Harris Georgian Syntax 1981 PDF
Harris Georgian Syntax 1981 PDF
Harris Georgian Syntax 1981 PDF
acquisition·
II MELISSA BOWERMAN: Early syntactic development·
ALICE C. HARRIS
Research Assistant Professor of Linguistics, Vanderbilt University
MELBOURNE SYDNEY
Published by the Press Syndicate of the University of Cambridge
The Pitt Building, Trumpington Street, Cambridge CB2 IRP
32 East 57th Street, New York, NYI0022, USA
296 Beaconsfield Parade, Middle Park, Melbourne 3206, Australia
Harris, Alice C.
Georgian syntax. - (Cambridge studies in
linguistics; 33 ISSN 0068-676x)
I. Georgian language
1. Title II. Series
499.996 PK9I 05 80-4I497
/1 2 -j � )
FOR JIM
Foreword
DAVID M.PERLMUTTER
Preface xv
Notes on presentation xvu
INTRODUCTION 1
I. Posing the problems 1
2. The approach taken: theoretical framework 4
3. Results of the investigation 8
4. Some necessary preliminaries 12
I. Tav-Reflexivization
2. Tavis-Reflexivization
3 . Person Agreement
4. Unemphatic Pronoun Drop
5. Summary
3 OBJECT CAMOUFLAGE
4 OBJECT RAISING 53
I. A description of the data in Georgian 53
xii Contents
2. An analysis 54
3. Facts concerning the nominative-nominal 56
4. Facts concerning the infinitive 60
5. Facts concerning the tvis-nominal 62
6. Object Raising as a test of direct-objecthood 63
7 PASSIVIZATION 103
.
I. The passive from the viewpoint of language universals 1 04
2. Indirect objects in passives 110
3. Theoretical issues 112
4. Interaction with other rules 1 14
5. Conclusion 116
8 INVERSION 1 17
I. The problem: the case marking differential in Series III 1 17
2. The proposal: a rule of inversion 1 18
3 . Preliminary arguments 120
4. Inversion verbs 1 27
5. Additional arguments 132
6 . The form of the rule 1 33
7. Interaction with other rules 1 38
8. Summary 141
Appendix A: Additional arguments for inversion verbs 141
Appendix B : Transitive inversion verbs with no overt suhject 144
Contents XIIl
12 TRANSITIVITY 181
I. Analysis 181
2 . Causative Clause Union 1 82
3 . Retired Term Marking 1 84
4. Conclusion and extension 186
5. Transitive and intransitive 1 86
Appendix: Real and apparent exceptions to transitivity 187
14 REFLEXIVIZATION 205
I. A review of the facts 205
2. A proposal 208
3. An alternative proposal 210
xiv Contents
EPILOGUE 275
1. The grammatical relations 'subject', 'direct object' and
'indirect object' 275
2. Simplifications in case marking 275
3 . Retired term marking 277
4. Rule interaction 277
5. Characteristics of grammatical relations 278
Notes
References
Index
Preface
3 v, w, f v
� z z
(» t' t
(') i, I 1
J � �
� 1 1
a m m
I) n n
ro 0 0
3 p p
-a z z
M r r
lJ 8 8
0 t t
::J u u
en p' p
k' k
a y y
�
� q q
s- - s
'a
c c
f)
(3 C(t8) C
d dz 3
xviii Notes on presentation
V 9 Us)
3 x
b x
� j j
h h
J
2 A note on glosses
The morphology of the Georgian verb. is very complex, and there is
no way to escape the use of examples with co mplicated verb forms. In
addition to the lexical meaning of the root, a single verb form may code
the following information:
person of subject
person of direct object
person of indirect object
number of subject
number of direct object
number of indirect object
tense
aspect (complete/incomplete, habitua l/non-habitual)
VOice
mood
direction and orien�ation
cilUsative/non-causative (cf. ch. 5)
version (cf. ch. 6)
etc.
In this work, the gloss of a verb will not i!lclude all this information
for two reasons: (i) A great deal of it is irrelevant to the topic of this
monograph and would simply overwhelm the reader with a mass of
material not necessary for interpreting pa rticlJlar examples. (ii) Although
Georgian is generally agglutinative, the information necessary to in
terpret particular examples cannot always be attributed to any par
ticular morpheme. Therefore, the principle behind my glosses is that
they should give only that information which will enable the reader to
understand the example in the context of the point under discussion.
Since one of the main concerns of this monograph is the correspondence
between case and grammatical relation, case is always dearly marked.
Since case varies with Series and Class of the governing verb, those
are always marked in the verb gloss . The grammatical relation is not
Notes on presentation XIX
2.1 Analytic gloss.In the analytic gloss, which is directly below the
Georgian example, the information carried by one word is hyphenated
together in English; e.g. darera 'he-wrote-it'. Analytic glosses are not
generally given for examples quoted in the text itself.
2.1.1 Nouns. Plurality is indicated by plurality in the English word,
2.2 Final gloss. The final gloss gives the sentence which corresponds
most closely in English, including the tense, which is the appropriate
one in English. In some instances, English has no corresponding syntac
tic rule, and the sentences in which the rule has applied and those in
which it has not have the same gloss.
Final glosses are sometimes omitted from ungrammatical sentences,
since those sentences do not always have a meaning. But if it would not
be clear what the ungrammatical sentences show, a final gloss is given
in parentheses, that corresponds to the intended meaning of the sentence.
3 Diagrams
This work uses the conventions of the 'network' diagramming developed
for relational grammar (Perlmutter & Postal 1 977). A clause and the
elements which constitute it are represented as nodes. Arrows connect
ing the clause node with other nodes indicate that the latter are depen
dents of the clause. Labels on the arrows indicate the grammatical
relations which various elements bear to the clause ; only the central
grammatical relations (cf. P.S below) are indicated in network dia
grams. The predicate relation is labeled 'P' . The subject relation is
marked 'I', the direct object relation '2 ', and the indirect object re
lation ' 3 ' . Benefactives are marked as 'B'. For example, network (I )
should be read as 'gela is the subject of the matrix clause (or of the verb,
codna "know"); and its direct object is the clause consisting of the predi
cate se�erva "sew", the subject nino, the direct objectperangi "shirt", and
the benefactive bavsvi "child".' This corresponds to the sentence (2).
4 The reader
The monograph is written primarily for syntacticians who know little
or nothing of Georgian. For this reason, material that will be of interest
only to the Georgian specialist is put, as often as possible, into notes or
appendices. The complexities of case marking and verb agreement
are introduced gradually.
The traditional names are used for cases; the names themselves are not
intended as a claim about the real structure of the language. The nomi
na tive case is marked by -i (0 after a vowel), the ergative by -ma (-m
after a vowel), and the dative by -so The three case marking Patterns
possible for a given clause are stated in (2).
Rules that change grammatical relations are of three types : (i) Revalu
ations include advancements and demotions. An advancement is the
promotion of a nominal dependent of a clause up the hierarchy (4). A
familiar example of this is Passivization, the advancement of a direct
object to subjecthood (Perlmutter & Postal 1974, 1 977, and ch. 7). A
demotion lowers a nominal dependent of a clause on the hierarchy (4).
One demotion is discussed here ; Inversion is the rule that demotes
subjects to indirect-objecthood (cf. Harris (to appear a), Perlmutter (to
appear b), and ch. 8). (ii) Ascensions raise a nominal out of an embedded
clause, making that nominal a dependent of the higher clause. A
familiar example of this is Non-Subject Raising ( Tough-Movement),
the ascension of an object from a sentential subject to subjecthood of
the higher clause (cf. ch. 4). (iii) Clause Union makes all dependents of
an embedded clause dependents of the matrix clause. Clause Union is
discussed here in relation to organic causatives (ch. 5) and inceptives
(ch. 1 6, §2.2.3 and n. I f).
Since the structure of a clause is the set of grammatical relations
holding between the elements of a clause, and since syntactic rules may
act on nominals to change grammatical relations, the theory recognizes
distinctions in derivational levels, or strata. It is found in this work that,
for Georgian, reference to initial and final levels is adequate for the
formulation of most rules. Intermediate levels, however, are also recog
nized. (The schematic representation of levels of derivation is discussed
in the introductory Note on Presentation.) An important principle of
relational grammar is that a particular grammatical relation is not
simply a bundle of properties ; rather, the syntactic rules of a particular
language predict precisely which properties are associated with initial
and which with final termhood (cf. Johnson 1 977). For example, we
will find that in Georgian the derived subject of a passive can trigger
Subject Person Agreement, a property of final subjects, and cannot
trigger Tav-Reflexivization, a property restricted to initial subjects.
The theory of relational grammar defined in the works cited above
puts a number of constraints on each of the types of rules that exist
and thus the types of human language that can exist. Three of these are
central and are stated informally here. Others are introduced in relation
to a specific rule as they become relevant. Some are not relevant to this
work and are not discussed here (cf. works cited above, especially
Perlmutter and Postal (to appear c), for a more detailed discussion of
constraints ).
2. Theoretical framework 7
3.2.3 Rules that refer to both -init-ial and final termhood. Chapters 14-16
establish that the rules of Tav-Reflexivization, Number Agreement, and
two of the rules that assign cases are stated on initial and final termhood;
that is, they are so-called 'global rules'. (Tav-Reflexivization is stated
more simply in ch. I and refined in ch. 14 on the basis of facts con
sidered in intervening chapters.)
In this section several rules and phenomena are introduced. The first
has no theoretical importance to this work, but it is necessary to the
understanding of some of the data presented below. The second is
Question-Word Question Formation; its importance derives from the
fact that it is used as a test of constituency. Suppletion is introduced
in §4.3; it will be the basis for arguments concerning termhood in later
chapters. Finally, a statement is made concerning word order in
Georgian.
4. I Aris-Cliticization
In Georgian, there are pairs of sentences like (5-8), where aris 'he/she/it
is' alternates with its c1itic form, -a.
(5) (a) es arts ceml r;tgni.
this is my book
'This is my book.'
(b) es cemi {:ignia.
book-is
'This is my book.'
(6) (a) sad aris cemi {:igni?
where is my book
'Where is my book?'
(b) sadaa cemi {:igni?
where-is
'Where is my book?'
(7) (a) es {:igni cemi aris. / es (:igni aris cemi.
'This book is mine.'
(b) es {:igni cemia.
'This book is mine.'
(8) (a) gela aris ekimi.
Gela is doctor
'Gela is a doctor.'
4. Some necessary preliminaries 13
As we can see by comparing (6) with (5) and (7-8), -a occurs after
vowels, -aa after consonants.
It is most natural for the cliticization rule to apply wherever its input
conditions are met, except when the word aris is emphasized. Because -a
is more natural than aris, most of the examples given in this work
use -a ; it will be glossed as in the (b) examples above.
I have called -a a clitic because it can never bear stress; nor can
there be a pause between the enclitic and the word to which it attaches.
This rule applies to reduce only the third person singular form only
i n the present tense; there are no clitic (i.e. unstressable) forms of the
other persons and numbers in the present or of any person in other
I ·nses. 6
( 1 0) ( a ) es noxi gar;mendilia.
this rug cleaned-is
' This rug has been cleaned.'
(b) es noxia gar;mendili.
rug-is cleaned
'This rug has been cleaned.'
14 Introduction
4.2 Question-word Question Formation
In this section I describe the formation of Q-word questions in Georgian.
Because of constraints described in §4.2.3 and 4-2.4, this rule serves as
a test of clause constituency (cf. §4.2.6).
For forming content questions, Georgian uses question pronouns, an
adjective romeli 'which', and a variety of adverbs.
phatic negative, or ver 'cannot' .6 (12) gives variant word orders ; only
(12a) is grammatical, with the order Q-word NEG VERB. Any other order
is ungrammatical.
(12) (a) sad ar ginda (:asvla?
where not you-want going
'Where do you not want to go?'
(b) "'ar sad ginda (:asvla?
(c) "'ar ginda sad (:asvla?
I will refer to the proper position of the Q-word as 'immediately pre
ceding the finite verb', in spite of the intercession of these particles.
occur in preverbal position in the same order that the full nominals for
which they substitute occur most naturally in clauses, that is, SUBJECT
UJJJECT. (13a) gives a declarative sentence; (13b) is its grammatical
ounterpart with Q-words as both subject and direct object. (*13c-e)
�ive variant word orders.
""J ECT.
l .1 '.
who-at you-go
'Tu whose house arc you going?'
16 Introduction
'Iol l l l rnatical meaning, then, is 'Where was Gela when he said that
I ' , I wr left ?' ( 1 8-19) show that in Georgian it is impossible to question
I ('I ! I 1st ituent of an embedded clause with a direct question.
' I ' l l ' mea ni ng intended for (*1 8b) may be expressed by (zoa) or (zob),
that (2ob) is. Each has two intonation peaks, one on each Q-word : ras,
. .
Vtn, rogor, vtn.
4.3 Suppletion
In Georgian there are certain verbs that arc supplctivl! for ani macy of
4. Some necessary preliminaries 19
1.,1 '0\1 'r, (aqvana, not m#ana, is used with the direct object bavsvi
tht child i fl krl or carried :
I l I l d ' , W I H.:tl lcr
20 Introduction
\ ' 1 " m e n t trigger (ch. IS ) . This can be seen in (32), where the animate
'
I Tav-Reflexivization
In this section I will examine the rule that governs the occurrence of
the reflexive pronoun tav-. It will be shown that this pronoun may be
coreferent only to the subject of its clause and that it therefore serves as
a test of subjecthood. 1
I.I Morphology
Reflexivization is expressed in Georgian with the pronominal element
tav-. This root also functions as an independent noun meaning 'head'
or 'source' ; it will be glossed here as 'self' when it functions as a
reflexive. The forms tavis- (singular) and taviant- (plural) serve as
possessive reflexives. The first person singular reflexive is cemi tavi
'my self' , the second seni tavi 'your self', and the third person reflexive
is tavisi tavi 'self's self', that is, 'himself, herself, itself', without dis
tinction of gender. 2 The possessive element, tavisi, is deleted from the
third person reflexive under certain circumstances (cf. ch. 6). The
reflexive phrase is grammatically third person and triggers third person
24 Syntactic tests for termhood
(2), (3), and (4) show that the reflexive pronoun is not coreferent with
the direct object, the indirect object or the object of a postposition.
In (2) tav- can refer only to mxatvari, not to vanos; the direct object,
vano, cannot trigger tav-. Similarly in (3a), tav- refers only to nino, the
subject, never to givis, the indirect object. The coreference that obtains
between the subject and tav- is independent of the relative positions of
the pronoun and antecedent, as can be seen in (3b) and (3c), which
differ from (3a) only in word order.
I.I Tav-Reflexivization 25
1 .3 Clausemate constraint
The rule of Tav-Reflexivization is governed by a clausemate constraint,
which requires that the pronoun and its antecedent be dependents of
the same clause. Thus in (sa) there is no ambiguity as to the meaning
reference of tav- ; it can refer only to nino, the subject of its own clause.
(s) (a) vano pikrobs, rom nino sa{mels
Vano-NOM he-thinks-I-3 that Nino-NoM food-DAT
amzadebs tavistvis.
she-prepares-it-I-I self-for
'Vanoi thinks that Ninoj is preparing food for herselfj .'
In spite of the fact that vano is also a subject, it cannot be the ante
cedent of tav-, since these two nominals are not dependents of the same
clause. We can see from (Sb) that this is independent of relative word
order, since here tav- precedes nino, yet nevertheless is unambiguously
coreferent to it.
(s) (b) vano pikrobs, rom tavistvis nino sa{mels amzadebs.
'Vanoi thinks that Ninoj is preparing food for herselfj.'
The lack of ambiguity in ( 5) shows that tav- in a postpositional phrase
cannot be triggered by the subject of a higher clause ; the ungram
maticality of (*6) shows that this is not possible, even where identifying
the reflexive pronoun with the embedded subject would produce
nonsense.
(6) *vano icvams axal �ostiums, romelic
Vano-NOM he-puts-on-it-I-I new suit-DAT which-NOM
tavistvisaa se�rili.
self-for-it-is-I-z sewn
(' Vano is putting on the new suit, which was made for him.' )
There i s no sensible grammatical reading for (*6).
26 Syntactic tests for termhood
(*7b) differs from (*7a) only in that in the latter the lower clause is
first, the pronoun thus preceding its antecedent.
Similarly, tav- in the higher clause cannot be coreferential to the
subject of the lower clause :
(*8b) differs from (*8a) only in that in the former the embedded clause
is first, and the trigger thus precedes the target of reflexivization. In
(*8) tav- in the matrix clause is triggered by vano, the subject of the
embedded clause.
1 . 5 Summary
Since the full pronoun tav- may have as its antecedent only the subject
of its own clause, it provides us with a test for subjecthood within a
single clause.
2 Tavis-Reflexivization 5
In this section, the rule that governs the occurrence of the possessive
reflexive pronoun will be described. The distribution of tavis-ftaviant- is
different from that of the full pronoun tav- ; we must therefore account
for these phenomena in terms of two distinct rules. It will be shown
that tavis-ftaviant- may be coreferent only with terms ; it therefore
serves as a test for termhood.
Any third person term can trigger tavis-ftaviant-. The subject is the
trigger in ( 10), the direct object in ( I I ), and the indirect object in ( 1 2) .
( 1 0) deda bans tavis svils.
mother-NOM she-bathes-him-I-I self's child-DAT
'The motheri is bathing heri child.'
( I I ) mno a3ievs bavsvs tavis dedas.
Nino-NOM she-gives-her-him-I-I child-OAT self 's mother-oAT ·
'Nino; is giving the childj to itsj mother.'
'Ninoi is giving the childj to herl mother.'
( t :.i.) ras acukebs gela ias tavis
what-oAT he-gives-her-it-I-I Gela-NOM la-oAT self's
dabadebisdyeze?
birthday-on
'What does Gelai give laj on herj birthday ?'
'What does Gela! give laj on his! birthday ?'
28 Syntactic tests for termhood
In ( I I ) and ( 1 2) either term may be the antecedent. Thus, any term may
be coreferential with tavis-.6
A non-term cannot trigger Tavis-Reflexivization, as established by
3 Person Agreement
( I S) Subject markers
singular plural
1 . person v- v-t
2. person " -t
3 . person -s/a/o -en/es/nen, etc.
. rhe dash represents the position of the verb stem, which includes the
root and various formants. The alternation between third person sub
ject markers is governed by the tense-aspect and Class of the verb form.
. fhe alternation in third person indirect object markers is phonologi
-ally determined. The latter alternation, first clarified in Shanidze
( [ 9 20), is also described in Tschenkeli ( 1 95 8 : vol. I, 370-2), Vogt
( [ 97 1 : 83), and Shanidze (1973 : 1 84-5) ' For some speakers the third
person indirect object marker is always zero, or has become part of the
vcrb root, thus making (16) and ( 17) nearly identical. Because the use
of the third person indirect object markers is not consistent, I will not
l i se the difference between them to argue that a given nominal is a
Compare the sets of sentences (24) and (25). The examples show that
mi(:era/mo(:era 'write' takes an indirect object, and is always interpreted
as having an indirect object even when there is none on the surface.
Da(:era 'write', on the other hand, does not take an (initial) indirect
object, and is interpreted as having a direct object, but not an indirect. 1 0
( 24) (a) (:erils miv(:er 3mas.
letter-DAT I-will-write-him-it-I-I brother-DAT
' I will write a letter to my brother.'
(b) (:erils miv(:er.
' I will write a letter to him.'
(c) miv(:er.
' I will write it to him.'
(25) (a) *(:erils dav(:er 3mas.
I-will-write-it-1- 1
('I will write a letter to my brother.')
(b) (:erils dav(:er.
' I will write a letter.'
( c) dav(:er.
' I will write it.'
The root of both verbs is (:er ; da-, mi-, and mo- are preverbs, the
presence of which indicates completed aspect. The difference between
mi- and mo- is determined by the grammatical person of the indirect
object (cf. ch. 6, example ( 19» ; first and second person require mo-,
third mi-. In addition, the difference between da- on the one hand, and
mi- and mo- on the other, corresponds to the glossed difference between
(24) and (25) ·
To account for the interpretation of the non-overt terms i n (24 25),
I4 Unemphatic Pronoun Drop 33
I I ngrammatical.
With the verb dafera, on the other hand, the indirect object agreement
marker must not occur. This is shown by the ungrammaticality of the
examples in (*28), which are parallel to (26), and the grammaticality of
the example in (29), which is parallel to (*27).
(28) (a) *vanom damfera ferili.
('Vano wrote me a letter.')
(b) *vanom dagfera ferili.
('Vano wrote you a letter.')
( c) *vanom dasfera ferili.
(,Vano wrote him a letter.')
(29) vanom dafera ferili.
'Vano wrote a letter.'
These four sets of sentences show that mifera/mofera must take an
indirect object marker and that dafera must not. This can be accounted
for in a straightforward manner if (i) these verbs are assigned the
initial terms claimed for them above, and (ii) Unemphatic Pronoun
Drop drops the pronominal indirect objects in (26) after they have
triggered Indirect Object Person Agreement. Since dafera will not
have an initial indirect object, nothing will trigger the rule, and the
sentences of (*28) will not be produced.
4.2.4 Conjoining. In (37), the subject is ia da gela 'Ia and Gela', not
ia, and not gela.
(37) ia da gela �avidnen mcxetali.
la-NOM and Gela-NoM they-went-II-2 Mtsxeta-in
'Gela and Ia went to Mtsxeta.'
I.4 Unemphatic Pronoun Drop 37
4.2.5 Possessives. The subject of (40) is givis 3ma ' Givi's brother'.
(40) givis 3ma cemi megobaria . .
Givi-GEN brother-NoM my friend-NoM-he-is-I-z
'Givi's brother is my friend.'
r f the referent of givi has been established in discourse, it may b e
pronominalized, but not dropped, as shown in (41). I n this case,
hecause of the necessity of interpreting the relation 'brother' as belong
ing to someone, the meaning of the sentence is entirely different if the
possessive pronoun is dropped ; but (4Ib), lacking the possessive, is not
natural.
(4 1) (a) misi 3ma cemi megobaria.
his
'His brother is my friend.'
(b) ?3ma cemi megobaria.
'My brother is my friend.'
38 Syntactic tests for termhood
4.3 Conclusion
We have seen that the rule of Unemphatic Pronoun Drop applies to
terms generally and not to a variety of nominals that are not terms ; I
conclude, therefore, Unemphatic Pronoun Drop can be used as a test
for termhood, since it applies only to terms. 14
5 Summary
In this section I will show that, in spite of the case marking differential,
the subjects of sentences containing Series II verbs correspond to the
intuitive notion of subject and to the subjects of Series I. These argu
ments are based on the tests for termhood established in ch. I . In each
instance I will demonstrate that for each verb Class, the ergative
nominal, the nominative-nominal, and the dative-nominal have all the
properties of the grammatical relation indicated for them in (5).
2. I Tav-Rejiexivization
J t was established in ch. 1 that tav- can be coreferential only to the
�ubject of its clause. In (6) we see that tav- is coreferential to the nom
inal bearing the case appropriate to the subject as stated in (5). That is,
fav- is coreferential to the same nominal as in Series I, though its case
2.2 Tavis-Reflexivization
In ch. I , §2, it was established that tavis- can be triggered by any
term. (8) shows that tavis- can be coreferential in Series II with those
nominals identified as terms in ( 5) above. In (8a) a subject triggers
tavis-, in (8b) a direct object, and in (8c) an indirect object. (8) 1S
parallel to (10-12) in ch. I , only the tense and cases being different.
(8) (a) dedam dabana tavisi svili.
mother-ERG she-bathed-him-II- I self's child-NOM
'The mother! bathed hert child.'
2.2 The analysis of case in Series II 43
zation. This is consistent with the fact that all and only terms can
trigger this rule.
(10) and ( I I), which repeat (19) and (20) of ch. I in tenses of Series II,
show that the nominals analyzed as objects in (5) are indeed those that
trigger Object Person Agreement.
. I ' l l ' subject properties established in ch. I are borne by the same
1 I 1 1 1 1 1 i nai in Series I and II, but the case marking is different in these
I W( I Series. The case pattern used in Series II, moreover, is governed
I . t he Class of the verb form. In ( 1 3) and ( 1 4) I combine the facts
1 I I I I I n arized in (5) above with those established for Series I i n ch. I .
46 Appendix : Constituent screeves
imperfect
conditional
present subjunctive
future subjunctive
Constituent tenses of Series II
aorist
optative (second subjunctive)
{;onstituent tenses of Series III
perfect (first evidential)
pluperfect (second evidential)
third subjunctive (third evidential)
The use of these tenses is described in Shanidze (1973 : 21 5-23), Tschenkeli
( 1 958 : chs. 8, 10, 1 2-13, 1 8-19, 39-40), and in other handbooks.
3 Object Camouflage
This chapter describes a rule that determines the surface form of first
and second person direct objects in clauses that also contain indirect
objects. This phenomenon will provide a further test for direct
objecthood and for the presence of an indirect object.
The sentences of (3) are like (*2) except that the expression seni tavi
'your-sG self' has been substituted for sen 'you-SG'.
I II (·5) and (6) we see that the same constraint that governs the occur-
1 l' 1 It' · of the second person also governs the first person ; (*5) and (6)
t l d l' ' I' from (*2) and (3), respectively, only in that the former have .first
1 '1 ' 1 !Ion direct objects instead of second person.
