflection and articles are developed for the expression of number (and gender),
which makes the suffixal markers of number and gender superfluous. As French
has the most consistent VO order, the grammaticalization of the free morphemes
at the left of the nominal or verbal head seems in an advanced stage. For
instance, the aoristic past tense (“passé simple”) is no longer used in oral
speech, the present perfect being the normal expression of the past tense. In
Spanish, on the contrary, where OV is still possible, the aoristic past tense con-
serves its vitality and the present perfect still has its full aspectual value. Italian
occupies once more the intermediate position and this in two ways, both as a
whole between French and Spanish and internally, because of important differ-
ences between the North and the Centre/South of Italy: whereas the Northern
variants tend to lose the aoristic past (similarly to French), the Centre and the
South favor the conservative alternative, viz. the use of the aoristic past tense
(“passato remoto”). The different stages of the development of the partitive in
the three Romance languages considered here and the regional differences
between North Italian with respect to Centre and the South can be accounted
for in the same perspective.
On the other hand, there is a correlation between the rigidification of word
order and the development of articles as markers of the discourse status of
referents: when word order becomes rigid and is put into service to mark exclu-
sively syntactic function, it can no longer be used to express the discourse status,
thematic or rhematic, of the nominal referents. This is the case for French which
evolved into a totally rigid SVO order. In these conditions, a full article system
which includes a partitive (indefinite) marker is a welcome bonus to mark the
discourse status of the nominal referents (indefinites being prototypically asso-
ciated with the rhematic position and definites with themes). In Spanish to the
contrary, word order is virtually free, so that the constituents can be moved in
the sentence for semantic and discursive reasons. As a consequence, the need
of a full article system is less stringent: zero determination is still widespread
and the prepositional partitive did not evolve into an article. Once again, Italian
occupies the intermediate position, between French and Spanish, and with an
internal contrast between the North, where the partitive article is common, and
the Centre and the South, where zero marking of the NP is still often preferred.
References
Adillo Rufo, Sergio. 2006. La partitividad en francés y español: un enfoque diacrónico. Actas
del XXXVI Congreso de la Sociedad Española de Lingüística (unpublished ms.).
Belletti, Adriana & Ur Schlonsky. 1995. The order of verbal complements: A comparative study.
Natural Language and Linguistic Theory 13. 489–526.
Bonvino, Elisabetta. 2005. Le sujet postverbal. Une étude sur l’italien parlé. Paris: Ophrys.
Brunet, Jacqueline. 1978–2008. Grammaire critique de l’italien. Paris: Université de Paris 8-
Vincennes.
Buridant, Claude. 1987. L’ancien français à la lumière de la typologie des langues: les résidus
de l’ordre OV en ancien français et leur effacement en moyen français. Romania 108: 20–65.
Carlier, Anne. 2001. La genèse de l’article un. Langue française 130. 65–88.
Carlier, Anne. 2007. From Preposition to Article: The Grammaticalization of the French Partitive.
Carlier Anne & Bernard Combettes (to appear, 2015). Typologie et catégorisation morphosyntax-
ique : du latin au français moderne. Langue française.
Carlier, Anne, Michèle Goyens & Béatrice Lamiroy. 2013. De: a genitive marker in French? Its
grammaticalization path from Latin to French. In Jean-Christophe Verstraete & Anne Carlier
(eds), The Genitive, 141–216. Amsterdam: Benjamins.
Clédat, Léon. 1901. La préposition et l’article partitifs. Revue de philologie XV: 81–131.
Combettes, Bernard. 1988. Recherches sur l’ordre des éléments de la phrase en moyen français.
PhD, Univ. of Nancy.
Company Company, Concepción. 1991. La extensión del artículo en el español medieval.
Romance Philology XLIV (4). 402–424.
Fogarasi, Miklós. 1969. Grammatica Italiana del Novecento. Budapest: Tankönyvkiadó.
Foulet, Lucien. 1916. Petite syntaxe de l’ancien français. Paris: Champion.
Geisler, Hans. 1982. Studien zur typologischen Entwicklung: Lateinisch, Altfranzösisch,
Neufranzösisch. München: Wilhelm Fink.
Goyens, Michèle, Béatrice Lamiroy & Ludo Melis. 2002. Déplacement et repositionnement de la
préposition A en français. Lingvisticae Investigationes 25(2). 275–310.
Greenberg, Joseph H. 1963. Some Universals of Grammar with Particular Reference to the Order
of Meaningful Elements. In: J. Greenberg Ed., Universals of Language, Cambridge/Ms.:
MIT Press, 73–113.
Guillaume, Gustave. 1969. Langage et science du langage. Paris: Nizet.
Harris, Alice C. & Lyle Campbell. 1995. Historical syntax in a cross-linguistic perspective. Cam-
bridge: Cambridge University Press.
Heine, Bernd & Tania Kuteva. 2002. The World Lexicon of Grammaticalization. Cambridge:
Heine, Bernd & Tania Kuteva. 2005. Language contact and Grammatical change. Cambridge:
Herslund, Michael. 2003. Le pluriel de l’article indéfini en ancien français. In Peter Blumenthal
& Jean Emmanuel Tyvaert (eds.), La cognition dans le temps. Etudes cognitives dans le
champ historique des langues et des textes, 75–84. Tübingen: Niemeyer.
Herslund, Michael. 2007. Roman et germanique: l’exemple de l’article indéfini. In Louis Bégioni &
Claude Müller (eds.), Problèmes de sémantique et de syntaxe. Hommage à André Rousseau,
221–230. Lille: Ed. du Conseil Scientifique de l’Université de Lille 3.
Hopper, Paul. 1993. On some principles of grammaticization. In Paul Hopper & Elisabeth C.
Traugott (eds.), Grammaticalisation, 17–35. Cambridge: Cambridge University Press.
Humbert, Jean. 19603. Syntaxe grecque. Paris: Klincsieck.
Koptjevskaja-Tamm, Maria. 2001. ‘A piece of the cake’ and ‘a cup of tea’: Partitive and pseudo-
partitive nominal constructions in the Circum-Baltic Languages. In Östen Dahl & Maria
Koptjevskaja-Tamm (eds.), The Circum-Baltic Languages: Typology and Contact, vol. 1–2.
Korzen, Iørn. 1996. L’articulo italiano fra concetto ed entità. Etudes romanes 26. Copenhague:
Museum Tuscalanum Press.
Lahousse, Karen & Béatrice Lamiroy. 2012. VOS Word Order in Romance. A grammaticalization
account. Folia Linguistica 2. 387–417.
Lamiroy, Béatrice & Walter De Mulder. 2011. Stages of Grammaticalization across Languages. In
Bernd Heine & Heikko Narrog (eds.), Handbook of Grammaticalization, chapter 24. Oxford:
Lapesa, Rafael. 1983. Historia del español. Madrid: Gredos.
Lazard, Gilbert. 1994. L’actance. Paris: PUF.
Lehmann, Christian. 1982/2002. Thoughts on Grammaticalization. München, Newcastle: Lincom.
Lehmann, Winfred P. 1973. A Structural Principle of Language and its Implications. Language
49, 47–66.
Leiss, Elisabeth. 2007. Covert patterns of definiteness / indefiniteness and aspectuality in Old
Icelandic, Gothic, and Old High German. In Elisabeth Stark et al. (eds.), 73–102.
Function in Language Research. Papers in honour of Christian Lehmann, 231–254. Berlin:
Mouton de Gruyter.
Luraghi, Silvia. 2013. Partitivi nel latino biblico. Studi Italiani di Linguistica Teorica e Applicata
42(1). 41–60.
Luraghi, Silvia. Unpublished ms. Partitives and differential marking of core arguments : a cross-
linguistic survey.
Marchello-Nizia, Christiane. 1992. L’évolution du français. Ordre des mots, démonstratifs, accent
tonique. Paris: Armand Colin.
Meillet, Antoine & Joseph Vendryes. 1927. Traité de grammaire comparée. Paris: Champion.
Melis, Ludo. 2003. La préposition en français. Paris: Ophrys.
Posner, Rebecca. 1996. Las lenguas romances. Madrid: Cátedra.
Real Academia Española / Asociación de Academias de la Lengua Española. 2009. Nueva
gramática de la lengua española. Madrid: Espasa.
Regula, Moritz & Josip Jernej. 1975. Grammatica Italiana descrittiva su basi storiche e psicholo-
giche. Berne: Francke Verlag.
Renzi, Lorenzo. 1982. Il vero plurale dell’articolo uno. Lingua Nostra 43. 63–68.
Renzi, Lorenzo, Giampaolo Salvi & Anna Cardinaletti. 2001. Grande Grammatica di Consulta-
zione, vol. 1, 371–437. Bologna: Il Mulino.
Sabatini, Francesco. 1985. L’italiano dell’uso medio: una realtà tra le varietà linguistiche italiane.
In Geunter Holtus & Edgar Radtke (eds.), Gesprochenes Italienisch in Geschichte und
Gegenwart, 155–184. Tübingen: Narr.
Schroeder, Christophe. 2006. Articles and article systems in some languages in Europe. In: G.
Bernini Ed. Pragmatic organization of discourse in the languages of Europe. Berlin: Mouton
de Gruyter.
Serbat, Guy. 1996. Grammaire fondamentale du latin. Leuven: Peeters.
Serianni, Luca. 1989. Appunti di grammatica storica italiana. Roma: Bulzoni.
Sornicola, Rosanna. 1994. On Word-order Variability: A Case-study from a Corpus of Italian.
Lingua e Stile 29(1). 25–57.
Spore, Palle. 1987. L’article partitif en italien moderne. Revue Romane 22(1). 30–39.
Stark, Elisabeth. 2007. Gender, number and indefinite articles. About the typological inconsis-
tency of Italian. In Elisabeth Stark et al (eds.), 49–71.
Stark, Elisabeth, Elisabeth Leiss & Werner Abraham (eds.). (2007). Nominal Determination:
Typology, Context Constraints, and Historical Emergence. Amsterdam: Benjamins.
Tekavčic, Pavao. 1980. Grammatica Storica dell’italiano. Bologna: Il Mulino.

Väänänen, Veikko. 1981. La préposition latine de et le génitif. Recherches et récréations latino-
romanes. Napoli: Bibliopolis.
Vennemann, Theo. 1974. Topics, Subjects, and Word Order: From SXV to SVX via TVX. In:
J. Anderson & Ch. Jones Eds. Proceedings of the 1st International Congress of Historical
Linguistics. Edinburgh, Sept. 1993. Amsterdam: Benjamins, 339–376.
Zamboni, Alberto. 1998. Dal latino tardo al romanzo arcaico: aspetti diacronico-tipologoci della
flessione nominale. In Paolo Ramat & Enzo Roma (eds.), Sintassi storica. Società Linguis-
tica Italiana 39, 127–146. Roma: Bulzoni. Bern: Francke.
Zubizarreta, María Luisa. 1999. Las funciones informativas: Tema y foco. In Ignacio Bosque &
Violeta Demonte (eds.), Gramática descriptiva de la lengua española, Vol. 3: 4215–4244.
Madrid: Espasa-Calpe.
VI Oceanic languages
Peter Budd
15 Partitives in Oceanic languages
This chapter provides a survey of morphological partitive forms in Oceanic lan-

guages and discusses the data in terms of semantics, syntactic distribution, and
diachrony. Oceanic partitive morphemes are divided into NP-dependent and VP-
dependent forms. NP partitives include articles and quantifiers, whilst VP parti-
tives, which occur almost exclusively in Vanuatu languages, are post-verbal
modifier forms that have semantic scope over either the object NP or the entirety
of the action denoted by the verb. In a very small number of languages there
are forms that straddle this syntactic division, functioning either as NP or VP
partitives. An indefinite accusative article *ta, reconstructed for Proto Malayo-
Polynesian, and which evolved into a more general marker of indefiniteness in
Proto Oceanic, is held to be the source of many contemporary Oceanic language
partitives (Lynch et al 2002); however, in numerous languages the historical origins
of the partitive forms remain unclear. Oceanic partitive forms may express a
number of other functions, including politeness and aspectual nuance. They
also feature in the expression of negation, entering into Jespersenian cycles as
emphatic minimisers and grammaticalising to varying degrees towards the status
of true negators.
Keywords: Oceanic, partitive, Proto Oceanic, indefinite article, quantifier
1 Introduction
This chapter presents a survey of partitive items in the Oceanic languages, the
large subgroup of the vast Austronesian language family.1 The Oceanic subgroup
stretches from Papua New Guinea (PNG) and the Micronesian islands in the west,
to New Zealand in the south, Hawai’i in the north, and Rapa Nui (Easter Island)
in the east. Lynch et al (2002: ix) estimate membership of the Oceanic subgroup
at between 450 and 600 languages, with the uncertainty over the figure largely
due to inadequate language documentation and description, particularly in parts
of PNG, Solomon Islands, and Vanuatu. Morphological partitives can be identified
1 I am very grateful for the comments and suggestions from an anonymous reviewer, which
have corrected some factual errors and improved the structure of this chapter; all remaining
errors are my own.
524 Peter Budd
in languages throughout the Oceanic subgroup and there is considerable syntactic

and semantic diversity. Efforts have been made to include data from as many
different areal and genetic groups as possible, but in parts there is a focus on
the languages of Vanuatu, a country which alone boasts around 100 distinct
languages. Here the growing number of detailed descriptions has made avail-
able data that allows certain diachronic paths to be illustrated.2
There are no previous Oceanic language studies devoted to partitives, and
whilst the present contribution aims to address this knowledge gap, it is very
much an initial step in what is a potentially huge area for research. Attention is
paid here to diachrony, syntactic distribution, aspectual functions, and the role
of partitive morphemes in negation.
An initial observation is that morphemes with partitive meanings in Oceanic
languages can be divided according to their occurrence as either noun or verb
dependents, though the latter are far more restricted in their occurrence across
the group’s languages. This broad division provides the structure of this chapter:
Following some preliminary discussion in (2), NP partitives are described in (3),
VP partitives in (4), followed by observations on historical origins and diachronic
pathways in (5), including the role of partitives in negation.
2 Background and Preliminaries

2.1 The partitive
The understanding of partitive adopted here starts from a semantic perspective
based on the notions of indefiniteness and quantification, and includes forms
which have scope either over the nominal phrase or the verb phrase.3 A partitive
is thus identified as any quantifying morpheme which denotes:
An indefinite subset of referents
Or,
An indefinite degree to which an action is carried out or to which a situation
pertains
2 The focus on Vanuatu language data also reflects the author’s own research familiarity.
3 The concept of verb phrase as a substitutable unit comprising V+O does not always apply in
Oceanic Languages, and authors of many grammars refer instead to a verb complex, a slightly
different unit which includes the verb plus subject marking, valence-changing, and object
marking affixes, but not the object NP.
Partitives in Oceanic languages 525
Within these broad definitions there are various subtypes, and it is also important
to stress that although a partitive form may clearly have a syntactic association
with the verb, its semantic scope may actually be over the verb’s object or over
the entire predicate.
Notable exclusions from the discussion (for reasons of space, and also for
consistency with other contributions to this volume) are the expression of part-
whole relations (meronymy) e.g. the bottom of the canoe, the branch of the tree
etc, and so-called pseudo-partitive constructions of the type a glass of wine, a
bottle of beer etc.
2.2 Overview of Oceanic subgrouping

