Frege, Gottlob - Russell, Bertrand - Stalmaszczyk, Piotr - Wittgenstein, Ludwig - Philosophy of Language and Linguistics - The Legacy of Frege, Russell, and Wittgenstein (2014, de Gruyter) PDF

Piotr Stalmaszczyk (Ed.
)
Philosophy of Language and Linguistics
Philosophische Analyse /
Philosophical Analysis
Herausgegeben von / Edited by

Herbert Hochberg, Rafael Hüntelmann,
Christian Kanzian, Richard Schantz, Erwin Tegtmeier
Volume / Band 53
Philosophy of
Language and
Linguistics
The Legacy of Frege, Russell, and Wittgenstein
Edited by
Piotr Stalmaszczyk
ISBN 978-3-11-034258-1
e-ISBN 978-3-11-034275-8
ISSN 2198-2066
Library of Congress Cataloging-in-Publication Data

A CIP catalog record for this book has been applied for at the Library of Congress.
Bibliografische Information der Deutschen Nationalbibliothek

Die Deutsche Nationalbibliothek verzeichnet diese Publikation in der Deutschen
Nationalbibliografie; detaillierte bibliografische Daten sind im Internet über
http://dnb.dnb.de abrufbar.
© 2014 Walter de Gruyter GmbH, Berlin/Boston

Druck und Bindung: CPI buch bücher.de GmbH, Birkach
∞ Gedruckt auf säurefreiem Papier
Printed in Germany
www.degruyter.com
Contents
Philosophy of Language and Linguistics: The Legacy of Frege,
Russell, and Wittgenstein. Preface
Piotr Stalmaszczyk ............................................................................... 1
Mapping the Ancient City: Historical Linguistics and
Conceptual Clarification
Joachim Adler .................................................................................... 11
Russell and Wittgenstein on Proposition, Judgement, and Truth
María Cerezo ...................................................................................... 29
How to Talk (Precisely) about Visual Perception? The Case of
the Duck/Rabbit
Paweł Grabarczyk .............................................................................. 53
Priority of Thought or Priority of Language
Arkadiusz Gut ..................................................................................... 71
On the Ambiguity in Definite Descriptions
Thomas J. Hughes .............................................................................. 99
Proceduralism and Ontologico-Historical Understanding in the
Philosophy of Language
Carl Humphries ............................................................................... 115
Quine’s Criticisms of Semantics
Gary Kemp ....................................................................................... 139
Who Wants to Be a Russellian about Names?
Siu-Fan Lee ...................................................................................... 161
Bradley, Russell, and the Structure of Thought
Gabriele M. Mras ............................................................................ 181
Logic and the Pursuit of Meaning
Jaroslav Peregrin ............................................................................ 193
Objects, Concepts, Unity
Ulrich Reichard ............................................................................... 213
vi Contents
The Legacy of Frege and the Linguistic Theory of Predication

Piotr Stalmaszczyk ........................................................................... 225
Russell, Wittgenstein, and the Notion of False Propositions
Piotr K. Szałek ................................................................................. 255
Categorial Grammar and the Foundations of the Philosophy
of Language
Mieszko Tałasiewicz ........................................................................ 269
Index ................................................................................................... 295
Piotr Stalmaszczyk
University of Łód
[email protected]
Philosophy of Language and Linguistics:

The Legacy of Frege, Russell, and
Wittgenstein. Preface
It is a task of philosophy to break the power of words over the
human mind.
Gottlob Frege, Begriffsschrift
All philosophy is a ‘critique of language’.

Ludwig Wittgenstein, Tractatus Logico-Philosophicus
Before we can understand language, we must strip it of its

mystical and awe-inspiring attributes.
Bertrand Russell, An Inquiry into Meaning and Truth
0. Introduction
The present volume investigates selected aspects of the legacy of Gott-

lob Frege, Bertrand Russell and Ludwig Wittgenstein in contemporary
philosophy of language and linguistics. These three philosophers are
considered to be the most important founders of analytic philosophy; at a
later stage they shaped and inspired various philosophical approaches to
the study of language.1 It would be difficult to imagine research on, for
example, truth, sense and reference, proper names, meaning and use,
presupposition, without the achievements of Frege, Russell, and Witt-
genstein. They have influenced technical discussions on such topics as
context, propositions and predication, definite descriptions, and philo-
1
For a concise overview, see Baldwin (2006) and García-Carpintero (2012).
2 Piotr Stalmaszczyk
sophical and linguistic inquiries into the limits of language and sense,
and the relations between language, mind, and the world.
Michael Potter has recently observed that the principal contribution of
Frege, Russell and Wittgenstein to the philosophy of language was not
so much connected with the fact that “they applied philosophical methods
to the study of language”, but rather that “they applied linguistic methods
to the study of certain problems in philosophy” (Potter 2013: 852). At
the same time, however, Frege, Russell, and Wittgenstein (both in the
Tractatus and in the Philosophical Inquiries) expressed highly critical
remarks about the nature of language, and considered spoken language
to be an instrument inadequate for the science of logic. Within this con-
text, Frege pointed to the need for creating a language made up of signs,
precise and clear of any double meaning (such as his Begriffsschrift),
and Russell postulated a hierarchy of languages.
Wittgenstein, in his Preface to the Tractatus, claimed that the prob-
lems of philosophy are posed because “the logic of our language is mis-
understood” (Wittgenstein [1922]: 3). Later he observed that “Philoso-
phy is a battle against the bewitchment of our intelligence by means of
language” (Wittgenstein [1953], §109), and that philosophers “bring
words back from their metaphysical to their everyday use” (Wittgenstein
[1953], §116). However, the trouble is that “this battle can be refought
only by language” (Arendt 1975: 115); hence the need for ‘reforming’
language for the purposes of philosophical and logical inquiries, either
through devising necessary formalism, or through elucidations of mean-
ing and focus on language use.2 The influence of Frege, Russell, and
Wittgenstein resulted in the development of at least three traditions: that
of formal logic, of ordinary language philosophy, and of contemporary
linguistics, even though the three philosophers were not interested in lin-
guistics and considered language only from the perspective of logical
and philosophical inquiries. Nevertheless, within contemporary linguis-
tics, formal and formalized approaches to language analysis, inspired by
the work of Frege, Russell, and Wittgenstein (and also by the next gen-
eration of scholars, such as Carnap, Tarski, Ajdukiewicz, Quine) are par-
2
For an introductory discussion on different turns in philosophy of language, see
respective prefaces in Stalmaszczyk, ed. (2010a, 2010b, 2011).
The Legacy of Frege, Russell, and Wittgenstein. Preface 3
ticularly characteristic of categorial grammars, Montague grammar, and

generative grammar.
Studies gathered in this volume aim to show that the results of the re-
search programs advocated and developed by Frege, Russell, and Witt-
genstein are still of utmost importance, and that the three philosophers
have significantly contributed to the linguistic turn in philosophy and the
philosophical turn in the study of language, contributing at the same time
to another movement in modern philosophy, a movement which began:
[…] when Kant exchanged the structure of the world for the structure of the
mind, continued when C. I. Lewis exchanged the structure of the mind for
the structure of concepts, and that now proceeds to exchange the structure
of concepts for the structure of the several symbol systems of the sciences,
philosophy, the arts, perception, and everyday discourse. This movement is
from unique truth and a world fixed and found to a diversity of right and
even conflicting versions or worlds in the making. (Goodman 1978: x)
1. Contents of the volume
The collection brings together contributions by philosophers, logicians

and linguists, offering an interdisciplinary approach to crucial problems
in philosophy of language and contemporary linguistics, concentrating
on different aspects of influence of Frege, Russell and Wittgenstein. Al-
so the mutual connections and further implications are discussed.
Joachim Adler observes that one of the most important claims of 20th
century philosophy of language was that the meaning of any linguistic
expression is its use. Wittgenstein and ordinary language philosophers
inferred from this that philosophical entanglements can only be dissol-
ved by clarifying the misleading concepts. Of course, such conceptual
clarifications invariably refer to the present use of a word and not to any
historical account of language. Except for a few remarks in Austin’s
Plea for Excuses, etymology has never been considered as relevant for
ordinary language philosophy. Adler’s aim is to establish historical lin-
guistics as a useful instrument for conceptual clarification.
María Cerezo presents the evolution of Russell’s theory of judgement
from 1903 to 1913, and discusses the problems that his theory encoun-
tered under its three versions in The Principles of Mathematics (1903),

On the Nature of Truth and Falsehood (1910) and Theory of Knowledge
(1913). Cerezo shows the stimulating effect that these problems had on
the development of some Tractarian ideas which can be interpreted as
something of a return to early Russell in certain aspects, together with
other crucial innovations. The chapter focuses on Wittgenstein’s revision
of the notion of logical form and its symbolic capacity, and on his truth-
conditional theory of sense.
Paweł Grabarczyk is concerned with visual perception and the cele-
brated case of the duck/rabbit. In Remarks on the Philosophy of Psycho-
logy Wittgenstein used ambiguous illusions to investigate the problem-
atic relation of perception and interpretation. Grabarczyk uses this prob-
lem as a starting point for developing a conceptual framework capable of
expressing problems associated with visual perception in a precise man-
ner. Throughout the chapter he explicates some of the common notions
associated with perception such as “to look at”, “to have an impression
of…”, “to react as if one had an impression of…”, “to convince oneself
that what one sees is…”. His principal aim is to create a precise and un-
equivocal conceptual framework capable of expressing problems and
solutions connected with the phenomenon of visual perception.
Arkadiusz Gut observes that Frege’s view concerning the thought-
language relationship contains inner tensions. They arise from the fact
that in Frege’s writings two inconsistent statements can be found, name-
ly that the structure of sentence serves as the structure of thought, and
that two structurally different sentences can express one and the same
thought. Gut starts with Dummett’s three partial claims leading to the
conclusion that Frege supports the hypothesis of the priority of language
over thought. Next, he shows that the three statements lead to a general
rule which, according to Dummett, says that for Frege and all the later
analytic philosophers the characteristics of a thought may be obtained
through a philosophical characteristics of a sentence (or more broadly –
of language). Gut also provides a set of arguments showing that the view
suggested by Dummett encountered many lines of criticism, and demon-
strates that Frege’s project advocates rather the thesis that two sentences
with different predicative structure can express one and the same
thought.
Thomas J. Hughes defends the thesis that definite descriptions should