The plurals of the first and second persons are also governed by this
constraint, but the relevant examples need not be given here.
It remains to be shown that this rule applies to direct objects only
when there is an indirect object in the clause. (7) and (8) give paradigms
that change for the person of the direct object in Series I and II, respec
tively. s It is clear from these that the simple pronoun forms of the first
and second persons as well as those of the third person occur as direct
objects, as long as there is no indirect object.
(7) (a) (me) mar�munebs.
me-DAT he-convinces-me-I-I
'He is convincing me.'
(b) (Sen) gar�munebs.
YOU-DAT he-convinces-you-I-I
'He is convincing you.'
(c) vanos ar�munebs.
VanO-DAT he-convinces-him-I-I
'He is convincing Vano.'
(8) (a) (me) damar�muna.
me-NOM he-convinced-me-I I-I
'He convinced me.'
(b) (sen) dagar�muna.
YOU-NOM he-convinced-you-II-I
'He convinced you.'
(c) vano daar�muna.
Vano-NOM he-convinced-him-II-I
'He convinced Vano.'
(5-6) and (2-3) show that Object Camouflage must apply to first and
second persons, respectively ; ( I) shows that it does not apply to third
persons. (7-8) show that it operates only if an indirect object is present
in the clause. 4
2 An analysis
Such an analysis is needed to account for the fact that first and second
person pronouns, which occur freely as subjects and indirect objects,
and which occur as direct objects when there is no indirect object, do
not occur only when there is an indirect object. Further, this rule is
needed to account for the fact that misi tavi 'his self' is not found in
this situation. 6
3. ( Person Agreement
I t was observed above that the verb agrees with the phrases cemi tavi
II1d seni tavi as third person nominals, not as first and second persons.
This can be accounted for by insuring that Person Agreement apply
\ 0 \ h<.: output of Object Camouflage. In chs. 4-8 I will establish that
( '(lSI' Marking
II 1'II 11100 db g ed object has the same case marking as any direct object
1 I \ l l d l an v ', as detcrmined by Series and Class of the governing
I h.
52 Object Camouflage
4 An alternative analysis
The expressions cemi tavi and seni tavi are used both as camouflaged
objects and as reflexive pronouns (cf. ch. I, § I. I) . Because of this, it
might be proposed that Object Camouflage and Tav-Reflexivization
are a unified phenomenon. However, these phenomena cannot be
accounted for with a single syntactic rule, since the rules apply under
different conditions, including the following :
(i) Tav� Reflexivization requires coreference between the trigger and
the target ; Object Camouflage has nothing to do with coreference.
(ii) Tav-Reflexivization applies to any nominal dependent of the clause
which contains its antecedent ; Object Camouflage applies only to
direct objects.
(iii) Object Camouflage applies only to first and second persons ; Tav
Reflexivization has no such constraint.
(iv) Object Camouflage applies only in the presence of an indirect
object in the same clause ; Tav-Reflexivization has no such con
straint.
I conclude that there is no way to unify the rules of Tav-Reflexivization
and Object Camouflage.
I' 'ason, Object Raising provides us with a test for the direct object in
( ; ·o rgi an .
Berman ( 1974) presents evidence that the for-nominal in
( o rigi nates, not as the subject of the embedded clause, but as a dependent
( o f t he matrix clause. She argues that this for-nominal is coreferential
" i t II the embedded subject, which gets deleted by Equi. In §s bel()w,
" l i d i l l ch. I I , I give two arguments to support a similar analysis for
:rorgian. The Georgian arguments are of a type very different from the
/)1 1(' w h i ch Berman presents for English. Chung ( 1 976b) argues for
.1 1 1 ill la!ogous structure in Bahasa Indonesia. The fact that languages
those in (2-4). In the (a) examples, the subject is a mas dar, or deverbal
noun, together with the nominals it governs. 2 The (b) examples contain
infinitives. Both masdars and infinitives are treated more completely
in ch. 10.
2 An analysis
(5)
3neli qOPn,�
' hard ''bi
a P
r�\
� ��
It 2\1
m03ebna me �argi magalitebi
, find ' 'I' ' good examples'
(6)
me �argi magalitebi
(7) With matrix verbs of the type, advili aris 'is easy' , the direct object
in a sentential subject may ascend.
Succession Law, not yet stated in this monograph, requires that the
:I:lccnding nominal (here �argi magalitebi) assume the grammatical
I' 'lation borne by its host (here ""me �argi magalitebi m03ebna), in this
instance becoming subject (cf. Perlmutter & Postal (to appear b» . The
Chomeur Law (cf. Introduction, §2. I) predicts that the displaced sub
i . ·t, the cmbedded clause, will become a chomeur. Since it is suggested
I h a t these processes are universal, it is not necessary to state them in a
3 . 1 Suppletion
The first step is to argue that �argi magalitebi 'good examples' of (4b)
is initially the direct object of the embedded clause, as in (5). The
arguments will be based on suppletion of the verb root for number or
animacy of the direct object. The arguments will be of the following
form :
(i) In simple clauses that have undergone no change of grammatical
relations, there is suppletion of verb roots such that one root is used
if the direct obj ect meets a certain criterion (on animacy or plu
rality), an d a different root is used if the direct object fails to meet
this criterion.
(ii) In object-raised sentences, the same suppletion is observed with
respect to the animacy or plurality of the nominative-nominal.
(iii) If the nominative-nominal is, at some stage of derivation, the direct
object of the suppletive verb, the simple sentences and the object
raised sentences can be accounted for by a single rule. If, on the
other hand, the nominal in question is not the direct object of the
suppletive verb, the grammar must contain two rules to account
for these facts.
(iv) The nominative-nominal must therefore be a sometime direct
object of the suppletive verb.
tv) Since it is shown below that the nominative-nominal is final sub
ject of aris 'be', and since it could not simultaneously bear a
grammatical relation to two verbs, it must be at the initial level of
derivation that it is direct object of the suppletive verb.
gadagdeba (root gd) is used with singulars, while gadaqra (root qr)
occurs only with plurals. Evidence to support this and a complete dis
cussion is given in the Introduction, §4.3.z.
The root alternation in this verb is also governed by the plurality
of the nominative-nominal in object-raised sentences, as shown in (8).
In (*8a) the final subject of 3nelia 'it is hard' is singular. (8b) differs
from (*8a) only in that the final subject of the former is plural, and the
sentence is grammatical.
3.6 Summary
It has been shown that the nominative-nominal in the (b) sentences of
(2-4) has the properties of the initial direct object of the verb which is
realized as an infinitive, and the properties of a non-initial ( §3.2) sub
ject of aris 'is'. The obvious way to account for these facts is with a rule
that makes the initial direct object of the embedded clause the subject
of the matrix clause.
4. 1 Case Assignment
In ch. 2 I showed that in Series I and II in Georgian, all terms are
marked with one of the following cases : nominative, dative, or ergative.
The infinitive bears none of these cases. This supports the proposed
analysis, according to which the infinitive is a chomeur, not a term of
the matrix clause. Nor could the infinitive be a predicate nominal, since
predicate nominals are marked with the nominative case in all Series.6
from that of the tvis-nominal in (4b). I assume that the former is the
initial indirect object of micema 'give'. The tvis-nominal of (4b) is dis
cussed below in § 5.
4.3 Summary
In this section I have shown that the infinitive behaves as predicted by
the analysis proposed in §2 : it is a non-term ( §+ I ) nominal governed
by aris ( §4.2) ; the nominals it governs initially are not finally governed
by aris ( §4.2).
of that clause. The fact that vistvis 'for whom' immediately precedes
the verb in
shows that it is governed by the verb, aris 'is'. The English translations
capture the difference in the meanings in Georgian ; ( 1 4) asks toward
whom gift-giving is difficult, ( 1 5) asks who finds gift-giving difficult.
English uses to in the former function, and for or to in the second. In
Georgian tvis is used for both purposes. Both ( 14) and ( 1 5) are unam
biguous in Georgian, because of the constituent structure ; but ( 1 6) is
ambiguous.
Since tvis marks nominals in the functions illustrated in both (14) and
( 1 5), ( 1 6) is ambiguous between these two uses. Only in questions does
the word order unambiguously reflect constituent structure.
I have shown here that the tvis-nominal is a dependent of the matrix
clause in derived structure. If it had originated as the subject of the
embedded clause, it should have remained a dependent of the infinitive,
as did visgan and vistvis in ( 1 3) and ( 1 4). In ch. I I an additional argu
ment is given that the tvis-nominal must have originated in the matrix
clause. This argument is based on the fact that this nominal does not
have the properties of a retired subject.
object from those same sentential subjects can raise. The (*b) and (*c)
s ' ntences differ from one another only in that the latter the direct
Object Raising is the first of the rules examined that change grammatical
relations. The main purpose in coming chapters will be to examine the
nature of other rules that change grammatical relations. A major con
cern will be the case marking of terms and the marking of retired terms,
both processes differing in Georgian from the equivalent ones in other
languages.
5 Causative Clause Union
In the studies cited above, it is shown that in this type of causative the
following is generally true across languages :
5. 1 An analysis 67
initial structures like (5) and (8) to complete networks (6) and �9);
respectively.
rr�
(5)
(6)
. .
mamam mtmacemtna vardebi dedistvis.
father-ERG he-caused-give-me-it-II-I roses-NOM mother-for
'Father had me give roses to Mother.'
(8)
CAUSE mama
, fathe r '
In the next two sections I will show that the initial structure of an
organic causative has as its direct object a clause, and that the relations
that nominals bear to the clause are the same as the relations borne in
a simple clause. I will show that the direct objects of embedded tran
sitives and the subjects of embedded intransitives behave like direct
objects of the united clause under Causative Clause Union, with
respect to all rules stated on final relations. That the subject of an
embedded transitive and the indirect object of an embedded intransi
tive behave like indirect objects will also be shown. In §4 I will show
that the initial indirect object of an embedded transitive, under Clause
Union, behaves like a non-term with respect to all rules stated on final
relations.
Aissen ( I 974a, b) points out that many syntactic rules fail to apply in
the embedded clauses of causatives. She proposes ways of accounting
for this, but there remain problems. The fact that causatives disallow
the application of many rules in complement clauses makes it difficult
to argue for the initial grammatical relations. While the arguments given
here are strong ones, particularly the last, they are not numerous. But
the claim made for the initial relations in Georgian is supported by the
fact that there is evidence of various types for such structures in many
languages t cf. references cited above).
5.2 Initial grammatical relations 71
2.2 Tav-Rejlexivization 3
In ch. I, §I I showed that Tav-Reflexivization can be triggered only
by subjects. In an organic causative, tav- can be coreferential either
with the matrix subject, or with the final indirect object in causatives of
transitives, or with the final direct object in causatives of intransitives.
In ( 1 2), the causative of an intransitive, tav- may be coreferential to
ekimma 'doctor' or to 'llano, the final subject and direct object, respec
tively. In ( 1 3), the causative of a transitive, tav- may be coreferential
to genom or rezos, the final subject and indirect object, respectively .
( 1 2) ekimma 'llano alap ara�a tavis tavze
doctor-ERG Vano-NOM he-caused-talk-him-II-I self's self-on
'The doctori got Vanoj to talk about himself! . j . '
(13) genom miatanina rezos rignebi
Geno-ERG he-caused-take-him-it-II-I Rezo-DAT books-NOM
tavistan.
self-at
'Genoi got Rezoj to take the books to his!, j place.'
Only if 'llano (in ( 1 2» and rezo (in ( 1 3» are initial subjects, and Tav
Reflexivization is stated on initial terms, are these sentences consistent
with the facts of Tav-Reflexivization established in ch. I . 4
2.3 Summary
I conclude that causatives like (4) and (7) must have complex initial
structures like (5) and (8). This analysis will account for the lexical
correspondences between causatives and the corresponding non
causatives in a simple and elegant way. This is the only analysis that
can account in a straightforward way for the fact that the final indirect
object can trigger Tav-Reflexivization. The arguments to support the
5.3 Derived grammatical relations 73
3 . 1 Case Assignment
In this section I will show that the initial embedded direct object and
the initial embedded subject of an intransitive have the case marking of
a direct object, dative in Series I and nominative in Series I I . I will
show that the initial embedded subject of a transitive and the initial
embedded indirect object of an intransitive have the case marking of an
indirect object, regardless of Series. This is very strong evidence that
the case marking of nominals in organic causatives is not arbitrary ;
rather, the nominals actually become direct or indirect objects.
Below I give an inventory of organic causatives in order to show the
case marking of each kind of nominal. Because of the case differential,
each example of an organic causative will be given in a Series I tense
and in a Series II tense. In each instance, the syntax of a causative verb
form will be compared with that of the non-causative corresponding
to it.
In Georgian all organic causatives are Class 1 forms, regardless of
the Class of the corresponding non-causative verb form.
3 . 1 . 1 The causative of an intransitive. Intransitive verbs may be in
Class I , 2, or 3 .
(14) (a) bavsvi i3inebs.
child-NOM he-sleeps-I- I
'The child i s sleeping.'
(b) ubeduri semtxveva xdeba baySi.
unfortunate accident-NOM it-happens-I-2 garden-in
'An unfortunate accident happens in the garden.'
(c) sp artsmeni varjisobs.
athlete-NOM he-exercises-I-3
'The athlete is exercising, training.'
74 Causative Clause Union
Compare the non-causatives of ( 17), which are in Series II, with the
catlsatives of ( I S) and ( I 9), which are in Series I and II, respectively.
I II each of the non-causatives, the subject is mzia ; in the corresponding
rallsatives, mzia is in the indirect object case, dative in both Series. The
I I i reet objects of the non-causatives, cecxli 'fire' and nardi 'backgammon',
are in the case appropriate to the direct object in each causative : dative
mission' - in the causative bears the case of the direct object ; that is,
it is dative in Series I and nominative in Series II. The indirect objects
of (20) also bear the case of the indirect object in causatives, that is, the
dative case in both Series.
(21) (a) deda asalmebs bavsvs deidas.
mother-NOM she-causes-greet-her-him-I-I child-DAT aunt-DAT
'The mother is making the child greet (his) aunt.'
(b) dire�tori �omisias astumrebs
director-NOM commission-DAT he-causes-visit-him-it-I- I
vanos.
VanO-DAT.
'The director is having the commission visit Vano.'
(22) (a) dedam miasalma bavsvi
mother-ERG she-caused-greet-her-him-II-I child-NOM
deidas.
aunt-DAT
'The mother made the child greet (his) aunt.'
(b) dire�torma �omisia astumra
director-ERG commission-NoM he-caused-visit-him-it-II-I
vanos.
VanO-DAT
'The director had the commission visit Vano.'
5.3 Derived grammatical relations 77
On the basis of the data discussed in §3. I . 1 and these data, we can
, I raw the following conclusion about the case pattern in the causative
of intransitives : The subject of the non-causative corresponds to the
direct object of the causative, and the indirect object, if there is one, to
t i t(; indirect object.
An analysis that did not treat the nominals being discussed here as
I -rms of the united clause would have to complicate the case marking
I . ':' Tav-Rejlexivization
' 1 ''' . of Tav-Reflexivization is governed by a clausemate constraint
I' Ilk
( , r. ·h. § 1 .3) ; tav- has as its antecedent the subject of its own clause.
I,
I IH' r 'ad i ng
'The doctor got Vano to talk about the doctor' for ( 1 2) is
' 1 1 i I 'IIt with the clausemate constraint on Tav-Reflexivization only
I kill/II/a 'doctor' and ta?!- arc final dependents of the same clause.
78 Causative Clause Union
3 .3 Person Agreement
In ch. 1 it was shown that only terms trigger Person Agreement ; In
chs. 3-4 it was established that this rule is stated on final terms.
The initial embedded subject does not trigger Subject Person Agree
ment in organic clauses ; it does trigger Direct Object Person Agreement
or Indirect Object Person Agreement, depending upon whether it is
the initial subject of an intransitive or transitive verb. That this is so
is shown by (23), which illustrates the causative of a transitive. The
(*a) sentences have first person Subject Person Agreement v-, the (b)
sentences, first person Object Person Agreement m- (cf. paradigms in
ch. I , §3 ).
(23) (a) *tamadam vamyera.
toastmaster-ERG I-caused-sing-him-II- I
('The toastmaster made me sing.')
(b) tamadam mamyera.
he-caused-sing-me-II-I
'The toastmaster made me sing.'
(24) (a) *ninom gavaflnendina saxli.
Nino-ERG I-caused-clean-her-it-II-I house-NoM
(,Nino made me clean the house.')
(b) ninom gamar;mendina saxli.
she-caused-clean-me-it-II I -
"how that the initial direct object and the initial subject of an intransi
I i v . can also drop ; the sentences are used naturally when the reference
, Ii Question Formation
\ hown in §4.2 of the Introduction, Q-words must occur immediately
hl' fc l r ' the verbs which govern them. Consider the sentences below.
• rhty 5how that all the nominals under discussion here must occur
1 1 1 1 1 \ \ 'd iately before the clause-united verb.
, 7 (a) Tin
• migacemina forti dedistvis?
whO-ERG he-caused-give-you-it-II-I cake-NOM mother-for
' Who made you give the cake to Mother ?'
'''vin torti migacemina dedistvis?
"'vin dedistvis migacernina torti?
" ) 1flamam vis miacemina forti dedistvis?
father-ERG who-DAT he-caused-give-him-it-II-I
' Who did Father make give the cake to Mother ?'
lfivis mamam miacernina dedistvis torti?
lfimamam miacemina vis dedistvis torti?
( .) 1IIomam ra migacemina dedistvis?
what-NoM he-caused-give-you-it-II-I
' What did Father make you give to Mother ?'
lfi rn mamam migacemina dedistvis?
3 .7 Summary
In this section and the preceding one I have argued that the initial
structure of an organic causative must be complex, and that its final
structure must be simplex. It has been shown that the initial structure
of an organic causative has as its direct object a clause in which the
relations that nominals bear are the same as the relations borne in a
simple clause. Furthermore, the direct objects of embedded transitives
and the subjects of embedded intransitives behave like direct objects
under Causative Clause Union, with respect to all rules stated on final
relations. The subject of an embedded transitive and the indirect object
of an embedded intransitive behave like indirect objects under these
conditions.
5.4 The tvis-nominal 81
• The tvis-nominal
In the causative of a transitive, the indirect object of the embedded
· 1<'I u s e is marked with the postposition tvis in all Series. (29a) illustrates
I his for Series I, and (29b) for Series II.
(29) la) mama maceminebs torts dedistvis.
father-NOM he-causes-give-me-it-I-I cake-DAT mother-for
'Father is making me give the cake to Mother.'
(b) mamam mimacemina torti dedistvis.
father-ERG he-caused-give-me-it-II-I cake-NOM mother-for
'Father made me give the cake to Mother.'
M ;rems 'gives', the non-causative corresponding to the verb form of
�( ), is a Class I verb. The causatives of Class 2 verbs show the same
pat tern of case marking ; (3oa) illustrates the causative of a Class 2
\' ' I' b in Series I, (30b) in Series II.
(33), where the person of the tvis-nominal changes, but the verb form
docs not.
(33) (a) mamam forti miacemina cemtvis.
father-ERG cake-NoM he-caused-give-him-it-I I-I me-for
' Father had a cake given to me.'
(b) mamam forti miacemina sentvis.
you-sG-for
' Father had a cake given to you (SG).'
( c) mamam forti miacemina genostvis.
Geno-for
' Father had a cake given to Geno.'
S i nce only final terms trigger Person Agreement (cf. chs. 3 and 4), the
lal:k of agreement in ( 33) is explained by the proposed analysis, accord
i ng to which the initial indirect object of a transitive verb is not a final
I ' .. 01 in the corresponding organic causative.
, h. I I.
5 Theoretical implications
In Georgian there are syntactic rules that create indirect objects; these
phenomena are traditionally referred to as kceva or 'version'. 1 Three
types are illustrated in ( 1-3) ; in the (b) sentences there is an indirect
object ; in the (a) sentences there is none.
( I) Benefactive Version
(a) gelam self.era axali sarvali fentvis.
Gela-ERG he-sewed-it-II- I new trousers-NOM you-for
' Gela made new trousers for you.'
(b) gelam segiljera axali farvali (fen).
he-sewed-you-it-II-I YOU-DAT
'Gela made new trousers for you.'
(2) Possessive Version
(a) mzia pnends dis pexsacmlebs.
Mzia-NoM she-cleans-it-I-I sister-GEN shoes-DAT
'Mzia is cleaning her sister's shoes.'
(b) mzia Ut;mends das pexsacmlebs.
she-cleans-her-it-I - I sister-DAT
'Mzia is cleaning her sister's shoes.'
( 3) Superessive Version
(a) gelam surati daxata lJ,edelze.
Gela-ERG picture-NoM he-painted-it-II-I wall-on
'Gela painted a picture on the wall.'
(b) gelam surati daaxata Ifedels.
he-painted-it-it-II-I wall-DAT
'GcIa painted a picture on the wall.'
There are the following significant differences between the (a) and (b)
Icntenccs in each pair :
88 Version
I An analysis
(5)
2 Benefactive Version
In this section I show that the analysis proposed in (4-5) makes the
correct predictions with respect to rules previously considered.
2. I Person Agreement
The basic facts of Person Agreement were presented in ch. I, §3. It
was shown in succeeding chapters that Person Agreement is stated on
final terms. The proposed analysis predicts, then, that the derived
indirect object triggers Indirect Object Person Agreement. This is
correct, though not necessarily immediately obvious to the reader who
does not know Georgian, since the markers differ somewhat from those
described in ch. I, for reasons explained below.
Earlier I noted the following paradigm, which represents an inter
section of person and number.
(6) Indirect Object Person Agreement
singular plural
I . person m- gv-
2. person g g-t
3 . person S/h/0- S/h/0-t
(7) illustrates the use of these markers m sentences corresponding
l ( I b).
(7) (a) §e-m-H�era.
'He sewed it for me.'
(b) §e-g-i �era .
-
(c) se-u-/f-era.
'Hei sewed it for himj .'
The object markers occur together with a 'version vowel' or 'version
marker' , which is introduced by the rule of Benefactive Version, ( 4).
In (7a, b) we can clearly distinguish the first and second person markers. 2
But in (7c) the object marker that is elsewhere realized as S-, h-, or '"
has combined with the i-marker by a synchronic morphophonemic
rule to give u. An i- prefix occurs in Georgian (i) as a marker of Benefac
tive or Possessive Version, (ii) as part of the marker of future, aorist,
etc. for Class 3 verbs (cf. ch. 1 6, §I), (iii) as part of the marker of the
first evidential for Class 1 and 3 verbs, and (iv) as an empty morpheme
with a few verbs. In any one of these functions it may combine with
the third person indirect object marker as the fusional morpheme u-.
(8) +h+i+ � + u +
The morpheme h- and a following morpheme i- combine to u-.
One example of this rule occurs in (9c), where i- functions as part of
the future marker for a Class 3 verb.
(9) (a) miqvirebs.
he-will-yell-me-I-3
' He will yell to me.'
(b) giqvirebs.
he-will-yell-you-I-3
'He will yell to you.'
(c) uqvirebs.
he-will-yell-him-I-3
'He will yell to him.'
Another example may be found in ch. 8, n. 5 illustrating the second
evidential. Thus, in spite of the occurrence of the u- variant in the
third person, we can see that (7) contains regular indirect object markers.
The fact that the dative-nominal in sentences like (I b) triggers Indirect
Object Person Agreement, while the object of the postposition tvis
in sentences like ( I a) does not trigger any Person Agreement, supports
the analysis of the former as an indirect object and the latter as a
(benefactive) non-term.
2.2 Case
In ch. 2 it was shown that the indirect object is marked with the dative
case in both Series I and Series II. In ( lob) merab is in the dative in
6.2 Benefactive Version 91
Series I I ; this is good evidence that merab is the indirect object in that
sentence. In ( l Oa), on the other hand, merab is not in the case of the
indirect object, but in the genitive with a postposition ; this is good
evidence that merab is not the indirect object in that sentence.
( 10) (a) gelam se�era axali sarvali merabisatvis.
Gela-ERG he-sewed-it-II-I new trousers-NOM Merab-for
'Gela made new trousers for Merab.'
(b) gelam seu�era axali sarvali merabs.
he-sewed-him-it-II-I Merab-DAT
'Gela made new trousers for Merab.'
( I I ) corresponds to (10), but is in a Series I tense.
( I I) (a) gela �eravs axal sarvals merabisatvis.
Gela-NOM he-sews-it-I-I new trousers-DAT
'Gela is making new trousers for Merab.'
(b) gela u�eravs axal sarvals merabs.
he-sews-him-it-I-I
'Gela is making new trousers for Merab.'
In ( I I a) merab is the genitive object ,Of the postposition -tvis, in ( I Ib)
it is in the dative. This is consistent with the latter, but not the former,
being an indirect object.
In ( 1 2C) tavi can have its more concrete meaning 'head', in addition to
'self' (cf. ch. I , S I ). For this reason, another expression, such as seni
partreti 'your portrait' is preferred by some speakers for the more
general meaning. But the fact that (*I2b), without Object Camouflage,
is ungrammatical, while ( I 2C), with Object Camouflage, is grammatical
shows that this rule must apply here. Since Object Camouflage applies
only when there is an indirect object in the clause, me 'me' must be an
indirect object.