The Oceanic subgroup of languages divides into three primary subgroups: the
Admiralties family, the Western Oceanic Linkage (WOc), and the Central/Eastern
Oceanic Linkage(CEOc).4 The data presented here are from the Western Oceanic
and the Central/Eastern Oceanic subgroups since the Admiralties family of lan-
guages is less well described.5 Tables 1 and 2 show the major internal subgroup-
ings within WOc and CEOc with information on the geographical areas and
some example languages which are cited in this chapter.
Table 1: Western Oceanic subgrouping
Western Oceanic Main locations where languages spoken Example languages

North New Guinea (NNG) Huon Peninsula; New Britain (PNG) Manam
Papuan Tip (PT) Southeast tip of Papua (PNG) Tawala, Motu
Meso-Melanesian (MM) North New Britain, New Ireland, Hoava, Kokota, Teop,
Bougainville (PNG); northwest Nakanai, Mono-Alu
islands of Solomon Islands
4 Possible other primary subgroups are the two-member St Matthias family, made up of Mussau
and Tench, and a single-member subgroup of Yapese. See Lynch et al (2002) for detailed descrip-
tion of Oceanic subgrouping.
5 Hamel’s (1994) description of Loniu is an exception.
526 Peter Budd
Table 2: Eastern Oceanic subgrouping
Eastern Oceanic Main locations where Example languages

languages spoken
Southeast Solomonic (SES) Southeast islands of the Toqabaqita, Kwaio, Bugotu
Solomon Islands
Utupua and Vanikoro (UV) Santa Cruz Islands Unavailable
(Solomon Islands)
Remote Oceanic (RO) Vanuatu; New Caledonia Mwotlap, Abma, Araki, Lewo,
Bierebo, Paamese, Neve‘ei, Avava,
South Efate, Sye, Lenakel, Bwatoo
Central Pacific (CP) Fiji, Polynesia, Boumaa Fijian, Māori, Samoan,
New Zealand Tuvaluan, Rapa Nui
Micronesian (Mic) Micronesia Mokilese, Kiribatese, Woleaian
3 NP-dependent morphemes with partitive

meanings
The following section is divided into descriptions of the syntax and semantics
of NP-dependent forms which express partitive meanings according to the defini-
tion presented above in 2. In the subsection on semantics a typology of three
identifiable partitive meanings is proposed, based on the Oceanic data reviewed.
3.1 Syntax of NP partitives

Oceanic languages display variation in terms of the word class, relative word
order within the NP, and pluralisation of partitive markers.
3.1.1 Syntactic status of partitives
NP forms which express partitive meanings are typically indefinite articles and
quantifiers. The precise semantic and syntactic distinctions between articles
and quantifiers are language specific, but a typical syntactic difference is that
quantifiers are optional elements in the NP and often bear distributional similar-
ities with numerals, such as the ability to function pronominally; articles on the
other hand are often obligatory elements in the NP and cannot function as
heads. Semantically, or functionally, articles typically code the identifiability or
specificity of a noun’s referent, whilst quantifiers serve to express the quantity

(countable or uncountable) of a noun’s referent. It will be seen that the partitive
concept cross-cuts these two notions and it is therefore not surprising that some
languages express partitive meanings through articles, some through quantifi-
ers, some through both, and some through forms that are not easily classified
as one or the other.
By no means do all Oceanic languages have articles, and there is consider-
able variation with regard to the functions encoded.6 In some cases there are
familiar definite versus indefinite contrasts, but elsewhere specificity, animacy,
and humanness may be the salient features. However, there are numerous ex-
amples of articles which clearly convey partitive meanings. To illustrate briefly,
Samoan is a language which employs articles, some of which have an evident
partitive function. There is a singular indefinite article se in Samoan, which
expresses the classic indefinite (non-identifiable) feature of the referent; then
like many other languages, indefiniteness for plural entities can be expressed
by the absence of an article:
(Samoan)
(1) a. ‘O se ta’avale
pres INDF.SG car
‘A car.’
b. ‘O ta’avale
pres car
‘(some) car.’
If the indefiniteness of the quantity of referents is to be emphasised in some

way, then there is a plural indefinite article ni which may be used, c.f.:
c. ‘Aumai se tusi
bring INDF. SG book
‘Bring a book.’
d. ‘Aumai ni falaoa
bring INDF. PL bread
‘Bring some bread’
(Hunkin Galumalemana 2009: 14ff)
6 In fact, reflexes of a Proto Oceanic definite article *na are probably more widespread than
reflexes of indefinite markers in modern Oceanic languages. Often, however, this form *na
has fused to the noun root and thus lost its referential function.
528 Peter Budd
There is a further article form, sina, which expresses a diminutive-partitive

meaning:
(Samoan)
(2) ‘Aumai sina wai
bring ART (PAR . SG ) water
‘Bring a little water.’
(Mosel 1992: 265)
Mokilese, on the other hand, is a language which expresses partitive mean-

ings through quantifier forms, such as ekij ‘some, a little’ and epwi ‘a few’.
Harrison (1976: 108) describes these quantifiers in terms of their similarity with,
and difference from numerals: they are “morphemes that describe the quantity
or amount of people or objects being discussed, without referring to a specific
number:
(Mokilese)
(3) Ngoah kin mwehuki ekij juke nehn nimoai koahpi
1SG usually like some sugar in my coffee
‘I like a little sugar in my coffee.’
(4) Mine epwi pwilei-n suhkoa pei-pei-pei nehn lammo

exist a.few log-CONST tree float-RED - RED in lagoon
‘There are some logs floating in the lagoon.’
(Harrison 1976: 108; glosses added)
The partitive-marking forms in some languages defy straightforward word-

class categorisation. Lewo (Vanuatu), for example, has a form tai, (glossed ‘ART ’
or ‘one’) which Early (1994: 248) describes as an indefinite article, based on its
primary function of marking indefiniteness and non-specificity, and also on its
position in the NP and co-occurrence with other NP modifiers. However, it has
other functions which are less article-like – as a nominal proform, and as a
post-verbal modifier with a partitive meaning. The following examples illustrate
the contrasts:
(Lewo)
(5) Lokuli keviu tai
dog big ART
‘A big dog.’
(Early 1994: 248)
(6) Tai ø-ka-n tai, tai ø-ka-n tai

one 3SG -eat-TR one one 3SG -eat-TR one
‘One person/some people ate one/some, another person/other people
ate one/some (i.e. everyone ate one/some each).’
(Early 1994: 249)
(7) Sisi malum nene ø-mai tai

child soft DEIC 3SG -sick ART
‘This delicate little baby was a bit sick (had a little sickness).’
(Early 1994: 271)
The multifunctionality of this form is revisited in Section 5 as it is insightful in

the analysis of the diachronic pathways of partitive markers.
To return to the definition of partitive adopted here – an indefinite subset of
referents – it can be seen that this designation essentially combines the semantics
of indefiniteness and quantification. And there are in fact Oceanic languages
where the duality of this notion is apparent in the composite nature of the
morpheme. Tuvaluan, for example, has a form niisi which Besnier (2000: 582)
describes as being “combined from the plural indefinite article ne and isi a
quantifier meaning ‘some, other’” and “because it results from the combination
of an article and a quantifier, niisi functions as both, and it can mean ‘(some)
other’ as well as simply ‘some’:
(Tuvaluan)
(8) Niisi e iinu ona ko te mmao keaattea mai
Some NPS drink because FOC the distant away from
olotou kaaiga
their family
‘Some start drinking because they are away from their families.’
(Besnier 2000: 583)
(9) Au e hanatu hanatu, koo ttuu mai mo niisi

1SG NPS go:DEIC go:DEIC INC stand DEIC with some
tamaliki tamaafine
child young.woman
‘I was walking along and [saw that she] was standing [there] with some
other young women.’
(Besnier 2000: 583)
530 Peter Budd
Similarly, Yapese has a form boech ‘some, a little’ which is analysed by Jensen
(1977: 160) as a combination of the indefinite article ba and the diminutive chi).
It occurs with a connector ee or ea in NPs to give a partitive meaning:
(Yapese)
(10) boech ee niig
Some CONN fish
‘Some fish.’
(Jensen 1977: 160)
Where languages like Yapese and Tuvaluan have complex fused forms, there are
also languages which exhibit combinations of independent words – articles,
quantifiers and nominals – or synchronically segmentable morphemes. Boumaa
Fijian combines a quantifier form vica with an article, Bwatoo uses a nominal
form with the meaning ‘a part (of)’ combined with a definite marker to express
a partitive meaning, whilst Kwaio can combine a nominal, gula ‘part’, with a
qualifier ta ‘some’ and a plural marker ni:
(Boumaa Fijian)
(11) vica tale a vula
some modif ART month’
‘Some months.’
(Dixon 1988: 142)
(Bwatoo)
(12) Ani bee xaabuit
DEF part person
‘Une partie des gens (some of the people).’
(Ehrhart & Rivierre 2006: 437)
(Kwaio)
(13) gula-ta-ni ‘ola
part-QLF- PL thing
‘Some things.’
(Keesing 1985: 92)
As evidenced by the examples shown so far, Oceanic languages also vary

with regard to the relative word orders of noun head and partitive marker, and
of the partitive marker and pluralising forms. In many Polynesian languages
of the Central Pacific subgroup the partitive marker, which is often an article,
occurs before the noun (e.g. Tongan ha, Māori he, etahi, Samoan sina, Tuvaluan
niisi etc); to the west, the picture is more mixed: pre-nominal partitive markers
are attested in e.g. Kwaio and Toqabaqita (South East Solomonic); in the Meso
Melanesian group Kokota also has a pre-nominal form keha but Mono-Alu has
post-nominal aiina; within Southern Oceanic many Vanuatu languages have
post-nominal partitive forms (e.g. Lewo tai, Avava tuan (ier), Neve‘ei tuan (ar),
Bierebo ta, Lenakel nɨvin), though there are other Southern Oceanic languages
where the markers occupy a pre-head slot in the NP, such as South Efate tete
and Mwotlap te and Bwatoo bee.
Some Oceanic languages have separate partitive markers for singular and
plural reference. For example, Central Pacific languages often have rich para-
digms of articles and the Samoan forms se, ni, and sina shown above are an
example of this. Elsewhere the indefinite marker does not necessarily indicate
singular or plural. In the following examples from Bierebo the reference is
singular in the first, but plural in the second:
(Bierebo)
(14) Te-te kia-da tawaga ta
1PL . INC - IRR :cut GCL-1 PL . INC . PSR canoe INDF
‘Let’s cut (make) ourselves a canoe.’
(15) No-pwar ka ta
1SG - R :carry PRG INDF
‘I’ve got some (cigarettes).’
However, if plurality is to be emphasised, then in some languages it is possible

to add a plural marker. Thus in Bierebo lala ~ nala, which precedes ta. is used:
(16) Manu lala ta

bird PL INDF
‘Some birds.’
(Author’s data)
In the nearby Malakula languages of Avava and Neve‘ei which also have post-
nominal partitive markers, the relative order of indefinite/partitive and pluraliser
is the opposite of Bierebo:
(Neve‘ei)
(17) Nemwen tuan ar at-tovu ‘out
man some 3 PL 3 PL . R- go ashore
‘Some men went ashore.’
(Musgrave 2007: 69)
532 Peter Budd
(Avava)
(18) Na-som-da abak tuan ier
1SG : R-see-hit turtle INDF PL
‘I found some turtles.’
(Crowley 2006: 63)
NP partitive forms also exhibit variation in terms of distribution: whether

limited as NP dependents, or whether they may also function as heads (i.e. pro-
nominally). In addition, there may be interaction with certain TAM verb operators.
For example, François (2001: 335) reports that the minus referential value of the
Mwotlap partitive marker te makes it most compatible with semantically irrealis
clauses. In Bierebo, by contrast, the indefinite form ta ‘one, some’ occurs in both
realis and irrealis marked clauses; it is distinguished from the true numeral ‘one’
by its behaviour with respect to realis and irrealis clauses. All Bierebo irrealis
clauses which feature a numeral must mark the numeral using an Irrealis-
inflected copula, c.f.:
(Bierebo)
(19) a. Ne-sa ø-ve tolu
1SG -eat 3SG - IRR :be three
‘I’ll eat three (lit. I will eat it will be three).’
b. Ne-sa tolu
1SG -eat three
‘I ate three.’
c. *Ne-sa ø-pe tolu

1SG -eat 3SG - R :be three
When an unspecified quantity is intended instead of a specific number, the form

ta is used, but it cannot be preceded by the irrealis copula ve, and so ambiguity
of reality status results:
d. Ne-sa ta
1SG -eat INDF
‘I ate some / I will eat some.’
(Author’s data)
The syntactic variation described above can in part be explained by the

historical origins and diachronic pathways of the forms in question which are
discussed in (5). Before that, however, a description of the semantics of NP-
dependent partitives is presented in (3.2) followed by an account of the verbal

partitives in (4).
3.2 A typology of meanings in NP partitive forms

Within the broad definition of partitive adopted here for NP partitives (a subset
of referents) there are three subtypes of meaning that can frequently be distin-
guished in Oceanic languages. They are represented graphically in Figure 1:
Figure 1: Subtypes of nominal partitive meanings in Oceanic languages
The partitive meaning in A is an indefinite subset of all possible members of that

class of entities; B refers to an indefinite subset of an identifiable group; and C is
one of a pair (or small number) of identifiable entities. The semantic distinction
between A and B can be expressed in English through some versus some of (a
given group), e.g. some men versus some of the men (standing over there), whilst
the semantics of Subset C is equivalent to ‘one’ / ‘the other’.
The distinction between some and some of is similar to the distinction high-
lighted by Luraghi (this volume: 4ff) between a partitive and a partitive nominal
construction (following Koptjevskaja-Tamm’s definitions 2006: 218): In the parti-
tive nominal, the construction is referential, inviting the hearer to select from an
available, previously mentioned or pre-supposed group. Luraghi exemplifies the
partitive nominal construction (this volume: 4) with the English a bottle of that
wine, the Dutch van de cake, ‘of the cake’ and the German ein bisschen vom Käse
‘a bit of the cheese’, where in each case the nominal that refers to the entity
from which the subset is drawn is established or accessible. However, these
partitive nominal constructions differ from the Subset B type of partitive defined
here for Oceanic languages in that the former can include examples such as a
bottle of that wine, in which there is a definite quantity specified. By contrast,
in the Oceanic Subset B the quantity of the subset is always an indefinite
amount.
The structural distinction between English some versus some of can be rep-
resented in the formulae below, and this is mirrored in some Oceanic languages,
such as Rapa Nui tetahi versus tetahi + no.
534 Peter Budd
QUANTIFIER (Subset A ‘some’)

QUANTIFIER + of + N (Subset B ‘some of N’)
Similarly, Tuvaluan contrasts niisi versus niisi + o:
(Tuvaluan)
(20) Niisi tino e see tii oota
Some person NPS NEG often eat-raw
‘Some people don’t eat raw [fish].’
(21) Kee fakamaagalo mai niisi o oku fakamaaseiiga ne fai

SBJV forgive DEIC some of my CAUS :bad:NMLZ PST do
‘Let some of the bad things I’ve done be forgiven.’
(Besnier 2000: 331)
More common, though, would appear to be the situation where there are two or
more seemingly unrelated forms reflecting the ‘some versus some of ’’ distinc-
tion. Thus, Boumaa Fijian, for example, has the forms vica and soo. Dixon
(1988: 143) describes vica as having “an indefinite sense of ‘some’ ”, illustrated
by the following:
(Boumaa Fijian)
(22) Ia, e rawa me tei ‘ina ia ‘ani-a e
well, 3SG be.able should planted prp:3sg well eat-TR 3SG
vica tale a vula i-na loma ni yaba’i mai muri
some MODIF ART month in:ART interior associated year come follow
‘It is possible for them to be planted then but we then can’t eat them until
some months into the following year.’
(Dixon 1988: 143)
By contrast, “soo has a rather different sense “some of (a given population)”.