receive an attributive semantics only in those instances where they fall
under the scope of certain semantic operators. The ambiguity defended
is not the one noted by Donnellan and Kripke, as it will not recognize an
ambiguity in one and the same proposition. In the approach advocated
by Hughes, each and every proposition receives a single semantics for
every occasion of use, and this semantics will be informed by the gram-
matical configuration of the expressions. Hughes combines semantic in-
vestigation with exploration of contemporary generative grammar.
Carl Humphries devotes his chapter to proceduralism and onto-
logico-historical understanding in the philosophy of language. Since
Frege, analytical philosophers have mostly construed language procedu-
ralistically, treating reference and assertion as largely uniform proce-
dures for re-identifying entities and proposing states of affairs as true.
Their conviction that these procedures make sense as such typically pre-
supposes a broadly Kantian intuition that they reflect some larger self-
validating normative sphere. This, according to Humphries, faces two
objections: that it misconstrues the role of ahistorical, ontologically sig-
nificant commitments/contexts, and that it ignores cases involving radi-
cally historical understanding. Each objection captures something, but it
seems that one cannot embrace both on pain of inconsistency, as they
entail conflicting readings of the existential quantifier. A third position,
explored by Heidegger and the ‘third’ Wittgenstein, rejects the disjunc-
tion between ahistorical-ontological and historical-contingent forms of
commitment and context, thus avoiding having to choose between these
two readings. This, however, involves a quietistic stance regarding the
distinguishability of ahistorical-ontological and historical-contingent
forms of commitment and context, which is sometimes counterintuitive.
Humphries sketches a possible alternative, involving the notion of on-
tologico-historical understanding.
Gary Kemp focuses on Quine’s criticisms of semantics. Kemp
demonstrates that Quine’s interest in semantics was subservient to his
epistemological agenda, and his overarching aim was to identify the real
presuppositions, and to sketch the main lines of a naturalistic and scien-
tific account of the whole of human knowledge. Quine had a number of
general criticisms of the discipline or science known as semantics. Kemp
tries to separate them into the interlinguistic and intralinguistic, suggest-

ing that semantics can survive the interlinguistic criticisms, but that
some of the more piecemeal intralinguistic criticisms remain.
Siu-Fan Lee discusses Russell’s theories of names. Russell had two
such theories and one theory of description. Logically proper names are
Millian names, which have only denotation but no connotation. Ordinary
names are not genuine names but disguised definite descriptions subject
to quantificational analyses. Only by asserting that ordinary names are
definite descriptions could Russell motivate his theory of description to
solve three problems for Millian names, namely, Frege’s puzzle, empty
reference and negative existentials. Whereas critics usually discuss Rus-
sell’s theories of names and his theory of description separately, Lee’s
paper takes a new perspective and presents a dilemma for the overall
project, arguing that it is hard to be a Russellian about names coherently.
The central issue is whether contextualisation is semantic or pragmatic
in nature, an issue very much alive in contemporary debates.
Gabriele M. Mras observes that Russell’s multiple theory of judge-
ment is commonly regarded as a failure. This is so because any attempt
to appeal to some entity as that in virtue of which an ascription of pro-
perties is true, inevitably invokes a regress. But Russell could have
known this all along. He was familiar with this objection from Bradley’s
work and he used a “regress argument” himself while he still was philo-
sophizing in the tradition of British idealism. Mras shows that the reason
why Russell, despite his own insistence that no ‘third thing’ could unite
the items of a sentence, ends up with a view that makes the appeal to
such a thing necessary has to do with what is commonly held against
him: Russell’s particular way to rely on the notion of structure led him
astray as a critique of Bradley.
Jaroslav Peregrin shows how the ‘linguistic turn’ of philosophy of the
twentieth century led to the overestimation of the role of logic in the pro-
cess understanding of meaning and in the consequent ‘dissolution’ of tra-
ditional philosophical problems. He stresses that the role logic can sensi-
bly play is the Wittgensteinian role of helping us build simplified models
of natural language (with all possibilities and limitations models have), not
the Carnapian role of reducing meanings, without a remainder, to logico-
mathematical constructs. Peregrin tries to shed some new light on this sit-
uation in terms of distinguishing two perspectives: the expression-as-ob-

ject perspective (looking at the relation between an expression and its
meaning as a contingent, a posteriori matter) and the expression-as-me-
dium perspective (looking at this relation as something necessary or a pri-
ori).
Ulrich Reichard discusses some aspects of Fregean ontology, in
which the most fundamental distinction is that between concepts and
objects. The famous ‘paradox of the concept horse’ has often been taken
to be devastating for Frege’s ontological distinction between objects and
concepts. Reichard argues that if we consider how the concept-object
distinction is supposed to account for the unity of linguistic meaning, it
transpires that the paradox is in fact not paradoxical.
Piotr Stalmaszczyk focuses on possible approaches to linguistic
predication inspired by the philosophy of language, and distinguishes
‘Aristotelian’ (or concatentaive) predication, and ‘Fregean’ (or func-
tional) predication. His chapter investigates the relevance of Fregean
semantics for contemporary linguistics, in particular generative gram-
mar. Though Fregean semantics is not concerned with natural language
categories, Frege’s line of reasoning may be fruitfully applied to analy-
zing predication understood as a strictly grammatical relation. Finally,
the paper offers a preliminary classification of predication types into
thematic, structural and propositional.
Piotr Szałek deals with one of the major puzzles for the Wittgenstei-
nian picture theory of language: if sense of propositions is determined by
picturing the possible state of affairs, how then can false propositions
have sense if their corresponding state of affairs does not exist? In order
to answer the question, Szałek reconstructs the main structure of the ar-
gument for the theory in the context of its relation to Bertrand Russell’s
view on judgments. The paper argues that the Wittgensteinian solution
to the problem of the false propositions relies on the notion of the intrin-
sic symbolic (syntactic) structure of the proposition in virtue of the anal-
ogy to the pictorial representation and possible configuration of its com-
ponents.
Mieszko Tałasiewicz addresses the issue of categorial grammar and
the foundations of the philosophy of language. He presents a new appro-
ach to explaining productivity of language, a feature that is crucial for
constructing a credible logic of natural language and elucidating many

key issues in the philosophy of language. The starting point of the pro-
posed approach is a combination of Fregean idea of functoriality and the
idea of bi-modal intentionality. These two ideas are dealt with in a way
inspired by Strawson, notably by his idea that categories are roles rather
than kinds of expressions, and that the logical syntax of language is to be
founded in some transcendental features of our thinking about the world.
As a result, the new approach reveals philosophical foundations of Cate-
gorial Grammar (far deeper than Ajdukiewicz ever explicitly acknow-
ledged) and shows, among other things, that Categorial Grammar should
not be considered as more or less accurate description of acceptability
judgments, which constitute the empirical base of linguistics, but rather
as a calculus of intentional structure of human cognition.
Acknowledgments
I am grateful to Mr Ryszard Rasi ski for comprehensive editorial assistance, and to

Dr Piotr Duchnowicz who prepared the final manuscript. I wish to thank Dr Rafael
Hüntelmann for professional advice and encouragement for the project, and espe-
cially warmly to Ms Olena Gainulina for the final formatting assistance.
References
Arendt, Hannah 1978. Thinking. New York: Harcourt Brace Jovanovich.

Baldwin, Thomas 2006. Philosophy of Language in the Twentieth Century. In: E.
Lepore and B. Smith (eds.), The Oxford Handbook of Philosophy of Language.
Oxford: Clarendon Press, 60-99.
Frege, Gottlob [1879] 1997. Begriffsschrift (translated by M. Beaney). In: M.
Beaney (ed.), The Frege Reader. Oxford: Blackwell Publishers, 47-78.
García-Carpintero, Manuel 2012. Editorial Introduction: History of the Philosophy
of Language. In: M. García-Carpintero and M. Kölbel (eds.), The Continuum
Companion to the Philosophy of Language. London and New York: Contin-
uum, 1-25.
Goodman, Nelson 1978. Ways of Worldmaking. Indianapolis: Hackett Publishing
Company.
Potter, Michael 2013. Frege, Russell, and Wittgenstein. In: G. Russell and D. Graff
Fara (eds.), The Routledge Companion to Philosophy of Language. New York
and London: Routledge, 852-859.
Russell, Bertrand 1940. An Inquiry into Meaning and Truth. London: George Allen
& Unwin.
Stalmaszczyk, Piotr (ed.) 2010a. Philosophy of Language and Linguistics. Volume
1: The Formal Turn. Frankfurt am Main: Ontos Verlag.
Stalmaszczyk, Piotr (ed.) 2010b. Philosophy of Language and Linguistics. Volume
2: The Philosophical Turn. Frankfurt am Main: Ontos Verlag.
Stalmaszczyk, Piotr (ed.) 2011. Turning Points in the Philosophy of Language and
Linguistics (Łód Studies in Language 21). Frankfurt am Main: Peter Lang.
Wittgenstein, Ludwig [1922] 1995. Tractatus Logico-Philosophicus (translated by
D. F. Pears and B. F. McGuiness). London and New York: Routledge.
Wittgenstein, Ludwig [1953] 2001. Philosophical Investigations, 3rd ed. (translated
by G. E. M. Anscombe). Oxford: Oxford University Press.
Joachim Adler
University of Zurich
[email protected]
Mapping the Ancient City: Historical

Linguistics and Conceptual Clarification
Abstract: It was one of the most important claims of 20th century philosophy of
language that the meaning of any linguistic expression is its use. Wittgenstein and
ordinary language philosophers inferred from this that philosophical entanglements
can only be dissolved by clarifying the misleading concepts. In order to do this, one
has to re-collect the ordinary use of the concepts in question. Of course, such con-
ceptual clarifications invariably refer to the present use of a word and not to any
historical account of language. Except for a few remarks in Austin’s Plea for Ex-
cuses, etymology has never been considered as relevant for ordinary language phi-
losophy. My aim in this paper is to establish historical linguistics as a useful in-
strument for conceptual clarification. After having outlined my notion of the meth-
od of conceptual clarification, I shall first examine why historical linguistics has
been constantly neglected in this very method of analytical philosophy. I will then
put forward two reasons why integrating a diachronic perspective on language can
advance the pursuit for linguistic and philosophical clarity. Finally, my conclusion
will show that diachronically enriched conceptual clarification does not lead philo-
sophy into mere empirical sciences.
Keywords: Wittgenstein, conceptual clarification, conceptual analysis, diachronic

historical linguistics, semantics, ordinary language philosophy
Our language can be regarded as an ancient city: a maze of lit-