2.5 Summary
The four phenomena discussed above show that the version nominal,
when it is in the dative case, is the indirect object, which supports the
final grammatical relations proposed in § r . The initial grammatical
relations are discussed in the following section.
3 Stative verbs
Holisky (1 978) shows that true stative verbs in Georgian consistently
do not occur with benefactive nominals, as illustrated in (* 1 3).
The same set of verbs do not occur with verSlOn objects, as (*14)
shows.
6.4 Direct generation 93
superessive relation do not exist in all languages, are derivative, and are
marked.
5 . 1 Analysis
I propose that ( 1 7) be accounted for in the following way. The rule of
Benefactive Version, as proposed in § I , does not refer to coreference. A
rule of Coreferential Version Object Deletion obligatorily deletes
reflexive version objects. (A similar analysis is proposed for Tzotzil, a
Mayan language (Aissen ms.).) There are two obvious ways to state
96 Version
grammaticality of both ( I Ia) and (l Ib) and of both ( 1 7a) and ( 17c). The
obligatoriness of rule ( 1 8) accounts for the ungrammaticality of (*17b).
Since rule ( 1 8) is stated on non-final grammatical relations, the deleted
version nominal does not trigger rules like Person Agreement, which
are stated on final termhood. This accounts for the difference between
the morphology of (l I b) and that of ( 1 7c) (cf. §2. 1 ).
The extended traditional analysis (Grammar B) :
a. Rule (23) : Coreferential Benefactive Deletion
A reflexive non-term benefactive nominal deletes.
b. Side Effect of rule (23) : Add i- as a prefix to the root of the verb
which governs the deleted nominal.
c. Rule (23) is optional.
d. Rule (23) is stated on initial terms.
e. Constraint (24) : Version may not apply to nominals coreferential
to the subjects of their clauses.
Description : ( I Ia, b) are accounted for as above. Rule (23), unlike
( 1 8) , applies directly to initial structures, to grammatical relations that
have not been changed by Benefactive Version. The optionality of (23)
accounts for the grammaticality of ( 1 7a). Rule (23) and its side effect,
applying to initial terms, account for the grammaticality of ( 1 7c).
(*17b) is derived by the application of Benefactive Version and Tav
Reflexivization, which must be included in any grammar of Georgian.
In order to block (* 17b), constraint (24) is added to the grammar.
Evaluation. Both grammars must include Benefactive Version, a state
ment of its side effect, and a statement of its optionality. Rules ( 1 8)
and (23) are of equal complexity. But the Extended Traditional Analysis
is significantly more complex and fails to capture two important gen
eralizations.
(i) The Extended Traditional Analysis requires the extra state
ment B(b). It fails to capture the generalization that the side effect of
rules (4) and (23) are the same. The proposed analysis, on the other
hand, does not repeat this side effect ; it captures this important gen
eralization by deriving ( I I ) and ( 17) through the application of Benefac
tive Version, with one side effect.
(ii) The Extended Traditional Analysis requires the extra condition
(24) in order to block (*17b). On the analysis proposed here, the non
occurrence of such sentences is an automatic consequence of two
rules that are independently needed, Benefactive Version and Coref-
6.6 So-called four-person verbs 99
Sentences of this type are generally not acceptable in the dialect de
scribed here, the modern Tbilisi dialect. But in the dialects where they
100 . Version
are acceptable, such sentences pose a problem for the Stratal Unique
ness Law, the claim that there can be at most one subject, one direct
object, and one indirect object in a clause (cf. Introduction, g2).
The verb da{ra 'cut' does not take an initial indirect object ; I assume
that the two dative-nominals (Svilebs and me 'me') were derived by
successive applications of Possessive Version, from a structure contain
ing the direct object cemi svilebis qel(eb)i 'my children's throat(s)'. When
me becomes an indirect object, svilebs becomes an indirect object
chomeur. 12
The only evidence that could be cited to show that svilebs 'children'
is an indirect object in (25) is the fact that it is in the dative case in
Series II. But this is not sufficient evidence, for in natural languages, a
nominal may become a retired term!, while still bearing the case marking
of a termi (cf. Perlmutter & Postal 1974). This does not occur in
Standard Modern Georgian, but it did occur in Old Georgian (cf.
Harris 1979) and apparently in the dialect represented by (25).
Notice that me 'me', but not svilebs 'children', has the following
additional properties of an indirect object. (i) It triggers Indirect
Object Person Agreement (m-marker). (ii) In the first and second
persons, it triggers Number Agreement.13 (iii) It can be dropped by
Unemphatic Pronoun Drop, as shown by (25). That svilebs cannot be
dropped is shown by the fact that qeli dami{ra means 'he cut my throat',
not 'he cut my someone's throat' (cf. Shanidze 1 973 : §402). I have
been unable to test additional aspects of this construction, as I have
had no informant who spoke this dialect.
Since Svilebs does not have the other characteristics of an indirect
object, I conclude that it is not an indirect object. The principal differ
ence between this dialect and the standard one is that in the non-standard
dialect, some former terms carry the case marking of the term relation
which they last bore. As shown in ch. I I , this is not true in the standard
dialect.
I have proposed that the Version rules are form�lated on the basis of
the relations benefactive, possessor, and superessive. A theory of gram
mar according to which rules were based on linear order and dominance
relations would be forced to formulate Version on the basis of the post
position or case with which the nominals concerned occurred, or on the
6.7 Relation-changing rules 101
basis of their position with respect to some other constituent. But both
alternatives present enormous difficulties, as I will show below.
Consider, for example, a formulation of Possessive Version on the
basis of the case of the possessor : 'Genitive nominals may become in
direct objects.' (26a) illustrates a grammatical sentence with a genitive ;
the ungrammaticality of (*26b) shows that this nominal cannot be an
indirect object in this sentence.
8 Summary
In Georgian there are rules that create indirect objects ; the derived
indirect objects have all of the syntactic characteristics of superficial
indirect objects. A grammar that derives version objects from nominals
bearing non-term relations is simpler than one which generates all
indirect objects directly. A grammar which includes a rule of Corefer
ential Version Object Deletion is simpler than one that deletes non
term benefactives directly. It has been shown that the so-called second
indirect object which is acceptable in some dialects does not have the
properties of an indirect object, aside from the case marking. Thus, the
Stratal Uniqueness Law and the Chomeur Law have been upheld in
this set of data. 1 4
7 Passivization
Rule (8) makes the correct predictions for (7) and for examples con
sidered in previous chapters, but it fails to account for the ungram
maticality of (*9), where the mier-nominal ( = initial subject) has trig
gered Tav- Reflexivization.
1 . 2.2 Unemphatic Pronoun Drop. I showed above that all and only
unemphatic pronominal final terms may be dropped. ( 13) is parallel to
(2b) and differs from it only in that the nominative-nominal has been
pronominalized. (1 3a) has an emphatic pronoun as final subject ; ( 1 3b)
has an unemphatic pronoun as final subject.
(13) (a) is dalf.benilia Jaylis mier.
he-NoM bitten-he-is-I-2 dog by
'He is bitten by a dog.'
(b) dalf.benilia Jaylis mier.
'He is bitten by a dog.'
This confirms that the nominative-nominal is a final term.
The mier-nominal cannot be dropped in the same way. In ( 1 2) we
see the pronoun and postposition remaining. (*14) is parallel, but the
pronoun has been dropped.
(14) *saxli gaqidulia mier.
house-NoM sold-it-is-I-2 by
('The house was sold by him.')
If the entire postpositional phrase were omitted, the resulting sentence
would be grammatical, but the meaning would not be that of ( I2) ;
saxli gaqidulia means 'the house is sold', not 'the house is sold by him' .
Since these pronominal initial terms cannot undergo Unemphatic
Pronoun Drop, and since terms can undergo this rule, these must not
be final terms. This is consistent with the conclusion drawn in earlier
chapters that Unemphatic Pronoun Drop is stated on final terms.
1 .2.3 Case Marking. As noted above, the verb qopna 'be', which is the
auxiliary used with passives, is a Class 2 verb according to the mor
phological criteria that define this Class (cf. ch. 1 6, Appendix A). In
all tense Series, Class 2 verbs govern Case Pattern B, which requires
that the subject be in the nominative case. This is evidence that the
nominative-nominal of passives is the subject.
The mier-nominal is not in a case associated with terms, · but is
marked with a postposition. Since Case Marking is stated on final term
hood (cf. chs. 4-6), and since the mier-nominal is the initial subject, it
must be finally a retired subject.
1 . 3 Summary
The nominative-nominal has been shown to be both initial direct
object and final subject, while the mier-nominal is initial subject and
final non-term. On these grounds I conclude that the construction con
sidered here is a passive in the general linguistic use of that notion. The
arguments presented in this section imply no particular analysis beyond
the initial and derived grammatical relations of each nominal involved.
The data given above also confirm conclusions drawn in earlier
chapters with respect to rule interaction and rule statement : Tav
Reflexivization must be stated on initial terms, and the rules of Person
Agreement, Case Marking, and Unemphatic Pronoun Drop on final
terms. §1 .2.4 confirms that Object Camouflage applies to the output
of rules that change grammatical relations.
1 10 Passivization
trigger Person Agreement. Nor does it bear the case marking of a final
subject, direct object, or indirect object, but rather that of a final non
term (cf. ch. I I ) . This nominal cannot undergo Unemphatic Pronoun
Drop. If the pronoun is dropped, leaving the postposition, the result is
ungrammatical, as in (*I 9a).
Dialect A
(20) (a) viracam acvena vanos surati.
someone-ERG he-showed-him-it-II-I Vano-DAT picture-NoM
'Someone showed the picture to Vano.'
(b) *surati iqo nacvenebi vanos.
it-was-II-2 shown
('The picture was shown to Vano.')
(c) surati iqo nacvenebi vanostvis.
Vano-for
'The picture was shown to Vano.'
(d) surati iqo nacvenebi.
'The picture was shown.'
1 12 Passivization
Dialect B
(20') (a)
(b)*
(c)*
(d)
Dialects A and B
(zr) (a) es 1!.acebi mivacviet musaobas.
these men-NOM I-accustomed-it-them-II-r work-DAT
'I accustomed these men to work.'
(b) es 1!.acebi arian cveuli mufaobas.
they-are-I-z accustomed
'These men are accustomed to work.'
(c) *es 1!.acebi arian cveuli musaobistvis.
work-for
('These men are accustomed to work.')
In Dialect A the verb 'show' is grammatical in the direct construction;
(zoa) , and in the passive, with or without an initial indirect object,
which is marked with tvzS «(zoe), (zod), respectively). In Dialect B, the
direct construction is grammatical, but the passive is possible only if
it has no initial indirect object (contrast (zo ' d) with (*zo ' b, c)). In both
dialects, the verb 'accustom', unlike 'show', is grammatical with a
dative indirect object (example (z 1 )) .
3 Theoretical issues
With respect to the behavior of the initial subject and direct object,
the Georgian passive is in no way remarkable, as shown in § r . But the ,
behavior of the initial indirect object is puzzling. At issue is the moti
vation of the final non-term status of at least some initial indirect objects.
Some linguists claim that 'spontaneous chomage' is possible in language
(Comrie 1 977 ; Keenan I975). If such rules must indeed be counten
anced, we can simply add a rule to the grammar of Georgian describing
what happens in passives. But Perlmutter and Postal, attempting to
characterize possible variations among languages, have proposed the
Motivated Chomage Law, a constraint on the class of possible gram
mars (Perlmutter 1978 ; Perlmutter & Postal I977, to appear c). This
law, formally stated in Perlmutter & Postal (to appear c), claims in
effect that nominals bear the chomeur relation only under the con-
7.3 Theoretical issues 1 13
(23) Passivization
In the presence of a subject, a direct object may advance to subject.
Neither (22) nor (23) describes the fate of the initial indirect object,
1 14 Passivization
but there is not sufficient evidence for doing so at present, for the
reasons stated above.
In this section I will consider the rules we have previously dealt with
and show that this analysis makes the correct predictions in each case.
4. I Object Raising
Object Raising and Passivization can never apply in a single clause.
(24a) is grammatical without Passivization ; but (*24b), where Passivi
zation has applied in the embedded clause, is ungrammatical.
(24) (a) es motxroba advilia fasa1!.itxavad.
this story-NOM easy-it-is-I-2 to-read
'This story is easy to read.'
(b) *es motxroba advilia fa1!.itxuli (sa)qopnad.
read to-be
('This story is easy to be read.')
This is predicted by the proposed analysis. Since the input conditions
to each rule require a direct object, and since the output of each con
tains no direct object, they are mutually exclusive in a clause.
4.3 Version
The Version rules cannot apply to the output of Passivization. This may
be due to the fact that the finite verb of passives is qopna 'be', which
cannot take a version object, as shown in (z6) and (z7).
(z6) (a) es vaSli aris sentvis.
this apple-NoM it-it-I-z you-for
'This apple is for you.'
(b) *es va§li garis.
it-is-you-I-z
(,This apple is for you.')
(z7) (a) es aris seni otaxi.
this it-is-I-z your room-NOM
'This is your room.'
(b) *es garis otaxi.
it-is-you-I-z
('This is your room.')
Since qopna never takes Version, we may conjecture that this also
blocks the application of Version to a passive sentence.
Passivization cannot apply to the output of Version. In (z8a), neither
rule has applied. In (z8b), Possessive Version has applied, making
zurab the indirect object. In (*z8c, d), both Possessive Version and
Passivization have applied. Since Version makes zurab the indirect
object, and Passivization causes it to retire (cf. §z and 3) , we would
expect it to be marked with tvis or the dative case, as in (*z8c) and
(*z8d), respectively.
(z8) (a) zurabis das (oraven.
Zurab-GEN sister-DAT they-gossip-her-I-I
'They are gossiping about Zurab's sister.'
(b) das u(oraven zurabs.
they-gossip-him-her-I-I Zurab-DAT
'They are gossiping about Zurab's sister.'
(c) *da ga?orilia zurabistvis
sister-NOM gossiped-she-is-I-z Zurab-for
('Zurab's sister is being gossiped about.')
(d) *da ga(orilia zurabs.
(,Zurab's sister is being gossiped about.')
(e) zurabis da ga(orilia.
'Zurab's sister is being gossiped about.'
116 Passivization
(28e) establishes that Passivization can apply with this verb, as long as
Version does not apply. Thus, (28) documents the inability of Possessive
Version and Passivization to interact.
We cannot determine whether Passivization can apply to the output
of Benefactive Version. Since the initial indirect object in a passive
construction and the non-term benefactive are both marked with tvis,
it is impossible to tell whether or not Benefactive Version has applied
in the derivation of
(29) axali sarvali se�erilia sentvis.
new trouserS-NOM sewn-it-is-I-2 you-for
'New trousers are made for you.'
The inability of Version to apply in passives does not follow auto
matically from any rule previously proposed. 8
5 Conclusion
3 Preliminary arguments
The form of the argument will be the same for both initial and final
grammatical relations : r will show how one of the phenomena whose
general characteristics in Georgian syntax have already been established
operates in sentences like ( za, c). If the grammatical relations are as I
have proposed, the phenomenon can be treated in a general manner
throughout the grammar. If not, special ad hoc rules will be needed to
account for each of these phenomena in inversion constructions. Final
grammatical relations are discussed first.
Person Agreement
-3 . 1
Paradigm (8) represents the intersection of the categories of person
and number for Subject Person Agreement, first introduced in ch. I.
8.3 Preliminary arguments 121
The dash represents the position of the verb stem, which includes the
root and certain formants. The alternations in the third person are
determined by the tense and Class of the verb.
The markers in (8) indicate Subject Person Agreement generally in
Georgian, as shown in ch. I , §3 . I . In subsequent chapters it was shown
that Person Agreement is stated on final termhood (cf. chs. 4, 5, 6,
and 7).
In Series III, it is the nominative-nominal which triggers Subject
Person Agreement. (9 ) gives a paradigm of the singular with the
markers of (8) in heavy type.
(14) and ( 1 5) show that Object Camouflage must apply in Series I and
must not apply in Series III.
I have shown in previous chapters that Object Camouflage applies
to the output of rules that change grammatical relations (chs. 5, 6, 7).
Inversion and Unaccusative are relation-changing rules. From this it
follows that Object Camouflage applies to the output of these two rules.
But Object Camouflage is a rule that affects direct objects ; there is no
direct object in the output of the obligatory rule of Unaccusative. The
input conditions for Object Camouflage are therefore not met wherever
Inversion and Unaccusative have applied (cf. network (16)).
( 1 6)
3.3 Tav-Rejlexivization
Tav-Reflexivization is triggered only by subjects (cf. ch. I), and the
subject-trigger condition on this rule is stated on initial termhood
(cf. ch. 5, §2.2, and ch. 7, § I . I .3).
In inversion constructions in Series III, only the dative-nominal can
trigger Tav-Reflexivization. In (17a) the nominative subject of a verb
8.3 Preliminary arguments 125
3.4 Suppletion
It has been established that there are verbs in Georgian that are sup
pletive for the number or animacy of their subjects or objects. Each of
those verbs supplies arguments in support of the analysis proposed in
§2 ; two will be discussed here.
3.4.2 Suppletion for animacy of the direct object. The verb 'bring' is
suppletive for the animacy of its direct object ; moitana 'he brought it'
is used only with inanimate direct objects, moiqvana 'he brought him'
only with animate direct objects. It has been shown in earlier chapters
that this constraint is stated on initial termhood.
The same condition on animacy governs the nominative-nominal in
Series III ; on the proposed analysis, these are initial direct objects.
The direct objects in (20) are inanimate, those in (21) animate.
(20) and (21) show that the nominative-nominal of the inversion con-
8-4 Inversion verbs 127
4 Inversion verbs
(25)
the rule of Inversion, the rules that relate syntax to semantic interpret
ation must be complicated to account for the pair of sentences in (26).
On the inversion analysis proposed here, on the other hand, the initial
subject of the inversion verb form is the same as the (initial and final)
subject of the non-inversion verb form. Therefore the semantic in
terpretation of both sentences can naturally be related to their initial
term dependents.
on final termhood (cf. chs. 3, 4, 5 , 6, 7 and 1 6). For (27) and (28),
then, we can substitute (27') and (28') :
(27') Final Subject Final Final
Direct Object Indirect Object
Pattern A ERGATIVE NOMINATIVE DATIVE
B NOMINATIVE DATIVE DATIVE
(28') Series I II III
Class I, 3 B A B
2, 4 B B B
This correctly represents case marking in relation to final termhood,
showing that Pattern B is general for Georgian, Class I and 3 verbs in
Series II being an exception to the general rule. The traditional state
ment, (27-28), is misleading in the sense that it represents case marking
in relation to initial termhood, while case marking rules are, in fact,
based on final termhood.9 The facts summarized in (27'-28') show that
if one accepts the final termhood proposed above for the inversion
construction, significant simplifications can be achieved in the rules of
case marking.
The marking of the tvis-nominal is also accounted for naturally on
the proposal outlined above. Since case marking in Series III is based
on final termhood, if the initial indirect object of a sentence like (2a)
is finally an indirect object chomeur as proposed, it should have mark
ing appropriate to a non-term. Tvis is used to mark non-term benefac
tives (cf. ch. 6) and delegatives, as well as certain retired terms (cf.
chs. 5 and 7). The systematic nature of the latter will be established
in ch. I I .
4.3 Summary
Inversion verbs, predicates that trigger Inversion in all finite verbal
categories, and inversion in Series III have been shown to involve the
same two rules - Inversion and Unaccusative. One argument, based
on suppletion, has been offered for the initial grammatical relations
claimed in (23), while a second argument provides support for the final
grammatical relations claimed in (6) for Series III and in (24) for
inversion verbs. This is based on the fact that the Inversion analysis
effects a significant simplification in the case marking rules : once
inversion is incorporated into the grammar of Georgian, the need for
Pattern C disappears.
132 Inversion
5 Additional arguments
of the causative, · mepe 'king', should be the final indirect · object (by
Inversion). Second, the direct object of the causative, sasaxle 'p�lace�,
should be the final subject of the inversion construction (by Unac
cusative). Third, · the indirect object of the causative should be an
indjrect object chomeur (by the Chomeur Condition). Each .0£ these
predictions of derived termhood is borne out in (30b), as shown by
the case marking and agreement facts.
Arguments presented above concern only the initial and final gram
matical relations in the inversion construction, without respect to the
form of the rules irivolved. We must consider in more detail what kind
of . rule or rules relate the initial grammatical relations to the final
ones.
Treatments of inversion phenomena written within the framework of
transformational grammar have generally proposed a rule that would
switch the relative line?r order of the subject and direct object (cf;
Rosenbaum 1 967 ; R. Lakoff 1 968 ; G. Lakoff 1970 ; Postal 1 970 & 1971).
This proposal has the problem · that a single rule cannot be used for
languages with different basic word orders. In addition, it makes the
wrong predictions about the word order of many languages, including
Hindi (Davison 1 969), Kannada (Sridhar 1976a, b), and Georgian,
where the initial subject of iriversion constructions and the subject of
134 Inversion
a single rule have two parts, as in A ? (Is there any other rule that has
two parts ?) Can a chomeur advance, as in B ? (Is there any other rule
that advances chomeurs ?) Can a term be demoted, as in c ? None of
these questions can be answered here. Nor is it possible at this time to
test the predictions outlined above against the facts of many natural
languages. We can choose between these three proposals only by taking
into account the properties of many languages. However, Georgian does
shed some light on the problem ; these data too must be taken into
account in determining the correct approach.
All of the hypotheses outlined above can account equally well for
inversion with transitive verbs. In Georgian, Inversion can also apply
with intransitive verbs, as shown by the examples that follow. (3 1) con
tains intransitive non-inversion verbs in the indicative ; (32) gives their
counterparts in the evidential (Series III). (33) contains intransitive
inversion verbs.
It is clear from (32) that intransitive verbs in the Classes that govern
Inversion in the evidential must undergo the rule, just as transitive
verbs do. Intransitive inversion verbs must also undergo Inversion in
all finite verbal categories.
In sentences of this sort, there appears to be no direct object to
undergo the first component of Inversion according to proposals A and B
above. Both proposals could be altered to accommodate the additional
data. On hypothesis A, we could make the direct object optional. That
is, if there is a direct object, it must undergo its part of the rule ; but
if there is no direct object, the second component of the rule could
still apply. On hypothesis B, part (ii) can apply only if the nominal has
been put en chomage (by the Motivated Chomage Law). A dummy
could be introduced to undergo the first rule, putting the subject en
chomage and thus implementing the application of part (ii).
Hypothesis c, on the other hand, can account for the intransitive
cases with no additional devices. Because it treats Inversion as just the
process of a subject becoming an indirect object, it is able to account
for (3 1-33) in a natural and simple way. Unaccusative will never be
involved in the derivation of sentences like these, since there is no
direct object. (3 1-33), then, would have no derived subject.H
(3 1-33) establish that Inversion applies independently of Unaccus
ative in Georgian. In ch. 1 3 it is shown that Unaccusative also applies
independently of Inversion. Unaccusative differs from Passivization in
that, while the latter applies only at a level of derivation where there is
a subject, the former applies only at a levelof derivation where there
8.6 The form olthe rule 137
ff
I 2
R-I I
x y y
I have not yet described the interaction of Inversion with three rules -
Passivization, Object Raising, and Inversion itself. Since there are two
distinct triggers of Inversion, there are two distinct ways in which
Passivization and Object Raising could each interact with Inversion ;
this means a total of five potential syntactic environments for inter
action. In each of these five environments only one of the rules can
apply in a given clause.
The input conditions for Inversion are met initially arid again ' after
Inversion and Unaccusative have applied. (47) represents a theoretically
possible network for such a clause ; (48) represents the clause as it
actually occurs.
7.6 Summary
We have seen that Inversion - triggered by a Series III or a Class 4
form - fails to apply to certain structures that meet the input con-
Appendix A : Additional arguments 141
ditions that have been specified thtlS far. In ch. 1 6, §3 it is shown that
this is part of a more general, and perfectly regular, phenomenon.
8 Summary
This appendix contains data to show that the arguments made in §3 above
are applicable, not only to Inversion in the evidential, but also to Inversion
'
triggere d by inversion verbs. There are no new arguments given here,. only
new data.
A. P E R S O N AGREEMENT
'
indirect object) will trigger Indirect Object Perso� Agreement.
As ( 1 ) shows, it is indeed the nominative-nominal that triggers Subject
Person Agreement. with inversion verbs. The bold-type subject marker in (Ia)
is from Pflradigm (8) above ; the example is in the aorist (Series II). In (Ib),
the. bold-type, agreement marker is from the set of secondary markers of
'
sub,j ect agreement (cf. n. 4) ; the example is in the preseht tense. (Series I). ,
,
142 Appendix A : Additional arguments
B. OBJECT CAMOUFLAGE
rule to apply to. Thus, the ungrammaticality of (*3) is accounted for with no
special devices.
C. TaV - R E F L E X I V I Z A T I O N
In support of the analysis proposed in §4 above, I will show that tav- can
be coreferential only to the initial subject. It has been shown that tav- can be
triggered only by the initial subject of its clause. In (4a) tav- is triggered by
the dative-nominal, temur, confirming that the latter is initial subject. In
(*4b), tav- is triggered by the nominative-nominal ( final subject) ; the
=
ungrammaticality of this sentence shows that the final subject is not the
initial subject. (*4c) is identical to (*4b) except that the order of dependents
has been changed, showing that the relative order of pronoun and ante
cedent is not responsible for the ungrammaticality of (*4b).