This is illustrated by e.g. i na soo a vei-ti’i-i Viti ‘in some parts of Fiji’ and the
following glossed example:
(23) E rawa ni sa-qei tuu e soo a suli-na. . .

3SG be.able that ASP-then be.at 3SG some.of ART sucker-3SG
‘It is possible to keep some of (the taro’s) suckers (and replant them. . .).’
(Dixon 1988: 144)
In Māori, Bauer (1997: 294) describes how he expresses ‘some’7 whilst (t)ētahi
‘a/some’ followed by o + NP expresses ‘some of’ (ibid: 491):
(Māori)
(24) Kawe-a atu he wai ki a au
Bring-pass away a water to PERS 1SG
‘Bring me some water.’
(25) ka haere tahi atu anō i a rātou ētahi o te

TAM move one away again at PERS 3PL some of the
tāngata whenua
people local
‘. . . some of the local people went with them as well.’
(Bauer 1997: 294)
In the Vanuatu language Avava the distinction between ‘some’ and ‘some of ’ is
also reflected in the presence of two different forms: tuan (ier) and tuut (ier)
respectively:
(Avava)
(26) Na-som-da abak tuan ier
1SG : R-see-hit turtle INDF PL
‘I found some turtles.’
(27) Komat-yan mwiniel i moroko-n tuut ier

1PL . EXC . R-eat taro INS rib-3SG some PL
‘We ate the taro with some of its ribs.’
(Crowley 2006: 63)
And in Toqabaqita ta ‘some’ and tootoqe ‘some, not all from a set’ contrast:
(Toqabaqita)
(28) Kuki ta kafo
cook some water
‘Boil some water.’
7 For discussion on the distribution and semantics of Māori he, see e.g. Polinsky (1992) and
Lyons (1999).
536 Peter Budd
(29) tootoqe kini bana

some woman only
‘Only some [of the] women.’
(Lichtenberk 2008)
It is on the basis of this kind of data that the distinction has been posited
between the two types of partitive subsets A and B.
Turning to the other distinctions (Subset A versus Subset C, and Subset B
versus Subset C), Bierebo is a language which illustrates each of the three parti-
tive meanings distinguished in Figure 1 with a different form: ta, an indefinite
marker; tiala ‘some (of)’; and kurua ‘one, other’. Firstly, ta, a post-nominal or
pro-form morpheme, marks indefiniteness, which, as we have seen, may be
either singular or plural:
(Bierebo)
(30) Nala chuwa ko-na a-ban a-do chel ta
3PL two spouse-3SG . PSR 3PL- R : go 3PL- R : stay place INDF
‘She and her husband went and lived at some place.’
(31) No-pwar ka ta
1SG - R :carry PRG INDF
‘I’ve got some [cigarettes].’
(Author’s data)
Secondly, tiala, which, like ta is also either a post-nominal modifier or a pro-

form, expresses the idea of ‘some of’ i.e. an indefinite quantity (of more than
one) of an established, identifiable group:
(32) Kumemi tiala a-m-yel

1PL . EXC some.of 3PL- R- walk
‘Some of us (exclusive of the addressee) left.’
(33) No-pwar-pla tiala a-pitove tano rui na

1SG - R :carry-remove some.of 3PL- R :go.down down already EMPH
‘I’d picked some (of the tree’s breadfruit) and brought them down already.’
(Author’s data)
Finally, for Subset type C, kurua ‘one, other’, which is formally identical to the
noun meaning ‘brother’, is used to refer to one of an established pair or small
group of entities. The following examples are from a traditional story about two
unnamed sisters who are each referred to using kurua to mean ‘one/the other’:
(34) Na kurua ø-pwia sa-na pwapwa

and one 3SG - R :squeeze ECL-3 SG . PSR laplap
‘One squeezed (coconut milk) onto her (own) laplap pudding.’
(35) Lala kurua-na, yirmwene ø-pwia sa-na pwapwa

3PL one-3SG . PSR man 3SG - R :squeeze ECL-3 SG . PSR laplap
‘As for the other, the man squeezed coconut milk onto the pudding.’
(36) Kurua ø-sa-le kurua sa-na

one 3SG -eat-try one ECL-3 SG . PSR
‘One tasted the other’s’
By contrast, in other languages a single morpheme can convey two or three of

these partitive functions. In Lenakel there is a form ker ‘one’ which can express
Subset A and Subset C types of partitive meaning:
(Lenakel)
(37) Nian ker, ker, rɨ-vɨn m-aamh ker
day one one 3SG -go and-see one
‘One day, one went and saw the other.’
(Lynch 1978: 40)
This contrasts with another Lenakel form, nɨvin ‘some (of)’ which in the
following example expresses the Subset B ‘some of’ partitive:
(38) N-eramim ka nɨvin k-n-ar-va ita

PL-person that some 3NS - PRF- PL-come already
‘Some of those people have already come.’
(Lynch 1978: 40)
(39) Ofa nɨmɨlh un nɨvin!

give.to.speaker orange that some
‘Give me some of those oranges (near you)!’
(Lynch 1978: 41)
This suggests that in Lenakel it is not referentiality, but rather, quantity or num-
ber, which is the feature that determines which indefinite form is used: nɨvin is
+referential and +plural, while ker is –plural but may be either + or –referential:
538 Peter Budd
Subset A Subset B / C
ker -referential, – OR +referential, -plural ‘one /
-plural ‘one day’ the other (of the Ns)’
nɨvin +referential, +plural ‘some of those Ns’
Another Southern Oceanic language of Vanuatu, South Efate, has a pre-

nominal quantifier form tete ‘some’ which in the following example expresses
an indefinite quantity (Subset type A):
(South Efate)
(40) Ra=to tu tesa tete nanromien
3D. R = STAT give child some present
‘They would give children some presents.’
(Thieberger 2009: 270)
Equally, it may occur in contexts where its partitive meaning is that of Subset
type B ‘some of’:
(41) Ra=pam tete nafnag ne

3D. R =eat:R some food this
‘. . . they ate some of this food.’
(42) A=mur-i-n na pwa-traem

1SG . R =want-TR-3 SG .OBJ say 2SG . IRR =try
ga preg tete nalkis gag
3SG . BEN make some medecine 2SG . PSR
‘I want you to make some of your medecine for him.’
Similarly, Kokota has a form keha which marks non-specificity and whose
meanings include that of Subset type A, as in the following examples, where
the referents of ‘kerosene’, ‘Maringe men’, and ‘Hograno men’ all represent the
total class of entities:
(Kokota)
(43) ara n-a-ke manahagi=di keha no-gu karesini
1 R-1 EXC . SBJ- PFV want-3PL . OBJ NSP GENP-1 SG . PSR kerosene
‘I needed some kerosene.’
(44) Ge mat la keha mane marihi keha mane hograno

NT-3 S come CND NSP man PNLOC NSP man PNLOC
‘If some Maringe man or some Hograno man came. . .’
But keha can evidently also refer to a Subset B partitive, as in the following
example, where the set of copies is an accessible, identifiable group to the
hearer:
(45) Keha pile di-re no-na bla tagi-na

NSP part 3PL . POSS -those GENP-3 SG . PSR LMT REFL-3 SG . PSR
‘Some copies will just belong to him.’
(Palmer 2009: 82)
It was shown in earlier examples above that Tuvaluan niisi can express both
Subset A and Subset C, whilst the expression of Subset B is achieved by merely
adding the preposition o; Lewo has a similarly multifunctional form tai, which
can express all three meanings. Subset types A and B are clear in the following
two examples respectively,
(Lewo)
(46) A-te-ro pur-kurki tai
3PL-cut-in.two tree-K .nut ART
‘They cut down a tree.’
(Early 1994: 248)
(47) Omami tai ø-mai

1PL . EXC ART 3SG -sick
‘One of us is sick.’
(Early 1994: 250)
Early (1994: 249) specifies that when tai is “placed in contradistinction with a
noun phrase quantified with tapwena ‘different, other’ or with another noun
phrase with tai, this article has a contrastive or distributive function, expressing
‘other/the other’ ”, thus illustrating Subset type C. Note also from the translation
that either singular or plural reference is possible:
(48) Tai ø-ka-n tai, tai ø-ka-n tai

one 3SG -eat-TR one one 3SG -eat-TR one
‘One person/some people ate one/some, another person/other people ate
one/some (i.e. everyone ate one/some each).’
(Early 1994: 249)
540 Peter Budd
Clearly this typology of three types of partitive meanings in Oceanic lan-

guages (some, some of, one/the other) is not exhaustive. Additional subtypes
might include specific or non-specific partitives, for example, or diminutive par-
titives, in which the indefinite subset is not simply designated as indefinite, but
rather, is specified as being relatively small in quantity. Examples of this include
Mono-Alu aiina, Samoan sina, and Yapese boech, all mentioned above. These
types and others may well transpire to be equally valid subtypes of partitive
meanings but they have not been proposed here since there was insufficient evi-
dence of systematic semantic distinctions reflected by formally discrete markers
across the languages reviewed.
4 Post-verbal forms with partitive meanings

This section describes ‘partitive’ morphemes that occur post-verbally. They are
included in the discussion since they have an identifiable partitive meaning in
the terms set out above in (2), expressing:
An indefinite, partial degree to which an action is carried out or to which a

situation pertains
Almost all the examples of post-verbal partitive forms come from Vanuatu lan-
guages, though there is also an example from Yapese, and further research may
well reveal the phenomenon to be more widespread.
Some details about the syntax of these forms are presented in 4.1 followed
by discussion of semantics in 4.2. It will be shown that, in addition to the core
partitive meanings, these forms can be used to express politeness and aspectual
nuance.
4.1 Syntax of post-verbal partitives

Syntactically, Oceanic post-verbal partitives are often described as verb modifiers,
encompassing clitics, particles and quantity adverbs. They are most often attested
with transitive verbs, but there are examples of occurrence with intransitives
too, as illustrated by the first of the Bierebo examples below:
(Paamese)
(49) [Ma-ani-tei] raise
1SG . IMM -eat-PAR rice
‘I’d like to eat some rice.’
(Crowley 1982: 145)
(Sye)
(50) [U-ovo-yau-wi] nacave
PL : IMP- BR :give-1SG - PAR kava
‘You all give me a little kava’
(Crowley 1998: 129)
(Bierebo)
(51) Ø-pavin ja=nga
3SG - R :go.up a.little=LMT
‘He went up just a little’
(52) Ne-vre-i ja histori na yo Bonkovio

1SG - IRR :say-about a.little history PRP place Bonkovio
‘I will tell a little of the history of Bonkovio’
(Author’s data)
By definition, these forms are restricted to a post-verbal position, in a slot

which may form part of the verb complex,8 as in the Paamese and Sye examples
above (where the bracketing indicates the boundaries of this unit), but alterna-
tively may fall outside this unit, as in Araki:
(Araki)
(53) Nam dogo na-sile-ko] n-re presin
1SG . IRR feel 1SG . IRR-give-2SG OBL-some present
‘I feel like giving you a present.’
(François 2002: 161)
The Bierebo form ja, in addition to modifying verbs, may occur after adjectival
forms9 and the negative form mara (though these are both of verbal origin):
8 It is recalled that the verb complex is a unit described in many Oceanic languages and is
usually defined in phonological, morphological, and syntactic terms, typically comprising the
head verb which may be affixed or cliticised by a variety of subject/object, transitivity, and
TAM markers, whilst the object NP is not included.
9 The criterion for verbhood in Bierebo is the occurrence of subject marking prefixes which are
not possible with the predicates in these examples.
542 Peter Budd
(Bierebo)
(54) Ø-je dupwa na ø-pe niemian ja=nga
3SG - R :see ANA CONJ 3 SG - R :be angry a.little=LMT
‘He saw this and became a bit angry.’
(55) na mara ja ø-sina-le ya chora nani wa

CONJ NEG a.little 3SG -eat-try LOC garden DEM yet
‘. . . she hadn’t quite yet harvested the garden (lit. hadn’t eaten from
the garden quite yet).’
(Author’s data)
In the case of transitive clauses, despite syntactic (and, in some cases, phono-
logical) association with the verb, these partitive forms would seem to have
semantic scope over the object noun in the same way that NP partitives do, by
delimiting a subset in some way. Crowley (1982: 144) states that in the Paamese
example above, “the referent of the object is an indefinite subset of the total
possible class of objects . . . the object raise ‘rice’ . . . refer[s] to . . . some indefinite
quantity of rice”.
It is important to re-iterate then, that the distribution of tei (and similar
forms in other languages) is quite different from the NP partitives described
above in (2). For example, tei cannot occur as a NP dependent to mean ‘some
rice’ (*tei reise), nor as a verb subject, nor in an existential clause, and indeed
cannot occur independently of the verb at all. The same applies for Mwotlap te,
which (François 2001: 336) reports can only occur in a slot between the verb and
its NP object; Bierebo ja is similarly restricted to this immediately post-verbal
position, as is the Lewo re.
As shown above, these partitive forms are prone to phonological binding
with adjacent forms, such as the transitiviser n- in the Araki example, and the
limiter nga in Bierebo. A feature of the usage of Mwotlap post-verbal te is its
frequent co-occurrence with possessive classifiers. Thus,
(Mwotlap)
(56) Nok so in te m-ek bē
1SG PRSP drink PAR DCL-1 SG water
‘I want to drink some water.’
4.2 Semantics of post-verbal partitives

Semantically the scope of post-verbal partitives is over the object NP (in bivalent
clauses) or over the entire predicate. The Bierebo post-verbal particle form ja
which occurs with both monovalent and bivalent verbs illustrates this:
(Bierebo)
(57) Ne-to ja wa
1 SG - IRR : wait some yet
‘I’ll wait a bit.’
(58) Ko-sa ja yuwi?