tle streets and squares, of old and new houses, of houses with
extensions from various periods, and all this surrounded by a
multitude of new suburbs with straight and regular streets and
uniform houses.
Wittgenstein, Philosophical Investigations, §18
12 Joachim Adler
0. Introduction
Due to the famous linguistic turn, philosophy finally took notice of lan-
guage. More than ever before, so-called linguistic philosophers now em-
phasised the importance of language. They strived to make progress in
their philosophical pursuits by concentrating on the linguistic form in
which it is couched. In the first half of the 20th century, linguistic philo-
sophy split into ideal language philosophy on the one side and ordinary
language philosophy1 on the other. And although Ludwig Wittgenstein
(together with his followers at Cambridge) is not to be counted as a pro-
ponent of either of the two, he was certainly much closer to the latter.
They shared important points, most notably in methodology. Contrary to
the ideal language movement, philosophy should not try to substitute or-
dinary language by some sort of logical calculus, but rather elaborate the
subtleties in our everyday parlance by examining and clarifying the in-
volved concepts (in the next section of this essay, I will sketch the meth-
od in more detail).
The heydays of ordinary language philosophy (as well as those of ide-
al language philosophy) may well be over. What has lasted to this very
day though is its method of conceptual clarification.2 In the last few
years, one could even speak of a renaissance, as hotly debated topics
(especially in the philosophy of mind) have seen promising attempts to
progress by clarifying the concepts in this manner. Apart from such good
news, this comeback has also reawakened former doubts about this
method. Since its early dawning, linguistic philosophy has repeatedly
1
I use the label ordinary language philosophy for the philosophical tradition
which goes back to Oxford in the 1960s and was championed by scholars like
John Austin, Gilbert Ryle and Peter Strawson. The important differences be-
tween their and Wittgenstein’s conception of philosophy notwithstanding, I will
neglect the differences in the following. For further reading, see Hacker (1996:
162-182) and Glock (2008a: 34ff).
2
The term conceptual analysis seems to be slightly more common. Despite this, I
shall refer to the method as conceptual clarification. While both terms are meta-
phors and hence carry the risk of confusion, I find the latter less captious. As
Ryle pointed out, philosophical problems cannot be analysed like chemicals. It is
better to compare the task of the philosopher with that of a cartographer than that
of a chemist or detective (Ryle 1957: 385).
Mapping the Ancient City: Historical Linguistics and Conceptual… 13
been attacked in various ways. Most of all, the objections focussed on

the allegedly intuitive basis of conceptual clarification and its specula-
tions in a lay quasi-linguistic fashion (see, for instance, Mates 1958 and
Gellner 1959 for two early criticisms). Quite a few commentators doub-
ted whether the grand questions of philosophy could really be answered
by observing the ordinary man’s speech. But even if ordinary language
could actually furnish the solution for philosophical entanglements, what
qualifies a philosopher to state how ordinary language is used? On what
basis does he pick out the concepts which are to be examined? And how
does he know whether his grasp of a certain concept is common and or-
dinary? Unfortunately, these reservations were apparently vindicated by
two of the most famous proponents of ordinary language philosophy.
Ryle and Austin independently from each other examined the use of vol-
untarily. But alas, their respective notions of the concept diverged es-
sentially. The mocking thus was foreseeable:
If agreement about usage cannot be reached within so restricted a sample as

the class of Oxford Professors of Philosophy, what are the prospects when
the sample is enlarged? (Mates 1958: 165)
Following this line, Daniel Dennett (2007: 82f) recently called the pro-
cedure of conceptual clarification “naïve auto-anthropology”. Observing
linguistic actions of a social group, he claims, would pass as anthropo-
logy. But to take only into account one’s own linguistic actions without
even assessing them critically, is, in his view, just naïve. Others (e.g.
New 1966: 380f) have claimed that fundamental features of language
such as changes of meaning are constantly neglected and that, in this
sense, the system of language is too dynamic to deduce any philosophi-
cal conclusions from it. At least, if one refers to past philosophers, one
has to be aware of the possibility that many a concept was used differ-
ently at that time. This danger of conceptual anachronism leaves us with
two alternatives: either one discards the history of philosophy in total, or
one has to discriminate all the different uses of a singular expression, al-
beit being formally identical (examples will follow below).
14 Joachim Adler
Thus, it seems that conceptual clarification would eventually turn into

a Sisyphean challenge.3 To be sure, both charges have a point. Since phi-
losophy is often conceived as a discipline which is only concerned with
the a priori, methodological autism is spotted here and there. In matters
of linguistic philosophy, however, such a restriction would be blatantly
unproductive. If language is seriously to be taken into account, one must
not ignore the discipline of which language is the object of study.
Among the different concerns about the method of conceptual clarifi-
cation, I will answer only two of them by introducing historical linguis-
tics as a tool for conceptual clarifications. The integration of a dia-
chronic perspective on language will prove to be an appropriate way to
enrich this method with linguistically coherent devices, so that the ob-
jection of lay linguistic methods can be repudiated. Moreover, this tool
enables us to deal with language change adequately, which will, as will
be shown, eventually turn out as grist to the mill for ordinary language
philosophy. This also helps us also refer to past philosophers considering
their use of the concepts in question. Before answering these objections
however, I will also examine why it seems fairly reasonable that propo-
nents of the method of conceptual clarification have hardly ever consid-
ered a diachronic perspective on language. I will then put forward two
reasons why integrating a diachronic perspective on language can nev-
ertheless advance the pursuit for linguistic and philosophical clarity. At
that point, I will have to forestall two misunderstandings which are quite
common in matters of language change in philosophy of language. Fi-
nally, my conclusion will show that diachronically enriched conceptual
clarification does not reduce philosophy to empirical linguistics.
Apparently, there are two crucial labels in this essay: historical lin-
guistics and conceptual clarification. What I mean by the latter shall be
discussed in the paragraph below. Historical or diachronic linguistics
seems to be a rather diffuse etiquette even for linguists,4 let alone philo-
sophers. The term diachronic linguistics was coined by Ferdinand de
3
For further discussion of the relevance of the history of philosophy, see Glock
(2008b).
4
Even though diachronic might seem more precise and less biased than historical,
the linguistic discipline I refer to is usually labelled as historical linguistics.
Speaking of this specific discipline, I shall thus follow this convention.
Saussure, and the locus classicus for its definition is to be found in the
Course of General Linguistics:
What diachronic linguistics studies is not relations between coexisting

terms of a language-state but relations between successive terms that are
substituted for each other in time. (Saussure 1959: 140)
Diachronic linguistics is concerned with the phenomenon of language

change. Every part of language changes: phonetically, as sounds change;
grammatically, as inflexional paradigms evolve or even collapse; syntac-
tically, as word orders vary; and, of course, semantically, as meanings
change. And since this is perhaps the most striking feature of language
change and easily observable also for laypersons within one lifespan,
historical linguistics is often reduced to historical semantics. Indeed, se-
mantic change is the main aspect I intend to focus on in the following.
For it is the meaning of philosophically contentious expressions that
conceptual clarification seeks to elucidate. However, if language change
is to be considered in the pursuit of dissolving philosophical problems,
one has to incorporate all other aspects of language as well. In fact, se-
mantic change cannot be isolated from phonetic changes. For the history
of a word and its meaning can only be traced back if one is able to re-
construct its phonetic shape through the different stages of a language.
1. Conceptual clarification as a method
At the beginning of the 20th century, language took centre stage in philo-
sophy. Although there have been and still are various disagreements on
what role exactly language should play, it has never lost its importance
since then. Arguably the most sustained and uncompromising turn to-
wards language – namely, as the focus of every philosophical venture –
was taken by Wittgenstein and, similarly, by ordinary language philo-
sophy. They agreed in the basic idea that philosophical problems arise
from linguistic delusions. And these arise when we become puzzled by
words used in an unordinary way so that we do not understand what they
mean in a certain context. In order to dissolve such conceptual entan-
16 Joachim Adler
glements, one has to clarify the misleading concepts by taking them back
to their ordinary use.
In Peter Hacker’s criticism of the neuroscientific terminology (see
Bennett and Hacker 2003), we find several examples for this method.
For instance, if we become puzzled by questions like “Can conscious-
ness be located in the brain?”, “Is consciousness essentially private?” or
“Does the phenomenon of consciousness resist naturalization?”, we first
have to work out what those question actually mean – and, relatedly,
what kind of answer there is to be given. After all, consciousness is not
just a technical term that has been invented and concisely defined, nor is
it only used in uncontentious situations without any philosophical rele-
vance. Conversely, while it is a word of ordinary language, it is also the
hot potato in the dispute between neuroscientists and philosophers. So
which context should be the reliable one in our search of the legitimate
use of consciousness? Since we do not seem to stumble over its defini-
tion in our everyday parlance, it is the ordinary context we should first
look to. We say He’s conscious when somebody has passed out and is
now opening her eyes again, or I’m perfectly conscious of the risks I’m
taking which means that I am well aware of the risks, that I am taking
the risks into account. We can hardly find any kind of substance which
these two sentences refer to, nor a certain area in our brain. Eventually,
we shall find that certain neurophilosophical claims do not use the term
as it is supposed to, and this is what often results even in plain nonsense.
We do not use consciousness as a name for some Cartesian inner theatre.
Thus, there seems to be something awry in the quest for consciousness.5
To sum up, the method discussed here is as follows. One collects the
problematic expressions in a given context, examines their use in ordi-
nary language, compares this use with the given context and delineates
the bounds of sense in respect of that very language game.6 Conceptual
clarification strives for an overview, for clarity (Übersichtlichkeit, as the
original reads, see Philosophical Investigations §122). Certainly, in our
5
I cannot put forward a sustained argument here, but see Bennett and Hacker
(2003: 237ff) and also Kenny (2009: 250-263).
6
Wittgenstein’s term “language game” (PI §23) is, for reasons of simplicity, to be
understood here as a certain context, or, in Austin’s diction, as an “area (of dis-
course)”. See Urmson (1967: 233).
everyday speech, competent speakers hardly become confused by lan-

guage, just as we are not lost in a city we have known since childhood,
although we might not be able to draw a map of that city. As expressed
by Wittgenstein in the quotation at the beginning, natural languages –
opposed to, say, Esperanto or the interpreted logical calculi envisaged by
‘ideal language philosophers’ – are not built only of rectangular streets
and uniform houses. In the alleys of language lurk many traps which we
are able to avoid simply because we have trained it more than anything
else in our life. Still, in our quest for philosophical insights we easily fall
into such traps or run into dead ends. In a remark from 1931, Wittgen-
stein’s metaphor runs as follows:
Language sets everyone the same traps; it is an immense network of easily

accessible wrong turnings. And so we watch one man after another walking
down the same paths and we know in advance where he will branch off,
where walk straight on without noticing the side turning etc. etc. What
I have to do then is erect signposts at all the junctions where there are
wrong turnings so as to help people past the danger points. (Wittgenstein
1980: 18e)
In the following, I do not wish to dwell on a defence for this method, as

much as it would deserve one. Rather, I shall explicate in which way it
can be enriched by historical linguistics. Armed with the vigour of lin-
guistic data, one may find the conclusions of conceptual clarification
even more compelling.
2. Why the past might not matter
At first sight, to combine the idea of conceptual clarification with a his-