(4) (a) temurs uqvars tavisi tavi.
Temur-OAT he-Ioves-him-I-4 self's self-NoM
'Temur loves himself.'
(b) *tavis tavs uqvars temuri.
self's self-oAT Temur-NoM
('Temur loves himself.')
(c) *temuri uqvars tavis tavs.
Thus, the dative-nominal of inversion verbs behaves like the dative
nominal ( initial subject) of inversion sentences in the evidential and like
=
D. SUMMARY
There is a small group of inversion verbs (Class 4 verbs) which pat,terp. like
meSinia 'I fear' in {r)Y
(1) (a) *meSinia sen / 3ayli /tvitmprinavi.
I-fear-it-I-4 YOU-NOM dOg-NOM airplane-NOM
('I am afraid of you/of the dog/of an airplane.')
(b) mesinia seni /3aylisa /tvitmprinavisa.
YOU-GEN dog-GEN airplane-GEN
'I am afraid of you/of the dog/of an airplane.'
Under certain circumstances,18 verbs of this Class require that their initial
direct objects be· marked with the genitive case, instead of the nominative,
which is usual for inversion verbs�
Consider the following hypothesis for accounting for (1) :
(2) Verbs of the sesineba 'fear' class mark their final subjects with the
genitive case.
I will show below that the genitive-nominal does not have the properties of
a final subject, and that (2) is therefore inadequate.
The first argument is, of course, the case itself. In Georgian, final terms are
never marked with a case other than the nominative, ergative, or dative. This
set of verbs would provide the only counter�example to that generalization.
Second, the genitive-nominal in (1 b) fails to trigger Person Agreement:
When the genitive-nominal is second person, the verb cannot be marked for
a second person (final) subject, as the ungrammaticality of (*3) shows. In (*3)
the verb bears a second person subject affix, but otherwise is identical to (1 b);
(3) *meSinixar seni.
I-fear-you-I-4
('I am afraid of you. ')
Nor does the genitive-nominal trigger Object Person Agreement. For these
reasons, I reject the analysis (2) for this sub-class of verbs.
A second possibility for accounting for sentences like (r) involves making
the genitive-nominal the complement of a head noun, which then gets
deleted or dropped. This would explain why these nominals are in the
genitive, and not some other case. The complement of verbs like 'hear',
'fear', etc. have special case marking in many languages. This hypothesis
fails to explain the fact that in languages like German and Turkish the special
case marking used for the complements of such verbs is not genitive.
A third proposal involves the insertion of a dummy, putting the initial
Appendix B : Verbs with no overt subject 145
direct object en chomage. This insertion might apply before or after Unac
cusative, the dummy being inserted as direct object or subject, respectively.
Other analyses fail to explain why the complement of this sub-class of verbs
is marked with the genitive and not some other case. But on the dummy
insertion analysis, their inarking is predicted from a more general process. In
ch. I I I will show that all direct object chomeurs are marked with the genitive
case. I conclude that the most reasonable way of accounting for sentences
like (I) is by a dummy insertion rule. It is to be hoped that investigations of
other languages will shed more light on the nature of such rules.
9 Why Pattern A is not reducible to
Pattern B
It was shown in ch. 8 that the case marking chart represented here as
(I) is unduly complicated and can be simplified to ( I ').
(I) Series I II III
Class I , 3 B A C
2 B B B
4 C C C
( I ') Class I , 3 B A B
2, 4 B B B
( I) refers to the patterns as stated below in (2), while (I ') refers to the
restatement in (2').
(2) Subject Direct Object Indirect Object
Pattern A ERGATIVE NOMINATIVE DATIVE
B NOMINATIVE DATIVE DATIVE
C DATIVE NOMINATIVE tvis-nominal
(2' ) Final Final Final
Subject Direct Object Indirect Object
Pattern A ERGATIVE NOMINATIVE DATIVE
B NOMINATIVE DATIVE DATIVE
The simplification in (2') is due to the fact that (2) was based on initial
grammatical relations, which is inappropriate, since case marking in
Series III is stated on final grammatical relations. Once this is realized,
and the rule of Inversion is incorporated in the grammar, C is reduced
to a special instance of B, and need not be included in the case marking
rules at all, as shown in ch. 8, §4.2.
It seems natural, then, to ask, (Is not Pattern A of (2') reducible in a
similar way to a special instance of Pattern B ?' The answer, unfor-
9.3 Arguments against the proposed analysis 147
tunately for the simplicity of our grammar, must be 'no'. The reasons
for this are given below.
with Class I or 3 verbs in Series II, since (3) obviates the direct object.
(7) shows that Object Camouflage does apply.
The fact that the nominative-nominal is tav- shows that it must be the
final direct object. (3) makes incorrect predictions in this respect.
3 Conclusion
In Georgian there are at least three types of non-finite verb forms : the
so-called 'masdar' l or nominalization, the infinitive, and the participle,
including so-called 'future (passive) participles', 'past passive par
ticiples', and 'active participles' (cf. Chikobava 1950 ; Shanidze 1 973 ;
Tschenkeli 1958 ; Vogt 1971 for details). The foci of this chapter are the
masdar and the infinitive, but most of the remarks made in §3-5 apply
equally to the participles. Space prevents my considering them in detail.
Non-finite verb forms are not well understood from the point of view
of linguistic theory or language universals. In particular, I know of no
theory that accounts adequately for the initial and final grammatical
relations borne by the nominals governed by non-finite verb forms.
Nor is there a theory that can predict when a subordinate clause will
be realized as a non-finite construction. Is a verb realized as a non-finite
form because of the application of a relation-changing rule or of some
other rule of syntax ? Or is the occurrence of a non-finite form pre
dictable from universal principles such as those that predict the occur
rence of a chomeur (cf. Introduction, §z) ? These questions will not be
answered here. I hope, however, that the data presented will help to
clarify the nature of non-finite verb forms and contribute to solving
some of those problems.
In this chapter, I wish to make the following points about non-finite
verb forms : ( § I ) So-called 'masdars' are derived nominals, not gerun
dives ; ( §z) forms characterized by the morphology sa-elad are syn
chronically infinitives, not participles ; ( §3) nominals governed by non
finite verb forms are marked in ways that are predictable on the basis
of the initial grammatical relations they bear ; ( §4) although these
nominals bear term relations initially, they are final non-terms. While
I am not proposing a complete analysis of non-finite verb forms in
Georgian, the facts presented here will have to be accounted for by
any analysis that might be proposed in the future.
152 Non-finite verb forms
This chapter and the one that follows are intimately related. On the
basis of material presented in §3 and elsewhere, I will propose in ch. I I
a general system of retired term marking. That, in turn, provides evi
dence for a clausal source for non-finite verb forms. This explains some
of the facts presented in §4 and at the same time represents a new kind
of argument for a clausal source for 'derived nominals', which masdars
are shown to exemplify in § I.
with it in case. I will show below that the so-called 'future participle in
the adverbial case' does not have these properties.
There is no doubt that the infinitive in Georgian bears the mor
phology of the future participle (sa-el) and that of the adverbial case
(-ad). I suggest, however, that synchronically there is a circumfix
sa-elad, which is probably derived diachronically from the conjoining
of the other two ; I shall argue that forms so constructed are infinitives,
not participles. 7
First, observe that the future participle in the adverbial case occurs
in two of the same uses as the infinitive in English. In (6) it is the
chomeur of the sentential subject in an object-raised sentence. In (7)
it represents a purpose clause.
(6) kargi magalitebi snelia mosasebnad.
good examples-NoM hard-it-is-I-2 to-find
'Good examples are hard to find.'
(7) (:avedi tqe§i datvis mosalJ.lavad. 8
l-went-II-2 woods-in bear-GEN to-kill
'I went into the woods to kill a bear.'
As far as I am aware, these are the only uses of this form in Georgian. 9
It cannot be used in all situations where an infinitive would be used in
English ; for example (*8) is ungrammatical.
(8) *minda (misi) gasalJ.eteblad. 1 o
l-want-it-I-4 it-GEN to-do
('I want to do it.')
Second, notice that the forms in (6-7) do not have the characteristics
of the participle in (5) :
a. In (5) the participle modifies a head noun ; in (6-7) the infinitive
does not.
b. In (5) the participle precedes its head noun ; in (6-7) the infinitive
is the head noun and in (7) is preceded by other words with which it
forms a constituent. 11
c. In (5) the participle shows case concord with its head noun ; the
case of the head noun itself is determined by its grammatical relation
to the governing verb. In (6-7) the form of the infinitive is invariant.
Third, the uses of the infinitive in (6-7) are not covered by any of
the regular uses of the adverbial case. 1 2 That is, it is not predictable on
any independent grounds that the adverbial, rather than some other
156 Non-finite verb forms
3 . 1 Direct object
The nominal that corresponds to the initial direct object of a finite
verb is, in construction with a mas dar or infinitive, in the genitive case.
This is so whether it is the only overt term, as in ( I I ), is with an overt
subject, as in ( 1 2), or is with both an overt subject and indirect object,
as in (1 3)'
( I I ) datvis mof$vla am !qeSi a�r3alulia.
bear-GEN killing-NoM this woods-in forbidden-it-is-I-2
'Killing bears in these woods is forbidden.'
3.5 Summary
Nominals corresponding to initial direct objects and subjects of intran
sitives are marked with the genitive case when they occur in construc
tion with masdars or infinitives. Similarly, initial transitive subjects
are marked with the postposition mier, and initial indirect objects with
160 Non-finite verb forms
the postposition -t'Vis. This marking is independent of what other nom
inals also occur in construction with the non-finite verb form.
In these examples, the preverb ca- 'down' has been substituted for
gada-. As the example shows, the same suppletion holds.
(28) (a) *ak amovedi didi kvis lasaqrelad.
here I-came-up-I1-2 big stone-GEN to-throw
('I came up here to throw a big stone.')
(b) ak amovedi didi kvebis lasaqrelad.
stoneS-GEN
'I came up here to throw big stones.'
An additional example is given as (8) in ch. 4.
An analysis that relates the genitive-nominal to the initial direct
object can account for the grammatical status of (26-28) with the same
rule needed independently to account for sentences with 'throw' that
have undergone no change of grammatical relations.
4.1 .4 Restrictions on the inventory of initial terms. The root rer 'write'
occurs with a variety of preverbs. When it occurs with the preverb da-,
the verb takes an initial subject and initial direct object ; it may have
an indirect object only derivatively. When it occurs with the preverb
mi- or mo-, on the other hand, the verb takes an initial subject, direct
object, and indirect object. This has been discussed in previous chap
ters, especially ch. 1 , §4, and ch. 6, §4.
The same distribution of nominals holds with respect to the masdars
and infinitives corresponding to the finite verb forms, except that, as
noted above, the subject does not appear overtly with an infinitive.
This distribution of initial terms is exemplified for infinitives in (33-H).
(33) (a) avedi cems �abine#i rerilis dasarerad.
I-went-up-II-2 my study-in letter-GEN to-write
'I went up to my study to write a letter.'
(b) avedi lems �binetli rerilis dasarerad uprosistvis
boss-for
'I went up to my study to write a letter for the boss.'
(H) (a) avedi lems �abinetli rerilis misarerad.
to-write
'I went up to my study to write a letter to him.'
IO.4 The nature of the nominals 163
(b) avedi lems l$abine#i {:erilis misacerad uprosistvis.
'I went up to my study to write a letter to the boss.'
The different inventories of initial relations of these two different verbs
result in two differences between (33) and (34). First, (34a) is interpreted
as having an initial indirect object that has been dropped ; this is
reflected in the different translations of (33a) and (34a). Second, the
tvis-nominal of (33b) is interpreted as a benefactive (tvis means 'for'),
since this verb form cannot take an initial indirect object. But the identi
cal tvis-nominal in (34b) is interpreted as an indirect object, since this
form of the verb requires an initial indirect object, which has the
meaning 'to'.
(35-36) show that the same distribution holds for masdars and the
nominals they govern.
(35) (a) cerilis da{,:era. . .
letter-GEN writing-NOM
'The writing of the letter. . .'
(b) Cerilis da{,:era uprosistvis.
. .
boss-for
'The writing of the letter for the boss. . ' .
4. 1 . 5 Selection restrictions. The verb fevs 'he is lying' requires that its
subject be animate, as shown in (37).
(37) (a) gela fevs taxtze.
Gda-NOM he-lies-I-z couch-on
'Gela is lying on the couch.'
(b) *(:igni fevs taxtze.
book-NOM
(,The book is lying on the couch.')
164 Non-finite verb forms
The same constraint governs the genitive-nominal of the corresponding
masdar, fola, as shown in (38).
(38) (a) gelas fola taxtze. . .
Gela-GEN lying-NoM
'Gela's lying on the couch . . .'
(b) *fignis fola taxtze. . .
book-GEN
(,The book's lying on the couch. . .')
This can be accounted for with the same statement independently
needed to account for (37), if figni 'book' and gela are the initial subjects
of the masdars in (38), as in (37).
him
'His copying over . . . '
IO.4 The nature of the nominals 165
It can be seen in (39) and (40) that there is no variation in the form of
the non-finite verbs for person of subject or object.
4.2.3 Object Camouflage. Direct objects that are first or second person
are camouflaged if there is an indirect object in their clause (ch. 3) ;
this rule applies to the output of rules that change grammatical relations.
Object Camouflage does not apply to nominals governed by. non
finite verb forms, as shown for masdars in (41) and for infinitives in
(42). In the (a) sentences the rule has not applied, in the (*b) sentences
it has.
4.3 Conclusion
The evidence amassed above leads me to conclude that (i) a term! of
a finite verb form corresponds to an initial term! of a masdar and
infinitive, but (ii) the initial term! of a masdar or infinitive differs
superficially from the term! of the corresponding finite verb. The facts
presented here can be handled by an initial clausal source to which
syntactic rules apply or by a lexical analysis with redundancy rules.
In ch. I I I will argue for a clausal source.
I I Retired Term Marking
are stated on initial termhood, and (ii) they behave like non-terms with
respect to rules like Person Agreement and Object Camouflage, which
are stated on the output of rules that change grammatical relations. In
Standard Modern Georgian, final terms and final non-terms are clearly
distinguished by a number of phenomena considered in previous
chapters.
However, while linguistic theory defines two types of retired terms,
Georgian grammar does not distinguish between them in any way I
have discovered. All retired terms behave the same with respect to
rules stated on final grammatical relations. The marking they are
assigned does not distinguish between chomeurs and emeritus terms,
as shown below.
I suggest that only those retired terms are chomeurs, whose occur
rence is predicted by the Chomeur Law and the Motivated Chomage
Law. Which nominals are chomeurs, then, is completely determined by
the analyses given in previous chapters. There are at least two chomeurs
in Georgian. The mier-nominal in the passive is a subject chomeur ; it
is put en chomage by the advancement of the direct object to subjecthood
(cf. ch. 7, §3). The tvis-nominal in the inversion construction is an
indirect object chomeur ; it is made a chomeur by the demotion of the
subject to indirect-objecthood (cf. ch. 8, §2 and 6). In §4.2 I will
argue that the genitive-nominal that occurs with certain inversion verbs
represents a third chomeur, a direct object chomeur (cf. ch. 8,
Appendix B).
Perlmutter and Postal analyze as an indirect object emeritus the
initial indirect object that is, under Causative Clause Union, a final
non-term (cf. ch. 5, § I). I suggest that those retired terms in Georgian
that have not been identified as chomeurs are also emeritus terms. It
is to be hoped that by studying the occurrence of these in Georgian
and other languages linguists can discover the universal principles that
govern their occurrence. In Georgian there are a number of retired
terms that are not chomeurs in the sense defined above. The tvis
nominal in the organic causative is an indirect object emeritus. In the
passive, the initial indirect object is a final non-term.
In Object Raising, the direct object of a sentential subject is raised
to matrix subject; the embedded subject is deleted on identity with the
tvis-nominal of the matrix clause. The entire sentential subject is put
en chomage by the raising of the direct object ; an indirect object depen
dent of that clause is a retired indirect object (cf. ch. 4). There is no
170 Retired Term Marking
absolute way in Georgian to determine what kind of retired term this
represents.
Initial subjects of infinitives of purpose are deleted. As shown in the
preceding chapter, all other nominals bearing initial term relations to
verbs realized as infinitives of purpose or masdars are themselves
realized as retired terms.
In the remainder of this chapter I shall not distinguish between
chomeurs and emeritus terms, but shall refer to both as retired terms.
The genitive case is also used to mark the initial direct object in
a subset of inversion verbs. (3) illustrates such a verb with the initial
indirect objects are marked in the same way. This shows that the final
marking of retired terms depends on the last held term grammatical
relation.
Retired indirect objects also occur in passives ; these too are marked
with lvis, as illustrated in (10).
( 1 0) valli micemulia masfavleblistvis.
apple-NoM given-it-is-I-2 teacher-for
'An apple is given to the teacher.'
( I I ) illustrates the marking of the retired indirect object of a masdar
with lvis.
( I I ) vallis micema masfavleblistvis. . .
apple-GEN giving-NoM
'Giving an apple to the teacher . . . '
With infinitives of purpose, too, retired indirect objects are marked
with tvis, as in ( 1 2).
( 12) valli viqide masfavleblistvis misacemad.
apple-NOM I-bought-it-II-3 teacher-for to-give
'I bought an apple to give to the teacher.'
Finally, in object-raised sentences, an initial indirect object of the
embedded clause is marked with lvis. Here the initial indirect object is
a retired dependent of the infinitive.
( 1 3) salukari 3nelia anzoristvis misacemad.
gift-NOM hard-it-is-I-2 Anzor-for to-give
'Gifts are difficult to give to Anzor.'
Thus, tvis is used in a wide variety of constructions to mark retired
indirect objects.
2.5 Summary
The examples above are drawn from chs. 4 to 1 0 ; additional exa�ples
are given in each of the relevant chapters. In each instance, evidence
has already been given that the nominals marked with the genitive case,
or with mier, or with tvis are term nominals that have been retired by
the application of some rule.
In Modern Georgian retired terms are marked regularly according
to the last term grammatical relation they bear. Georgian does not dis
tinguish between chomeurs and emeritus terms by marking : it assigns
174 Retired Term Marking
retired term marking on the basis of the last term relation held, not on
the basis of the rule that retires the nominal. One can imagine a number
of other principles upon which final non-term marking might be
assigned j in §3 some of these will be explored.
3 Alternative analyses
Below I will discuss some other principled bases for marking retired
terms but in each instance I will show that this is not the best analysis
of Georgian.
Without further data, we might assume that at least the following
language types exist :
A. Retired terms are individually marked by the rule that retires them.
B. Retired terms are marked by a general set of rules on the basis of
objects (Harris 1979) and there are some petrified forms in Modern
Georgian that still mark on this basis (cf. ch. 5 , §4.2 ; ch. 6, §6). But
Standard Modern Georgian does not use this marking system pro
ductively (but cf. ch. 7, §2.3).
The putative type E language would mark former terms, not on the
basis of their termhood, but on the basis of semantics. 4 This would
mean, for example, that agents are marked with mier. This can be
shown to make the wrong predictions for Georgian. Agents are gen
erally limited to animate or personified nominals (Fillmore 1 968). But
( 1 7) illustrates the use of mier with nominals that are neither animate
nor personified.
( 1 7) (a) Nazmnari saxelis mier saxelta martva udur
deverbal noun by nOUnS-GEN governing-NoM Udi
enasi
language-in
'The governing of nouns by deverbal nouns in the Udi
language.'
Title of Panchvidze ( 1960)
(b) . .procesi
. ebiani mravlobitis mier zmnis setanxmebisa.
process-NoM plural by Verb-GEN agreeing-GEN
' . . . the process of the eb-plural agreeing with the verb.'
Imnaishvili (1957 : 664)
(c) . . .inpinitivis mier pirdapiri obiektis martvis
infinitive by direct object-GEN governing-GEN
semtxvevebi . . .
instances-NoM
' . . .instances of the governing of the direct object by the
infinitive. . .'
Chxubianishvili ( 1972 : 78)
There is evidence of another kind, too, that mier is not just an agent
marker. The initial subject of 'affective' (inversion) verbs is an
experiencer, not an agent. Yet this experiencer is marked with mier in
masdars, as shown in ( 1 8).
( 1 8) (a) tkveni megobris davifqeba cems mier didi
your friend-GEN forgetting-NoM me by big
uzrdelobaa.
rudeness-it-is-I-2
'It was very rude of me to forget your friend.'
178 Retzred Term Marking
I conclude from the evidence in ( 17) and (18) that mier does not mark
only agents, but rather marks retired subjects. More generally, I con
clude that the markers of retired terms are assigned, not on the basis of
semantics, but on the basis of grammatical relations. 5
In addition to the five possible language types discussed above, there
are many possible mixed systems ; for example, a language might choose
to mark all chomeurs with one case or positional and mark all emeritus
terms with another case or positional. But to use any mixed system to
describe Standard Modern Georgian would be to miss the generaliz
ations made in §2 above.
It has been shown here that type B above represents one possible
language type. It is hoped that further research in other languages will
reveal more about what types of marking exist and more about the nature
of retired terms, especially about the poorly understood emeritus
relations.
4. 1 Object Raising
In ch. 4 I proposed an analysis of Object Raising according to which
the tvis-nominal of (19) originates in the matrix clause, not as the
subject of the embedded clause. At that time I discussed evidence for
this position from other languages, and evidence from Question For
mation in Georgian, which shows that the tvis-nominal is not a con
stituent of the embedded clause.
Version, Passive, and Inversion ; it will have to state the same set of
generalizations a second time as redundancy rules in the lexicon for
masdars. Thus, a grammar that inserts masdars lexically as nominals
will miss the generalization that the retired terms of masdars are
marked like the retired terms of other verbs.
12 Transitivity
The verbs fers ' he is writing it' (Class 1) and tamaSobs 'he is playing'
(Class 3) are alike in that Series II counterparts of both govern ergative
case subjects, and Series III counterparts of both govern Inversion. In
this chapter I will discuss some syntactic differences between these
two verbs. This short chapter has three purposes : (i) to show that some
verbs have an initial direct object obligatorily, while others have an
initial direct object optionally or not at all, (ii) to relate this difference
in inventories of initial terms to the difference between Classes 1 and 3 ,
and (iii) t o discuss and define the notions 'transitive' and 'intransitive',
which will be important in later chapters.
I Analysis
I suggest that inventories of initial terms are lexically determined, and
that a given verb may govern a particular term relation either obli
gatorily, optionally, or not at all. In Georgian, a 'verb' in this sense
must be understood as a verb root with a specific preverb, if that root
takes preverbs. In ch. 1 it was shown that the verb da(:era 'write' takes
an initial subject and direct object, while the verb mi(:era 'write' takes
an initial indirect object, as well as an initial subject and direct object.
Thus, for the purposes of stating initial term inventories, da(:era and
mifera must be considered different verbs, even though they share the
root fer. l
I propose the following specific requirements for some verbs in
Georgian :
final indirect object in the causatives, (4). (*5) shows that the causative
would be ungrammatical if vano were instead the direct object ; (* 5)
is identical to (4a), except that vano is case-marked as the direct object. 3
(5) *vano davaferine.4
Vano-NOM
('I got Vano to write.')
With the verb tamaSi 'play' and other verbs of its type, on the other
hand, the derived termhood of the initial subject in an organic causative
depends upon whether or not an initial direct object of tamaSi is overtly
expressed. The causatives in (7) correspond to the simple sentences
in (6). In the (a) sentences, no initial direct object of tamasi is overtly
expressed ; in the (b) sentences, the initial direct object sami partia
'three rounds' is expressed.
(6) (a) vanom itamaIa.
Vano-ERG he-played-II-3
'Vano played.'
(b) vitamaset sami partia.
we-played-II-3 three rounds-NoM
'We played three rounds.'
(7) (a) vano vatamase.
Vano-NoM I-caused-play-him-II-I
'I got Vano to play.'
(b) bebiam (even) gvatamasa
grandmother-ERG we-DAT she-caused-play-us-it-II-I
sami p artia.
'Grandmother got us to play three rounds.'
Case marking and Person Agreement reveal that here vano is the final
direct object in (7a), that 'us' is the final indirect object in (7b), and
that sami partia is the direct object. (8) shows that it would be ungram
matical for vano to be instead the final indirect object of the causative
in (7a).
(8) *vanos vatamaIe.3
VanO-DAT I-caused-play-him-II-I
('I got Vano to play.')
The simple sentences in (3) and (6) are apparently alike ; the causa
tives in (4b) and (7b) are apparently alike. But the causatives in (4a)
184 Transitivity
and (7a) are different. The initial subject of dOfera 'write' is the final
indirect object of the corresponding organic causative, regardless of
whether an initial direct object of dafera is overtly expressed. The
initial subject of tamali 'play', on the other hand, is the final indirect
object of the corresponding causative if an initial direct object of tamaSi
is overtly expressed - (7b) ; the initial subject is the final direct object
of the causative if no initial direct object of tamasi is expressed - (7a).
This difference between dar;era and tamasi can be handled most
simply by (i) assigning to these verbs initial direct objects as specified
in ( 1 ) , and (ii) recognizing that Unemphatic Pronoun Drop does not
apply to nominals which are optional in the inventory of initial terms. 5
The marking noted above is not optional. (*13) shows that the
retired subject of da(:era may not be marked with the genitive instead
of the postposition mier. (*13) differs from (IOa) only in this respect.