2SG -eat a.little yam
‘Will you eat some yam?’
(Author’s data)
When post-verbal partitives occur with an object NP there are in theory two
possible readings, which are distinguishable to varying degrees: In the first
reading, the object NP’s referent is partially affected (Figure 2), and in the
second reading, only part of the NP’s referent is affected (Figure 3):
Figure 2: Partial affectedness of an entity
Figure 3: Affectedness of part of the entity
The following Bierebo example illustrates a case of an entity being partially

affected:
(59) mwigo na a-piling ja kelerigo dupwa=ni

then EMPH 3PL- R :turn a.little troll ANA =DEM
‘Then they moved the [sleeping] troll a little.’
(Author’s data)
In the story’s context the sleeping creature is turned (from his side onto his back),
i.e. the whole body is partially turned, rather than just a part of the creature’s
body being turned.
544 Peter Budd
This is contrasted with another Bierebo example in which the speaker states
his intention to tell some of the second part of a story. One would therefore
analyse this as a complete accomplishment of one part of the whole:
(Bierebo)
(60) ne-vre-i ja namba tu pat
1SG - IRR :say-about a.little number two part
‘I’ll tell some of part two of the story.’
(Author’s data)
Similarly, this second reading –of part of the entity being affected– is illus-
trated by the following Abma example. Schneider (2010: 166) states, that the use
of te in Abma “refers to some portion of a whole NP, expressing an indefinite
quantity (for example, ‘some’). . . , inclusion of te ‘partitive’. . . highlights the
fact that the amount of bwet ’taro’ to be grated is imprecise”, i.e. a part of the
whole is wholly affected:
(Abma)
(61) Ba nanong, ba na=ma sawiri=te bwet si=ah
COMM now COMM 1SG = PRSP grate=PAR taro POL = EMPH
‘But now I’ll grate some of this taro first.’
(Schneider 2010: 166)
To judge from the available data it seems unlikely that there are separate
morphemes which systematically encode these distinct meanings. Rather it ap-
pears that the reading is inferred, based on other factors, such as the semantics
of the verb and of the object NP, as well as the discourse context; potentially,
there is therefore a degree of ambiguity in some cases.
4.3 Aspectual and other effects of post-verbal partitives

Expression of incomplete accomplishment of an action can lead to interpreta-
tions of aspectual nuance. Schneider (2010: 167) makes precisely this analysis
of the “partial execution of an event” for Abma (Vanuatu). In the following
example she states that “an attempt to break is marked with the partitive,
whereas a successful execution of the verb takes regular transitive marking”
[emphasis added]:
(Abma)
(62) Kaa=ga, u=bma ne-bwah=te
2PL = LMT ADD =come CONN -break=PAR
‘Now you guys come and try to break it.’
(63) Mwa=bwah-a, ra=mwa bwah-a vet nong

3SG . IPFV =break-TR 3PL = IPFV break-TR stone this
‘He breaks it, they break this stone.’
And similarly, the same author compares the following example, in which a
series of ‘tentative’, incomplete actions is expressed by the partitive (in conjunc-
tion with the repetition of the verb), with an example of the Finnish partitive
which expresses unbounded action, and therefore an imperfective interpretation:
(Abma)
(64) Nema siba=te ba, nema siba=te ba
3SG . PRSP peel=PAR COMM 3SG . PRSP peel=PAR COMM
Mabonmwel nema siba i biri bu

Mabonmwel 3SG . PRSP peel INS small knife
‘She’ll peel it, she’ll peel it, Mabonmwel will peel it with a small knife.’
(Schneider 2010: 166ff)
(Finnish)
(65) Tyttö lakaisi lattiaa
girl:NOM swept floor:PAR
‘The girl was sweeping the floor.’
(Lyons 1999: 101)
For Sye/Erromangan (Vanuatu), Crowley (1998: 129) also states that for intran-
sitive verbs the partitive -wi can be used to express the fact that the action de-
noted by the verb is performed only to a limited extent. However, as no examples
are provided it is impossible to tell if similar atelicity can be conveyed by -wi.
In Araki there are examples of the partitive marker re apparently expressing
the opposite aspectual meaning i.e. perfectivity. As well as its prototypical func-
tion of marking indefinite quantity for object NPs (in Irrealis clauses), where re
occurs post-verbally, it also functions as an aspectual marker in Realis clauses,
occurring in the regular pre-verb TAM slot. Here, François reports that it signals
that a certain amount of the verb’s action has been accomplished:
546 Peter Budd
(66) Nam dogo Na inu re hae

1SG . R feel 1SG . IRR drink some kava
‘I feel like drinking some kava.’
(67) Nam re inum-i-a hae

1SG . R PFV drink-TR-3 SG kava
‘I have drunk the/some kava.’
François (2002: 61) argues that in the Realis occurrences of re the perfective
meaning has emerged from the fact that the marker has scope over the verb
and its object, such that semantically . . . inu re hae is not so much ‘carry out
the action of drinking upon a certain amount of kava’, but would more accu-
rately be glossed as ‘do a certain amount of kava drinking’. In turn, the meaning
develops from ‘there is a certain amount of such an action’ to the perfective
‘such an action has been done’ (ibid: 62). (For further discussion of Araki, see
Luraghi this volume: 29ff ).
Similar logic can arguably be applied in the following Bierebo example.
Here the post-verbal partitive ja modifies the meaning of kititi ‘run’ to mean not
‘run a little’, but to ‘have a race’. In other words, ja is not expressing an
unbounded or tentative performance of an action as in the Abma examples, but
rather is expressing the action as a bounded and complete event:
(Bierebo)
(68) E peni te-kititi ja wa
hey tomorrow 1PL . INC - IRR :run a.little yet
‘Hey, let’s have a race tomorrow.’
In a similar vein, the use of ja with the verb che ‘look’ seems to modify its mean-
ing to a more punctual action of ‘having a look’:
(69) Me-sidom me-vitove me-che ja

1PL . EXC -want 1PL . EXC - IRR :go.down 1 PL . EXC - IRR :see a.little
ya na a-mwaka ka=nia
what EMPH 3PL-make INS =3OBJ
‘We wanted to go down and take a look at what they were doing there
(lit. making with it).’
(Author’s data)
For the Araki example, the fact that re has moved to a position normally occupied
by aspect markers in the language provides some grounds for analysing it in the
terms that François does, though it would be useful to see more examples to
support the claim that it has grammaticalised into a perfective marker.10
For other Vanuatu languages the presence of a partitive marker with scope
over the verb or the verb + object NP clearly has a semantic effect which can
sometimes be construed as having some aspectual nuance; yet there is insuffi-
cient evidence to draw general conclusions at this stage on what that nuance
could be predicted as, and it seems that, again, much depends on the lexical
semantics of the verb and the object in determining these meanings.
Early (1994: 271) provides some interesting examples of Lewo tai and re: The
indefinite marker tai can occur with intransitive verbs to indicate a “reduced
instance of the situation expressed by the verb”. In the two examples given, tai
conveys that the subjects are not “‘fully sick’ or ‘asleep for the night”, but rather
have ‘a minor ailment’ or are ‘taking a short nap’ respectively”:

(71) Ne-m-ligan ø-mwe ke ø-mo-malio ke tai nini

1SG - R-put 3SG - R .stay TA 3SG -lie-sleep TA ART DEIC
‘I put the child to stay sleeping a little here (for a little sleep).’
(Early 1994: 271)
For re, he describes its effects in the following terms: “This particle indicates
that the action of the verb is only partly accomplished, or limited in application
or effect in some way. This includes actions that might only have been carried out
occasionally . . . or actions whose importance or significance is de-emphasised . . . .
(72) Ø-vano-n Ø-pisu re terasi lala a-kla kie-la maru lala

3SG -go-TR 3SG - R : see PAR shoot PL 3PL-make POSS -3PL copra PL
‘Because he had sometimes seen (happened to see) the older teenage
boys making their copra.’
10 One anonymous reviewer also pointed out the possibility of this occurrence of re being an
unrelated homophonous form.
548 Peter Budd
(73) Ø-piyun ko Ø-sape sine-na-n Ø-sape Ø-visu ko re

3SG - R :ask 2SG 3SG -say gut-3SG . P-TR 3SG -say 3SG - I :see 2SG PAR
‘He asked for you saying that he wants to see you (just wants to have a
bit of chat with you).’
(Early 1994: 261)
Early (1994: 541ff) also explains how the use of the partitive re conveys a non-
assertive tone such that “an element of politeness [is] associated with the use
of re. In requests and commands particularly, it has the effect of making the
demand seem less impertinent or onerous”.
(74) o-vietava pulu-tava na ne-vamasu re lo-yumwa wa

2SG -open hole-door CONJ 1SG - IRR :come PAR inside-house yet
‘Please open the door and I’ll just come inside to you a little.’
(75) Ø-sape Ø-ka-n re ka-mimi kinan-ena Ø-kom-in

3SG -say 3SG -eat-TR PAR POSS -1 PL . EXC eat-NMLZ 3SG -pass-TR
Ø-sape imimi na kia-mimi puru-vi na naga
3SG -say 1PL . EXC EMPH POSS -1 PL . EXC stump-banana EMPH 3SG
‘He said he wanted to eat (asked to eat) some of our food because he
said he was a brother of ours.’
5 Historical perspective – sources and diachronic

pathways of Oceanic partitives
A full historical analysis of partitive forms in Oceanic is beyond the scope of this
chapter, but a few observations can usefully be made. Descriptions exist for a
growing number of Oceanic languages and much work has been done on recon-
structing the lexicon and grammar of Proto-Oceanic (see e.g. Ross 1988, Lynch
et al 2002 for extensive discussion and lists of sources, as well as the series of
thematically organised Proto-Oceanic (POc) lexicons by Ross et al 1998–2011).
Of particular relevance to the present study are the following reconstructed
forms:
– The Proto Polynesian (PPN) indefinite marker *ha
– A POc “indefinite common non-human article” *ta (Lynch et al 2002:71)
– Reconstructions of the POc numeral ‘one’: *ta-sa, *sa-kai, *tai, and *kai
Figure 4 provides an abbreviated diagram of the relations between the proto

languages referred to here:
Figure 4: Oceanic Subgrouping and Proto languages
5.1.1 Reflexes of PPN *ha
Cross-linguistically the numeral meaning ‘one’ is a common source of indefinite

markers, (Latin unus ‘one’ being a well-known example). For the Polynesian lan-
guages which express partitive meanings through an indefinite or non-specific
article, their provenance seems relatively straightforward then: PPn *ha was first
reconstructed by Pawley (1967) and is held to reflect PMP *esa ‘one’ (see
e.g. Lynch et al 2002: 73). PPn *ha is reflected in many modern Polynesian
550 Peter Budd
languages, e.g. in West Futuna Aniwa11 and Tongan non-specific articles sa and
ha respectively, Samoan se, and in Māori he ‘some’:
(76) Kawe-a atu he wai ki a au

bring-pass away a water to PERS 1SG
‘Bring me some water.’
(Bauer 1997: 294)
However, it is also clear that in some Polynesian languages it is not a reflex

of *ha, but rather, a plural article which is the source of the partitive markers.
Tuvaluan niisi ‘some/other’,which was exemplified above in 3.1. is analysed as
the combination of the plural indefinite ne + a quantifier isi.
5.1.2 Reflexes of POc terms for ‘one’ and the POc indefinite article *ta
Outside the Polynesian languages there are also many examples of the POc
numeral ‘one’ as the source of indefinite and partitive markers; yet the situation
is somewhat confusing, for two main reasons: first, there are multiple reconstruc-
tions of the POc form for ‘one’, and second, the similarly shaped POc indefinite
marker *ta is also held to be the source of some indefinite articles in modern
Oceanic languages.
Lynch et al (2002) reconstruct four proto forms for the numeral ‘one’: *ta-sa,
*sa-kai, *tai, and *kai – and have the following remarks on the subject:
The reconstruction of the POc term(s) for ‘one’ is more difficult as several morphemes and
morpheme combinations occur. These combinations presumably differed semantically and
functionally in the same way as English a/an, one, single, and only, but these differences
have yet to be sorted out. The morpheme *sa . . . reflects PMP *esa ‘one’ and is reflected in
the proto-Polynesian indefinite singular article *ha but was often apparently bound in
POc, where *ta-sa and *sa-kai are also reconstructable. We take *ta in *ta-sa to be the
indefinite article. . . reflexes of *ta are used alone as numeral ‘one’ in a number of WOc
languages, but we take this to be a derived usage. There are also widespread reflexes of
*tai ‘one’ but in the absence of evidence to the contrary we take this to be a morpheme
separate from *ta; we also take *tewa ‘one’ reflected in languages of north and central
Vanuatu and the Central Pacific to include a reflex of *tai. (Lynch et al 2002: 73)
11 West Futuna Aniwa is one of several Polynesian outlier languages spoken in Vanuatu, a
consequence of westward back-migrations from Polynesia, which occurred much later than
the original settlement of the archipelago.
For example, the Lamen language has the form sai, which functions both as
numeral ‘one’ and as an indefinite marker. Closely related Vanuatu languages
with apparently cognate forms are listed below in Table 3:
Table 3: Vanuatu language partitive/indefinite markers
Lamen sai
Bierebo ta
Paamese tei
Lewo tai, re
Abma te
Ske tewa
Araki re
Mwotlap te
South Efate tete
Anejom tah
Whether these forms are all reflexes of the indefinite marker *ta or alterna-
tively of one of the POc forms for ‘one’ is not always clear. Lynch et al (2002: 71)
cite Anejom tah ‘indefinite’ and Mwotlap te ‘partitive’ as reflexes of the POc
indefinite article *ta, whereas François (2001: 335 fn1) holds that Mwotlap te
reflects Proto North and Central Vanuatu (PNCV) *tea ‘one’, which in turn
reflects one of the POc forms for ‘one’ rather than the indefinite article.12 Some
information on the latter is provided below.
Lynch et al (2002: 59; 71) suggest that POc *ta (like two other common non-
human articles) is descended from an “article-like morpheme” in Proto-Malayo
Polynesian (PMP), an ancestor language of POc. They propose that there were
three such morphemes in PMP, one indefinite and two definite:
*a nominative (definite)
*na genitive (definite)
*ta accusative (indefinite)
The two definite morphemes reflect the different grammatical relations of

their head noun to the verb. It is widely established in Austronesian linguistics
that PMP exhibited various types of voice in verbal clauses, including a direct
passive and an active voice, each of which had two core arguments. Following
Dixon’s (1979) framework of describing grammatical relations, in the passive
12 It is noted that a Proto North and Central Vanuatu (PNCV) language is no longer posited as
an exclusively shared ancestor of the NCV languages (see e.g. Ross et al 2007), but the term’s
use here reflects the subgrouping current at the time of the various authors’ analyses.
552 Peter Budd
voice the O argument is marked nominative and the A argument is marked

genitive;13 in the active voice nominative marking occurred on the A argument,
whilst the other core argument received accusative case marking. Lynch et al
(2002: 58) claim that, based on characteristics of modern languages which retain
this PMP system, the active voice “was in a sense the less transitive voice as the
nominative was typically definite and the accusative typically indefinite”. Their
reconstructed examples illustrate this:
(77) a. *kaRat-ən na manuk a wai

bite-DIRECT. PASS GEN chicken NOM mango
‘The chicken is biting the mango OR the mango is being bitten
by the chicken.’
b. *k<um>aRat ta wai a manuk