torical perspective on language seems highly problematic.7 For when
Wittgenstein famously claimed that “the meaning of a word is its use in
the language” (PI §43), one clearly has to understand use here as present
use. The use of a word changes constantly; most words are not used the
7
This impression may partly derive from a common prejudice against analytic
philosophy as being a- or even anti-historical. See Glock (2008b) for a detailed
discussion of the mismatch between analytic philosophy and history.
18 Joachim Adler
way they were several hundred years ago. Even middle-aged people
know of some expressions that are used differently nowadays than in
their youth. But clearly, one cannot arrive at the actual meaning by ex-
ploring etymology. For instance, the fact that etymology derives from
Greek étymos ‘true’ does not imply that etymology is about elucidating
the true meaning of a word. Today, etymology is the discipline which
describes the origins of a word, without any normative power. Despite
its name, etymology has shed the idea of a true meaning.
Still, academia apart, etymology is consistently exploited for language
policies, in some quarters language change is still conceived as some
sort of decay. But how – and why – should one want to turn the clock
back in matters of meaning, anyway? Consider someone prompting,
“Meet me at noon”. Another might know that noon derives from Latin
nona (hora), originally ‘the ninth hour from sunrise’ (see ODEE 614),
but that would, of course, hardly justify his showing up only at 3 p.m.
Regardless of its meaning until the 14th century, noon is used nowadays
only to refer to 12 o’clock in the day; one cannot use it for any other
hour without being misunderstood.
Admittedly, meanings do not always change as unambiguously as in
the case of noon. Former uses very often loom into the present through
secondary use or by tinging the primary. These cases, of course, are the
interesting ones for the method discussed here.
Saussure and Wittgenstein compared linguistic acts with moves in
chess.8 Both activities are guided by rules which have a constitutive
power. Chess rules are constitutive in two ways. On the one hand, they
constitute the game that is played as chess. You only play chess if you
move your pieces according to the rules of chess. And indeed, these
rules have evolved through the centuries just as linguistic ones, but there
is only one set of rules that you have to stick to if you play chess these
days. If one would bring dice into the game, one could not be said to
play chess anymore, although it was played with dice in the beginning.
On the other hand, chess rules also constitute the individual pieces. It is
not the shape that defines them, but rather the rules that confine their
8
Saussure (1959: 22); PI § 31. Further interesting parallels between Saussure and
Wittgenstein can be found in Harris (1988).
movements: the piece that is allowed to move only horizontally or verti-

cally through any number of unoccupied squares is called the rook.
These rules build a synchronic, ahistorical system, which has to be ac-
cepted as a whole when one wants to engage in a chess-game.
Similarly, in order to take part in a language community, one has to
follow the rules that build the synchronic, ahistorical system of every
language.9 So, of course, if one wants to understand the meaning of a
word, one simply has to look at its present use. Thus it seems that, how-
ever perplexing a philosophical problem may be, the former meanings of
the expressions involved simply do not play any role both for the reasons
and the solution of our problem. Why then should we bother about his-
torical linguistics at all? I want to present two reasons why we should.
3. First reason: clarifying the starting point
The first reason leads us back to the initial phase of the clarification pro-
cess and also to a weighty objection against this method. In this section,
I will answer the demurs about the lay linguistic attitude which has been
rightly detected in several works of ordinary language philosophers.
When we are to describe the relations and dependencies between con-
cepts in a certain area of discourse we have to struggle mostly with the
synchronic disorder in our everyday language. The sheer number of
near-synonyms, derivations and loanwords makes it very difficult to col-
lect all the relevant expressions, let alone to examine their subtle dif-
ferences and nuances. Taking the diachronic perspective into account,
we can understand the reasons behind those oddities much more easily
and have the area of discourse arranged more appropriately.
Consider, for example, knowledge. The conceptual entanglements are
well known. Philosophers have wrestled with the problem of an ade-
quate definition for ages; dozens of books and presumably hundreds of
articles have been published about it. Several linguistic philosophers
(Wittgenstein, Ryle, and Austin) dealt with it too. The heterogeneity of
9
For reasons of space, I will not discuss the well-known qualms whether or not
language is a rule-governed activity. As a defence for the position that is taken
here, see Glock (2008c).
20 Joachim Adler
the expressions involved is mentioned quite regularly – which actually

means, grasping all the relevant concepts, confining the area of dis-
course, turns out to be at least one of the fundamental parts of the prob-
lem. For, as many a failed definition shows, knowledge is related to con-
cepts like consciousness, conscience, wise or witty, which are likely to
be overlooked but may yet reveal insights into the language game in
which knowledge is embedded (see, for example, Bennett and Hacker
2003: 148ff, and Hanfling 1985). Not least, because of these ties, the
concept of knowledge has turned out to be too ramified for any concise
definition. Rather, the different uses of knowledge are a decent example
of a Wittgensteinian family resemblance: “a complicated network of
similarities overlapping and criss-crossing” (PI §66).10 But how to cope
with all these strands? Synchronically, the formal variety of the afore-
mentioned expressions entails the danger of limiting the investigation to
knowledge and to know. However, the diachronic perspective readily un-
covers the relations to conscience and consciousness, as they trace back
to Latin scire ‘to know’ and con-scire ‘to share knowledge with some-
body’ or ‘to be privy with another or oneself’ (see ODEE 205f, 508f). A
few etymological insights further, one has to recognise that any clarifica-
tion of knowledge must not neglect consciousness, and vice versa. While
dealing with consciousness, Bennett and Hacker state:
Transitive consciousness lies at the confluence of the concepts of

knowledge, realization (i.e. one specific form that acquisition may take),
receptivity (as opposed to achievement) of knowledge, and attention caught
and held, or given. (Bennett and Hacker 2003: 253)
And as complex or even desperate as this statement may sound, the in-
terrelated histories of these expressions not only prove it to be correct,
but explain also how this came about.
Behind every confusing arrangement of expressions lies a bundle of
word histories that are well capable of being explained clearly. Of
course, there are still plenty of etymologies that linguists have been una-
ble to discover. But due to the ongoing progress of linguistic research,
10
Ernst (2002) does not share this view in his monograph about knowledge. Refer-
ring to Hanfling (1985), he basically distinguishes two major uses.
more and more expressions can be traced back to their roots. Moreover,
the still obscure prehistoric stages with no written accounts are of low
interest for the issue discussed here. For since western philosophy
emerged centuries after the invention of writing, these stages could not
have had an immediate impact on philosophical concepts. And as for any
indirect traces, there is just no way of telling. By contrast, the more rele-
vant stages, covering the last few centuries, provide us with a vast num-
ber of written records, at least in the case of the major European lan-
guages.
What is more, etymological pathways cross language borders inces-
santly. Tracing back word histories across different languages does not
just affect conceptual clarification in one language, but in many. Despite
their different developments, genetically related languages like English
and German still share a considerable amount of cognate words. Certain-
ly, many of them are so-called false friends, which means that, in spite
of their similar form, the two related words have reached different mean-
ings in the respective languages. Still, how a word is used in German, for
instance, is often highly informative for the clarification of its cognate in
English. For instance, some former meaning may have survived in one
language while it has been abandoned in the other, but it may still be rel-
evant for the understanding of the specific development of use. Apart
from cognates and loanwords, languages with similar cultural back-
grounds can also simply serve as objects of comparison (Wittgenstein
occasionally mentioned the use of a Vergleichsobjekt, see, for instance,
PI §131). For what is formally muddled in one language can be much
clearer in another, whereas the conceptual framework beneath is always
the same. The aforementioned relation between knowledge and cons-
ciousness may well serve as an example here. Since the German words
Gewissen ‘conscience’ and Bewusstsein ‘consciousness’ were not, as it
is the case for English, borrowed from Latin, the close connection to
wissen ‘to know’ is even formally evident. And this, of course, reinfor-
ces the supposed relevance of the relation between these two concepts.
Language-crossing kinship thus often reveals interesting insights too,
and is the object of so-called external reconstruction, which is a standard
procedure of historical linguistics in order to reconstruct poorly docu-
mented language stages or to track nebulous etymologies.
22 Joachim Adler
4. Second reason: understanding language change
The second point shows another advantage of a diachronically enriched