( 1 3) *vanos da(:era . . .
VanO-GEN
('Vano's writing . . . ')
Similarly, (* 14) shows that the retired subject of tamasi may not
optionally be marked with mier when there is no expressed initial
direct object.
( 14) *givis mier tamaSi. . .
Givi by
('Givi's playing . . . ')
These sentences show that da(:era, but not tamasi, is treated as a transi
tive verb by the rules of Retired Term Marking, regardless of whether an
186 Transitivity
In §2 and 3 I have argued that some verbs have initial direct objects
obligatorily and some optionally. Verbs like qiqini 'croak' and xitxiti
'giggle' never occur with a direct object, and there is no reason to
believe that they ever permit one.
It is important to observe that da{:ers 'he will write it' is typical of
Class I verbs in that it takes an obligatory initial direct object. Tamasobs
'he plays' and qiqinebs 'he croaks' are typical of Class 3 in that they take
an initial direct object optionally or not at all. It is not the case, however,
that all Class I verb forms take an obligatory direct object. Nor is it
the case that all Class 3 verb forms take a direct object optionally or
not at all. Nevertheless, this characteristic is widespread enough for us
to state the following generalization : Most Class I verb forms take an
obligatory initial direct object ; most Class 3 verb forms take an initial
direct object optionally or not at all.
This appendix gives examples of verbs that are real or apparent exceptions
to the generalization that Class 1 verbs are transitive and Class 3 intransitive.
§I deals with intransitive and apparently intransitive Class 1 verbs ; §2 dis
cusses transitive Class 3 verbs.
One sub-class of Class 1 verbs never takes a direct object ; this group includes
sei{q#a 'he peeked', daaboqina 'he burped', daaxvela 'he coughed', and
gakusla 'he hurried away'.
Another small group of verbs take a direct object optionally. When they
have a direct object they are Class 1 verbs ; when they lack a direct object,
they are irregularly members of Class 2. For example, seagina 'he cursed him'
(Class I), but seigina 'he cursed' (Class 2) ; �bens 'he bites it' (Class I), i�bineba
'he bites' (Class 2).
A handful of verbs occur with a direct or indirect object, but not with
both. For example, the case marking in the sentences below shows that for
this verb, persons are indirect objects, while inanimates are direct objects. 6
(I) (a) *vano gavige.
VanO-NOM I-heard-him-II-I
(b) vanos (*vano) gavuge.
VanO-DAT I-heard-him-it-II-I
'I heard, understood Vano.'
(2) vanos ambavi /*ambavs gavige.
VanO-GEN news-NoM/news-DAT I-heard-it-II-I
'I heard, understood Vano's news.'
A verb with similar characteristics is dare�va 'ring' , although for this verb
the indirect object is derived by Locative Version. 7 For both of these verbs,
additional examples show that the inanimate behaves as direct object and
the animate as indirect object with respect to Passivization (3-4), Inversion
(5-6), Retired Term Marking (7-8) and Causative Clause Union (9).
(3) zari dare�ilia.
bell-NoM rung-it-is-I-2
'The bell was rung.'
188 Appendix : Real and apparent exceptions
One subclass of Class 3 verbs takes an obligatory initial direct object ; this
group includes ipovna 'he found it', isova 'he got it', iqida 'he bought it',
itxova 'he borrowed it', isesxa 'he borrowed it'. The initial direct objects of
these verbs behave regularly with respect to case marking, agreement,
Passivization (16), Inversion (17), Retired Term Marking (18), and all
other rules.
( 1 5) ipove satvale?
you-found-it-I1-3 glasses-NoM
'Have you found your glasses ?'
(16) cemi satvale sanagve qutsi napovni iqo.
my glasses-NoM trash can-in found it-was-I1-2
'My glasses were found in the trash can.'
(17) turme gipovnia satvale.
apparently you-found-it-I1I-3 glasses-NoM
'Apparently you have found your glasses.'
( I 8) satvalis povna
glasSeS-GEN finding
'Finding the glasses'
190 Appendix : Real and apparent exceptions
This class of verbs behaves like Class I verbs syntactically, but like Class 3
verbs :Plorphologically. As discussed in Holisky ( 198oa), these verbs are
intrinsically perfective in aspect and are characterized by the suffix -ulob in
Series 1.
Another small group of Class 3 verbs also take an obligatory initial direct
object. Examples are upasuxa 'he answered it to him' and uqvira 'he yelled
it to h im'. The examples below show that upasuxa behaves like mi(:era, and
that its initial objects are regular with respect to case marking and agree
ment ( 19), Passivization (20), Retired Term Marking (21) and (22), and
Inversion (23) and (24) ·
( 1 9) vupasuxe galt:vetili mas(:avlebels.
I - answered-him-it-II-3 lesson-NoM teacher-DAT
'1 answered the lesson to the teacher.'
In ch. 12 it was shown that some verbs, such as d{1fera 'write' and
daxrcoba 'drown' , have an obligatory initial direct object ; even when
the direct object is not overtly expressed, it must be posited as an initial
relation because of the meaning and because of the syntax of organic
causatives and the marking of retired terms. In this chapter it will be
shown that a subset of such verbs, while they have an initial direct
object obligatorily, take an initial subject optionally.
When verbs of this type lack an initial subject, their (obligatory)
initial direct object becomes the subject. This is illustrated in (Ib).
(I) (a) vano axrcobs rezos.
Vano-NOM he-drowns-him-I- I Rezo-DAT
'Vano is drowning Rezo.'
(b) rezo ixrcoba.
Rezo-NoM he-drowns-I-2
' Rezo is drowning.'
Verb forms like that in (I b) have traditionally been called 'passives' by
Georgianists. I have argued elsewhere (Harris, to appear b) that forms
of this type must be sharply distinguished from true passives, those of
the type discussed in ch. 7. However, because of the traditional name
and the similarities between the two, I will refer to forms like that in
( Ib) as 'synthetic passives' ; true passives, those discussed in ch. 7,
will henceforth be referred to as 'analytic passives' .
The purposes of this chapter are to distinguish between the direct
construction, the analytic passive, and the synthetic passive in Georgian
and to relate them to those notions in general linguistics.
grammatical relations that characterize them and the rules that derive
them. These are discussed in §3. In this section, the three construc
tions are illustrated and delimited.
I . 1 Examples
All three constructions are illustrated in (2-6). Additional examples of
direct forms, analytic passives, and synthetic passives are given in
ch. 16, Appendix A as group A of Class I , and groups D and E of Class 2,
respectively. In each example below, the direct (a) and (b) sentences
have undergone no rule that changes grammatical relations. In the (b)
sentences, the subject and direct object are coreferential; in the (a) sen
tences they are not. The (c) sentences are synthetic passives, and the
(d) sentences analytic passives.
(7) and (8) show two of the ways that the direct construction differs
syntactically from the analytic and synthetic passives. These differences
are due to the single fact that only in the direct construction is the final
subject an agent.
The sentences of (2-6c) are identical to those of (*9), except that the
former lack the mier-nominal. The grammaticality of (2-6c) shows that
the ungrammaticality of (*9) is caused by the presence of the mier
nominal.
The sentences of (2-6d) could, of course, occur without an expressed
mier-nominal ; in that case the nominal is unspecified (cf. ch. I , n. 14) .
The sentences cited above show that the analytic passive differs from
the synthetic passive in that the former may have a mier-nominal.
In this section I have pointed out some of the ways in which these
three constructions differ semantically and syntactically.
11\
daxre- 'vano rezo daxrc- 'vano rezo daxrl- rezo
verbs. According to ( 1 1), the analytic passive, but not the synthetic
passive, has such a nominal. This accounts for the fact that analytic
passives occur with a mier-nominal, as in (2�6d), while synthetic passives
do not, as shown in (*9).
Additional arguments to support the initial and final grammatical
relations proposed in ( 1 I) are presented in §4 and 5 , respectively.
4. 1 Suppletion
Shanidze observes that the verb 'kill' is suppletive for the number of
its direct object, such that mo/f-vla is used for the singular and daxoca
for the plural (Shanidze 1973 : 504). He cites the examples ( 1 2) ; the
glosses are added.
( I 2) (a) mgeli mov/f-ali.
wolf-NOM I-killed-it-II- I
'I killed the wolf.'
(b) mglebi davxoce.
wolveS-NOM I-killed-it-II- I 9
'I killed the wolves.'
From his exposition we can infer that this alternation is obligatory.
In the corresponding synthetic passive, the same suppletion 1S
governed by the nominative-nominal. Shanidze cites these examples :
(13) (a) sma momi/f-vda . 1 0
brother-NOM he-died-me-II-2
'My brother died (on me).'
(b) smebi damexoca.
brothers-NoM he-died-me-II-2
'My brothers died (on me).'
If the nominative-nominal of the synthetic passive is its initial direct
object, and if suppletion is stated on initial termhood as proposed in
200 Synthetic passives
We must add a rule to the grammar to account for the use of the preverb
da- with plural direct objects. The same rule will account for the use
of da- with plural subjects of intransitives if (i) final subjects (nomina
tive-nominals) of synthetic passives are initial direct objects, and (ii)
the da- rule is stated on initial termhood. If the nominative-nominal of
the synthetic passive were not its initial direct object, two syntactic
rules would have to be stated, missing the relevant generalization.
6 Conclusion an d extension
(I) ikna-Passive
saxli ikna seyebili �itlad.
house-NOM it-was-II-2 painted red
'The house got painted red.'
(2) Zustandspassiv
m/veri aparia iatakze.
dust-NOM it-spreads-on-I-2 floor-on
'Dust is spread on the floor.'
Table I
A B C D
Analytic Analytic Synthetic
(iqo-) Passive (ikna-) Passive Zustandspassiv Passive
Initial I & 2 Initial I & 2 Initial 2 Initial 2
Stative Non-stative Stative Non-stative
The analytic iqo-passive (discussed in ch. 7 and in this chapter) and the
analytic ikna-passive have the same initial inventories of terms ; both are
derived by Passivization.13 The Zustandspassiv and synthetic passive, on the
other hand, have initial direct objects, but no initial subjects ; they are derived
by Unaccusative.
The two analytic passives are alike formally, except in the auxiliary they
use. The A-type above uses the verb qopna 'be', while the B-type uses the
204 Appendix: On four 'passives'
verb 'become'. The latter is generally used only in the aorist. The ikna
passive has not been discussed at length here because it is used only in the
written language.
The iqo-passive differs from the ikna-passive in that the former is stative
and the latter non-stative. This is shown by the fact that the ikna-, but not
the iqo-passive, can co-occur with adverbs like sam saatSi 'in three hours',
which limit the duration of the action, as illustrated in (3).14
(3) (a) saxli sam saatli ikna Ieyebili.
house-NOM three hour-in it-was-II-z painted
'The house was painted in three hours.'
(b) *saxli sam saatSi iqo seyebili.
it-was-I I-z
('The house was painted in three hours.')
Note that saxli iqo seyebili 'the house was painted' is grammatical without
the limiting phrase sam saatIi.
The synthetic passive and the Zustandspassiv also differ from one another
in stativity, as stated above.
Finally, the analytic (iqo- and ikna-) passive is necessarily perfective. It
cannot be used to express the dynamic and necessarily imperfective passive of
English, such as the house is being painted. The participle, which constitutes
a part of the analytic passive, must always occur with a preverb, which is in
general an indication of perfectivity in Georgian. The synthetic passive, on
the other hand, may be either perfective or imperfective ; that is, it has forms
both with and without a preverb.
14 Rejlexivization
The pronoun tav- must be coreferential with the subject of its clause
(cf. ch. I). The Tav-Reflexivization rule is stated on initial terms ; that
is, the rule can be triggered only by initial subjects (cf. ch. 5, §z.z ;
ch. 7, §I. 1.3 ; and ch. 8, §3.3). But some facts have been left unaccounted
for. In object-raised sentences, there cannot be coreference between
the derived subject and the matrix tvis-nominal. (*la, b) show that
coreference is not possible between these two nominals if either is
reflexivized ; (*IC, d) show that co reference is also impossible if neither
nominal is reflexivized. (*IC, d) are grammatical on the non-coreferent
reading.
(I ) (a) *( tavisi) tavi ;;nelia gelastvis gasatavisupleblad.
self's self-NOM hard-it-is-I-z Gela-for to-free
cf. *'Himself is hard for Gela to set free.'
(b) *gela ;;nelia tavistvis gasatavisupleblad.
Gela-NoM self-for
cf. *'Gela is hard for himself to set free.'
(c) *is1 ;;nelia gelastvisi gasatavisupleblad.
he-NOM
cf. *'Het is hard for Gelat to set free.'
(d) * Gela t ;;nelia mistvis! gasatavisupleblad.
him-for
cf. ·'Gelal is hard for him! to set free.'
206 Rejlexivization
The grammaticality of (2), where the derived subject and the matrix
tvi>-nominal are not coreferent, shows that the ungrammaticality of (*1)
must be due to coreferentiality.
But the active sentence (4) is fully grammatical, even though there is
coreference.
In (""'6) the initial subject is coreferential with tav-. Yet in the corre
sponding direct construction, (7), the initial subject can trigger Tav
Reflexivization.
Thus, the ungrammaticality of (""'9) must be due to the fact that the
derived subject triggers Tav-Reflexivization.
In ch. 6 we saw that either the initial embedded subject or the initial
matrix subject can trigger tav- in an organic causative. This accounts for
the ambiguity of ( I I), where vano and ekimi 'doctor' are the initial
embedded subject (= final direct object) and initial matrix subject,
respectively.
208 Reftexivization
2 A proposal
I propose the following restatement of Tav-Reflexivization to account
for all the facts enumerated above :
(14) The antecedent of tav- must be
a. the initial subject of a clause of which tav- is a dependent,
and
b. a final term.
I4.2 A proposal 209
I will show below how ( 1 4-) correctly predicts the facts of Tav
Reflexivization.
In object-raised sentences, the initial embedded subject is deleted
on coreference to the matrix tvis-nominal (cf. ch. 4- and ch. I I , §4-.I ).
Since it is deleted, it is not a final term, and therefore cannot be an
antecedent of tav-.2 The derived subject cannot trigger tav-, as shown
in (""Ib), since it is not an initial subject, which (14-) requires an ante
cedent to be.
(14-) also correctly predicts the ungrammaticality of (""2a), and (""9)
where the derived subject of a passive triggers tav-. The initial subject
of a passive cannot be the antecedent, as shown in (""2b) and (""6), since
it is not a final term.
( 14-) permits both readings of (I I). The pronoun tav- is an initial
dependent of the embedded clause and final dependent of the matrix
clause. Both of the nominals ekimi 'doctor' and vanQ satisfy the trigger
conditions in ( 1 4-), since both are initial subjects of clauses of which
tav- is a dependent, and both are final terms subject and direct object,
respectively.
At the same time, the rule also accounts for the Inversion data. In
(I2a) and ( 1 3a) the dative-nominal is the initial subject and a final
term (indirect object). In (""ub, c) and ( "" 1 3b , c) the nominative
nominals, gela and vano, cannot trigger tav-, since they are derived
subjects, and are not initial subjects.
I have not yet accounted for the ungrammaticality of (""IC, d) and
(""3C, d). We can see from the grammaticality of (4-) that there is no
initial-structure constraint blocking (""3). The grammaticality of ( 1 5)
shows that (""IC, d) also break no constraint on initial structures ; ( I S)
is derived from the same initial structure as (""I) (cf. ch. 1 0).
(IS) tavis tavisupleba 3ne1ia gelastvis.
Self-GEN freeing-NoM hard-it-is-I-2 Gela-for
'It is hard for Gela to set himself free.'
We can therefore not account for the ungrammaticality of (""IC, d) and
(""3c, d) in terms of an initial-structure constraint. We can account for
them correctly with a convention, ( 1 6).
(16) Coreference involving an initial subject that is a clausemate of the
nominal coreferential to it must be resolved by reflexivization. 3
Examples (""IC, d ) and (""3C, d ) break this convention. The grammatical
sentences above, (4-), (7), (I I), \ I 2a), and ( 13a), do not violate ( 1 6).
210 Reflexivization
3 An alternative proposal
Postal ( 1971 ) proposes to account for sentences in English similar to
the ones we have been discussing in Georgian with a Crossover Con
straint. He proposes that coreferent nominals cannot cross one another
in a derivation. 4
A crossover approach makes sense only with respect to ordered
strings. If we adopt a linear order analysis, instead of the relational
analysis proposed in this work, the Crossover Constraint will correctly
predict the ungrammaticality of the object-raised sentences (""l a, b)
and the passive sentences (""3a, b). But it could not account for the
grammaticality of the inversion sentences (I 2a) and ( 13a). In order to
state that under Inversion the initial subject becomes the final indirect
object and that the direct object becomes the final subject, any gram
mar in which rules operate on structures in which constituents are
linearly ordered must have a rule in which these two nominals cross one
another in the course of derivation. The Crossover Constraint would
then predict incorrectly that ( I 2a) and ( 13a) are ungrammatical.
In order to save Crossover, we might propose to account for inversion
sentences without a crossing rule. But it would greatly complicate the
statement of the rules of Case Marking, Subject Person Agreement,
Indirect Object Person Agreement, and Object Camouflage in such a
grammar not to have these nominals cross in Inversion.
I conclude that a Crossover approach is inadequate for Georgian
because it misses the generalizations captured simply in (14) and (16).
15 Number Agreement
In this chapter I will consider the rule of Number Agreement and the
conditions under which it applies.1 It will be shown that certain facts
can be accounted for best with a rule sensitive to both initial and final
grammatical relations. The primary purpose of the chapter is to provide
an example of the relevance of initial grammatical relations to agree
ment. Until recently it has generally been assumed that all agreement
rules are stated on final grammatical relations. This chapter contributes
an example of a process in which initial grammatical relations are one
of several determining factors in agreement.
The approach will be this : I will give data and make a generalization
that describes it, then give more data and revise the generalization to
account for the additional facts. Several increments of this sort will be
necessary. It is first established that any term grammatical relation can
trigger Number Agreement in Modern Georgian. Then it is shown
that third person nominals trigger Number Agreement only if they are
subjects. Third, it is final termhood that is relevant to the trigger con
dition among third person nominals. Fourth, it is shown that among
final third person terms, that nominal which is the first subject triggers
Number Agreement. This correctly excludes initial subjects that are
not final terms, includes derived subjects of passivization, and includes
the initial subject of the inversion construction. As a last refinement, it
is shown that this condition holds unless the third person nominal is
outranked (on the relational hierarchy) by a first or second person
nominal in final structure. It will be shown that this rule accounts for
a wide variety of constructions in Modern Georgian. In �7 some
alternative analyses will be considered and shown to be inadequate.
I Polypersonalism
( I I) (a) m-icnob?
you-knovv-me-I-l
'Do you knovv me ?'
(b) gv-icnob?
you-knovv-us-I- I
'Do you knovv us ?'
In the examples above, the sentences have undergone no rule that vvould
change the grammatical relations in them. When vve look at analytic
or synthetic passives or at object-raised sentences, vve find that (7)
must be stated vvith respect to final termhood.
Analytic passives are derived by Passivization. vvhich promotes
direct objects to subjecthood ; the nominal vvhich bore the subject
relation becomes a chomeur, according to the Chomeur Condition (cf.
chs. 7 and 13). In ( 12). the verb agrees vvith the derived subject, not
vvith the initial subject.
The verb agrees in number vvith its third person final subject. The
derived subject vvas the initial direct object and, being third person,
could not trigger Number Agreement according to (7) ; only in its
derived capacity as subject could the third person nominal trigger
Number Agreement according to (7). Further, in ( 12C), the initial sub
ject, policieleb- 'policemen' , fails to trigger Number Agreement ; only
216 Number Agreement
With (7') in mind, consider the sentences of ( I S), which are in the
evidential mode.
IS4 First SIlbjects 217
The sentences of (17) show the need for a further condition on the state
ment of Number Agreement. ( 17) is analogous to ( I S). In ( I S) both
nominals are third person ; in ( I7) the initial direct object ( = final
subject) is second person.
(17) differs from (IS ) only in the person of the final subject initial
,
direct object) and in Number Agreement. In ( 17) student(eb)s 'student(s)
does not trigger Number Agreement as it does in ( I S). In (17) the final
subject does trigger Number Agreement, while in ( I S) it did not. In
fact, if there is a first or second person term in a clause, a third person
ISo5 Relational hierarchy 219
It was shown above that the rule proposed here as ( IS) accounts for a
wide variety of constructions, including simple sentences (clauses in
which no change of grammatical relations takes place), analytic and
synthetic passives, object-raised sentences, and inversion constructions.
In this section, I will discuss some additional constructions - causatives
and version - which completes the list of major syntactic constructions
of Modern Georgian. In each instance I will show that ( I S) correctly
accounts for Number Agreement. I conclude that ( I S) accounts for
Number Agreement generally in Modern Georgian.
6.1 Causatives
If an embedded clause is intransitive, its subject becomes the direct
object of the corresponding causative ; (19) illustrates this type.
(19) (a) ekimma gv-alap ara�-a (even).
doctor-ERG he-caused-talk-us-II-I we-NOM
'The doctor got us to talk.'
(b) ekimma alapara�-a bi{ebi.
he-caused-talk-him-II-I bOYS-NOM
'The doctor got the boys to talk.'
The initial grammatical relations in (19) are established in ch. 5.
(i) Ekimma 'doctor' is the initial subject of the matrix clause. (ii) The
predicate of this clause is 'cause'. (iii) Its direct object is the clause con
taining the verb lap ara�- 'talk'. (iv) even is the initial subject of that
clause in ( I9a), and bi?ebi 'boys' is in ( I9b).
According to (ISa), the first person final direct object in (I9a) should
trigger Number Agreement. It does, and this is reflected in the first
person plural direct object marker gv-. According to ( ISb), ekimma
' doctor' in ( I9a, b) should trigger Number Agreement, since it is the
final subject of the causative and is (i) the first subject of that verb
that is a final term, and (ii) not outranked finally by any nominal. The
number of this nominal is reflected in the -a suffix, which is a marker
of third person singular subjects. Bi?ebi 'boys' in (I9b) does not satisfy
criterion (I 8b-i) and is therefore not a Number Agreement trigger. It
does not satisfy (I8b-i) because, although it is a final term of the
causative, it is never its subject. Although it is the initial subject of
lap ara�- 'talk', it is not a final term of that verb. As predicted by ( I 8b),
the number of bi{ebi is not reflected in the verb form.
I5.6 Applicability of rule (r8) 221
6.2 Version
Benefactive Version promotes a benefactive-nominal to indirect-object
hood. Number Agreement in this construction is illustrated in (21).
(21) (a) jadosanma (tkven) ga-g-i'!J.et-a-t jadokruli
witch-ERG yOU(PL)-DAT she-made-you(PL)-it-II-I magic
sar!J.e?
mirror-NoM
'Did the witch make a magic mirror for yOU(PL) ?'
222 Number Agreement
7. 1 A case-based analysis
It might be proposed that the determination of the triggers of Number
Agreement should be stated, not on grammatical relations, but on cases.
Because of the peculiarly complex system of case marking in Modern
Georgian, a solution of this kind would be particularly complicated.
Stating Number Agreement on cases would entail incorporating the
case marking differential « 1 '-2') of ch. 9) into the Number Agreement
rules. It would require making at least the statements of (2 3 ) .
(2 3) (a) First or second person nominative-, dative-, or ergative
nominals trigger Number Agreement.
(b) Third person nominative-nominals trigger Number Agree
ment in Series I, verbs of Class I , 2, or 3 .
(c) Third person nominative-nominals trigger Number Agree
ment in Series II or III, verbs of Class 2.
(d) Third person ergative-nominals trigger Number Agreement
in Series II, verbs of Class I or 3 .
(e) Third person dative-nominals trigger Number Agreement in
Class 4 verbs in all Series and in Class I or 3 verbs in Series
III, just in case the nominative-nominal is also third person.
This analysis fails to draw any relevant generalizations. (23 ) appears
to be a random set of unrelated facts, while ( 1 8) presents general prin
ciples for determining Number Agreement triggers. Thus, ( 1 8),
although complex, is more general than a case-based analysis.
compare this hypothesis with that proposed in this chapter we must con
sider some additional aspects of Georgian morphology. With the con
vention that object agreement suffixes attach to the right of subject
agreement suffixes,10 the relevant morphophonemic rules may be
stated as (24) and (25).
The morpheme t in both rules is a plural marker (cf. paradigms ( 1-3 /3'» .
Rule (25) accounts for the non-occurrence of the following sequences,
where 'V' represents the verb stem : *V-en-t, ""'V-an-t, *V-es-t, ""'V-nen-t,
and ""'V-t-t. None of these sequences is ruled out by a phonological
rule.
The morphology-based analysis would require the following state
ments to account for simple sentences : (1-2), (3') to replace (3) of the
proposed grammar, and the morphophonemic rules (24-25).
(27) The third person plural subject marker is not -en, -es, etc., but
rather -a or -s, just in case the clause has undergone Inversion and
the final object is third person.
Rules (26) and (27) are ad hoc, and this analysis fails to show that
Number Agreement is based in part on non-final grammatical relations.