<ACTIVE >bite ACC mango NOM chicken
‘The chicken is biting a mango’
(Lynch et al 2002: 59)
François (2001: 335 fn1) suggests that both the Araki partitive re and Mwotlap
te reflect PNCV *tea ‘one’ and the same would apply for Paamese tei, Lewo re
and Ske tewa. The confusion occurs when one considers the following situation:
All of these languages have reflexes of *tea or *tewa ‘one’ (which themselves are
reflexes in daughter languages of POc *tai ‘one’), and yet in a language like
Lewo, along with the form re, there is also tai which must be an unaltered reflex
of POc *tai ‘one’; in Lewo both forms can express partitive meanings, and co-
occurrence is attested:
(Lewo)
(78) mwara ne-ka-n re tai
mate 1SG : S -eat-TR PAR ART
‘Mate, I’ll just eat a piece/one/some (ie. Can I have a small piece please?).’
(Early 1994: 261)
It seems, therefore, that for Lewo at least, two different reflexes (one more
conservative than the other) of the same proto-form *tai have developed over
13 As Lynch et al (2002: 58) explain, the nomenclature is rather unconventional and the term
‘genitive’ is used to reflect the fact that the case marking was identical to the possessor marking.
time with slightly different distributions and functions: a post-verbal reflex re

and a post-nominal tai.14
5.2 Diachronic pathways

Of more interest for the present discussion is the fact that in a language like
Lewo there is a partitive form which straddles the two main syntactic environ-
ments of the partitive items that have been reviewed here: the form tai functions
both as an indefinite marker occupying a slot in the NP, but also as a post-verbal
modifier:
(Lewo)
(79) Lokuli keviu tai
dog big ART
‘A big dog.’
(Early 1994: 248)

(Early 1994: 248–249)
This contrasts with the language’s other partitive form re, which, like its equiva-
lent (and sometimes, cognate) forms in a number of other languages, is typically
restricted to a post-verbal slot. However, at least one other Vanuatu language
has a form which similarly spans the two environments: the Avava form tuut
was described as an NP quantifier form in this example:
(Avava)
(81) Komat-yan mwiniel i moroko-n tuut ier
1PL . EXC . R-eat taro INS rib-3SG some PL
‘We ate the taro with some of its ribs.’
(Crowley 2006: 63)
14 In terms of Lewo sound correspondences with ancestor languages both /t/ and /r/ reflect
*t, though *t > /r/ has only occurred intervocalically, suggesting that *te(a) > re is a con-
sequence of post-verbal position and close phonological association with the (vowel-final)
verb complex.
554 Peter Budd
Yet, as Crowley (2006: 102) states, “this form can also be used to express a par-
titive meaning with a transitive verb, indicating that the referent of the object is
only partly affected by the action expressed in the verb. Thus:
(Avava)
(82) I-yan tuut emer ki
3 SG : R-eat PAR eel DEM
‘He ate some of the eel.’
(Crowley 2006: 102)
Although the resulting meanings are very similar, it is recalled that there is a dif-
ferent word order here: tuut occurs after the verb and before the NP object in the
second example, in contrast with its post-nominal position in the previous
example, meaning ‘some of its ribs’. For both Lewo and Avava it seems most
likely that the post verbal function of the form has evolved from the NP article/
quantifier function, as will be demonstrated below.
Outside the Vanuatu group of languages, Yapese (Micronesia) is another lan-
guage with a partitive marker occurring in the same two environments. Recall
that this language has a form boech ‘some, a little’ which is analysed by Jensen
(1977: 160) as a combination of the indefinite article ba and the diminutive chi. It
occurs with a connector ee or ea in NPs to give a partitive meaning:
(Yapese)
(83) boech ee niig
some CONN fish
‘Some fish.’
(Jensen 1977: 160)
However, it also functions as a post-verbal partitive modifier, as does the related

form boechquw ‘a little’:
(84) Qii marwee boech

‘He worked some.’
(85) Qii marwee boechquw

‘He worked a little.’
(Jensen 1977: 160)
Given Jensen’s synchronic analysis that boech and boechquw include the indefinite
article ba, it seems most likely that the pre-nominal NP partitive functions of these
forms pre-date their verb modifying functions. To argue for the reverse process
in Yapese, would entail that an indefinite article ba has evolved by splitting
from a post-verbal adverb boech, which seems highly implausible.
For the Vanuatu languages discussed above, the grammaticalisation path-
way of the post-verbal partitives must in many cases also involve a shift from
an article-like form, which presumably occurred in NPs of any grammatical func-
tion (subject, object, existential subject, attribute of equational clause etc), and
which then became restricted to the post-verbal position with scope over the
object NP. Speculating further, in some languages, from this post-verbal position
its function appears to have generalised, such that it can be used even with
mono-valent verbs to mark a partial accomplishment of the action. This stage
of the grammaticalisation pathway is substantiated by the fact that of the lan-
guages which exhibit post-verbal partitives, there are some in which the forms
are restricted to transitive clauses (e.g. Avava tuut and Mwotlap te), some in
which the forms occur in both transitive and intransitive clauses (e.g. Sye -wi,
Bierebo ja, Abma te), but no examples of languages in which the post-verbal
partitive occurs exclusively with intransitive verbs.
There is possibly a further explanation for the presence of post-verbal parti-
tive forms in some Vanuatu languages. Bierebo ja seems likely to be an accreted
form of gicha, a variant form which occurs in some more conservative dialects of
the language. In turn, gicha can be diachronically analysed as gi-cha, where gi is
a serial verb meaning ‘take time’ which is still productive in the contemporary
language and cha is perhaps a phonologically-conditioned variant of ta. What
seems plausible here then, is that a modified serial verb with a literal meaning
of ‘take a little time’ has developed into a more general VP partitive meaning
‘do a little’. It remains to be seen whether this analysis would be valid for other
Vanuatu languages with VP partitives.
5.3 The role of partitives in negation

One area in which both NP and post-verbal partitives appear is in the expression
of negation. It is a well-established cross-linguistic diachronic tendency for parti-
tives and other minimisers to enter into negative constructions as emphatic de-
vices15 through Jespersenian cycles. As a result of their occurrence as emphatics
in negative constructions such minimising forms may take on negative semantics
themselves and develop into fully grammaticalised negators. In some Oceanic
languages partitive markers have followed this path and grammaticalised to
15 French pas is the most well-known example.

556 Peter Budd
varying degrees. This point has received some attention in the typological litera-
ture (see e.g. Kahrel & van den Berg (eds) 1994; Miestamo 2005; Auwera 2010,
Horn (ed.) 2010), and examples are also provided elsewhere in Miestamo (this
volume) and Luraghi (this volume). Most of the examples cited are from
Vanuatu languages and it is still unclear how widespread the phenomenon is
across the whole Oceanic subgroup. The phenomenon is briefly illustrated below.
In Samoan, the partitive article sina functions to reinforce the negative, in
conjunction with another emphatic lava:
(Samoan)
(86) 0 ua leai lava sina ofi
PFV not.exist EMPH ART ( PAR . SG ) room
‘There was absolutely no room at all.’
(Mosel 1992: 265)
Similarly, in Abma, which has bipartite negation of the form ba. . .=nga, the
addition of the partitive –te after the verb adds further emphasis:
(Abma)
(87) Ko=t ih bamte abma nanib?
2SG = PFV hit make.die what yesterday
‘What did you kill yesterday?’
(88) Na=t=ba if bamte abma=nga

1SG = PFV =NEG 1 hit make.die something=NEG 2 16
‘I don’t kill things.’
(89) Na=t=ba if bamte=te abma=nga

1SG = PFV =NEG 1 hit make.die=PAR something=NEG 2
‘I didn’t kill anything.’
In Paamese the form tei occurs in negated clauses but is still analysed by
Crowley (1982: 144) as the partitive marker:
16 A number of Vanuatu languages have discontinuous negation marking and the glossing con-
vention is to mark each negator with a numeral indicating its relative order in the construction.
(Paamese)
(90) a. Longe-nV ree-ku
3SG . R .hear-COMN voice-1SG
‘He heard my voice.’
b. Ro-longe-tei ree-ku
3SG . R . NEG -hear-PAR voice-1SG
‘He didn’t hear my voice.’
(Crowley 1982: 144)
In Lewo the cognate re is treated in Early’s (1994) description as two separate

forms, a partitive and a negator, which do not co-occur, i.e. re is now analysed
as an obligatory marker in negative expressions:
(Lewo)
(91) Ve a-tol inu re
NEG 1 3PL . SBJ-touch 1SG NEG 2
‘Don’t touch me!’
(Early 1994: 78)
In Bierebo the same has occurred: re is fully grammaticalised as a negative

marker (and there is in fact no homophonous partitive marker. It is glossed NEG 2
to reflect its relative position in the maximally tripartite negative marking, but in
Irrealis clauses it is the sole and exclusive negator, leaving no doubt that it has
moved from an emphatic to a genuine negative marker:
(Bierebo)
(92) Ko-teng re
2 SG - IRR : cry NEG 2
‘Don’t cry.’
Further evidence of the cycle can be seen in Lewo and Bierebo: With the post-
verbal partitive marker re having fully grammaticalised into a negator in both
languages, other partitive minimisers are pressed into service as emphatics in
negative constructions. The indefinite markers tai or ta are used for the expres-
sion of ‘no one’, ‘nothing’ etc:
(Bierebo)
(93) Mara ta
NEG INDF
‘There’s no-one/none.’
558 Peter Budd
(94) Ta ø-to re
one 3SG - IRR :be.at NEG 2
‘Nothing will remain.’
(Author’s data)
(Lewo)
(95) Tai pe ø-pa re poli
one NEG 1 3SG - R :go NEG 2 NEG 3
‘No-one/none went.’
(Early 1994: 249)
Not only this, but the post-verbal partitive ja in Bierebo also occurs as a negative
emphatic:
(Bierebo)
(96) Mara kama ø-pinim ja Bonkovio rui
NEG 1 EMPH 3SG -R :come PAR Bokonvio already
‘He never ever came to Bonkovio.’17
(Author’s data)
6 Conclusion
This chapter has provided an initial synthesis of findings on partitive morphemes
in Oceanic languages. Given the large size of the Oceanic subgroup and the
significant number of undocumented languages, much work remains to be
done in order to form a more definitive picture. From the languages investi-
gated, the following general statements can be offered: it appears that partitive
meanings are most frequently expressed by NP dependents in the form of in-
definite markers and quantifiers. In some of the languages of Vanuatu and also
in Yapese there are also post-verbal particles and other quantifying modifiers.
There is no evidence to suggest that the NP and VP partitives are mutually
exclusive and one must assume that there are simply different sources for each.
Kittilä and Luraghi (this volume) propose a diachronic pathway of case-marking
morphemes grammaticalising into indefinite markers, and according to Lynch
17 Vanuatu’s National language, Bislama, an English-lexifier creole has equivalent constructions

using lelebet (from English little bit) as a minimiser to emphasise negation e.g. hem i neva kam
lelebet long ples ia ‘he never ever came to this place’.
et al (2002: 58ff), the indefinite markers of many modern Oceanic languages can
indeed be traced back to the Proto Malayo-Polynesian accusative article *ta.
PMP is notable for its voice systems, including a distinction between a direct
passive and an active voice, both of which were used transitively with two core
NP arguments. The accusative case marker occurred in the ‘low transitivity’
active voice, associated with indefiniteness, giving rise to the Oceanic reflexes
of *ta as an indefinite marker. However, the formal similarity of the recon-
structed POc terms meaning ‘one’ with the POc indefinite *ta make it difficult
to say with certainty which forms modern partitive markers reflect.
Post-verbal partitive markers are identifiable in a number of languages of
North and Central Vanuatu. The data presented here demonstrate that for transi-
tive clauses they express partial affectedness of object NPs in two senses: either
part of the entity is affected (to any degree), or all of the entity is affected to a
lesser degree. When they occur with intransitive verbs, they express a limited,
partial, reduced, or incomplete instance of the verb’s action or situation. A
number of tangential meanings have arisen, such as low assertivity or politeness,
and various aspectual nuances. Based on the available data it seems most plau-
sible that the post-verbal partitive markers have developed from the NP partitive
morphemes rather than vice-versa. In some cases the distribution of the post-
verbal partitives appears to have spread from marking object NPs to also mark-
ing intransitive verbs. In the case of Araki the grammaticalisation has gone further
still and it now occurs in multiple environments with a variety of functions.
A reasonably well-known function of partitive morphemes in Oceanic lan-
guages, and one which fits with cross-linguistic tendencies, is the role they
play in negative marking. Their minimizing function serves as an emphatic
negator, but partitives may develop into negative markers proper. Various stages
along this grammaticalisation pathway have been demonstrated for Oceanic
languages and it is interesting to note that both VP and NP partitives enter into
the Jespersenian negative cycle.
560 Peter Budd
7 Appendix
Table 4: Partitive markers in selected Oceanic languages
WOc Lamen sai