clarification, but also rules out a misconception of language change
which I shall mention first. Critics have occasionally argued that the
method of conceptual clarification must fail due to semantic change. The
attack runs as follows. The ordinary language philosophers’ pretence of
distinguishing uses of expressions that make sense from those that do
not turns out to be mere linguistic conservatism. Meanings change con-
stantly and therefore it is principally impossible to fix the rules for the
use of an expression. This strand of objection was taken up by Dennett
(2007) in his reply to Bennett and Hacker (2003), even though they had
already discussed it in this very book. Dennett claims that linguists
would hesitate in stating any grammatical rules as these would virtually
change the moment they are written down. And since continuous change
is conceived as an essential feature of any living language, linguistic
conservatism would either be in vain or even restrict the required devel-
opment of a certain vocabulary.
Dennett may be right in his statement that at least some ordinary lan-
guage philosophers have undeniably been inattentive towards language
change (but Hacker is certainly not to be included).11 However, he is
wrong in his view both of language change as some sort of chaotic de-
structive force and of linguistics throwing in the towel in the face of lin-
guistic change. Dennett claims that, as an example, linguists would re-
frain from calling The cat climbed down the tree an abuse of the verb to
climb (Dennett 2007: 84). Given that it originally expressed an ascend-
ing motion, should the combination with down count as a mistake or as a
semantic change? Pace Dennett, linguists do not have any difficulties in
describing the phenomenon (e.g., see Wierzbicka 1990: 363ff).
The term to climb down is not be taken as a contradictio in adjecto
making linguists shudder; it is simply a nice example of a semantic ex-
tension. To climb is a cognate of German (er)klimmen, which is indeed
11
It has to be noted that in his famous and programmatic Plea for Excuses, Austin
explicitly took etymology into account as part of the method he envisaged (see
Austin 1956). In spite of that, only a few of his disciples followed his example.
only used to express an ascent. And without the adverb down, the Eng-
lish verb is restricted to the very same use. The cat climbed the tree
means that the cat is now up in the tree, not that it just descended it. In
combination with down, the meaning of climb has been generalised from
‘move horizontally upwards (by hands and feet)’ to ‘move horizontally
(by hands and feet)’. What is more, contrary to Dennett’s contention,
this is not an example of recent semantic change; the expression to climb
down is documented since the early 14th century, and the evidence is
readily accessible in the OED.
To be sure, much more dramatic meaning changes are legion. Con-
sider, as a single example, English nice, deriving from Latin nescius ‘ig-
norant’; one seemingly has to conclude that semantic change is a com-
pletely arbitrary process. But what would that mean for ordinary lan-
guage philosophy? Irritating expressions like My brain is conscious –
notably the bone of contention between Hacker and Dennett – could
simply be justified by referring to spontaneous meaning change. And
this would make it possible to ascribe the term conscious not only to
persons, but also to brains, computers and anything you like.
Before giving in, however, the “arbitrariness” of this process is worth
looking at more closely. For if one does not just settle for shuffling the
starting and ending points, but rather aims at a reconstruction of the his-
tory step by step, the emerging picture is that of a much more compre-
hensible process. The Latin nescius evolved into Catalan neci ‘ignorant’
which was incorporated into Middle English as nyce ‘foolish’. By the
15th century, the meaning had become ‘coy, shy’ and via ‘fastidious,
dainty’; the present meaning of ‘agreeable, delightful’ was eventually
reached in the 18th century (see ODEE 609). It is the instruments of his-
torical linguistics that provide an insight into the black box of language
change. Seemingly arbitrary developments like the aforementioned re-
sult from the alignment of well-known processes such as specialisation
of meaning, generalisation, amelioration, and so on. Impressions of ar-
bitrariness and unpredictability often result from the scope of the dia-
chronic perspective: over the centuries, meanings may indeed take im-
pressive and surprising leaps. But if we dwell on the details of a conti-
nuous process, there is no mysterious lottery anymore, no etymological
miracle. At the very most, there are unexplainable meaning shifts be-
24 Joachim Adler
cause of word histories that have not yet been uncovered. But semantic
change itself is not a quodlibet. With the exception of explicit stipula-
tions, meanings do not leap, they move slowly. For semantic change is
always constricted by synchronic comprehensibility: if the majority is
unable to make sense of a new use, it cannot be established.
At this point, I shall try to forestall two misunderstandings. First, lan-
guage change is for the most part an unpredictable process – the achie-
vements of post festum diachronic linguistics concerning the explanation
of changes in the past notwithstanding. However, while we cannot tell
how the meaning of a certain word may evolve, it is quite reasonable to
predict what shift in meaning an expression will not undergo, consi-
dering its development and present use. Secondly, I am not suggesting
that language change is or should be subject to any normative claims.
Linguistics is not a normative discipline. Intentionally transferring a
word to an unfamiliar context is by no means to be condemned; it is evi-
dence of a living language. Even more to this point, metaphors, as much
as loanwords, are the language’s fountain of youth. If scholars manipu-
late ordinary language though, we have to be cautious. Whatever hap-
pens to language within laboratories and philosophy departments may be
jolly good, whether their jargon contains technical terms, everyday ex-
pressions or codes. Still, even though scientists claim to understand each
other’s terminology, many an example shows that, due to conceptual en-
tanglements, they seem to draw the wrong conclusions from their data.
For scientific language, as elaborated as it may be, is impregnated with
ordinary language, and so are the questions about, say, consciousness.
The conceptual confusions vitiate the speaking and thinking of experts.
However, it is even more harmful if this idiosyncratic lingo should con-
vey scientific findings to laypeople. For whenever scholars want to make
their results and findings available to the public, they have to bridge a
gap of knowledge, and this is not to be done by introducing linguistic
innovations without explaining them. Consciousness may be a shibbo-
leth for whatever neuroscientific theory, but if neuroscientists talk in the
same way in popular scientific articles and books, it should be made
clear that this cannot possibly mean the same as it does in ordinary lan-
guage.12
Considering the historical depth of language must not lead us to any
sort of linguistic inertia. But, as so often, history provides us with a criti-
cal reminder of recent trends. And so we might then reject certain me-
tonymies like “thinking brains”, because they do not fill a gap in our
everyday language, nor do they explain anything. Rather, they cause
confusion.
5. Conclusion: philosophy as an appendix of empirical linguistics?
In the previous two sections, I tried to rebut two objections against con-
ceptual clarification by showing the advantages of a diachronic perspec-
tive. I argued that intuitive reasoning can and should be replaced by
proper linguistic assessments, including an adequate historical perspec-
tive on the concepts which are to be clarified. The second objection con-
cerning language change turned out to be a strong argument for the inte-
gration of historical linguistics, provided that we understand the princi-
ples of meaning change. Of course, in order to profit from these linguis-
tic tools, substantial empirical data is indispensable.
Ever since Russell’s early criticism, ordinary language philosophy has
been suspected of dissolving philosophy into empirical linguistics. Fur-
thermore, others have claimed that if philosophy has to bite the bullet of
turning empirical it should at least abstain from armchair reasoning and
engage in proper empirical linguistics. In this vein, hotly debated ex-
perimental philosophy has tried to prove or disprove – if not to improve
– the former results of ordinary language philosophy (e.g., see Knobe
and Nichols 2008).13
12
It is well noteworthy that the communication between science and the public has
increasingly attracted attention. The journal Public Understanding of Science is
just one outcome of this movement, several publications and conferences have
been devoted to this hotly debated topic.
13
Sandis (2008) has set out vividly the differences between ordinary language phi-
losophy and experimental philosophy and why conceptual clarification cannot be
replaced by polls.
26 Joachim Adler
The fears are baseless. Etymological investigations can no more repla-

ce philosophy than sociolinguistic surveys can. Empirical data typically
cause frowns in philosophy conceived as referring only to the a priori.
But just as in epistemology or philosophy of biology, philosophy of lan-
guage also must not ignore the empirical basis. Of course, data cannot
replace any kind of conceptual clarification or philosophical argument.
But it is probable that they help us set the agenda for a certain philoso-
phical pursuit. With regard to conceptual clarification, linguistic data
may circumscribe the scope of the area which is to be examined. Diffe-
rent linguistic tools, of which I consider the historical ones only as some
among others, provide us with data or premises for philosophical delibe-
ration. When Austin habitually read through the whole dictionary (see
Urmson 1967: 234), he collected data for his premises. Thus, what we
find in ordinary language, be it that of today or of yesterday, be it by
armchair reasoning or by philology, “is not the last word”, as Austin fa-
mously declared, but “it is the first word” (1970: 133).14
References
Austin, John L. 1956. A Plea for Excuses. Reprinted in 1970 Philosophical Papers,
Oxford: Clarendon, 123-152.
Bennett, Maxwell and Peter M. S. Hacker 2003. Philosophical Foundations of Neu-
roscience. Malden: Blackwell.
Bennett, Maxwell, Daniel Dennett, Peter Hacker and John Searle (eds.) 2007. Neu-
roscience and Philosophy: Brain, Mind and Language. New York: Columbia
University Press.
Dennett, Daniel C. 2007. Philosophy as Naïve Anthropology. In: M. Bennett et al.
(eds.), 73-95.
Ernst, Gerhard 2002. Das Problem des Wissens. Paderborn: Mentis.
Gellner, Ernest 1959/2005. Words and Things. London: Routledge.
Glock, Hans-Johann 2008a. What Is Analytic Philosophy? Cambridge: Cambridge
University Press.
Glock, Hans-Johann 2008b. Analytic Philosophy and History: a Mismatch? Mind. A
Quarterly Review of Philosophy 117, 867-897.
14
I wish to thank Hanjo Glock, Stefan Riegelnik and Peter Hacker for comments
and discussions on several topics that appear in this essay.
Glock, Hans-Johann 2008c. Meaning; Rules and Conventions. In: D. Levy and E.
Zamuner (eds.), 156-178.
Glock , Hans-Johann and John Hyman (eds.) 2009. Wittgenstein and Analytic Phi-
losophy. Essays for P. M. S. Hacker. Oxford: Oxford University Press.
Hacker, Peter M. S. 1996. Wittgenstein’s Place in Twentieth Century Analytic Phi-
losophy. Oxford: Blackwell.
Hanfling, Oswald 1985. A Situational Account of Knowledge. The Monist 68, 40-
56.
Harris, Roy 1988. Language, Saussure and Wittgenstein. London and New York:
Routledge.
Kenny, Anthony 2009. Cognitive Scientism. In H.-J. Glock and J. Hyman (eds.),
250-262.
Knobe, Joshua and Shaun Nichols (eds.) 2008. Experimental Philosophy. Oxford:
Oxford University Press.
Levy, David and Edoardo Zamuner (eds.) 2008. Wittgenstein’s Enduring Argu-
ments. London: Routledge.
Mates, Benson 1958. On the Verification of Statements about Ordinary Language.
Inquiry 1: 1, 161-171.
New, Christopher 1966. A Plea for Linguistics. Mind, New Series, 75 (299), 368-
384.
Onions, Charles T. (ed.) 1966. The Oxford Dictionary of English Etymology. Ox-
ford: Clarendon Press. [ODEE]
Rorty, Richard (ed.) 1967. The Philosophical Turn. Recent Essays in Philosophical
Method. Chicago and London: The University of Chicago Press.
Ryle, Gilbert 1957. The Theory of Meaning. Reprinted in 2009, Collected Papers,
Vol. 2. London and New York: Routledge, 363-385.
Sandis, Constantine 2010. The Experimental Turn and Ordinary Language. Essays
in Philosophy 11, 181-196.
de Saussure, Ferdinand 1959. General Course in Linguistics. Translated from the
French by Wade Baskin. New York: Philosophical Library.
Tsohatzidis, Savas (ed.) 1990. Meanings and Prototypes: Studies in Linguistic Cat-
egorization. London: Routledge & Kegan Paul.
Urmson, James O. 1967. J. L. Austin. In: R. Rorty (ed.), 232-238.
Wierzbicka, Anna 1990. ‘Prototypes Save’: On the Uses and Abuses of the Notion
of ‘Prototypes’ in Linguistics and Related Fields. In S. Tsohatzidis (ed.), 347-
367.
Wittgenstein, Ludwig 1980. Culture and Value. Translated by Peter Winch. Oxford:
Blackwell.
Wittgenstein, Ludwig 2009. Philosophical Investigations. Revised 4th edition by P.
M. S. Hacker and J. Schulte. Malden: Blackwell.
María Cerezo
University of Murcia
[email protected]
Russell and Wittgenstein on Proposition,