The hypothesis proposed in this chapter, on the other hand, needs
statements ( 1-3), (24-25), and (18). While ( 1 8) is complex, it draws
linguistically significant generalizations and avoids ad hoc rules.
I5.7 Alternative analyses 225
8 Conclusions
1.2 What are the implications of the Ergative Hypothesis for Georgian?
We have established the existence of two case marking Patterns in Mod
ern Georgian, as shown in ( I a). A large amount of evidence has been
amassed to support the correlations between the cases and the gram
matical relations stated in ( I a) ; this evidence is presented particularly
in chs. 2-9 and 1 3-16. In particular, these data show that nominals
marked with the ergative in Pattern A and with the nominative in
Pattern B are subjects, while those marked with the nominative in
Pattern A and with the dative in Pattern B are direct objects.
The ergative analysis of Georgian Series II identifies the 'ergative'
marked subject with transitive verbs and the 'nominative' -marked sub.
ject with intransitive verbs. According to the Ergative Hypothesis, in
Series II, transitives govern Pattern A and intransitives govern Pattern
B. We have seen the following distribution of Patterns A and B with
respect to different Classes in Series II.
(2) Distribution of Patterns in Series II
Class 1 Pattern A
Class 2 Pattern B
Class 3 Pattern A
Class 4 Pattern B
Grammarians writing within the framework of the Ergative Hypothesis
have made the following statements about these verb Classes :
(3) Class 1 is transitive and governs Pattern A.
Class 2 is intransitive and governs Pattern B .
Class 3 is neither transitive nor intransitive and governs Pattern A.l
I6.I The Ergative Hypothesis 231
which defines an ergative case marking system. For example, there are
a great many final transitive-intransitive pairs, like those listed in (4).
All of the transitive verbs in (4) govern Pattern A in Series II, and all
of the intransitive ones Pattern B. In addition to pairs like these, verb
forms like mivida 'he went', movida 'he came', iqo 'he was', and darla
'he stayed, he remained', are intransitive and govern Pattern B. Further,
as established in ch. 8, inverted verbs are finally intransitive, with Pat
tern B marking. Thus, the Ergative Hypothesis makes the correct
predictions for a large number of verbs in Georgian.
(i) The first such explanation is embodied in (3), namely that Class 3
verbs are neither transitive nor intransitive. In the sense defined in
ch. 1 2 for these notions, this claim has no apparent meaning. In the
work of linguists who make this claim about Georgian, the terms
'transitive' and 'intransitive' have lost the meanings that they have in
general linguistics and have become the names of particular verb
Classes, regardless of the syntactic properties of the verbs. Thus, 'tran
sitive' has become the name of Class I , and 'intransitive' the name of
Class 2. This nomenclature has led to calling Class 3 verbs 'neutral'
and stating that they are neither transitive nor intransitive. Clearly this
statement is meaningless if the notions 'transitive' and 'intransitive'
are also to be used with reference to languages other than Georgian.
(ii) A second approach is to play down the importance of this large
class of counterexamples by labeling it 'irregular'. Indeed, the Class has
even been defined as the set of irregular verbs (cf. Holisky 1980a for a
history of the analyses of this Class). The implication seems to be that
if a group of verbs is morphologically irregular, it cannot be expected
to be regular from the point of view of the case Pattern it governs. The
label 'irregular' seems to have been applied on the grounds that verbs
of Class 3 build their future/aorist fonns2 with a formant different from
that used by verbs of Classes I and 2. This Class is, however, regular
in the sense that its members regularly use one particular formant to
build the tenses in question. (5) lists the formants regularly used to
form these tenses by verbs of Class I and 2 on the one hand, and of
Class 3 on the other.
(5) Classes I, 2 Class 3
Future Preverb- i-eb
Conditional Preverb-(0)d-i i-eb-(0)d-i
Future Subjunctive Preverb-(o)d-e i-eb-(0)d-e3
Aorist Preverb-e/i i-e
Optative Preverb-ofa t--o
In (5), dashes indicate the position of the verb root. The preverbs used
by Classes I and 2 are chosen from a set which includes mi-, mo-, a-,
amo-, ca-, lamo-, da-, ga-, gamo-, fa-, famo-, etc. and are lexically
determined. In the aorist and optative, respectively, most of the Class I
and 2 verb forms use the -e and -0 suffixes ; some, however, use -i and -a,
respectively. All of the Class 3 verbs on the other hand, use -e and -0
regularly. In order to illustrate the striking regularity of this Class, I
I6.I The Ergative Hypothesis 233
have listed the verbs in the future in Appendix A to this chapter. While
there exist irregularities in this Class as in every other, the overwhelm
ing majority of verbs in Class 3 show a consistent morphology. These
verbs have 'irregular' morphology, then, only in the sense that their
morphology is different from that of Class I and 2 verbs. 4-
The 'irregularity' of this Class has also been understood to mean
that it is arbitrary, that is, that the verbs which constitute the Class are
an arbitrary set. S Holisky (I98oa) shows that Class 3 is semantically
regular, as well, constituting the Class of dynamic verbs that are
inherently of durative aspect (cf. n. 4). Even a glance at the representa
tive list of Class 3 verbs in Appendix A will reveal that the verbs which
constitute Class 3 are not an arbitrary set.
In earlier chapters we have seen that Class 3 verbs exhibit a high
degree of syntactic regularity. In particular, all verbs which satisfy the
morphological definition of Class 3 (cf. Appendix A), also have the fol
lowing syntactic characteristics : (a) they trigger Inversion in Series III,
and (b) they govern ergative case subjects (Pattern A) in Series II.
This Class of verbs is strikingly regular, then, in its morphology,
syntax, and semantics. Labeling these verbs 'irregular' in no way
accounts for the case Pattern they govern in Series II. Labeling them
in this way simply disguises the fact that linguists have not been able
to account for the case Patterns governed by a large and systematic
Class of verbs in Georgian.
(iii) A third explanation that has been offered for the case marking
of Class 3 verbs in Series II is that these verbs ' do not have their own
Series II forms' and that the Series II forms used are 'borrowed from
the corresponding' Class I verbs. (This entails that the future forms
are also 'borrowed', since futures and aorists are similar morpho
logically for all regular verbs in Georgian ; cf. (5) above.) Thus, for
example, the aorist of vbatonob 'I reign' vibatone 'I reigned' - is said
to be 'borrowed' from the Class I transitive verb : gavibatone cemi tavi
'I made myself ruler'. Since the form is borrowed, it is implied, so is
the case Pattern. Thus, an intransitive verb, vbatonob ' I reign', has an
ergative case subject in Series II because the morphology of the verb
is 'borrowed' from a transitive, which naturally governs an ergative
subject.6 Thus, by claiming that these verbs have 'no S eries II forms
of their own' , linguists have tried to explain the fact that these intransi
tive verbs govern an ergative case subject in an ergative system of case
marking.
234 The nature of the Georgian verb classes
There are at least two problems with such an explanation. The first
is that this theory does not explain why syntactic characteristics sho,uld
be 'borrowed' along with morphological ones. In a language with an
ergative case system, case marking is determined by the transitivity of
the verbs, rather than by the origin of their morphology. If a language
has ergative case marking, it will apply the principles of that system to
all verbs alike, whether or not they are derived or borrowed. Until the
borrowing of morphological formants can be shown to be related sys
tematically to the borrowing of syntactic characteristics, this explanation
for the use of the ergative case by intransitive verbs is completely ad hoc.
A second fault of such a hypothesis is that it offers no explanation of
why some verbs (Class 3) should 'borrow' their morphology while
others (Class I , 2) do not. Even if we accept the notion that 'borrowed'
morphology entails 'borrowed' syntax, we still have no answer as to
why Class 3 is different from Class 2.7
To summarize, neither the notion 'neither transitive nor intransitive',
the label of 'irregular', nor the concept that Series II forms of Class 3
verbs are 'borrowed' offers a satisfactory explanation of the systematic
use of Pattern A by many intransitive verbs in Series II. We must
recognize that there is no way adequately to incorporate Class 3 verbs
into the ergative analysis of Georgian.
There are other problems for the claims made in (3). Intransitive
verbs in Class I include daaboqina 'he burped' , daamtknara 'he yawned',
and daacxi�va 'he sneezed'. The transitive verbs in Class 2 are dis
cussed in Appendix B to this chapter. While the existence of intransi
tives in Class I and transitives in Class 2 is contrary to the claims of (3),
it does not pose a serious problem for the Ergative Hypothesis ; these
verbs are few in number and may be treated as simple exceptions. The
existence of the large, systematic Class 3, however, shows the real
inadequacy of the ergative analysis for Georgian.
At the beginning of this chapter I stated a problem for general
linguistics and Georgian grammar, as it has traditionally been viewed
by Georgian specialists. Looking at the distribution of cases in (2), we
see that it is appropriate to pose this question in a different way as well :
in Series II ?
2. How can we predict which verbs govern Pattern A in Series II ?
I6.2 The Unaccusative Hypothesis 235
2. ILanguage universals
Sapir ( 1 9 17) was perhaps the first to make a clear typological division
of languages on the basis of case marking systems. He observed the
existence of five distinct types. Of particular interest to us here is the
clear distinction he made between ergative (transitive vs. intransitive)
and active vs. inactive systems, deploring the usual inclusion of these
two types under one rubric. He represents these two case systems
schematically as in (8), which also includes his representation of an
accusative language for comparison ; the names given to the types of
case systems are not used by Sapir (op. cit.). 8
(8) Object Subject Subject Type of
transitive intransitive transitive case system
Inactive Active
A A B Ergative
A A B B Active/Inactive
A B B Accusative
Although Sapir does not go into detail about the active/inactive type,
we may infer something about its nature from Dakhota, the language
he cites as its archetype. It has often been said that Dakhota and the
other Siouan languages distinguish between stative verbs and dynamic
verbs. In Lakhota, a Dakhota dialect, sentences like those below suggest
that wa marks first person singular in dynamic verbs, while ma marks
the same person and number in statives.
wakte' 'I kill it'.
thawa'sose 'I spit.'
mahli'ske 'I am tall'.
However, examples like
mahi'xpaye 'I fall down.' (Examples from Van Valin ( 1977»
show that ma is used not only by statives but also by intransitive
dynamic verbs that are non-agentive, involuntary, or non-controllable.
236 The nature of the Georgian verb classes
2.2. I Class I . Class I contains two derivational types : basic verb forms
(A)lO and organic causatives (B).l1 Dafers 'he will write it' is a basic
238 The natur e of the Georgian verb classes
Class 1 verb (direct construction) with initial subject and direct object,
as established in ch. 13. The structure of this form is represented in
network (10).
(10)
dar;era dar;ers.
' write ' ' he ' , it ' ' He will write it. '
This Class includes a few verbs which take an initial intransitive sub
ject ; one is represented in ( I I ) .
(I I)
gakusvla gakusla.
' gallop away ' ' he ' ' He galloped away. '
Daaferina 'he caused him to write it' is the causative of a transitive ; its
derivation is represented in ( 1 2) (cf. ch. 5).
( 12)
daar;erina.
dar;era ' He made him
' write ' ' he ' , it ' write it. '
As observed in ch. 12, most Class 1 verbs are transitive. All Class 1
verbs are characterized by an initial subject, as illustrated in networks
( 1 0--12).
2.2.2 Class 3. Class 3 also contains two derivational types : basic verb
forms (M) and denominals (N) . Tamasobs 'he plays' is an example of the
former. As established in ch. 12, it takes an optional initial direct
object ; ( 13) and (14) give the two constructions possible for this verb
form.
I6.2 The Unaccusative Hypothesis 239
tamaH itamasa.
' he ' ' He played. '
(16)
daxrcoba daixrco.
' drown' ' he ' ' He drowned. '
tetri tetrdeba.
' white ' , it ' ' It becomes white.'
tirili a#rdeba.
, '
cry ' he ' ' He begins to cry. '
B (R) . Basic Class 4 verbs are mostly initial transitives, but there are a
few intransitives. These are represented in (20) and (21), respectively.
(20)
qvareba miqvarxar.
' ' you ' 'I love you. '
, love 'I'
�
H
civeba mciva .
' feel cold ' 'I ' 'I feel cold. '
Generalization 1
In Series II, Class I and 3 verbs are those verbs which have an
initial subject that is also their final subject.
intransitive verbs of Classes 1 and 3, we can see that their initial sub
jects are also their final subjects. Class 2 verbs, on the other hand,
either have no initial subject (networks ( 16-19» , or their initial subject
is a final chomeur (passive network ( I S» . Verbs of Class 4 have an
initial subject ; but having obligatorily undergone Inversion, their
initial subject is not a final subject, but a final indirect object (networks
(20-21» . A solution to Problem 2 will be given in the following section.
Most of the derivations discussed here have been justified in detail
in earlier chapters. The only new claims presented here are (a) initial
termhood governed by basic Class 2 verbs (network (1 7» , (b) the
derivation of the inceptive (networks (18- 1 9» , and (c) the generaliz
ations concerning the nature of each Class. The first two of these new
claims are supported by the additional evidence which is presented in
§3 and 4. The generalizations made in this section follow logically from
all of the derivations which have been individually justified.
lIb correctly marks the nominals listed above and prevents retired
direct objects from being marked with the nominative case.
In Series II the dative case marks indirect objects that have undergone
no change, as well as indirect objects derived by Version or Inversion :
II If the verb is in Series II,
c. a final indirect object is marked with the dative case.
Unlike rules IIa and lIb, rule IIC need not refer to initial grammatical
relations. Yet because of the rule inventory of Georgian, there is no
overlap between the rules as stated. These five term case marking rules
account for cases for all of the constructions of Georgian syntax.
These five rules could be written in a variety of other ways, some of
them as succinct as those given here. 1 6 But the generalizations embodied
in them must be preserved by any valid grammar of Georgian.
The five rules stated above do not refer to verb Classes as ( I b) does.
This means that with respect to the rules of case marking, it is unnecessary
to list Classes for each verb form in the lexicon. Rather, we can predict
case marking on the basis of rules I and II, together with the semantics
of the verb. Thus, rules I and II effect an important simplification in
the grammar, in terms of not requiring that Class be listed in the
lexicon.
sense of §2. 1 . Her data are an exhaustive list of the hundreds of Class 3
verbs, which she culled from standard dictionaries and tested with
informants. Similarly, the semantically active verbs of Class 2 occur in
closed subclasses ; these include subclasses J and K (discussed in Appen
dix B to this chapter) and subclass L (discussed briefly in Harris (to
appear b) and Harris 1976). The productive subclasses of Class 2 -
inceptives, analytic passives, synthetic passives, object-raised con
structions are, without exception, inactive in the sense of §2. I .
A second important point about these exceptions is that some cause
confusion among native speakers. (22) gives two variants for the verb
�etva 'drip' an inactive verb formally in Class 3 . In the (a) sentence,
Pattern A is used, in the (b) sentence, Pattern B.
(22) (a) macivridan fqalma i�eta.
refrigerator-from water-ERG it-dripped-II-3
'Water dripped from the refrigerator.'
(b) macivridan fqali i�eta.
water-NOM
'Water dripped from the refrigerator.'
Both Patterns are used for this verb, though (a) is considered more
literary and is preferred by prescriptive normsY Case marking, com
plex as it is in Georgian, does not generally cause any difficulty for the
native speaker. The fact that it does cause a problem with this verb
confirms its status as an exception, rather than a counterexample.
Examples of confusion of case marking with active verbs formally in
Class 2 are discussed in more detail in Appendix B.
Finally, although these frozen forms are retained as exceptions in the
literary language of the metropolis, many or all of them are regularized
in a number of dialects, including Gurian and Imerian (western
Georgia), the mountain dialects (north central Georgia), and Pereidnian
(Iran). 18 The examples most frequently cited in the literature on these
dialects are of the type �acma favida 'the man-ERG went' and �acma
adga 'the man-ERG got up ' , where the subjects are in the ergative, as
would be expected for active verbs ; in the dialect of Tbilisi, these and
several other very common verbs are irregular, taking Pattern B in
Series II. It is well known in linguistics that a literary dialect tends to
preserve irregularities, particularly with very common words.
I6.3 Inversion 247
In earlier chapters it was noted that the Class of the governing verb
determines two distinct syntactic phenomena case marking in Series
II and Inversion in Series III. In the preceding section it was shown
that the rules of case marking do not need to refer to the notion of
' Class'. Rather, case marking can be predicted on the basis of the initial
and final termhood of nominals, which is, in turn, determined by the
semantics of the verb. If it were possible also to predict which verbs
undergo Inversion in Series III without reference to Classes, then the
notion 'Class' would not be needed in the syntax. In this section, I shall
show that this is possible.
In ch. 8 it was shown that Inversion applies under two conditions :
(23) Inversion is triggered by
(a) Class 4 (inversion) verbs
(b) Series III (evidential) forms.
(23b) must be further restricted by (24).
(24) Class I and 3 verbs trigger Inversion,
Class 2 does not.
Chapter 8, §7 describes five environments in which Inversion does not
apply in a clause with some syntactic rule. Whether triggered by Class 4
or Series III forms, Inversion fails to apply to the output of Passiviz
ation, Object Raising, or Inversion and Unaccusative. Each of the
impossible derivations discussed there involves the application of Inver
sion to a derived subject. If clauses of the types represented by networks
( 1 6-19) of the present chapter are put into Series III, they too fail to
undergo Inversion. In each instance, Inversion fails to apply to a non
initial subject, just as was the case with the examples discussed in ch. 8.
In fact, Inversion never applies to a non-initial subject ; further, every
clause to which Inversion fails to apply (if triggered by (23» contains a
non-initial subject. I therefore propose to replace the ad hoc statement
(24) with
(24') The Initial Subject Constraint on Inversion
Inversion applies only to initial subjects. 1 9
In the framework proposed here, (24') prevents Inversion from applying
to analytic passives (network ( I S)), synthetic passives (network ( 1 6» ,
basic Class 2 verbs (network ( 1 7» , inceptives of both sorts ( 1 8-19),
248 The nature of the Georgian verb classes
Thus, on the analysis proposed here for Class 2 verbs (crucially involv
ing the notion of initial intransitive direct objects), the correct pre
dictions are made for Series III by the rules of Inversion, U naccusative,
Passivization, (23) and the Initial Subject Constraint, all of which are
independently necessary.
The proposal made in §2 enables us to effect three simplifications in
the grammar : (i) the rules of case marking can be stated in a general
way, (ii) a general statement can be made of which verbs govern
Inversion in the evidential, and (iii) because neither case marking rules
nor Inversion need refer to verb Class, the lexicon can be simplified. 20
In addition, the Initial Subject Constraint on Inversion, in the gram
mar proposed here, explains the otherwise arbitrary fact that Class 2
verbs do not undergo Inversion in Series III. These economies and this
explanation can be made only with the analysis presented in §2.
In this section I will discuss two productive pairs of verb forms, where
one member of the pair belongs to Class 2 and the other to Class 3 . The
purpose of this section is to show that in each instance the Unaccusative
Hypothesis makes the correct predictions concerning Class membership.
Or, differently stated, this hypothesis makes the correct predictions
250 The nature of the Georgian verb classes
concerning case marking and Inversion. Other productive formations
have already been discussed in detail elsewhere, including the analytic
and synthetic passives. The two types discussed here particularly stand
out because of the fact that in each pair both members are intransitive
(finally and initially). This means that the Ergative Hypothesis will
make the wrong predictions concerning the case marking and Inversion
for one member of each pair. The fact that these formations are pro
ductive emphasizes the inadequacy of treating one member ( Class 3
form) of each pair as an exception, as the Ergative Hypothesis does.
Holisky ( 1978) points out the difference between pairs like tamarma
imepa 'Tamar reigned' and tamari mepe iqo 'Tamar was (a/the) monarch'.
Both predicates are based on the root mepe 'monarch'. The first is
deliberate, active, controlled by the subject ; the second is stative, non
deliberate, and out of the control of the (final) subject. On the basis of
the semantics of the verb, we know that the first takes an initial subject ;
the second takes an initial direct object, as qopna 'be' always does. The
final subject of the first is its initial subject, and is correctly marked
ergative by rule IIa. The final subject of the second is its initial direct
object, and is correctly marked nominative by lIb. Derivations are
represented in (28) and (29).
(28)
tamarma imepa.
mepoba tamar tamar-ERG she-reigned
' reign' ' Tamar reigned .'
The initial subject in (30a), but not the derived subject in (30b), under
goes Inversion.
Thus, the Unaccusative Hypothesis makes the correct predictions
concerning the case marking and Inversion of each member of this
type of pair, without reference to Class, but rather on the basis of
initial grammatical relations, which are predictable on the basis of the
semantics of the verb. Since both predicates are intransitive, the
Ergative Hypothesis cannot distinguish between them, except by treat
ing them as exceptions and referring to a lexical listing of Class. Without
referring to Class, the Ergative Hypothesis would incorrectly predict
that both verb types govern nominative case subjects in Series II and
fail to undergo Inversion in Series III.
A second productive type is illustrated in (3 1 ) and (32). (31) is for
mally an inceptive ; it is inactive, and non-controllable.
of the two will reveal that the Unaccusative Hypothesis is more general,
and therefore superior. In §5 . 2 I will briefly consider and dismiss the
proposal that Georgian rather exhibits a low 'degree of ergativity'.
5 . 1 Hypotheses A and B
5 . 1 . 1 Case Marking. Hypothesis A ( The Ergative Hypothesis). According
to the Ergative Hypothesis, in Series II the case marking of nominals
governed by verbs of Classes I and 2 is predictable on the basis of the
transitivity of the verb. The grammar must, therefore, include the
following rules of case marking.
n If the verb is in Series II.
a'; the subjects of transitive verbs are marked with the ergative
case.
n If the verb is in Series II,
b'. the subjects of intransitive verbs and the direct objects of
transitive verbs are marked with the nominative case.
In addition, it will include rules Ia and b and rule nc, all stated above
in S2.3. These rules ( na ' , b') are inaccurate for the regular systematic
group of Class 3 verbs, which are mostly intransitive, but do not fit the
ergative generalization. For them, we must include another rule,
n If the verb is a Class 3 verb in Series II,
d. the subject is marked with the ergative case and the direct
object with the nominative.
Few Class 3 verbs have direct objects, but the last clause must be
included for those that do.
Hypothesis B (The Unaccusative Hypothesis). According to the pro
posal introduced here, in Series II, the case marking of nominals
governed by verbs of all Classes is predictable on the basis of initial
and final termhood. The transitivity of the verb and the secondary
grammatical relations, ergative and absolutive, play no part in term
case marking. The grammar must include rules I and II (five rules).
Conclusion. When we compare the statements (rules) which must be
included in each grammar, we find that Hypothesis B is significantly
more general. Hypothesis A is unable to include Class 3 in the general
ization that covers the rest of the grammar ; this is reflected in the
inclusion of rule lId, which is unnecessary in Hypothesis B. There are
254 The nature of the Georgian verb classes
5. 1.4 Reference to Classes. Hypothesis A. Since rules lId and (35) refer
specifically to Class 3 , each verb of this Class must be indicated in the
lexicon. I n addition, exceptions must be indicated.
Hypothesis B. There are also exceptions to this theory and they must
be listed in the lexicon. Class 3 is not referred to as such by any rule,
and need not be indicated in the lexicon. Indeed, there is no need to
make any special indication of the syntax of regular verbs discussed
above.
'ergative language' and use notions that can be precisely defined, such
as 'ergative rule'.
Second, Georgian has, to the best of my knowledge, only two rules
that might be considered ergative. Retired Term Marking is truly
ergative ; retired ergatives are marked one way (mier) and retired
absolutives another �genitive case). Causative Clause Union is also
ergative, in the sense that it treats ergatives one way, making them
matrix indirect objects, and absolutives another way, making them
matrix direct objects. 23 Interestingly, this type of causative occurs in
languages that apparently have no other ergative rule, such as French,
Turkish, and Japanese (cf. ch. 5, § 1). As shown above, neither the term
case marking rules, nor the agreement rules have anything to do with
the secondary grammatical relations ergative and absolutive. 24
Term case marking in Georgian might be mistaken for a 'partially
ergative' rule because of the fact that marking in an active/inactive
system differs from that in an ergative system only with respect to
initial intransitive subjects. On the other hand, marking in an active/
inactive system differs from that in an accusative system only with
respect to initial intransitive direct objects.
Series II of Georgian belongs to a case marking type that was recog
nized at least as early as 1 917. Sapir deplored the 'inclusion under one
rubric of transitive versus intransitive, and active versus inactive' (Sapir
1 9 1 7 : 85) ; yet sixty years later, the same fallacy is perpetuated in
analyses of Georgian, and it is labeled an 'ergative language' . To main
tain that Georgian is ergative or even 'partially ergative' would be to
ignore the need for accurate description and to do violence to the
integrity of the active/inactive system as a distinct type.
6 Theoretical implications
The Georgian data presented in this and earlier chapters are important
for the study of universal aspects of active type rules. These data bring
particularly strong evidence to bear on three interrelated questions :
(a) What level of derivation is relevant to the active/inactive distinction ?
(b) Can active type rules be stated directly on semantic notions, rather
than on syntactic relations ? (c) Why is active distinguished from inactive
in intransitive verbs, but not in transitive verbs, in (8) ? The evidence
from Georgian in each respect is summarized briefly below.