North New Guinea South Efate tete
Manam alu; sesu; suka; muku Sye -wi
Anejom tah
Papuan Tip Lenakel ker; nɨvin
Tawala gehou-na Canala bwere; bwete
Motu taina Bwatoo bee
Grand Couli pweře
Meso Melanesian
Nakanai isahari Central Pacific
Mono-Alu aiina; aabau Boumaa Fijian vica; soo
Kokota keha Samoan se; ni; sina
Hoava kiqa West Futuna Aniwa sa
Teop peha; sa Tongan ha
Māori he; (t)ētahi
EOc Tuvaluan niisi
Southeast Solomonic Hawaiian kekahi
Kwaio ta-; gula- Rapa Nui he; hai; tetahi no
Toqabaqita ta; tootoqe
Bugotu si na Micronesian
Mokilese ekij; epwi
Remote Oceanic Ponapean ekei
Mwotlap re Woleaian se-
Abma te Kiribatese te; taian
Araki re
Neve‘ei tuan ar Yapese boech; boechqw
Ske tewa
Avava tuan ier; tuut
Paamese tei
Lewo tai; re
Bierebo ta; tiala; kurua-; ja;
References
Auwera, Johan van der. 2010. On the diachrony of negation. In Horn (ed.).
Bauer, Winifred. 1997. The Reed reference grammar of Māori. Auckland: Reed books.
Besnier, Niko. 2000. Tuvaluan: a Polynesian language of the Central Pacific. London: Routledge.
Crowley, Terry. 1982. The Paamese language of Vanuatu. Canberra: Pacific Linguistics.
Crowley, Terry. 1998. An Erromangan (Sye) grammar. Oceanic Linguistics Special Publication
No. 27. Honolulu: University of Hawai’I Press.
Crowley, Terry. 2006. The Avava language of central Malakula (Vanuatu). Canberra: Pacific
Linguistics.
Dixon, R.M.W. 1988. A grammar of Boumaa Fijian. Chicago: University of Chicago Press.
Early, Robert. 1994. A grammar of Lewo, Vanuatu. Unpublished PhD dissertation. Canberra: The
Australian National University.
François, Alexandre. 2001. Contraintes de structures et liberté dans l’organisation du discours.
Une description du mwotlap, langue océanienne du Vanuatu. PhD dissertation, Université
Paris-IV Sorbonne.
François, Alexandre. 2002. Araki: A disappearing language of Vanuatu. Canberra: Pacific
Linguistics.
Hamel, Patricia J. 1994. A grammar and lexicon of Loniu, Papua New Guinea. Pacific Linguistics
C-103. Canberra: The Australian National University.
Harrison, Sheldon, R. 1976. Mokilese reference grammar. Honolulu: University Press of Hawaii.
Horn, Laurence R. (ed.). 2010. The expression of negation. Berlin: Mouton de Gruyter.
Hunkin Galumalemana, Afeleti. 2009. Gagana Samoa: A Samoan language coursebook.
Honolulu: University of Hawai’i Press.
Jensen, John Thayer. 1977. Yapese Reference Grammar. Honolulu: University of Hawai’i Press.
Kahrel, Peter & Rene van den Berg (eds.). 1994. Typological studies in negation. Amsterdam:
Benjamins.
Keesing, Roger M. 1985. Kwaio grammar. Canberra: Pacific Linguistics.
Koptjevskaja-Tamm, M., 2006, Partitives, in K. Brown, ed., Encyclopedia of Languages and
Linguistics, Second Edition, Oxford: Elsevier, 218–221.
Lichtenberk, Frantisek. 2008. A grammar of Toqabaqita. Berlin: Mouton de Gruyter.
Lynch, John. 1978. A grammar of Lenakel. Canberra: Pacific Linguistics.
Lynch, John; Malcolm Ross & Terry Crowley (eds.). 2002. The Oceanic Languages. Richmond:
Curzon Press.
Miestamo, Matti. 2005. Standard negation: The negation of declarative, verbal main clauses in
a typological perspective. Berlin & New York: Mouton de Gruyter.
Musgrave, Jill. 2007. A grammar of Neve‘ei, Vanuatu. Canberra: Pacific Linguistics.
Mosel, Ulrike & Even Hovdhaugen. 1992. Samoan Reference Grammar. Oslo: Scandinavian Uni-
versity Press.
Mosel, Ulrike & Evan Hovdhaugen (eds.). 1999. Negation in Oceanic languages. München:
Lincom Europa.
Pawley, Andrew. 1967. The relationships of Polynesian outlier languages. JPS 76:259–296.
Rivierre, Jean-Claude, Sabine Ehrhart & Raymond Diéla. 2006. Le bwatoo et les dialectes de la
région de Kone (Nouvelle Calédonie). Leuven: Peeters.
Ross, Malcolm. 1988. Proto Oceanic and the Austronesian languages of Western Melanesia.
Canberra: Pacific Linguistics.
Ross, Malcolm. 1998. The Lexicon of Proto Oceanic: Material Culture. Canberra: Pacific Linguistics.
Ross, Malcolm, Andrew Pawley & Meredith Osmond (eds.) 1998–2011. The Lexicon of Proto
Oceanic: The culture and environment of ancestral Oceanic society, vols. 1–4. Canberra:
Pacific Linguistics.
Schneider, Cynthia. 2010. A grammar of Abma: A language of Pentecost island, Vanuatu.
Canberra: Pacific Linguistics.
Subject index
Note: The terms ‘genitive’ and ‘partitive (case)’ are not included in this index on
account of their high frequency, which would have made them uninformative, as
they occur in virtually every page.
Abessive 113, 119, 120, 121–123, 132, 134, 232–234, 236–240, 242–252, 294–
137 296, 307–308, 311, 325–326, 330, 365,
Ablative 1, 8, 33, 35, 48–55, 71, 89, 110, 113, 379, 382, 384, 388, 392, 394, 396,
115–121, 123, 130, 134, 137–138, 145, 417–422, 424–427, 429–434, 437–
267, 323, 325, 333–342, 365, 370, 372, 440, 443–445, 447–449. 454, 457,
419, 444, 467–468, 483 459–460, 464–465, 472–473, 477,
Accusative 7–8, 10, 19–20, 28–31, 37–45, 483–485, 492, 515, 551–552, 559
47, 52–53, 64–65, 70–71, 92, 97–98, Article
105, 111, 113–115, 120–12, 123, 126– – Definite 10, 23–25, 72–73, 291, 298,
128, 130–135, 139–143, 191–192, 195, 300–301, 303–304, 306–308, 310–
201, 226, 249–250, 281, 347, 356, 367, 316, 325, 327, 341, 447, 457, 481, 497,
379, 382–384, 386–387, 391, 393– 500, 505, 515
394, 396, 399–401, 405–413, 419, – Indefinite 7, 23, 55, 71, 72–74, 82, 259,
421–422, 424–432, 434, 436, 438– 301, 327, 478, 488, 503–504, 509,
440, 443, 445, 447, 449–450, 452– 526–530, 550–551, 554–555
454, 457–461, 463, 467–473, 479, 481, Aspect 4, 7–10, 18, 38–40, 58, 61, 64, 66,
486, 490–491, 501, 505, 523, 551–552, 70–71, 76, 83, 90, 92–93, 95–96, 111,
559 126, 139–141, 143, 146, 174, 178, 184–
Adessive 119–120, 134, 137, 159–160, 162 185, 188–193, 199, 201, 206, 208–209,
Adjunct 92, 133, 230, 278, 481, 483, 492 218, 220, 237, 240, 252, 257–265, 270,
Adposition 6, 11, 17, 20, 22, 46, 48, 112, 118, 275–276, 280–281, 379–380, 384–
133, 138, 279, 280, 466–469, 473, 385, 387–392, 394–396, 404, 434–
477–488 435, 547
Affectedness 1, 6–7, 17–18, 40–44, 56–58, Asymmetry 8, 64–65, 81–82, 135, 137
60–61, 128, 139, 220, 247, 367, 371– Atelic 108, 139, 141, 143, 218, 260–261,
372, 420, 429, 431, 447, 449–453, 461, 263, 278, 299, 317, 425, 429, 433–435,
469, 486, 496, 543, 559 439–440, 453
Agreement 7, 22, 26–28, 59, 93, 132, 154– Authority, verbs of 417, 422, 430–431
156, 158–159, 164, 167, 170, 174, 177,
180, 182–183, 186–187, 200, 205, 243, Bare nouns 89, 105, 126, 145, 225, 227, 229,
245–247, 249–251, 341, 349, 371, 383, 293
404, 454–456, 464 Bounded 38–39, 41, 107, 110, 128, 139, 194,
Alignment 28, 52, 63, 68, 80, 114, 179, 184, 218, 260–261, 269, 276, 279, 385–386,
188, 226, 248–250 405, 546
Allative 113, 115, 119, 137, 341–342, 467–
468 Change of state 42–43, 56, 58, 142, 159,
Argument 7, 9, 11, 36, 52, 58–59, 78–79, 260, 417, 420, 422, 424, 429, 431, 434,
103, 112–113, 123, 127–128, 132–135, 447–453, 461, 472
167–169, 177–189, 191–192, 195–207, Co-text 273, 281
209–218, 220–221, 224–226, 230,
564 Subject index
Comitative 113, 121, 123, 274 Indefiniteness 1, 2, 4–7, 11, 17–19, 22, 23,
Count 29, 31, 56, 72–73, 128–129, 154–157, 28, 29, 31, 36–38, 53, 56–61, 93, 95,
164, 168, 171–172, 191, 193–196, 241, 105, 111, 130, 139, 157, 164, 171, 173,
269, 299–301, 310–311, 381, 451, 469– 237, 264, 323, 383, 388, 389, 391, 396,
470, 488, 503 443, 446, 447, 452, 453, 457, 458, 472,
Cross-categorial 113, 130–135, 137, 143, 145 487, 488, 497, 501, 503, 504, 518, 523,
524, 527–529, 536, 559
Definiteness 5, 19, 50, 57, 71, 74–76, 82, 96, Individuation 248, 355, 486, 496
106, 113, 126, 140, 145, 157, 184, 188, Information
201, 224, 238–242, 246, 251, 264, 272, – Given 156, 224, 232
296, 301, 325, 339, 406, 500–501, 505– – New 11, 156, 232, 443
506 – Structure 11, 156, 158, 170, 185, 230, 249,
Delimitation 379, 394–396 384, 477, 500
Determiner 1, 6, 23, 56, 71–73, 75, 80, 82, Ingestion, verbs of 56, 417, 422–426,
225, 293–295, 297, 298, 302, 311, 323– 452
325, 327, 335, 344, 383, 477, 480, 485, Instrumental 113, 115, 123, 393, 419, 449,
490, 492, 502, 505, 510 465, 473, 478
Diminutive 271, 276, 350–351, 359, 361, Intransitive 4, 9, 22, 45, 52, 71, 79, 129,
363, 374, 381, 528, 530, 540, 554 156–157, 168, 177–179, 181–183, 188,
201, 204–205, 210, 218–219, 231–232,
Elative 2–3, 19, 89–91, 100, 102, 107, 110, 244, 247, 249–251, 268, 279–280, 308,
113, 115–118, 120–122, 124, 126, 127, 325–326, 336, 393, 460–461, 540, 545,
134–135, 137, 145, 159, 163, 267, 364– 547, 555, 559
365, 370 Intransitivity, fluid 177, 179, 183, 248, 250,
Epistemic 89, 95, 106, 108–109, 139–143, 252
146, 165
Evidential 37, 60, 89, 91–92, 106–109, 131, Layering 481
138–141, 143–146 Lineage 483
Existential
– Clause 5, 9, 58–59, 61, 70, 153, 155–160, Mass 5, 18, 29, 31–32, 56, 82, 99–100,
163, 165, 167–168, 173–174, 231, 251, 128–129, 154, 156–157, 160, 164, 168,
542 184, 194, 198, 207, 297, 299–301, 304–
– Construction 27, 31, 153, 155, 158, 161– 305, 307, 310–311, 314, 358–359, 367,
162, 173, 177–178, 181, 205, 221, 486 371–372, 374, 380, 382, 401, 470, 485,
– NP 9 488, 493
– Subject 6–7, 27, 154–156, 159–160, 162– Message packaging 177–179, 224, 239,
165, 167–168, 173–174, 555 243–244, 246–247, 250–252
Extraction 483, 502, 507 Motivation 26, 54, 63, 66–67, 74, 80–83,
97, 179, 207–209, 217, 372, 389, 399,
Grammaticalization 10–11, 35, 49–50, 53, 401, 412, 427, 444, 450
57–58, 82, 89, 106, 130, 323, 325, 335,
336–339, 342, 447, 456, 477, 481, Negation 4, 6–8, 36–38, 58, 61, 63–83, 93,
488–489, 499–500, 502, 504, 515 107, 131, 154–155, 158, 160, 163–165,
167–170, 172–173, 180, 186, 206, 209,
Illative 113, 115, 121, 135, 265 217, 235, 237, 271, 324, 326–329, 348,
Impersonal 45–46, 179, 192, 206, 250, 328, 368, 372, 384, 400, 453, 460–461, 472,
362, 382–384, 443, 454, 461–464, 472 509, 523, 555–556, 558
Subject index 565
Nominative 7, 9–10, 20, 26–28, 32, 38–39, 291, 293–294, 297–300, 304–307,
46, 55, 58–59, 69–71, 76, 113, 115, 119– 311–316, 318, 323, 333–334, 339, 347,
121, 123, 125, 134, 137, 140, 153–159, 350, 356, 380, 382–383, 386, 392, 451,
162, 164–166, 168–174, 178, 183, 186– 455, 459, 469–470, 488, 500, 503–505,
187, 189, 192–196, 200–201, 203–206, 509, 514, 526–527, 529–531, 536–539,
208, 210–212, 214, 217, 219, 226–229, 550
240–241, 247, 249–250, 329, 347–348, Possessive 38, 54, 96, 159–163, 165, 185,
356, 379, 382–384, 387, 394, 396, 222, 279, 349, 417–419, 428, 480, 542
418–419, 422, 431, 445, 454–456, Post-verbal 53, 178, 205, 230, 248, 250,
461–464, 472, 492, 551–552 459, 523, 528, 540–546, 553–555, 558–
Non-bounded 139, 386 559
Non-finite 60, 80, 89, 92, 106, 108, 111, 113, Pragmatic strengthening 487, 497
119, 131–135, 137–138, 143, 145, 178, Praising, verbs of 422, 432–433
180–181, 329, 337 Prefix 10, 75–76, 262, 384, 392–393, 399–
Number 22, 26, 34, 50, 69, 76, 100, 103, 405, 407, 412, 541
104, 122, 155–156, 167, 179, 186–188, Preposition 6, 10, 11, 22–24, 46–47, 52–53,
192, 196, 210, 213–217, 227, 243–244, 72–73, 101, 104, 327, 340, 347–348,
251, 263, 293, 304, 305, 307, 311, 314, 352, 354, 357, 362–366, 368, 370, 372,
324–325, 339, 347, 454–456, 464, 500, 376, 381, 443–445, 467, 469–471, 477,
504, 514–515, 528, 532, 537 480–487, 490, 492, 509, 515, 539
Progressive 133, 228, 268–271, 275, 379,
Oblique complement 483–484 385, 388–391, 396, 435, 439, 481, 486,
Overlap stage 512 498
Prolative 113, 115, 120, 134
Paradigm 24, 115–116, 120–121, 131, 133, Prosecutive 113, 115, 338–339, 341–342
134, 137, 186, 187, 348, 350, 356, 387,
481, 484, 488, 497–498, 507, 509, 531 Quantification 8, 18, 64, 66, 70, 82, 96, 138,
Paradigmatization 498 196–198, 201, 207, 209, 251, 348, 355,
Part-whole 1–2, 17–19, 32–33, 52–55, 89, 363–364, 524, 529
91, 106, 139, 145 Quantifier 1, 6, 28, 33–35, 61, 73, 91, 105,
Participle 143–144, 279, 295, 310, 335, 225, 295, 309, 316, 331, 333, 362, 384–
337–338, 456, 499, 510, 515 385, 388–390, 393, 477, 523–529, 534,
Partitive construction 1–3, 18, 22–25, 32– 538, 550, 553–554, 558
35, 48, 51, 53–58, 61, 73, 92, 101, 110,
124, 325, 445–447, 452, 453, 457, 480, Referentiality 5, 8, 64, 66, 73, 77–83, 196–
525 197, 391, 537
Perception, verbs of 44, 99, 108, 142, 267– Resultative 7, 9, 43, 139, 159–160, 257–260,
268, 281, 407–408, 417, 422, 425–429, 262, 264–265, 267, 271–272, 274, 279,
433, 444, 447, 449, 452–453, 461 335
Perfectivity 57, 70, 140, 263, 379, 387, 396,
545 Secondary case 10, 347, 352, 371
Pseudopartitive 8, 32, 33, 62, 91, 101, 112, Spatial meanings 47–48, 159, 162, 468,
123, 124 481, 483
Plural 5, 9, 18, 23–24, 27–29, 31–32, 39, Specificity 5, 8, 64, 66, 71, 78, 188, 527
56, 59, 70, 72–73, 105, 121, 128, 130, Subject 1, 4–7, 9, 11, 17, 20–22, 25–28,
153–158, 160, 164, 168, 174, 187, 189, 31–32, 36, 43, 45–46, 48, 55, 58–59,
192, 194, 198, 200–207, 213–217, 251,
566 Subject index
68–71, 76, 92–94, 96, 105, 107, 141, Translative 113, 115, 119, 123, 132, 137, 141,
153–174, 177–183, 185–188, 192, 194, 269–270, 273
196–197, 200–201, 203–206, 210–212,
221, 224, 226, 230–234, 244, 246–247, Unaccusative 4, 21–22, 58–59, 195, 382–
249–252, 263–264, 267, 271, 274, 281, 382, 438–440, 458–460, 472, 486,
296, 302–303, 306, 308, 325–326, 501, 507
328, 336, 365, 367–368, 372, 379– Undbounded 4, 18, 31–32, 38–39, 105,
380, 382–384, 386–387, 394, 396, 107–108, 111, 127–128, 166, 200, 218,
417–419, 421–422, 426, 433–434, 260, 264, 276, 278, 385–387, 395, 433,
436–440, 443–445, 447, 453–464, 446, 545–546
472, 478, 486, 488, 492, 494, 501–502, Unergative 4, 458, 486
505–507, 509, 524, 541–542, 547, 550,
555 Vocative 113, 115, 348, 371, 418–419
Subjecthood 59, 95, 177, 179, 197, 204, 249 Voice morphology 436, 437
Subjectivity 264, 281
Word order 11, 69, 155–159, 164, 178, 180,
Telic 108, 111, 127, 139, 140, 218, 260, 260, 183, 185, 196, 203, 224, 227, 230–232,
263, 428–429, 453 243–245, 250–251, 273, 275–276, 282,
Telicity 111, 113, 139, 140–141, 145, 261, 420 296, 477, 498, 500–501, 505–506,
Temporal 128, 132–133, 140, 158, 332, 335– 514–516, 526, 530, 554
336, 338, 402, 406–407, 419, 445, 466,
483 Zero 36, 101, 126, 178, 183, 186, 201, 210,
Terminative 113, 119–120, 134, 137, 266 224, 225–226, 229, 231, 233, 237, 245,
Transitivity 7, 11, 17–18, 39–40, 43–44, 56– 348, 380, 477, 483, 485–486, 488,
58, 82, 178–179, 219–220, 223, 244– 493, 495–498, 500, 502–503, 507,
245, 249, 420, 437, 439, 443, 447, 509–510, 512–513, 516
449–450, 452–453, 472, 486, 496, 541,
549
Author index
Abeillé, Anne 316 Bromley, Sofja V. 348, 354
Abondolo, Daniel 92 Brown, Sue 381, 384
Ackerman, Farrell 139 Brugmann, Karl 444 fn, 448, 449, 462
Adelaar, Willem F. H. 131 Brunet, Jacqueline 509, 512
Adillo Rufo, Sergio 502 Bubenik, Vit 467 fn, 468
Aikhenvald, Alexandra Y. 131, 132, 143, 339 Budd, Peter 11, 24, 26, 33, 34, 53, 78
fn, 454 fn Bulatova, Lidia N. 348, 354
Almqvist, Ingrid 69, 70, 82 Buridant, Claude 498
Alves, Flávia de Castro 80 Butt, Miriam 131, 177
Ambrazas, Vytautas 70 Bybee, Joan L. 133
Amon, Marri 266
Andoni Lakarra, Joseba 342 Campbell, Lyle 58, 59, 107 fn, 143, 493
Anttila, Arto 93, 125 Carey, Susan 96
Ariste, Paul 70, 114 fn Carlier, Anne 5, 7, 11, 24, 52, 57, 60, 73, 316,
Artiagoitia, Xabier 292, 293, 294 fn, 295, 327 fn, 459, 478, 482, 483 fn, 485 fn,
298 fn, 311, 314 487, 489, 496, 503 fn, 515
Askonen, Ebba 257, 264, 265, 267, 276, 281 Carlson, Greg 297, 298, 304
Auwera, Johan Van der 83, 143, 556 Casenave-Harigile, Junes 302, 317
Azkue, Resurreción M. 308 Chafe, Wallace L. 168, 177, 210, 211, 224,
233
Babby, Leonard H. 382 Chantraine, Pierre 445, 467 fn, 469
Baños, José Miguel 454 fn Chelliah, Shobhana 413
Barðdal, Johanna 249, 413, 454 fn, 461, 462 Chesterman, Andrew 5, 156
Bartning, Inge 316 Cheung, Johnny 423
Bauer, Brigitte 461 fn Chierchia, Gennaro 297, 298 fn, 305
Bauer, Winifred 535, 550 Cienki, Alan 267
Beavers, John T. 420, 424 Clark, Ross 53
Belletti, Adriana 514 Cole, Peter 454 fn
Berg, Rene van den 556 Collinder, Bjőrn 110
Besnier, Niko 529, 534 Combettes, Bernard 498, 515
Bhaskararao, Peri 454 fn Company Company, Concepción 505
Bickel, Balthasar 180, 181, 182, 184, 188, Comrie, Bernard 83, 184, 191, 192 fn, 201,
192, 200, 201, 202, 225, 248, 249 261, 454 fn
Bielec, Dana 70 Conti, Luz 4, 6, 7, 11, 27, 41, 42, 44, 46, 47,
Bilbao, Gidor 326 fn 48, 55 fn, 59, 423, 438, 452, 453 fn, 454
Blake, Barry 1, 6, 17, 60, 131, 267, 445 fn, 455, 460, 461, 462 fn, 463, 478
Bleam, Tonia 300 Corbett, Greville 65, 83, 305, 347 fn, 348,
Bobrova, Tat’jana A. 104 fn 351, 371
Bonvino, Elisabetta 508, 513 Coyos, Battittu 291 fn, 302
Bosque, Ignacio 298 fn, 300, 306 Creissels, Denis 74, 75, 83, 179, 250, 323 fn
Bossong, George 461 Croft, William 180, 181 fn, 182, 184, 185,
Bosveld-de Smet, Leonie 298 fn, 386 361, 372
Brattico, Pauli 4, 93 Crowley, Terry 26, 78, 79, 532, 535, 541, 542,
Bresnan, Joan 301 fn 545, 553, 554, 556, 557
568 Author index
Cuzzolin, Pierluigi 461 fn, 462 Fara, Delia Graff 297