Judgement, and Truth*
Abstract: In this paper I present the evolution of Bertrand Russell’s theory of
judgement from 1903 to 1913 and discuss the problems that his theory encountered
under its three versions in The Principles of Mathematics (1903), On the Nature of
Truth and Falsehood (1910) and Theory of Knowledge (1913). I intend to show the
stimulating effect that these problems had on the development of some Tractarian
ideas which can be interpreted as something of a return to early Russell in certain
aspects, together with other crucial innovations. In particular, I will focus on Witt-
genstein’s revision of the notion of logical form and how it symbolizes and on his
truth-conditional theory of sense.
Keywords: Russell, Wittgenstein, judgement, truth, falsehood, sense, logical form,

propositional unity, bipolarity, assertion
0. Introduction
Recent work on the objection that Wittgenstein raised to Russell’s theory

of judgement in the unfinished version of Theory of Knowledge (1913,
henceforth TK) has revived discussion on the particular problems that his
theory had. The received view was that the problem derived from its
conflict with the Principia’s theory of types (Griffin 1985, 1985-1986),
but subsequent work by Hochberg (1996), Stevens (2004, 2006), Hanks
(2007), Carey (2007), Landini (2007), Pincock (2008) and Connelly
(2011-2012) has brought new hypotheses into the picture. Hanks looks
*
Special thanks are due to the late Angel d’Ors, with whom I discussed some of
the issues of the paper during many years. I also want to thank for the funds re-
ceived from the Spanish Ministry of Science and Innovation and the Ministry of
Economy and Competivity [FFI2009-13687-C02-01/FISO].
30 María Cerezo
for a difficulty internal to the Russellian theory of judgement itself, ra-

ther than one that derives from its conflict with other Russellian theories.
He claims that Wittgenstein’s objection was a version of the classic
problem of the unity of a proposition. Stevens also thinks that Russell’s
problem was related to this unity and the need to account for the right
constituents that can be combined to form a proposition. This point is
also stressed by Hochberg when he offers his interpretation of Wittgen-
stein’s criticism, but Hochberg also insists on the need to account for the
connection between a representing and a represented complex. Carey’s
detailed analysis of the writing and contents of TK attempts to show that
Russell abandoned it because he was not able to account for the bipo-
larity of a proposition within the framework of his theory of judgement.
This line is also followed by Pincock, who focuses on what he calls “the
correspondence problem”, that is the question about what must be the
case for a belief to be true and what must be the case for it to be false.
Landini thinks that Wittgenstein’s difficulty had to do with the presence
and role of logical forms in the analysis of judgement. Finally, Connelly
defends a reading of the objection in line with Griffin’s reading, but with
a different interpretation of the conflict internal to Russell’s proposal, a
conflict deriving from the requirement of a significance constraint on
judgements together with some facts about logical inference.
Most of these discussions approach the relation between Russell’s and
Wittgenstein’s ideas from the perspective of the paralyzing effect that
Wittgenstein’s criticism had on Russell. In adopting this point of view,
the discussion centres on determining which difficulty caused the para-
lysis. In this paper, my perspective is different. I will not focus on the
paralyzing effect of Wittgenstein’s criticism on Russell, but rather on the
stimulating effect that Russell’s difficulties had on Wittgenstein’s own
development. On the one hand, a similar perspective has perhaps been
attempted recently in Stevens approach (2006). Even though he centres
his attention on Russell’s theory, he also seeks to account for the way in
which the picture theory tried to solve the problems encountered in the
Russellian theories. However, Stevens concentrates mainly on the fea-
tures of the general picture theory (TLP 2.1-2.225), and does not pay
sufficient attention to the particular way in which that theory is applied
to propositions in the Tractatus. Hochberg’s reading (1996) has a similar
Russell and Wittgenstein on Proposition, Judgement, and Truth 31
limitation: he thinks that Wittgenstein did not provide a complete solu-

tion to the problem of correlation of representing with represented com-
plexes, since the correlation of the constituents was not sufficient to ac-
count for it. However, once one takes into account not only the general
picture theory, but its application to truth-functional propositions, new
light can be shed on the issue. On the other hand, due to their stress on
bipolarity, which is a crucial feature of the Tractarian doctrine, I do
share much of Carey’s and Pincock’s flavour, but I also pay attention to
the contribution that the Tractarian truth-functions theory makes in tack-
ling Russell’s problems.
In Section 1, I focus on the evolution of Russell’s theory of judgement
and its problems. Section 2 presents and briefly comments on the texts in
which Wittgenstein explicitly mentions and criticizes Russell’s theory.
Wittgenstein’s moves to solve these problems are displayed in Section 3
and the last section summarises the problems and moves and provides
some final textual evidence for my approach.1
1. The evolution of Russell’s theory and its problems
Russell’s clearest difficulty in developing his theory was the necessity to

account for how the sense of a proposition is independent of its truth.
This required accounting for the truth and falsehood of a proposition and
for its meaning in different ways. In general, his proposals were deve-
loped along two lines:2
Line 1: To consider truth and falsehood as properties of propositions

which are conceived as some sort of independent complex entities
whose unity should be ontologically guaranteed. This is the view in
The Principles of Mathematics (1903, henceforth PoM), and I will
refer to it as the PoM-Theory.
1
Table 1 at the end of the paper is provided to help the reader follow the presenta-
tion of Russell’s problems and Wittgenstein’s moves.
2
For a useful account of the evolution of Russell’s views about judgement, see
Candlish (1996). Stevens (2006), Hanks (2007) and Johnston (2007) also offer
clear reports of that evolution.
32 María Cerezo
Line 2: To consider truth and falsehood as properties of judgements,

so that the judging mind was in charge of unifying the objects of be-
lief. The unity of true and false judgements was then accounted for
similarly, since it was grounded in the act of judging. The existence/
non-existence of complex entities corresponding to beliefs was then a
means of accounting for truth/falsehood. This idea is realized in dif-
ferent ways in both versions of Russell’s multiple relation theory of
judgement in On the Nature of Truth and Falsehood (1910, hence-
forth OTF), which is in general terms the same in Principia Mathe-
matica (1910) and The Problems of Philosophy (1912), and in the
unpublished TK (1913). I will refer to these two views as the OTF-
Theory and TK-Theory, respectively.
It is crucial that in Line 1, propositions are primitive with respect to psy-

chological relations. Russell offers a theory of objective true and false
propositions, and explains judgement as a psychological relation of the
mind to propositions. The problems to which this theory gives rise make
Russell invert the account in Line 2: psychological acts are primitive. As
we shall see, Wittgenstein’s Tractatus can be viewed as a return to Line
1, with a more complete theory of propositions, and a revision of the
analysis of psychological relations.
PoM-Theory (1903)
There are three important features of Russell’s conception at this early

stage of development: the ontological nature of propositions and the re-
lation of language to them in terms of indication; the nature of truth and
falsehood, and finally the issue of the unity of a proposition.
Russell conceived the relation between language and world as one of
indication, that is, a direct relation between linguistic items and world
entities. Words indicate terms and sentences indicate propositions. There
are two kinds of terms: things and concepts, which are indicated by
nouns and by verbs and adjectives, respectively. Propositions are com-
plex entities, whose constituents are terms. Thus, for example, the sen-
tence (1) indicates the proposition (2) below (I will represent entities by
placing them in square brackets):
(1) Obama is younger than Bush

(2) [Obama is younger than Bush]
where (2) is composed of the following constituents: the things [Obama]

and [Bush] and the concept [being younger than]. But these constituents
can be combined in different ways to give rise to different propositions,
like the different propositions that are indicated by (1) above and (3) be-
low:
(3) Bush is younger than Obama
This is the way of combination of constituents problem (Problem 1),

which has two sides: the way in which the constituents combine (Prob-
lem 1.1) and their being actually combined (Problem 1.2).
Russell thinks that there is only one kind of relation of language and
world, indication. There is no difference in the relation between a word
and a term and the relation between a sentence and a proposition, except
for the fact that the latter is a complex entity. This fact generates the
need to account for truth and falsehood in a different way with respect to
reference and lack of reference. If the relation between names and terms,
and sentences and propositions is the same, then it seems that if a sen-
tence has meaning (if it indicates a particular complex entity in the
world), then there is no way to differentiate between the sense of a sen-
tence and its truth, and thus to account for falsehood (Problem 2: false-
hood problem).
Of course, one can account for truth and falsehood in other ways, and
that is what Russell does: there are objective true propositions and objec-
tive false propositions. The difference stems from an important extra
primitive quality that true propositions have, namely, assertion. Russell
distinguishes between psychological assertion and logical assertion. Ac-
cording to Russell, psychologically, propositions can be merely con-
sidered or thought of, or they can be actually asserted. True and false
propositions can be equally asserted in this psychological sense, but only
true propositions are logically asserted:
34 María Cerezo
True and false propositions alike are in some sense entities, and are in some
sense capable of being logical subjects; but when a proposition happens to
be true, it has a further quality, over and above that which it shares with
false propositions, and it is this further quality which is what I mean by as-
sertion in a logical as opposed to a psychological sense. (Russell, PoM:
§52)
Both true and false propositions, as entities, have being, but in the case
of the former their being is being true (that is, being asserted). The pro-
position indicated by (1) is true, its constituents are actually related by
the relation [being younger than], and this is what assertion consists in.
Assertion, discussed below, will play further roles in Russell’s proposal,
giving rise to new tensions.
Therefore, there are two different complex entities (propositions) indi-
cated by (1) and (3) above. Both have the same constituents ([Obama],
[Bush], [being younger than]), but the former is asserted and the latter
unasserted. The relation of each of these propositions to their truth or
falsehood is internal. Russellian propositions are monopolar: they are
either true or they are false, but it is not so that they can be true and they
can be false.
Logical assertion is also responsible for the unity of propositions with
that quality and it thus solves Problems 1.1 and 1.2, but only in the case
of true propositions. Due to (2) having the quality of being asserted, it is
the relation [being younger than] that actually relates the constituents.
Russell insists on the idea that logical assertion, even if it is a quality of
propositions, is not one of its constituents. If assertion were a further
constituent in charge of relating the two terms and the relation in (2),
there would arise a risk of a possible infinite regress, since a new rela-
tion would be necessary to relate such constituent (assertion) to the rest
of the constituents. The peculiar character of assertion, which is not a
further element, can join together the constituents, without running such
a risk. Conversely, the analysis of (2) into its constituents, insofar as a
proposition dissolves into its parts and stops being an asserted pro-
position, makes the relating relation into a non-relating relation, [being
younger than].
The monopolarity of propositions raises two further problems. The
first one is explicitly recognized by Russell. Since both true and false
propositions are entities, the difference between a proposition being ac-

tually true from what it would be as an entity if it were not true lies in its
being asserted when actually true. And then we have Problem 3: asser-
tion-truth problem:
[…] if assertion in any way changed a proposition, no proposition which

can possibly in any context be unasserted could be true, since when asserted
it would become a different proposition. (Russell, PoM: §38)
In order to grasp the problem at which Russell is pointing here, we need

to pay attention to his further development of the notion of logical asser-
tion. Assertion is the quality belonging to propositions in verbal form,
like [Caesar died] as opposed to propositions in verbal-noun form, like
[the death of Caesar] or [that Caesar died].3 The latter can appear in con-
texts like “p implies q”, as for example in (4):
(4) [That Caesar died implies that a new emperor was needed]
Therefore, the proposition indicated by “Caesar died” when it is not em-