The many rules that change grammatical relations in Georgian make
258 The nature of the Georgian verb classes
it clear that both initial and final levels of derivation must be taken into
account in the following way :
(i) Initial grammatical relations are determined by semantic
notions 'agent', 'experiencer', 'patient', etc. (cf. §2. I).
(ii) Syntactic rules may change grammatical relations.
(iii) Case marking in Series I and Inversion in Series III are
sensitive to grammatical relations at more than one level (cf.
§2·3 and 3).
Case marking in Series II cannot be stated directly on semantic
notions, without the intervention of the syntactic relations, as stated
above. Although agents in sentences like (10-14) are marked with the
ergative case, it is not true that all and only agents are so marked. For
example, the initial subject of a passive is a semantic agent but is not
marked with the ergative. On the other hand, the subject of (34) below
is non-agentive but is marked with the ergative.
Georgian also provides data that show why Sapir did not distinguish
active from inactive in transitive verbs. Like intransitives, transitive
verbs may be semantically active or inactive. The initial subjects of
active transitives are agents, those of inactive transitives are experi
encers. If inactive transitives undergo no change in grammatical re
lations, they will have a final subject and direct object, as in (34).
(34) (a) kartulma da�arga 'vin' da 'ra' �tegoriata. . .
Georgian-ERG it-lost-it-II-1 who and what categorieS-GEN
farmoeba.
formation-NOM
'Georgian lost the formation of the human and neuter cat
egories.'
(Example from text, Chikobava 1 940 : 1 6.)
(b)
kartul
' G eo rgi an '
r;armoeba �ategoria
' formation' ' catego ry '
Appendix A : Sample lists of verb classes 259
(b)
dakargva zmna v
CLASS 4 : a. The future/aorist tenses are formed with the character vowel e-.
b. In the future tense, the suffix -a marks third person singular
subjects.
c. In the aorist , the third person plural final subject is marked
as a singular (cf. ch. 15).
The lists following show that (g), unlike (d-f), is a tendency, not an absolute
correlation.
Some of the semantic correlates of Class are discussed in Holisky (I98oa).
It must be noted that these are Classes of derivational forms, not of verb
roots ; a particular root may be represented in more than one derivational
type. The root fer, (A) dat;ers 'he
for example, occurs as a basic Class I verb
will write it', a causative (B) daat;erinebs 'he will cause him to write it', an
analytic passive (D) dat;erili ikneba 'it will be written', an ikna-passive dat;erili
ikna 'it got written', a Zustandspassiv (H) efereba 'it will stand written'. The
root cur 'swim' occurs as a basic Class I verb (A) gacuravs 'he will swim it',
and as a basic Class 3 verb (M) icuravebs 'he will swim'. Ce'f!,v 'dance' occurs
as a causative (B) ace'f!,vebs 'he will cause him to dance', as an inceptive (F)
ace�vdeba 'he will begin to dance', as a basic Class 3 verb (M) ice'f!,vebs 'he
will dance', and as a desiderative (Q) ece�veba 'he feels like dancing'. Tetri
'white' occurs as a causative (B) gaatetrebs 'he will make it white', and as an
inceptive (F) gatetrdeba 'it will become white'. Each of these forms represents
a regular, productive derivational category. Without exception, all causatives
(B) belong to Class I ; all synthetic passives (E), analytic passives (D), incep
tives (F), ikna passives (G), and Zustandspassivs (H) to Class 2 ; and all desider
-
atives (Q) to Class 4. Class 3 contains all the members of the -ob denominals
(N) discussed in Holisky (I 98oa).
The lists are presented here to illustrate the verb types that constitute each
Class and to illustrate the verbs which constitute each derivational type
referred to in §2.2. The verbs selected are typical; they illustrate the general
izations made in the text above. At the same time, the lists contain examples
of every kind of exception to the generalizations, both to the claims of tra
ditional analyses and to the proposals made here.
S A M P L E L I S T OF C L A S S I V E R B S
,
gacuravs 'he will swim it (e.g., river)
daus1Vens 'he will whistle it'
S A M P LE L I S T OF C L A S S Z V E R B S
S A M P L E L I ST O F C L A S S 3 VERBS
I T R A N S I T IVE V E R B S IN CLASS 2
This type has the following ambivalent properties : (i) My analysis predicts
that they would belong to Class I , since they have an initial subject and
direct object, the initial subject being also the final subject. But they have the
morphological characteristics of Class 2.28 (ii) Syntactically, they behave
either like Class I or Class 2 verbs. This group includes the following verbs :
mouqva 'he told it to him', seelf-itxa 'he asked him it', sehpirda 'he promised
it to him', seevedra 'he begged him for it' , and seexveca 'he asked him for it'.
An example of this group in Series I is provided in ( I ) .
We can see that these verbs are transitive through their behavior under
Appendix B: Ambivalent exceptions 269
various rules. Under Causative Clause Union, for example, the nominals
behave as we would predict for an embedded subject, direct object, and
indirect object, respectively ; that is, the embedded subject is the indirect
object of the causative, the embedded direct object the direct object of the
causative, and the embedded indirect object is the retired indirect object of
the causative (cf. ch. 5) :
INITIAL Matrix 1 Embedded 1 Embedded 2 Embedded 3
CAUSATIVE 1 3 2 Retired 3
(2) illustrates the causative of ( I ) in Series II. Recall that all organic causatives
are in Class 1 and thus govern Pattern A in Series II. (2) shows that in the
causative of qola 'tell', each of the terms is case marked as expected for a
causative according to the chart above.
The fact that the nominal vano is marked with mier in (3) shows that it is
the retired subject of a transitive verb. The fact that be{edi 'ring' is marked
with the genitive in (3) supports the view that it is the retired direct object.
A further argument that these are transitive verbs comes from Passiviz
ation. In ch. 7 it was shown that Passivization is a rule that promotes a direct
object to subject. The fact that moqola 'tell' can undergo Passivization sup
ports the view that it has a direct object initially and is a transitive verb. The
passive of qola is illustrated in (4) .
(4) motxroba iqo moqolili myelvarebit.
story-NoM it-was-II-2 told excitement-INST
<The story was told with excitement.'
2 70 Appendix B : Ambivalent exceptions
Not aU of the verbs of this group can passivize, however. For example, there
is no Georgian sentence corresponding to ' ?\Vater was begged for' . 2 9
Thus, causatives, masdars, and passives provide evidence that these verbs
are transitive. My analysis predicts that their initial subject would undergo
Inversion in Series III, and that in Series II the initial subject, being final
subject, would be marked ergative. Yet these verbs unexpectedly have the
morphology of Class 2. Sentences (s) and (6) show that uqveba 'he tells it to
him' may have the syntactic characteristics of either Class I or 2 .
(s) Class I
(6) Class 2
(a) mama ojaxs motxrobas mouqva.
father-NOM story-nAT
'Father told the story to the family.'
(b) turme bip saxels sely.itxia gogos.
boy-NOM name-nAT he-asked-her-it- III girl-nAT
'Apparently the boy (has) asked the girl her name.'
In (sa) Pattern A is used, and in (6a) Pattern B. In (Sb) the verb has under
gone Inversion, in (6b) it has not. The difference in the verb forms in (S-6b)
corresponds to the different morphology required for verbs that have under
gone Inversion and verbs that have not.
I have shown that the verbs of this group are ambivalent in the following
way : (i) On semantic grounds, they should belong to Class I , but they have
the morphological characteristics of Class 2. (ii) Syntactically, they may
behave like Class I or 2 verbs ambivalently.
::>. A C T I V E I NT R A N S I T I V E S I N C L A S S 2
These verbs are ambivalent in the following way : (i) my analysis predicts
that they all have an initial subject, some an optional initial direct object ;
.
that is, they are Class 3 verbs. Morphologically, they have some character-
istics of Class 2 and some of Class 3. (ii) With respect to case marking in
Series II, these verbs may have the characteristics of Class 2 or 3, under cir
cumstances that are specified below. With respect to Inversion, they are
strictly Class 3 verbs. Verbs of this group include elapara�eba 'he is talking
Appendix B : Ambivalent exceptions 271
to him', etamaseba 'he is playing with him', estumreba 'he is visiting him',
esaubreba 'he is conversing with him' , and others.
Each of the verbs in this group corresponds to a regular Class 3 verb :
(c) (d)
Based on the semantics of the verb, we would predict that each of the above
types represents a Class 3 verb. However, the first two examples are expressed
by ob forms, and the last two by e-ebi forms. It is not clear why the differ
ence between having or not having a final indirect object should mean a
difference in Class, especially as it does not elsewhere in the grammar
(cf. ch. 6).
Although normative gr ammars indicate that e-ebi forms belong to Class 2,
that does not tell the whole story. Let us consider each of the morphological
characteristics of Class listed in Appendix A, with reference to these verb
272 Appendix B: Ambivalent exceptions
forms. The first (a) is not applicable, since these forms have no future distinct
from their present and form their aorists differently from any of the types
listed in Appendix A. According to characteristic (b), these verbs are strictly
in Class z, since they use -a/-an as the third person markers in the future
( = present) :
With respect to criterion (c), on the other hand, my informants accepted -es
(typical of Class 3) or -nen (typical of Class z) equally:
( 1 0) elaparalJes
(a)
(b) e Iapara.1men
}' they t. alked to h1m
" .
For some speakers, the ergative is used as subject case only when there is a
direct object, and the nominative only when there is not :
3 ARSBBOBS
The verb arsebobs 'it exists' is ambivalent in the following way: (i) On the
basis of the semantics of the verb, my theory predicts that it belongs to
Class 2. Morphologically, it is a member of Class 3. (ii) Syntactically, it
behaves like a verb of Class 2 or 3 ambivalently, under circumstances that
are stated below. Not surprisingly, normative grammars classify it, according
to its morphology, in Class 3 only.
(16) shows that arseboba 'exist' may trigger Inversion only if a location is
stated. Triggering Inversion is a characteristic of Class 3 verbs, not of
Class 2 verbs.
274 Appendix B: Ambivalent exceptions
4 S U M M A R Y AND C O N C L U S I O N S
The purpose o f this appendix has been t o show that the strict classifications
of normative grammars and dictionaries do not always reflect the true use
of the verbs in the speech of present-day dwellers of Tbilisi. In each of the
examples cited above, my theory predicts that a particular verb belongs to
one Class, and normative grammars classify it as belonging to another Class.
The truth in each instance is somewhere in between.
Epilogue
In this epilogue, I will try to draw together related results that have
been discussed in separate chapters. In addition, I will give very brief
summaries of the most important conclusions reported here.
Some linguists have expressed doubts that these notions play a role in
the syntax of a language like Georgian, where the apparent subject may
be marked with any one of three cases and may trigger either of two
sets of agreement markers. I have shown that a large number of rules
refer to these syntactic notions, regardless of the case of the nominal.
In particular, the rules of Tav-Reflexivization, Subject Person Agree
ment, Causative Clause Union, Inversion, Retired Term Marking, and
Number Agreement refer crucially to the grammatical relation subject.
The rules of Direct Object Person Agreement, Object Camouflage,
Object Raising, Causative Clause Union, Passivization, Retired Term
Marking, and Unaccusative are among the rules that refer to the gram
matical relation direct object. And the rules of Indirect Object Person
Agreement, Object Camouflage, Causative Clause Union, the several
Version rules, and Retired Term Marking refer to the grammatical
relation indirect object. For Georgian, none of these rules can be
stated in a simple way on the basis of case marking or linear order.
In the Introduction and early chapters of this work, it was shown that
case marking appears to be very complex in Georgian. Three Patterns
were identified,
276 Epilogue
At the same time, these rules show that the term case marking rules in
Georgian are not ergative, in the generally accepted meaning of that
notion.
A retired term! is a nominal which bears the i-relation as its last term
relation, and which is a final non-term. At least six constructions in
Georgian involve retired terms : object-raised clauses, organic caus
atives, passives, inversion clauses, masdars, and infinitives of purpose.
In ch. II it is shown that all of the retired relations involved in these
disparate constructions can be accounted for with three simple rules :
4 Rule interaction
The investigation reported here takes into account a large body of facts
and shows the systematicity and consistency in apparently disparate
rules. The data assembled here do not support the notion of a 'squish'
of grammatical relations (cf. Keenan (I976) and Johnson ( 1 977) for two
sides of this problem). The system analyzed here supports instead the
idea of a distinction between levels of derivation (or initial and final
termhood) and rules which take these into account, in something like
the manner summarized in §4. Taken as a whole, the system worked out
Characteristics of grammatical relations 279
Introduction
1
A problem with Chikobava's and Aronson's analysis is noted in ch. I S , n. I I .
2 Johnson ( 1977) discusses the difficulties of arbitrarily selecting criteria which would
define the notion 'subject'.
3 Some traditional works do recognize levels of derivation to a certain extent. They
must intervene in this order : Q-word, NEG, undo, VERB. It is not clear whether undo
is synchronically a verb (auxiliary or modal) or a particle in this use.
7 'Emphatic -a' is added to personal pronouns in sentence-final position, as in len-a,
and under certain other circumstances.
s Both (.I6b) and (*I 6c) can have a grammatical reading. Each requires a heavy
pause to separate figni from the rest of the sentence, and is equivalent to left- and
right-dislocation in English, in that the isolated element represents a fore- or after
thought with respect to the sentence as a whole.
D Relative clauses are an exception to this generalization. When the head of a relative
clause is questioned , the relative clause is necessarily separated from its head, as
shown in
10
It might alternatively be proposed that the suppletion described in §4.3.I and 4.3.2
is conditioned by the animacy/number of the semantic patient, rather than that of
the direct object. These positions are not distinct, however, since in chs. 4, 5 , 7.
and 8 it is established that it is the initial direct object that governs suppletion, and
in ch. 16 it is shown that all semantic patients are initial direct objects.
Notes to Chapter I 281
11 Although prescriptive norms call for the use of jdoma only with singular subjects,
in fact it is used Vl<ith both singulars and plurals. Sxdoma, however, is used only
with plurals, as my examples show. Cf. Shanidze (1 973 : 5 01-2) ; Tschenkeli (1958 :
z86-7) ; Tschenkeli (1960-73).
Chapter :l
1 The dialect described here and referred to throughout this monograph is the more
restrictive of two dialects I identified with respect to Tav-Reflexivization. Both are
described in Harris (ms.).
2 These same forms occur as emphatic pronouns and as camouflaged objects (cf.
Like its English counterpart, (v) may in context mean either that some particular
people (referent identified) say this, or that in general it is said (referent not ident
ified). In the latter case the nominal is unspecified ; in the former, it is an unemphatic
anaphoric pronoun.
Chapter 3
In (i-ii) the dropped indirect object triggers Object Camouflage obligatorily. This
constitutes additional evidence for a rule of Unemphatic Pronoun Drop, since
without such a rule the ungrammaticality of (+i) would be unexplained.
, The person of the indirect object does not affect the rule in the dialect reported
here. In (Z-3) and (5-6), the indirect object is third person ; in (iii) below, it is
first person, and in (iv) second.
Notes to Chapter 4 283
(iii) (a) ·vanom (sen) semada1'a (me).
YOU-NOM he-compared-me-him- I I - r me-DAT
('Vano compared you to me.')
(b) vanom seni tavi semadara (me).
'Vano compared you to me.'
(iv) (a) ·vanom (me) segadara (sen).
me-NOM he-compared-you-him-II-r you-DAT
(b) vanom cemi tooi segadara (Sen).
'Vano compared me to you.'
(1) shows that this rule does not apply to third person direct objects with a third
person indirect object. (v), which differs from (1b) only in the person of the indirect
object, shows that the third person direct object is unaffected, regardless of the
person of the indirect object.
(v) vanom anzori segadara (sen).
'Vano compared Anzor to you.'
There exists another dialect in which Object Camouflage applies only if the
clause contains an indirect object of the first or second person. That dialect is
described in Boeder ( 1 968 : § I .6) and in Harris ( 1 976 : Appendix to ch. 3). I found
that dialect only among speakers not native to Tbilisi.
5 The phrase tooisi tooi 'self's self' occurs, but only as a reflexive (cf. ch. r, § I ) .
Therefore,
means, not 'Vano! compared himj to me', but 'Vano compared himself to me'.
6 'Personal pronouns' must here be understood to include es, igi, and is, which also
function as demonstrative pronouns.
Chapter 4
1 I found one older speaker who did not have Object Raising at all. All of my younger
informants did have the rule, though object-raised constructions do not occur
frequently.
2 I have glossed the masdars with English -ing forms ; this is not intended as a claim
that the two are the same syntactically.
3 A refinement of this analysis is made in ch. 1 6.
, See Introduction, §4.1 on Aris-Cliticization. An optional form of ( I I ) has aris cliti
cized to ra-
The genitive in (i) cannot occur with the infinitive, as predicted by the analysis
in §:a:
Chapter 5
After Object Raising applies, there is no direct object ; the input conditions for
Object Camouflage are therefore not met, and the rule cannot apply.
8 That Causative Clause Union can apply interative1y is shown by (iv), which is the
causative of the causative sentence (r I).
8 Some speakers do not accept these forms as double causatives at all. For such
speakers, Iemaimevina means 'he fed it to me'; that is, it is a single causative. Those
speakers I interviewed who did accept this fonn as a double causative considered
it ambiguously a single causative. (See also n. 6.)
9 The comparison of (vi) with (35) shows that this nominal does not trigger Person
Agreement or Number Agreement.
(vi) dedam tkven papa Iema{mevina.
mother-ERG yOU(PL)-DAT gruel-NOM she-caused-feed-me-it-II-1
'The mother made me feed gruel to YOU(PL).'
The fact that changing the initial embedded indirect object to second person plural
causes no change in the verb shows that this nominal fails to trigger both agreement
rules. It can also be seen in Shanidze's and Vogt's additional examples that in other
dialects this nominal fails to trigger the agreement rules.
If tkven 'YOU(PL)' is omitted from (vi), the meaning is changed :
(vii) dedam papa Iema{mevina.
'The mother fed me gruel.' (Cf. n. 8.)
This shows that the nominal cannot undergo Unemphatic Pronoun Drop.
10
In Old Georgian, retired indirect objects in all constructions were regularly marked
with the dative, as in these idioms (Harris 1 979). See ch. 1 1 for marking of retired
indirect objects in Modem Georgian.
Chapter 6
1 There is confusion in the traditional tenninology. I use the word version only with
reference to rules that create indirect objects. The marker that they trigger in the
verb I call 'version vowel' or 'version marker'. It is important to distinguish this
from other uses of the so-called 'character vowel', or vowel prefix which occurs
immediately before the verb root. Character vowels are used as (a) markers of
version, (b) parts of the productive markers of synthetic passives, such as icvleba 'it
changes' (cf. ch. 1 3), (c) parts of the markers of certain tenses including future and
aorist, for Class 3 verbs, such as imuIava 'he worked', (d) parts of the markers of
causatives, such as antebinebs 'he causes him to light it', (e) parts of the markers of
286 Notes to Ckapter 6
Series III forms for verbs of Class I and 3, such as dauferia 'apparently he has
written it' (cf. ch. 8), and (f) morphologically empty markers which are attached to
finite forms of some verbs, e.g. ifqebs 'he begins it', icinis 'he is laughing'.
2 The same is true in the plural :
(i) fe-gv-i-1!.era.
'He sewed it for us.'
(ii) fe-g-i-1!.era-t.
'He sewed it for YOU(PL).'
(iii) fe-u-1!era.
'He sewed it for them.'
(iii) is not distinct from (7c) ; that is, plurality is not marked in the third person.
This is explained in ch. I S.
s Examples (*'13) and (*'14) are respectively (*'34) and ("36) from Holisky (1978) ;
transliteration and glosses have been changed to correspond to the conventions
adopted here. The verb in (*'I3-14a) is highly irregular and not easily assigned to
a Class.
4 In Harris (1976) I proposed a general constraint on Version, such that it cannot
create a chomeur. While this constraint is supported by many examples, there are
exceptions to it. In (iv), gela has been put en chomage by the advancement o f 'me',
as shown in (v). Gela is marked with the dative case in (iv a) and with a positional
in (Iv b). (As discussed in §6, indirect object chomeurs may be in the dative case
in non-standard dialects.) Both sentences are marginal, but each is accepted by
some speakers. The grammaticality of these examples depends on the version
nominal being first or second person (cf. ch. 5, §4.2 for a similar situation).
(v)
ferili gek
wr
ite , secretary '
' '
'letter ' ' me '
While there are exceptions like (iv), some form o f this constraint must b e included
in the grammar to account for the non-occurrence of (I sC) and many other examples.
fi The counterpart of (*'I7b), with the version vowel i- instead of u- is also ungram
matical :
7 It can also happen that the version nominal, the subject of its clause, and the
direct object are all coreferential. In such a case, the version nominal and direct
object are reflexivized, as in (vii).
Version may optionally apply, with Coreferential Version Object Deletion, giving
Thus, the rules already proposed correctly account for these phenofl,1ena.
8 There is no form -mivi"ere ('I wrote a letter to myself') corresponding to (19).
While ilapara�a and imyera, corresponding to (20) and (:21) exist as forms, they
do not mean 'he talked to himself', 'he sang to himself', but simply 'he talked' and
'he sang', respectively.
9 There is no -gairmendina.
10 Among the analyses known to me, only Boeder ( 1 968) and Nebieridze (1976) have
recognized that synchronically sataviso kreva is a special instance of Benefactive or
Possessive Version and that the i- is a marker of Benefactive or Possessive Version,
whether or not there is coreference between the subject and version nominal.
1 1 This discussion has been limited to Benefactive Version for the sake of simplicity.
To state the revised version of (23) with 'reflexive' therefore would constitute the
false claim that (x b) and similar sentences express coreferentiality. Therefore (23)
288 Notes to Chapter 7
must be restated as (xi), which is more complex in this sense than ( I S) and consti
tutes an argument against the Extended Traditional Analysis.
marker.
(xiv) ' • rom migvifero
• . ferili.'
that you-write-us-it-II-I letter-NOM
' . . . that you may write a letter for us.'
This is quoted in Shanidze ( 1973 : §4oz), from Vazha Pshavela, an author who wrote
in the Pshav dialect.
1 4 Seiter (1979), however, presents data that pose a serious challenge to both Laws.
Chapter 7
1 Direct construction is used in this work instead of active, in order to avoid the mul
tiple ambiguity of the latter.
2 Passivization is an optional rule. If a clause undergoes the rule, the verb it contains
Notes to Chapter 8 289
is a Class 2 verb. If the same clause does not undergo the rule, the verb it contains
is a member of another Class.
3 The formation of the past passive participle, but not its use in the passive con
Without a first or second person direct object, the input conditions for O bj ect Camou
flage are not met. Thus, the proposal made above accounts for the failure of Object
Camouflage to apply in Series I I I , with Class z verbs as well as with Class I and 3
verbs.
7 In both types, Unaccusative applies only if its input conditions are satisfied, of
course. This is discussed below in §6.
B Not all traditional descriptions of Georgian take the position summarized in (27-28).
Shanidze (1 963), for example, takes the view that the relations indicated in (Z4) are
the only grammatical relations. From this position, case marking could be simplified
as in (Z7'-28'), but the facts observed in §4. 1 , 5 . 1 , and Appendix A cannot easily
be accounted for.
9 This statement is refined in ch. 16.
10
Alternative forms for some speakers are moamIives and mOQmIies; the meaning and
grammatical relations are the same as in (29b).
11 In the causative of a transitive inversion verb, the experiencer is the derived indirect
obj ect, and the stimulus is the derived direct object. This shows nothing, however,
since the derived indirect-objecthood of the experiencer could arise either from its
being the subject of a transitive or from its being an indirect object. Similarly, the
final direct-objecthood of the stimulus could arise either from its being a direct
object or from its being an intransitive subject. Since these are precisely the pairs of
relations that are in question, the causative of a transitive inversion verb sheds no
light on the problem.
12
There is a systematic ambiguity in examples of this type, which derives from the
fact that tvis not only marks retired indirect objects, but also marks benefactives.
Thus (30b) could also be glossed 'Apparently the king has had the palace cleaned
292 Notes to Chapters 9-IO
for you,' with the initial subject of 'clean' left unspecified. The alternative word
order, tUNne mepes gaupnendinebia sasaxle Ient'Vis is also, for many speakers, accept
able and ambiguous.
13 Example (45) below appears to provide additional support for the initial direct
languages, cf. Perlmutter (to appear b) and Perlmutter & Postal (to appear c).
15 The inability of Passivization to apply to the output of Inversion, on the other hand,
is accounted for automatically by the fact that the output of Inversion (or Inversion
and Unaccusative) fails to meet the input conditions on Passivization. Passivization,
as stated above in §6, may only apply in clauses containing a subject and direct
object. The output of Inversion (only) contains no subject. The output of Inversion
and Unaccusative contains no direct object. Therefore, it is correctly predicted that
Passivization cannot apply to the output of Inversion. The ungmmmatical sentences
that would result cannot easily be illustrated, since the fonns are lacking.
The same is true of the other interactions discussed in §7.2-7.5, as the reader
can confinn for himself.
16 This example is from Tschenk�li (1958 : vol. II, 268, no. 3).
17 At least the following verbs require a genitive under certain circumstances : mjera 'I
believe it', mesmis 'I hear, understand it', melinia 'I fear it', mrfams 'I believe in it',
mrcxvenia 'I am ashamed, shy of it', mexatreba 'I am shy of it', mSurs 'I envy it'.
1 8 It is not clear under exactly what circumstances the complement must be genitive.