Cyffer, Norbert 79 Fennel, Trevor G. 70
Fernandez-Vest, M.M. Jocelyne 4, 9, 257,
Dahl, Eystein 4, 11, 21, 40, 42, 43, 44, 59, 259, 263, 266, 267, 280, 282
419, 420, 425, 427, 430, 434, 435, 448, Fischer, Susann 4
454 fn Foley, William A. 80, 258
Dahl, Östen 65, 97, 261, 263 Fong, Vivienne 93, 125
Dambriunas, Leonardas 70 Forest, Robert 65
Daniel, Mikhail 10, 22, 50, 371 Foulet, Lucien 491, 493, 503 fn
Dasher, Richard B. 57 Frajzyngier, Zygmunt 78
Dayal, Veneeta 297, 305 François, Alexandre 34, 76, 77, 83, 532, 541,
Debrunner, Albert 445 542, 545, 546, 547, 551, 552
Delbrück, Berthold 417, 444 fn, 448, 449 Franks, Steven 380, 381, 384
Delfitto, Denis 305 Freeze, Ray 301 fn
Denison, Norman 93, 263
Depraetere, Ilse 260 Garde, Paul 349, 351, 352, 354, 358, 361,
Dickey, Stephen M. 405, 406 362, 364, 365, 368, 371
Dik, Simon C. 454 fn, 459 fn, 463 Geeraerts, Dirk 402
Dixon, R. M. W. 188, 189, 192, 196, 197, 201, Geisler, Hans 499
210, 250, 339 fn, 454 fn, 530, 534, 551 Gelsen, Henry 70
Dobrovie-Sorin, Carmen 300 Gershevich, Ilya 423
Doetjes, Jenny 305, 316 fn Giannakidou, Anastasia 295 fn
Doke, Clement M. 75 Givón, Talmy 73, 75, 81, 165, 173, 200, 201,
Donohue, Mark 250 fn 202, 233, 238, 239, 249
Dowty, David R. 218, 258 Goenaga, Patxi 294 fn
Dryer, Matthew S. 66, 182 Goldberg, Adele E. 180, 181, 185, 258
Gómez, Ricardo 323 fn, 333 fn, 338
É. Kiss, Katalin 135 fn Goyens, Michèle 483 fn, 484
Early, Robert 33, 79, 528, 529, 539, 547, Greenberg, Joseph H. 498
548, 552, 553, 557, 558 Grimm, Scott 420
Eguren, Luis 293, 294 fn, 295 fn, 298 fn, 299 Grzegorczykowa, Renata 405
Ehrhart, Sabine 530 Guérin, Valérie 77
Englebert, Annick 316 Guillaume, Gustave 503 fn
Erelt, Mati 70, 128 fn, 139, 143, 158, 159,
160, 181 fn, 185, 189, 190, 191, 192, 193, Hagège, Claude 279
198, 210, 218, 249 Hajdú, Péter 130
Erteschik-Shir Nomi 234, 239 Hakanen, Aimo 156, 157, 159
Espinal, Maria Teresa 300 Hakulinen, Auli 64, 161, 162, 163, 261
Etxebarne, Jüje 302, 317 Hakulinen, Lauri 70
Etxeberria, Urtzi 4, 9, 10, 20, 31, 34, 35, 71, Hamel, Patricia J. 525 fn
292, 293, 294, 295, 298 fn, 299, 300, Harbert, William 454 fn
301 fn, 302, 305, 306 fn, 307 fn, 311, Harris, Alice C. 493
316 fn, 317, 325, 458, 460 Harrison, Sheldon R. 528
Etxepare, Ricardo 292, 301 fn, 307 fn, 309, Haspelmath, Martin 3, 59, 66, 93, 94, 95,
325 fn, 331, 342 fn 137, 371, 455 fn, 461 fn
Author index 569
Hayek, John 340 Janhunen, Juha 83, 113 fn

Heath, Jeffrey 75, 76 Janowska, Aleksandra 403, 404, 405, 411
Heinämäki, Orvokki 4, 93, 260 Jensen, John Thayer 530, 554
Heine, Bernd 2, 3, 49, 51, 337, 338, 340, Jernej, Josip 509
512, 515 Jespersen, Otto 305
Helasvuo, Marja-Liisa 59, 154, 161, 178, 192, Jodłowski, Stanisław 400
203, 211, 226, 233, 244, 247 fn
Helimski, Eugen 113 fn, 115 fn Kahrel, Peter 556
Hermann, Eduard 462 Kaleta, Zofia 400
Hermon, Gabriella 454 fn Kangasmaa-Minn, Eeva 113 fn, 261, 262,
Herslund, Michael 503 268
Hettrich, Heinrich 424, 426, 433 Karlsson, Fred 97, 161, 162, 261, 263
Hewson, John 467 fn, 468 Keenan, Edward 454 fn
Heyd, Sophie 298 fn, 316 Keesing, Roger M. 530
Hiietam, Katrin 93, 185 Kehayov, Petar 143
Hoeksema, Jacob 91, 327 Kempf, Zdzisław 45, 401, 402, 407, 412
Holst, Jan Henrik 70 Kiefer, Ferenc 135 fn
Honda, Isao 65 Kiparsky, Paul 4, 93, 139, 261, 385, 386,
Honti, László 114 fn 387
Hoop, Helen de 5, 23, 91, 187, 188, 201 Kittilä, Seppo 8, 82, 105 fn, 124, 178, 368,
Hopper, Paul J. 40, 43, 82, 83, 219, 220, 458, 487, 504, 558
244, 248 fn, 263, 420, 453 fn, 481, 486, Klaas. Birute 261
496 Kleiber, Georges 297, 298 fn
Horn, Laurence R. 556 Klenin, Emily 391, 395
Hovdhaugen, Even 24, 78, 86 Kont, Karl 139
Hualde, José Ignacio 71, 324, 334 fn, 335 fn, Koptjevskaja-Tamm, Maria 1, 2, 3, 48, 61,
336, 339 fn 65, 67, 68, 70, 71, 91, 101 fn, 124, 194,
Humbach, Helmut 429, 432 325, 328, 445, 447, 477, 533
Humbert, Jean 478 Korzen, Iørn 485, 512
Hunkin Galumalemana, Afeleti 527 Krajewska, Dorota 335 fn
Huumo, Tuomas 4, 5, 6, 8, 21, 35, 36, 37, 39, Krasovitsky, Alexander 64, 71, 82
59, 70, 89 fn, 93, 141, 154, 161, 169, Kratzer, Angelika 105 fn
178, 185, 211, 218, 230, 233 fn, 261, Krifka, Manfred 91, 143, 297, 305
282, 291 fn, 323 fn, 458 Kroeker, Menno 74
Hyman, Larry M. 79, 83 Kupferman, Lucien 316
Kuroda, Sige-Yuki 384
Iggesen, Oliver A. 113 Kurzowa, Zofia 404
Ikola, Osmo 107 fn Kuteva, Tania 2, 3, 49, 51, 337, 338, 340,
Ingram, Andrew 340 513, 515
Insler, Stanley 437
Irigoien, Alfonso 292, 310 Laanest, A. 143
Itkonen, Esa 81 Laca, Brenda 298 fn
Itkonen, Terho 260, 263 Lafitte, Pierre 308, 316
Iva, Sulev 114 fn Lafon, René 341 fn
Lahousse, Karen 498, 501, 505, 514
Jackendoff, Ray 92, 94 Laka, Itziar 6, 295, 310, 324, 328 fn
Jakobson, Roman 348, 379, 380, 382, 385, Lakarra, Joseba A. 323 fn, 338 fn, 339, 340,
388, 390, 391 341 fn
570 Author index
Lakoff, George 48 Maslov, Jurij S. 263