bedded in another and the proposition indicated by “Caesar died” in (4),
where it is unasserted, are two different propositions. Similar reflections
apply to “Obama is younger than Bush” embedded in the context “John
judges that …” in (5) below.
(5) John judges that Obama is younger than Bush
The fact that logical assertion accounts for the unity of propositions and
also for the truth of true ones makes the problem of false propositions
stronger (Problem 2). The proposition indicated by (3) is false. It is thus
unasserted, and hence its constituents are not unified as they are in (2).
The second problem that arises as a consequence of the monopolarity
of a proposition is related to negation. In denying a sentence, like “aRb”,
3
This is why Russell is sometimes read as if one of the constituents (the verb)
were responsible for the unity of the proposition. The verb as a verb (in verbal
form) is precisely logical assertion, but the verb is also present, as constituent (a
concept) in verbal-noun form expressions, which are not asserted.
36 María Cerezo
which indicates a true or false proposition, its truth-value changes, but

we do not have any clue to help us figure out how the world is according
to the negating sentence “~aRb”. In other words, Russell can account for
the way in which negation affects the truth-value of the sentence, but not
for the way in which it affects its meaning (the indicated complex enti-
ty). In particular, if “aRb” is true, Russell cannot account for the mean-
ing of “~aRb”, because due to the asserted character of “aRb” its con-
stituents happen to be related in such a way that there is no complex en-
tity indicated by “~aRb”(or at least we do not know how to determine
it). And, if “aRb” is false, it is not asserted, and therefore its constituents
are not related, but there is no way of determining the complex entity
indicated by “~aRb”. This is the yes-no direction problem (Problem 4),
which is closely related to Problems 1-3.
At this stage judgement, as a psychological relation, is a dyadic rela-
tion for Russell: a relation between a subject and an unasserted propo-
sition or propositional concept. Russell’s view about judgement at this
time might be symbolized as follows:
(6) J (S, aRb)
where “S” stands for the subject, “J” for the judgement relation and
“aRb” for the judged (unasserted) proposition.
Once all these elements are taken into account, it is possible to see
that, in the case of false propositions, Problem 1 and 2 have not really
been solved, that Problem 3 appears in judgement, as in (5) above, and
that Problem 4 has been left untouched. The root of the problems is that
being asserted has come to be identified with being true, and therefore it
is not possible to account for sense with independence from truth.
OTF-Theory (1910)
The series of problems just described led Russell to try to account for
truth and falsehood in a different direction, namely, as properties not of
propositions, but of judgements or beliefs, and to make the subject re-
sponsible for their unity, if either true or false. Truth then was corre-
spondence with a further complex entity.
In his 1910 version of the multiple relation theory of judgement, Rus-

sell denied the existence of a proposition as an entity independent of the
judging act, and stated that the judgement relation itself established the
unity of the belief. Russell’s view might be symbolized as follows:
(7) J (S, a, R, b)
where “S” and “J” stand for the subject and judgement relation, and
where “a”, “R” and “b” stand for the objects of judgement, which are
entities of the world. Beliefs are entities whose unity is psychological,
and therefore the unity of true and false judgements is accounted for in
the same way. Truth is then conceived as the fact that there is a complex
corresponding to the judgement.
In this way, Russell thinks he has solved the problem of false proposi-
tions (now false judgements), in the specific forms in which it appears in
Problems 1, 2 and 3.4 Notice however that there are no elements in the
OTF-Theory to solve Problem 4.
In addition, a new problem arises. Given that [Obama], [being young-
er than] and [Bush] are actual terms, world entities, with which the mind
is acquainted, and it is the mind that unites them into a belief or judge-
ment, then relations are assimilated to things, since they are not in
charge of actually relating the relata in judgement. [Obama], [Bush] and
[being younger than] are all on equal terms in their relation to the unity
action of the subject. This generates a new problem: the right-consti-
tuents problem (Problem 5). [Obama], [Bush] and [Clinton] are on the
same terms in their relation to the unifying action of the mind as
[Obama], [Bush] and [is younger than] are.
4
Candlish (1996) thinks that in the OTF-theory Russell did not solve Problem 2. I
will not discuss this issue here. It is sufficient for my purposes that the
OTF-theory did not solve all the problems, and this is even truer if Candlish is
right. Furthermore, the new version of the multiple theory of judgement to be de-
veloped in 1913 shows that Russell was not satisfied, and Wittgenstein’s criti-
cisms of both versions confirm that the solution was inadequate. More on this
later.
38 María Cerezo
TK-Theory (1913)
In 1913, Russell has recourse to logical forms to solve the problems in-
herited from his previous theories (in particular, Problems 1, 2 and 5):
Suppose now that someone tells us that Socrates precedes Plato. How do we
know what he means? It is plain that his statement does not give us ac-
quaintance with the complex “Socrates precedes Plato”. What we under-
stand is that Socrates and Plato and “precedes” are united in a complex of
the form “xRy”5, where Socrates has the x-place and Plato has the y-place.
It is difficult to see how we could possibly understand how Socrates and
Plato and “precedes” are to be combined unless we had acquaintance with
the form of the complex. (Russell, TK: 99)
We can now formulate Russell’s difficulty more clearly: the sense of the
belief is not established by the determination of the meaning of its parts.
Russell thinks that logical forms can play the role of combining the con-
stituents into something that can correspond to an actual combination in
the world when the belief is true, and logical forms can also relate con-
stituents which are not so related in the world (Problem 2).
Logical form is what remains when all constituents have been remo-
ved from a complex. In (2), for example, if [Obama], [Bush] and [being
younger than] are removed, a logical form remains, which in this case is
the form of a “dual complex”, represented as “xRy”. Russell insists that
logical forms are not further constituents of the complex to avoid the risk
of infinite regress. His view about judgement can now be symbolized as
follows:
(8) J (S, a, R, b, xRy)
The constituents designated by “a”, “R” and “b” are real objects, con-
stituents of actual complexes with which the subject is acquainted; the
5
In this Section I use the italic “R” in “xRy” to express the fact that here “R” is a
variable that stands for any particular relation [R]. Russell did not use the italics,
only “R” to express this, and he used particular examples of relations, like “simi-
larity”, “precedes” to express particular relations between particular objects like
[Socrates] and [Plato] or [a] and [b].
logical form represented by “xRy” is that common to all dual complexes

with which the subject is also acquainted.
The recourse to logical forms allows us to solve Problem 2, but only
partially, since old related problems reappear and new ones arise. The
difficulty to account for the way in which constituents combine (Prob-
lem 1.1) reappears because it is necessary to specify the position of “a”
and “b” in “xRy” in order to determine what is being judged. Russell of-
fers a technical account of how this could work, whilst being fully aware
of the limits of his explanation (Hanks 2007, Carey 2007). Furthermore,
Problem 5 remains: judgement has an undesirable freedom which allows
it to combine different objects with logical forms, and there is nothing in
the theory that prevents the combination of any two things with any rela-
tion, unless further requirements are specified. The status of logical
forms, on the other hand, remains unexplained and problematic, and
Wittgenstein reacts to it, by insisting that if it is not a further constituent,
it can be neither named nor expressed. For if named, it would be a con-
stituent; if expressed, it would be a proposition, and the problem of ac-
counting for its logical form would arise again (N 20.11.14). (Problem
6: status of logical forms problem)
In addition, Problem 4 remains as it was: the recourse to logical forms
does nothing to explain why a proposition can be true and can be false,
or in Wittgenstein’s words in the Notes on Logic (1913, henceforth NL),
we must not only know that p implies “ “p” is true”, but also that ~p im-
plies “ “p” is false” (Wittgenstein, NL: 94). Russell’s struggle with posi-
tive and negative facts in Appendix B.I, “Props” of TK verifies that this
was one of his concerns.6
2. Wittgenstein on Russell’s Theory of Judgement
There are three series of texts in which Wittgenstein explicitly reacts to

Russell’s theory of judgement. The first of these is the letter written by
Wittgenstein to Russell in June 1913, after visiting him, which is said to
have paralyzed Russell.
6
For further details of the content of TK, see Carey (2007).
40 María Cerezo
Text 1:
I believe it is obvious that, from the prop[osition] ‘A judges that (say) a is
in the Rel[ation] R to b’, if correctly analysed, the prop[osition] ‘aRb v
~aRb’ must follow directly without the use of any other premiss. This con-
dition is not fulfilled by your theory. (Wittgenstein, 1913)
An immediate reaction is to notice that since “aRb v ~aRb” is a tauto-

logy, from the perspective of the theory of inference, it does follow from
any proposition, and therefore it should also follow from ‘A judges that
a is in the Rel[ation] R to b’. But this is too obvious a fact for Wittgen-
stein to have overlooked it. Fortunately, in NL, Wittgenstein offers a clue
to help understand what he meant.
Text 1.2:
I understand the proposition “aRb” when I know that either the fact that
aRb or the fact that not aRb corresponds to it; but this is not to be confused
with the false opinion that I understood “aRb” when I know that “aRb or
not aRb” is the case. (Wittgenstein, NL: 104)
Wittgenstein is pointing out Problem 4 which is precisely the problem

that not only remains unsolved but is not properly addressed in the three
Russellian theories presented above.
Problems 1 and 5 seem to be the target in the explicit criticism of Rus-
sell’s theory in the Tractatus:
Text 2:
The correct explanation of the form of the proposition, ‘A makes the
judgement p’, must show that it is impossible for a judgement to be a piece
of nonsense. (Russell’s theory does not satisfy this requirement.) (Wittgen-
stein, TLP 5.5422)
Again NL helps us here, since Wittgenstein claims that Russell’s theory

cannot explain what makes it impossible for us “to judge that this table
penholders the book” (NL: 103). Wittgenstein claims that some condi-
tions are required on the structure of “p”, so that for example, “Obama
Bush”, “is younger than multiplying” or “Obama multiplies Bush” are
excluded.
A further explicit mention of Russell’s theory of judgement in NL is
the following:
Text 3:
There is no thing which is the form of a proposition, and no name which is
the name of a form. Accordingly we can also not say that a relation which
in certain cases holds between things holds sometimes between forms and
things. This goes against Russell’s theory of judgement. (Wittgenstein, NL:
105) (See also N 20.11.14.)
Here Wittgenstein points to Problem 6, and thus challenges Russell’s