In (x) both animate and inanimate nominals must be in the genitive, but cf.
Chapter 9
1 Analyses similar to this have been proposed by several scholars, including Marr
(1925) and Schuchardt (1895). Chikobava (I96r) considers in detail this analysis of
Georgian and several others and argues eloquently in traditional terms that the
ergative-nominal in Series II is both the 'real' (initial) subject and the 'morpho
logical' (final) subject.
Chapter 10
9 Some speakers find the infinitive acceptable in (iii), but most of my informants
prefer (iv), with a true participle.
Chapter II
1 Retired transitive subjects of participles are also usually marked with mier (cf.
ch. 10, n. 1 3), but a detailed analysis of participles has not been proposed here.
There is a second postposition that is sometimes used in place of mier. (i) illus
trates this use of the postposition -gan 'from', apparently marking a retired transitive
subject.
(i) es vasli micemulia bavlvisagan.
this apple-NOM given-it-is-I-z child-from
'This apple is given by ( ?from) the child.'
The use of -gan can be predicted from the fact that the nominal is a retired subject,
together with the fact that the governing verb is micema 'give'. This verb permitll
the use of -gan to mark its retired subjects generally.
In tills matter there is considerable variation among speakers, which I will
characterize in terms of dialects. In one dialect, -gan is acceptable only with a very
limited number of verbs, including micema of (i). Speakers of this dialect find (ii)
unacceptable.
(ii) gelasagan (:erilis dtJfera . • .
It is probable that in this dialect -gan has only the meaning 'from' and does not
mark retired subjects, even in (i). In a second dialect, -gan is used more widely,
but it is still restricted to a particular set of verbs. Speakers of this dialect find (ii)
acceptable. In a third dialect, -gan is sometimes preferred to mier : (iii) is preferred
over (5).
296 Notes to Chapter II
(iii) bavfvi dalt.benilia 3aylisagan.
child-NOM bitten-it-is-I-z dog-from
'The child is bitten by the dog.'
But in the intransitive (v), vano is not the initial subject; mier is unacceptable.
{
while meiveobit is acceptable.
"'mier
(v) vardi gaizarda vanos
mesveobj't.
v
it-grew-II-z
'The rose grew because of Vano./The rose grew through the intervention of
Vano.'
(The nature of constructions like (v) is discussed further in ch. 13.) I conclude that
meiveobit means 'through the intervention of' rather than having a purely gram
matical function. It is interesting that the same meaning attached to mier in Old
Georgian. though today it is not used that way in the standard dialect (cf. Harris
1979),
S Because I had not included sentences like (9)/(16) in my 1974-75 field work in
Georgia, in Harris (1 976, 1977) I reported only (r6), on the basis of work with only
one informant in the United States. Based on ( 16) and other examples of that type,
I concluded that marking was as in ( I S). Subsequent field work in Georgia in 1977
revealed that the dialect originally reported is not widespread.
4 One fact seems to support this possibility : a very few verbs govern a particular post
position other than tvis or mier for marking a specific retirement relation. For
example, miYfeva 'reach' governs -mde 'until' fpr its retired object : miueva hal
akamde 'reaching the city'.
Notes to Chapter IZ 297
£ The relationship between grammatical relations and semantic relations is explored
further in ch. 16.
• Mier in Standard Modern Georgian is used only to mark retired subjects. Tvis, on
the other hand, while it marks retired indirect objects, also marks other non-terms
in the general meaning 'for'. Thus, the fact that the nominal in question in the
object-raised sentences is marked with tvis does not indicate that it is a retired
indirect object.
A. Synthetic passives are illustrated and characterized in the body of this chapter.
B. Potentials, like synthetic passives, are related to direct forms, in the sense that
they are formed from the same abstract verb. For example, es xe ixerxeba 'this
Notes to Chapter I3 299
tree/wood cuts well', a potential, is related to viyac xerxavs am xes 'someone
is cutting this wood/tree', a direct form. Potentials may be distinguished from
other verb forms with the same morphology by the fact that they have no
forms outside Series I.
As far as I am aware, all of the characteristics discussed in this chapter with
respect to synthetic passives apply equally to potentials. I have excluded them
because of lack of data.
c. Certain other verb forms which share the morphology of synthetic passives
do not share some of the other characteristics discussed in §2. This group
includes icereba 'it is written', iyeba 'it is received', and ifreba 'it is cut'.
D. Forms like vardeba 'he falls', as stated above, share the morphology of syn
thetic passives, but may be distinguished from them by the fact that they do
not have corresponding direct forms.
Traditionally these sets of verbs have been described together, since they share
morphology. The arguments adduced in this chapter do not necessarily apply to
types B-D. In ch. 16 the similarity of these four types and others of Class 2 is
explored.
There are also several verb forms which are morphologically like the synthetic
passive, but which are not regularly derived. These include verb forms like igineba
'he curses', imaleba 'he hides'. Some of the problems presented by these irregular
verb forms are discussed in Harris (to appear b) and Harris (1976), Appendix C.
3 This definition rules out the inversion construction. This could be considered as a
fourth contrasting construction, but it has already been discussed in detail in ch. 8.
;1 Notice that animacy plays no role in distinguishing between the analytic and the
synthetic passive ; both may have animate or inanimate initial direct objects. These
possibilities are illustrated in the examples of § 1.
;; Some Indo-European middles correspond to the forms I refer to as 'synthetic
passives', others to some of the other forms which share that morphology (cf. n. 2).
Still other Indo-European middles correspond to the Georgian Benefactive or
Possessive Version with Coreferential Version Object Deletion, as observed in
Schmidt (1965).
While most Georgianists have called synthetic passives and the other forms
described in footnote 2 'passives', Blake ( 1 932 : 234) observes that they are 'in
reality middle'.
6 According to Holisky, these diagnostics test for the 'presence of an agent' ( 1 978 :
I SO, 1 52). There is no doubt that the adverbs of intention and the command
imperative are impossible in sentences lacking an agent, but neither are they possible
in all sentences containing an agent. They may in general occur in sentences with
semantic agents only if the agent is a final subject. As an exception to this, ganzrax
'intentionally' may marginally occur with analytic passives if the initial subject is
specified and ganzrax is associated with it by word order, as (·i) and (ii) show ; the
word order of (.i) is the unmarked order.
support the initial grammatical relations proposed in §3. In the direct construction,
(iii a), the version object possesses the direct object; in the synthetic passive, (iii b),
it possesses the final subject. However, the constraints on exactly which nominals
may be interpreted as possessed by the version object are still poorly understood.
No argument can be based on sentences like (iii) until this fundamental property
of Possessive Version is determined.
Case marking and Inversion as evidence for the analysis proposed here are con
sidered at length in ch. 16.
13 In the ikna-passive, like the iqo-passive, the initial indirect object is realized as a
final non-term , as shown by (iv) (cf. ch. 7, §z).
Notes to Chapters I4-I5 301
(iv) es vardi dedastvis ikna micemuli.
this rose-NOM mother-for it-was-II-z given
'This rose was given to Mother.'
The retired indirect object, deda 'mother', is marked with tvis.
14 The use of such phrases as tests of stativity in Georgian is established in Holisky
(1978).
Chapter I4
Chapter IS
Holisky ( 1980a) emphasize the regularity of this group. Nozadze (1974: 30ff.) also
traces the historical development of the morphology that sets Class 3 apart from
other verbs. Holisky shows that the difference is due to the fact that the preverb in
Georgian is associated with punctual aspect, and this category is inapplicable to
medial verbs because they are atelic.
5 An example of this is the statement below. Emphasis is added .
active/inactive rule in Sapir's sense, three Georgian linguists have observed that
the use of the ergative or nominative as the subject case in Series II depends upon
whether the verb is active or inactive (Jajanidze 1970 ; Jorbenadze 1975 : 219ff. ;
Topuria 1923 : 1 I 5 and 120, and 1954: no. 7).
0
1 The notion 'basic' will not be defined ; d. lists of examples in Appendix. 'Basic
Class I verbs' may be taken to be Class 1 verbs that are not causative, etc.
11
The type atetrebs 'he makes it white' from the adjective tetri' white' is also a causa
tive. The type xerxavs 'he is sawing it' (xerxi 'a/the saw') may represent an ad
ditional derivational type. In any case, its term relations are identical to those in (10).
1 2 Perlmutter & Postal (to appear c) propose a Final 1 Law, requiring every clause
to have a final subject. If this law is correct, it obviates the necessity of making a
language-particular statement that Unaccusative is obligatory.
13 Not only noun subjects, but also sentential subjects of basic Class 2 verbs are
initial direct objects. This includes sentential subjects of qopna in sentences in which
304 Notes to Chapter I6
Object Raising applies. This analysis is summarized in the network below (cf. (6)
of ch. 4).
qr-areba semiq'Z·arlieba.
" love' 'I' ' he ' • I will fal l i n love with hill!. ,
The analysis presented here is supported by the following facts. The initial
grammatical relations represented in the dependent clauses of ( 1 8), ( 1 9), and (i),
are the same as those which the unmarked form of each verb takes in an independent
clause. The initial grammatical relation of the embedded clause to the abstract
predicate COME ABOUT is the same as that of a non-sentential dependent to the
concrete verb axda, moxda, gaxda 'it came about, happened, became'. The final
grammatical relations are consistent with Person and Number Agreement and case
marking (cf. (I» . In each of the three structures (1 8), ( 1 9), and (i), the initial gram
matical relations are correctly related to the final grammatical relations by the rules
of Clause Union and Unaccusative, withno special apparatus. (Cf. Aissen & Perlmutter
1976 and Frantz 1976 on Clause Union in constructions other than causatives.)
Finally, to the best of my knowledge, this analysis violates no rule or principle
of Georgian or universal grammar.
15 Basic Class 2 verbs (c) may correspond to causatives (B), which are always Class I
verbs, e.g., daarlina 'he left him, he caused him to remain', Such causatives do not
provide an argument to distinguish between initial intransitive subjects and initial
intransitive direct objects, since Causative Clause Union treats direct objects and
intransitive subjects alike.
Notes to Chapter I6 305
IS
I have generalized the rules for Series I and III (rule I) but the rules for Series I I
apply equally well for Series III, as the reader can verify for himself.
17 One of my informants made a revealing remark about example (22). "''hen pre
sented with the alternative of (22a) or (22b), he replied that he knew that (22a) was
right (normative), but that it sounded strange, as though the water were dripping
on purpose, as though the water were active (ak/iuri).
18 Although the use of the ergative case with verbs that irregularly govern Pattern B
in the literary dialect has been much discussed in the literature on Georgian dialects,
only a few linguists have observed that this phenomenon is limited to semantically
active verbs that are formally in Class 2 (Jajanidze 1 970 : 258 ; Topuria 1 923 : 1 20,
and 1954 : 455). Topuria, in the works cited, also shows that the use of this case is
not due to the influence of Mengrelian, but that it is a process of regularizing
(unipi�acia).
19 'With respect to complex structures, (24') must be interpreted to mean that Inversion
may apply in a clause only to the initial subject of that clause. In clause union and
raising constructions, Inversion applies only to initial subjects of matrix clauses. In
sentences that are finally multi-clausal, Inversion may apply in an embedded clause
to the initial subject of that clause. For example, in the saying
Inversion applies in the dependent clause only (because of the negative), demoting
the initial subject of that clause (aravis 'no one') and marking the verb of that
clause with Series I I I morphology.
20 Nor does the grammar need to refer to Class 4. (23a) can be restated on the basis
of semantics as
The sentences in (iv) and (v) are parallel to (33), in the same order.
is ergative. This is based on the claim that the 'intransitive subject' and the 'direct
object' trigger the same (bold type) agreement marker in sentences like (vi) and (vii).
(vi) me vqopilvar.
I-NOM I-am-I1I-z
'I am.'
(vii) me gavgzavnivar.
I-NOM he-sent-me-III-l
'He (has) sent me.'
In ch. 8 it was shown that the markers of Person Agreement here reflect the final
grammatical relations, 'I' in both sentences being the final subject. Numerous
arguments for this analysis have been amassed in chs. 8, I S and in §3 of this chapter.
85 In (3Sa) the initial subject ( final indirect object) triggers the plural marker -t,
Notes to Chapter I6 307
as predicted by the rule of Number Agreement formulated in ch. I S . I am grateful
to Dee Ann Holisky for bringing this example to my attention.
26 These characteristics were selected instead of some other possibilities because of
(a) their high degree of mutual consistency, and (b) their high degree of correlation
with syntactic characteristics (d-f). Other morphological properties include
These properties are less regularly correlated with the syntax and semantics than
are those noted in the text.
27 The verb codna 'know' is an absolute exception. Unlike any other verb in Modern
Georgian, it uses Pattern A in some tenses of Series I ; in others it undergoes
Inversion.
28 Of the three morphological characteristics listed in Appendix A, only character
istic (c) (third person plural suffix in the aorist) was tested with informants. Infor
mants consistently rejected the Class I suffix -es for all four verbs tested. In addition,
they accepted only the imperfect formant -od-, which is used for Class 2, not
Class I, verbs.
30 I assume here that the indirect object is an initial comitative and advances to
indirect object by a rule of Comitative Version, essentially similar to Benefactive
Version (cf. ch. 6). I make this assumption for the following reasons :
I. The meaning of this indirect object differs consistently from that of initial
indirect objects, such as those with the verbs cems 'he gives it to him' or
mifeTa 'he wrote it to him'.
2. The comitative-nominal may occur either as an indirect object, in the dative
case and triggering Indirect Object Agreement, or as a non-term, with the
postposition -tan and not triggering agreement:
However, the assumption that the final indirect object is an initial comitative, not
an initial indirect object, is not crucial to the remarks that follow.
81 Non:p.ative grammars deprecate this use of the ergative (Tschenkeli I9S8: vol. 1, 430).
32 The ungrammaticality of (·ISb) is sometimes 'explained' by saying that the form
does not exist in Series III. Regular Class 2 verbs with indirect objects do have
Series III forms of exactly this type, e.g., dasfqebia 'it began for him'. If it is claimed
that (·ISb) is ungrammatical because the form does not exist, the non-existence of
the form must still be explained. I believe that the form does not exist because the
failure of Inversion to apply to the initial subject (gela in (-I Sb» results in ungram
maticality .
33 Since normative grammars do not recognize that arseboba fails to trigger Inversion
308 Notes to Chapter I 6
when its meaning is 'exist (generally)', it seems imperative to quote examples from
literary Georgian :
(x) mis gverdit unda arsebuliqo is mr:l$rivic.
it-GEN side-INST MODAL it-existed-III-z that screeve-NoM-too
'That screeve apparently must have existed beside this.'
(xi) . . .romelic rogorc cans, martlac arsebula 3vel kartuW.
which-NOM as it-seems truly it-existed-III-z Old Georgian.
'. • •which, as it appears, really did exist in Old Georgian. '
Kiknadze 1 96 7 : 186-7
In neither example has Inversion applied. Notice that (xi) is particularly similar to
(I6b) ; but the verb here is not controllable and does not undergo Inversion.
R eferences
Abbreviations
Notes referred to here are cited by chapter ; notes for all chapters begin on
p. 280. Small capitals refer to sub-classes listed in Appendix A of ch. 1 6, beginning
on p. 259. Entries in bold type indicate pages where a definition, explanation, or
statement of a rule can be found.
1 5 1, n. 2, 5 , & 6 (ch. 10), 1 76, 1 94, Subject Person Agreement. see Person
199-200, n. I I (ch. 13). 227. 229, 259 Agreement
Sheintuch, G., n. 7 (ch. 7) Subject-to-Object Raising, n. 1 2 (ch. 10)
Shibatani. M . • 8, 66 superessive relation, 5, 1 00, 102
Silverstein, M. 230 Superessive Version, see under Version
Siouan languages. 235. 236 suppletion, 1 8-21 . 168; for animacy of
Sommerfelt. A., 3 direct object, If)-20. 5 7-8, 105, 1 2 6-7 ;
spontaneous chomage, I I 2 for number of direct object. 20. 56-7,
Sridhar, S. N., 9, 68, 1 3 3 104-5, 1 25-6. 1 60-1, 1 99-200 ; for
stative/dynamic, 92-3, 204, 233. 2 3 5 , 250 number of subject, 21, 1 6 1 ; for tense,
stimulus, 128, 236 21, 1 2 9-30
strata, see levels of derivation synthetic passive, see under passives
Stratal Uniqueness Law, 7, 100
subject, 5. 275 -tan. n. 4 (ch. 6), n. 7 (ch. 1 2), n. 30 ;
agreement markers, 29; see also (ch. 1 6)
Number Agreement and Person Tanamedrove kartuli saliteraturo enis
Agreement normebi, n. 5 (ch. 7)
in analytic passives, 1 04-9, 139, 1 69, Taqaishvili, A., n . I (ch. 5)
r 7 1 , 206, 269-70 tav-, see pronouns, reflexive and
with Aris-Cliticization, 1 3 pronouns, camouflaged
case, 39-46, 62, 74-7. 243-4, 250-2, tavis-/taviant-, see pronouns, reflexive
270 Tavis-Reflexivization, see under
in causatives, 67-9, 182-4, 207-8, 220, coreference
269 Tav-Reflexivization, see under
and Coreferential Version Object coreference
Deletion, 95-9 passim tense, xx, xxi. 103 ; -aspect categories, 2,
diagramming, xxi 2 1 , 29, 46-7, 129-30, 1 47 ; sequence
first, 256, 278 of, 18
identifying, 2-4, 8, 23-38 . 45-6, I I9, term relations, 5, 23-38, 2I I-1 3 , 275 ;
147-50 constraints, 208-9. 2 16-1 8 ; see also
intransitive : of causatives, 70, 73-4, direct object, indirect object, inventory
79, 1 8 2-6, 257 ; rctired, 1 5 7-8, 1 59, of initial grammatical relations, and
1 6 1 , 1 7 1, 1 74, 179 subject
in Inversion, 1 1 7-21, 124-36. 1 40-5 Topuria, V., n. 10 (ch. 1 3), n. 9 & 1 8
passim, 169, 177. 1 79, 208, 247-8, (ch. 16)
270 Tough-Movement, see Object Raising
and Number Agreement, 2 13ff. transitive inversion verbs with no overt
in object raised constructions, 58-60, subject. I44-5, 169. 1 70 - 1 . 179. 241-2
62, 63-4, 1 40, r 69 transitives. 186, 1 89-90, 194. 248, 258 ;
in questions, 1 5 case marking governed by. 3 9-46,
retired, 169, 171, 174, 1 78, 277 ; with 1 30-1 , 23 5-46 ; causative of, 69-70,
non-finite verb forms, 1 5 7-8, 16 1-7 75-6, 8 1-5, 1 82-3, 221 ; inactive, 2 5 8 ;
passim, 1 77. 1 84-6, 269 ; in object inversion of, I I7-31 , 1 3 5-6, 242, 2 5 8 j
raised constructions, 5 5 , 6 1 , 63 ; in Retired Term l\1arking with, 1 70-1 ,
passives, 1 07-8 174, 1 84-6 ; see also transitivity
in synthetic passives, 1 97-202 transitivity, 181-90, 230, 236, 253, 257-
and Tav-Reflexivization, 24-5, 41-2, 259, 26 1 , 268-70 ; and Retired Term
72, 105-6, 124-5. 205- 10 Marking, 1 7 1 ; and Term Case
and Tavis-Reflexivization, 27-8, 42-3 Marking, 39-4 1 , 229-40, 249-53 ,
transitive : of causative, 70, 75, 79. 257-9; see also intransitives and
1 82-4 ; retired, 1 58, 1 7 1 , 174, 178, transitives
1 79. 1 84-6 transliteration, xvii-xviii
and transitivity, 41. 182-6. 236 Trubetskoj . N. S., 230
Unemphatic Pronoun Drop of. 3 5, Tschenkeli, K., 2, n. I I (Intro.), 29, 47.
78-9 n. I (ch. 5). 97, n. 3 (ch. 7), 1 I 8, n.
see also under suppletion 3 & 16 (ch. 8), 1 5 1 , 1 94, 200, n. I I
Index 327
(ch. 1 3), n. 9 (ch. 1 5), 229, n. 6 & 3 1 governed by lexical items, n. 1 8 (ch.
(ch. 1 6) 8) ; in marking retired terms, I I I
Turkish, 68, 1 14, 1 44, 257 I I 2, n. I & 4 (ch. 1 1) ; in marking
turme, l I S terms, 189, 246, 268-74
-mis in meaning of nominalizations, 1 5 3
assignment of, 171-3, 174, 176, 277 i n rules, 5 3 , 68
with obliques ; in object raised verbs
constructions, 1 8 , 61-2, 62-3, 1 69, gloss of, xix-xx
178-9. 205-6 ; other, 90, 1 0 1 , 1 3 1 morphology of, n. I (ch. 6), n. 5 (ch.
with retired indirect objects : in 8), 1 74 ; analytic passives, 103 ;
causatives, 81-5, 1 6 9 ; with non causatives, n. I (ch. 5) ; Class, 190,
finite verb forms, 1 5 9, 162-3, 1 6 5 ; 2 32-4, 25�o, 268, 270-2 ;
in passives, 1 10-12, 1 1 3, I I 6, n. 1 3 infinitives, 1 5 5 ; inversion, 1 3 9-40 ;
(ch. 1 3 ) ; in Series I I I , I , 1 22-3, Number Agreement, 223-4 ;
1 30, 1 3 1 synthetic passives, 1 9 4 ; version, 88,
Tzotzil, 95 90, n. 6 (ch. 6), 97
see also Class, inventory of initial
Udi, 256 grammatical relations, Number
Unaccusative ; with intransitive direct Agreement, Person Agreement,
objects, 1 9 1-204, 2 1 6, 240, 24 1 , 244, preverbs, selection restrictions,
248; with Inversion, 1 1 9. 1 3 3 . 1 3 4-7, Series, tense, and transitivity
1 79, 1 86, 242, 247 ; Passivization Version, 87-102, I I 5-1 6, 221-2, 244,
compared with, 1 36--7 , 197-8. 203 ; 245 ; Benefactive, xxii, 87ff., I I 6, n. 5
see also passives, synthetic (ch. 1 3 ) ; Comitative, n. 30 (ch. I 6) ;
Unaccusative Hypothesis, 235-46, 249- double, 99-100; Locative, 1 8 7 ;
256, 268 Possessive, 87-9, n . 1 1 (ch. 6), 100- 1 ,
Unemphatic Pronoun Drop ; conditions I I 5 , 1 93 , n . 5 & 1 2 (ch. 1 3 ) ;
on, 5 1 , 99, 1 84 ; failure to apply to Superessive, 87-9
retired terms, 62, 82, 88, 9 1 , 108, I I I ; version marker. n. I (ch. 6), 88. 89. 90.
principles, 32-8, 1 56, n. 2 (ch. 1 4) ; n. 7 (ch. 1 2)
with terms, 3 , 26-7, 45, 78-9, 85, 88, version object (version nominal), 88, 92,
9 1 , 100, 1 49, n. 7 (ch. 1 2) 95, 96, 1 0 1
universals, 7-8, I o-I I , 1 74-5 ; of version vowel, see version marker
agreement, 227 ; of case marking, 235- Vogt, H . , 2, 3, 22, 29. n. 5 & I I (ch. I ) .
237, 256-9 ; of causatives, 66--8 , 70 ; of 83 , 84, n. I & 6 (ch . 5), 97, n. 3 & 6
indirect objects, 94 ; of Inversion, 1 34- (ch. 7), 1 18, n. 3 (ch. 8). l S I , n. 4
1 37, 248 ; of Object Raising, 5 3 , 5 5 , (ch. 10). 1 94 , n. I I (ch. 13), n. 9
62; of non-finite verb forms, l S I , 1 60 ; (ch. 1 5), 229. n. I (ch. 1 6)
of passives, 104ff., I I 2- I 4 ; of Retired voice, 194; see also direct construction
Term :Marking, 1 75-8 ; of retirement and passives
relations, 1 1 2- 1 3 , 1 68-9 ; of rule
alignment, 257-9 ; of subjects, 3 ; of word order. 7-8, 1 54, n. 6 (ch. 1 3 ) ;
voice distinctions, 1 98 disambiguation by, n. I (ch. 3), n. 1 2
unspecified nominals, n. 1 4 (ch. I ) , I I I , (ch. 8) ; with respect to inversion
170, 1 93 , 1 9 7 constructions, 1 3 3 - 4 ; in questions,
14-16, 1 8 , 6 1 , 63. 1 66-7 ; with respect
Van Valin, R. D., Jr, 2 3 5 to reflexivization, 24-5, n. 6 (ch. I ), 1 2 5 ,
variation
1 43 ; unmarked, 1 3 , 22. 1 0 1 , I 3 3-4,
dialectal, n. I (ch. I ), n. 4 (ch. 3),
1 55 , n. 7 (ch. I S )
n. I (ch. 4), 1 8 9 ; in marking retired
indirect objects, 84-5, n. 4 (ch. 6),
99-100, 1 1 1 -12, 1 76 ; in marking Xolodovich, A. A .• 66
retired subjects, n. 1 3 (ch. 10), n. I
& 3 (ch. I I ) ; in verb agreement Ziv, Y., 9
29. n. 5 & 9 (ch. 1 5 ) ; in marking Zorrell, F., 227
subjects, 246, 272-3 Zustandspassiv, see under passives