Lambrecht, Knud 181, 203, 232, 234, 237 Matushansky, Ora 293 fn, 295 fn
Lamiroy, Béatrice 5, 7, 11, 24, 52, 60, 327 fn, Mayrhofer, Manfred 432 fn
459, 483 fn, 484, 498, 501, 505, 514, 515 McNally, Louise 300
Langacker, Ronald W. 162, 168, 180, 405, Meillet, Antoine 478
406 Melis, Ludo 484
Lapesa, Rafael 501 fn, 505 Metslang, Helena 4, 9, 39, 44, 60, 70, 166,
Larjavaara, Matti 28, 50, 107 fn, 143 185, 188, 193, 194, 195, 196, 200, 207
Larramendi, Manuel 308 Metslang, Helle 93, 139, 159, 160, 178, 185,
Larsson, Lars-Gunnar 93, 139, 262, 281 260, 262
Lass, Roger 51 Miestamo, Matti 8, 35, 64, 65, 66, 67, 79,
Lazard, Gilbert 486, 496 80, 81, 83, 93, 327 fn, 368, 460, 556
Lazdiņa, Terẽza B. 70 Mihkla, Karl 158, 166
Lees, Aet 93, 139 Miklosich, Franz 458, 460
Lehmann, Christian 484, 488 Miller, Jim 266
Lehmann, Winfred P. 498, 499 Milsark, Gary 301 fn
Leino, Pentti 260, 263 Mitxelena, Koldo 323, 326 fn, 338, 341 fn,
Leinonen, Marja 263 342
Leiss, Elisabeth 500, 505 Moore, John 139
Letuchij, Alexander 368 Moravcsik, Edith 5, 6, 51, 65, 127, 263, 402
Levelt, Willem Johannes Maria 258 Moreno, Juan Carlos 461 fn
Levin, Beth 183, 420, 458 Mosel, Ulrike 24, 78, 528, 556
Lichtenberk, Frantisek 536 Mulder, Walter De. 515
Lindström, Liina 5, 9, 36, 37, 70, 93, 185, Musgrave, Jill 531
203, 211, 224, 230 Mustajoki, Arto 356
Lõhmus, Maarja 98 fn Muysken, Pieter C. 131
Luján, Marta 301 fn
Luraghi, Silvia 4, 6, 7, 8, 11, 29, 41, 42, 44, Nachmanson, Ernst 452
46, 47, 48, 55fn, 59, 61, 82, 83, 89fn, Napoli, Maria 25, 423, 424, 425, 447, 453 fn,
105fn, 124, 178, 291fn, 323fn, 368, 423, 461 fn, 462
426fn, 427, 438, 445, 449fn, 450, 452, Naughton, James 71
454fn, 458, 463, 467fn, 468, 470fn, Nedyalkov, Igor 71 fn
471, 478, 486fn, 487, 488, 501, 504, Neidle, Carol 380, 382, 384
505fn, 506, 509, 510fn, 513, 533, 546, Nelson, Diane 4, 104
556, 558 Nemvalts, Peep 93, 158, 166, 178, 185, 193,
Lynch, John 523, 525 fn, 537, 548, 549, 550, 195, 207
551, 552, 558 Newmark, Leonard 73
Lyons, Christopher 5, 184, 238, 459, 535 fn, Newmeyer, Frederick J. 95
545 Nikiforidou, Kiki 48, 54
Lytkin, Vasilij I. 113 fn Nordlinger, Rachel 131
Maling, Joan 4, 93, 105 fn Oihenart, Arnaud 329

Manterola, Julen 292, 293 fn, 294, 295 fn, Oliveira, Christiane Cunha de 80
299 Onishi, Masayuki 173, 249
Marchello-Nizia, Christiane 498, 501, 505 Orav, Heili 137
Markianova, Ludmila 114 fn Ortiz de Urbina, Jon 334 fn, 335 fn, 336,
Marty, Anton 384 fn 339 fn
Author index 571
Paducheva, Elena V. 368 Santazilia, Ekaitz 292, 293 fn, 295 fn, 333 fn
Pajusalu, Renate 137 Sarasola, Ibon 326
Pakendorf, Brigitte 71 fn, 83 Sasse, Hans Jürgen 455
Pärn, Priit 99 Sausa, Eleonora 449 fn, 467 fn
Partee, Barbara H. 5 Schmid, Maureen Alicia 83
Pastuchowa, Magdalena 403, 404, 405, 411 Schmitt, Cristina 301 fn
Paus, Charles 380 Schneider, Cynthia 544, 545, 556
Pawley, Andrew 549 Schroeder, Christophe 515
Paykin, Katia 10, 30, 38 fn, 59, 195, 380 fn, Schroten, Jan 305
382 fn, 383, 384, 392 fn Schwyzer, Eduard 444
Payne, John R. 65 Seefranz-Montag, Ariane von 454 fn, 455
Penttilä, Aarni 260 Selkirk, Elisabeth O. 3
Plungian, Vladimir 143 Senn, Alfred 458
Posner, Rebecca 504 Serbat, Guy 478
Press, Ian 71 Serianni, Luca 507
Pugh, Stefan M. 71 Seržant, Ilja A. 5, 171, 354 fn, 379 fn
Shakhmatov, Alexei A. 352, 354, 361
Rajandi, Henno 93, 260 Shanskiy, Nikolaj M. 104 fn
Rannut, Lehte 158, 166 Shay, Erin 78
Rappaport Hovav, Malka 183, 420, 458 Silva, Maria Amélia Reis 80
Rätsep, Huno 195, 196, 207, 209 Sornicola, Rosanna 514
Reed, Ann M. 92 Spencer, Andrew 131, 348, 351, 371
Regula, Moritz 509 Spore, Palle 509
Reichelt, Hans 417 Sridhar, S.N. 454 fn
Renzi, Lorenzo 509, 511 Stark, Elisabeth 505, 514
Riaño, Daniel 453 fn Subbarao, Karumuri Venkata 454 fn, 461
Riese, Timothy 113 fn Sulkala, Helena 139
Rijk, Rudolf P. G. de 308, 309, 310, 323, 332, Surányi, Balázs 135 fn
335 fn Swan, Oscar E. 70
Ritter, Ralf-Peter 93 Świdziński, Marek 400, 406
Rivierre, Jean-Claude 530 Szimoncsics, Péter 113 fn, 115 fn
Rix, Helmut 432 fn Szober, Stanisław 400
Rodriguez, Sonia 294 fn, 299
Roodenberg, Jasper 297 Tabakowska, Elzbieta 4, 10, 45, 412
Ross, Kristiina 133 Tael, Kaja 178 fn, 230
Ross, Malcolm 52, 548, 551 fn Talmy, Leonard 266, 282, 469
Rudzka-Ostyn, Brygida 400, 403, 405, 408, Tamm, Anne 1, 4, 8, 26, 31, 48, 93, 105 fn,
412 111 fn, 125, 130, 133, 134, 139, 140, 143,
Rullmann, Hotze 305 185, 192 fn, 260, 261, 262
Ruttkay-Miklián, Eszter 123, 125 Taraldsen, Knut Tarald 75
Tauli, Valter 93
Sabatini, Francesco 513 Tekavčic, Pavao 506, 509, 511, 513
Sadler, Louisa 131 Tereščenko, Natalija Mitrofanovna 130
Salminen, Tapani 113 fn, 115 fn Terrill, Angela 79
Saloni, Zygmunt 400, 406 Thompson, Sandra A. 40, 43, 82, 219, 220,
Sands, Kristina 59 244, 248 fn, 420, 453 fn, 486, 496
Sang, Joel 158, 169, 171 Ticio, Emma 294
572 Author index
Toivonen, Ida 103, 121, 122 Verkuyl Henk 261

Tommola, Hannu 260, 282 Viitso, Tiit-Rein 114 fn
Torn-Leesik, Reeli 185 Vilkuna, Maria 31, 36, 57 fn, 58, 158, 167,
Trask, Robert Lawrence 293, 294 fn, 309, 194, 215, 232, 238, 239, 241, 247, 248
324, 341 fn
Traugott, Elizabeth C. 57 Wackernagel, Jacob 462
Tsunoda, Tasaku 420 Wade, Terence 71
Tsvetkov, Dmitri 114 fn Wälchli, Bernhard 65, 67, 68, 70, 71, 83,
Tveite, Tor 92, 139 143, 194, 328
Txillardegi ( José Luis Álvarez Wiedemann Ferdinand J. 262
Enparantza) 292, 302 Wierzbicka, Anna 388
Wilhelm, Andrea 305
Urbina, Jon Ortiz de 71, 334 fn, 335 fn, 336, Witzlack-Makarevich, Alena 181, 182, 183,
339 fn 188, 196, 201, 210, 250
Uribe-Etxebarria, Myriam 310 Wolff, H. Ekkehard 78
Urquizu, Patricio 333 fn
You, Aili 305
Väänänen, Veikko 479, 480, 483 fn, 490
Vaillant, André 370 Zabala, Igone 293 fn, 295 fn
Vainikka, Anne 4, 93, 105 fn Zaicz, Gábor 113 fn
Vaiss, Natalia 93, 185, 190, 191, 198, 205, Zalizniak, Andrej A. 348, 351, 370
218 Zamboni, Alberto 505 fn
Van Peteghem, Marleen 30, 379 fn, 380 fn, Zamparelli, Roberto 297, 298 fn
382 fn, 383, 384, 392 fn Zavala, Roberto 340
Van Valin, Robert D. 183, 218, 258, 259 Zribi-Hertz, Anne 316
Večerca, Radoslav 460 Zuazo, Koldo 292
Vendler, Zeno 218, 260, 386 Zubizarreta, María Luisa 505
Vendryes, Joseph 478
Vennemann, Theo 498

(Empirical Approaches To Language Typology 54) Silvia Luraghi., Tuomas Huuomo (Eds.) - Partitive Cases and Related Categories-De Gruyter Mouton (2014) PDF

Uploaded by

Copyright:

Available Formats

(Empirical Approaches To Language Typology 54) Silvia Luraghi., Tuomas Huuomo (Eds.) - Partitive Cases and Related Categories-De Gruyter Mouton (2014) PDF

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

(Empirical Approaches To Language Typology 54) Silvia Luraghi., Tuomas Huuomo (Eds.) - Partitive Cases and Related Categories-De Gruyter Mouton (2014) PDF

Uploaded by

Copyright:

Available Formats

What is the book about?

What is the book about?

What topics are covered in the book?

What topics are covered in the book?

Silvia Luraghi and Tuomas Huumo (Eds.

Library of Congress Cataloging-in-Publication Data

Bibliographic information published by the Deutsche Nationalbibliothek

6 2014 Walter de Gruyter GmbH, Berlin/Boston

Silvia Luraghi and Seppo Kittilä

Tuomas Huumo and Liina Lindström

Borja Ariztimuño López

V Historical perspectives on Indo-European languages

Luz Conti and Silvia Luraghi

Anne Carlier and Béatrice Lamiroy

Subject index 563

A the most actor-like argument of a transitive verb

e–NP the sole argument of the existential clause

GCL general possessor classiﬁer

O object; the ‘not most actor–like argument’ of a transitive verb

S single argument of canonical intransitive verb

WOC Western Oceanic

Dutch (constructed example)

Finnish (from Heine & Kuteva 2004: 32–33)

Finnish (from Koptjevskaja-Tamm 2001)

Finnish (personal knowledge)

Finnish (personal knowledge)

German (from Heine & Kuteva 2004: 33)

Finnish (from Koptjevskaja-Tamm 2001: 523)

Crucially, pseudo-partitives are non-referential. As shown in (7), non-referentiality

Finnish (personal knowledge)

(8b) Houkuttel-i-n sinu-t ulos.

In any case, non-referentiality is often connected with partitive cases, as

extension of partitive case markers to subjects seems to start with existential

Finnish (from De Hoop 1998: 204)

(10) Tunne-n ruotsalais-i-a

According to De Hoop (1998), (9) has a strong (referential) interpretation, while

Importantly, Moravcsik remarks that the marking diﬀerence brought about

Finnish (personal knowledge)

Similarly, in several Indo-European languages the partitive genitive is in free

The ﬁfth section, on Indo-European languages, also contains three chapters.

Partitive cases constitute a rather heterogeneous category. They may be deﬁned

Keywords: Partitive case, partitive genitive, indeﬁniteness, low transitivity,

specific meanings, such as partial affectedness, indefiniteness, non-telicity (see

Finnish (personal knowledge)

b. nais-i-a tul-i koti-in

(2) a. Aino sö-i leivä-n

b. Aino sö-i leipä-ä

In (1b), the Finnish partitive expresses indeﬁniteness; the group of women in

be followed by a thorough functional typology of partitive markers based on the

2 Formal typology of partitive markers

2.1 Case markers

Finnish (personal knowledge):

Basque (Etxeberria, this volume)

Partitive markers typically contrast with nominative/accusative/absolutive

unaccusative verbs as in (1), and even, occasionally, with subjects of transitive

Finnish (Tuomas Huumo p.c.)

Avestan (in Yašt 7.4; Dahl, this volume)

Ancient Greek (Arist. HA 612a24)

Modern Russian takes a position in between the two groups, as it features a