solution in TK, which required unification in thought to consist in the
joining of forms and constituents.
Bearing in mind the Tractarian theory of a proposition, helps to under-
stand Wittgenstein’s criticism better. The following text, also from NL,
is an example of an attempt to address Problems 2, 3 and 4 and amend
Russell’s theory along the lines pointed out in texts 1 and 1.2:
Text 4:
In “a judges p” p cannot be replaced by a proper name. This appears if we
substitute “a judges that p is true and not p is false”. The proposition
“a judges p” consists of the proper name a, the proposition p with its two
poles, and a being related to both of these poles in a certain way. This is
obviously not a relation in the ordinary sense. (Wittgenstein, NL: 95)
Wittgenstein develops this idea in a longer text in NL. He even draws a

diagram of the relation of the subject to the two poles of a proposition.7
Text 4.1:
When we say “A believes p”, this sounds, it is true, as if here we could sub-
stitute a proper name for “p”; but we can see that here a sense, not
a meaning, is concerned, if we say “A believes that ‘p’ is true”; and in order
to make the direction of p even more explicit, we might say “A believes that
‘p’ is true and that ‘not-p’ is false”. Here the bi-polarity of p is expressed,
and it seems that we shall only be able to express the proposition “A be-
lieves p” correctly by the ab-notation; say by making “A” have a relation to
the poles “a” and “b” of a-p-b. The epistemological questions concerning
7
The similarity of this diagram with the diagrams in the manuscript “Props”, Ap-
pendix B.1 of TK has been stressed by Hanks (2007) and Carey (2007), who of-
fer it as evidence that Wittgenstein’s difficulty to TK was related to the bipolarity
of the proposition.
42 María Cerezo
the nature of judgement and belief cannot be solved without a correct ap-
prehension of the form of the proposition. (Wittgenstein, NL: 106)
In any case, in NL Wittgenstein often insists that in order to understand a

proposition we must know, not only how the world is when it is true, but
also how it is when it is false, and he refers to this feature as the sense of
the proposition. Therefore, Russell’s difficulties provided the required
stimulus to develop the truth-conditional notion of sense proper to the
Tractatus.
3. Wittgenstein’s moves under the stimulus of Russell’s problems
Russell had explained the relation of judgement as a relation between a

subject (a mind, a simple object) and further objects (a proposition in
PoM-Theory, constituents of propositions in OTF-Theory, and consti-
tuents of propositions and logical forms in TK-Theory). Wittgenstein’s
crucial idea to overcome the series of problems to which Russell’s theo-
ries had given rise is to understand the relation of judgement as the rela-
tion of saying, that is, the relation between a proposition and what a
propositions says, and to give an account of propositions and of what
they say in terms of depiction. As he pointed out to Russell in his letter
of 22nd July 1913, the objection he had raised to Russell’s theory of
judgement could “only be removed by a correct theory of propositions”.
The (Russellian) relation of indication between language and world is
replaced by two more complex relations: that of naming, between names
and objects, and the relation of depiction, between propositions and what
they say. The Tractarian truth-functions theory offers the required con-
ceptual framework to account for sense in terms of truth-conditions. By
means of being a picture and a truth-function, a proposition can account
for all the desiderata that Russell’s difficulties involve. Let us now revise
Wittgenstein’s moves in more detail.
In Text 4, a pre-Tractarian text, Wittgenstein also conceives the rela-
tion of judgement as a relation between an object, a, and a further entity,
p, and he focuses his attention on a particular characteristic that the lat-
ter, p, must have so that the theory of judgement can avoid Problem 4. In
the Tractatus, Wittgenstein takes a further step, making a crucial move
in overcoming Russell’s difficulties. Instead of relating an object (the

subject or the mind) with further objects, depiction is a relation between
facts (propositional signs) and depicted facts (what is declared to be the
case), which are the sense of propositions.8 Wittgenstein’s move is then
summarized as follows:
It is clear, however, that ‘A believes that p’, ‘A has the thought that p’, and
‘A says p’ are of the form ‘ “p” says p’: and this does not involve a corre-
lation of a fact with an object, but rather the correlation of facts by means of
the correlation of their objects. (Wittgenstein, TLP 5.542)
Text 2 is precisely one of the reasons why the judgement relation must
be accounted for as depiction: if the judgement relation is explained as
depiction, it shows why it is impossible for us “to judge that this table
penholders the book” or any other piece of nonsense.9
But the move made by Wittgenstein, namely, to conceive of jud-
gement as a relation between facts, is a complex one, a move that is
composed of further novelties and adjustments. The first novelty is a re-
course to the pair possibility/factualness to account for the determination
of sense with independence of truth. Instead of having the same constit-
uents in judgement and in judged fact with two different ways of com-
bining them in thought and reality (like in OTF-Theory and TK-Theory),
leaving sense undetermined, Wittgenstein distinguishes between constit-
uents of the depicting and the depicted fact and requires them to share
their logical form, that is, their possibility of combination into states of
8
A further contrast between Russell’s and Wittgenstein’s theories concerns the
belief statement. In Russell’s accounts, the judgement relation (“J” in (6), (7) and
(8) above) is an external relation whose holding can be meaningfully reported in
a statement of the form “A believes that p”. However, the depiction relation is an
internal relation: in the Tractatus the relation between a depicting and depicted
fact is not a further fact of the world, and thus cannot be depicted. Belief state-
ments are therefore pseudo-propositions, since they attempt to say what cannot
be said (depicted). For further details on this issue and a revision of the literature
on this matter, see d’Ors and Cerezo (1995).
9
As we shall see below, this is solved by the Tractarian idea of isomorphism, i.e.
the idea that representation takes place by means of reproduction of structure.
See also Hochberg (1996).
44 María Cerezo
affairs (TLP 2.014-2.0141). Depicting facts and depicted facts are corre-
lated because they are in a particular internal relation: even if they are
composed of different objects, they have the same form.
The Tractarian isomorphism thesis establishes that there are two net-
works of objects that share their form, their possible combinations to
other objects (TLP 2.16-2.161). This relation of identity of form between
language and world grounds the connection between the constituents of
language and those of the world. A combination of names (a propo-
sition) can thus present a possible combination of objects, since the con-
nection of the propositional components is a possible connection for the
represented things (N 5.11.14; see also N 30.9.14; 29.10.14).
Language constituents, names, contain all its possibilities (all possible
combinations of names into propositions) and, given isomorphism, any
combination represents a possible way in which the corresponding ob-
jects could be combined in the world, even if they are not so combined.
This allows us to account for the sense of false propositions, and to solve
Problem 2; but it also solves Problem 5, because now logical form is in-
ternal to language, and any permissible expression represents a possible
state of affairs (TLP 5.473-5.4733). Isomorphism is thus the conceptual
tool in solving the problems which Wittgenstein pointed out in Text 2.
Since as a consequence of isomorphism any possible combination of
names represents a possible combination of objects, it is impossible for
us “to judge that this table penholders the book” or any other piece of
nonsense. “A thought contains the possibility of the situation of which it
is a thought. What is thinkable is possible too” (TLP 3.02).
Furthermore, being internal to language, logical form is neither named
nor represented in a proposition; it is rather shown (mirrored, displayed)
in language (TLP 4.12-4.121). As a consequence, Problem 6 is avoided,
and the Tractarian theory can escape the difficulty indicated in Text 3.
However, the elements of the picture theory so far introduced cannot
solve Problem 1, and in particular Problem 1.2. Given isomorphism, any
proposition or possible combination of names represents how things can
be combined in the world, and not how they are combined. In order to
account for judgement, it is necessary to account for its aim at truth, for
its assertive character. After having descended into simples and possi-
bilities to ground logical form, Wittgenstein needs to ascend back to fac-
tualness. Propositions are the possibilities of language, but they are actu-
alized when a propositional sign is used and projected onto the world.
Propositions do not contain their sense, but only the possibility of ex-
pressing it by means of a propositional sign, which is not a possibility,
but a fact (TLP 3.1-3.144).
Since propositional signs are facts, their constituents (the words) stand
in determinate relations to one another (TLP 3.14). Propositional signs
are not sets of words, but they are articulate (TLP 3.141-3.142). And it is
the fact that their constituents are related in a determinate manner that
says that the corresponding objects are also thus related (TLP 3.1432).
Problems 1.1 and 1.2 are solved because the propositional signs “aRb”
and “bRa” are two different facts (Problem 1.1) and because they are
facts (Problem 1.2): in “aRb” and “bRa” the constituents a, R and b
stand in (Problem 1.2) determinate (Problem 1.1) relations to one anoth-
er.10 Notice also that Problem 3 does not arise in the Tractatus since as-
sertion is not a property exclusive to true propositions. Any pro-
positional sign that declares that something is the case is asserted and it
can be true and false. Russellian monopolar propositions are replaced in
Wittgenstein’s theory by bipolar ones.
There are two components of sense, which I refer to as sense-1 and
sense-2. By isomorphism a proposition (a picture) represents a possible
state of affairs by reproducing its structure; it shows that other objects
can be combined in the same way as the elements of the picture are
combined (sense-1). The second component of sense accounts for the
assertive character of depiction. Pictures depict facts: they do not only
represent possibilities; they also say (sense-2) that the world is as it is
represented (sense-1) in the picture. This duality is captured in TLP
4.031-4.0311 by means of the difference between representing (darstel-
len) and presenting (vorstellen) a state of affairs, and it is explicitly de-
clared in T 4.022:
A proposition shows its sense.

A proposition shows how things stand if it is true. And it says that they do
so stand. (Wittgenstein, TLP 4.022)
10
This is the part of Wittgenstein’s move on which Hanks (2007) focuses in his
analysis.
46 María Cerezo
Pictures present how things can be combined (sense-1), but propositions

can say how the world is and how the world is not (sense-2). In the Note-
books (henceforth N), Wittgenstein is aware that a picture, on its own, is
not able to account for the yes-no direction proper to propositions:
Can one negate a picture? No. And in this lies the difference between pic-
ture and proposition. The picture can serve as a proposition. But in that case
something gets added to it which brings it about that now it says something.
In short: I can only deny that the picture is right, but the picture I cannot
deny. (Wittgenstein, N 26.11.14)
Wittgenstein conceives affirmation and negation as the two ways of

symbolizing (Bezeichnungsweise) that must be attached to the picture for
it to determine sense-2. This is a first step in order to solve Problem 4:
the state of affairs that is represented by “aRb” and “~aRb” is the same,
even if what they say (sense-2) is different.
If a picture presents what-is-not-the-case in the aforementioned way [nega-

tive way], this only happens through its presenting that which is not the
case. For the picture says, as it were: “This is how it is not”, and to the
question “How is it not?” just the positive proposition is the answer. (Witt-
genstein, N 3.11.14; see also TLP 4.023)
But how is sense-2 determined? In the Tractatus Wittgenstein describes

the determination of sense-2 as a demarcation of a place in logical space
(TLP 3.4-3.42), and the truth-functions theory offers him the tool to ac-
count for it. The idea that a proposition is a truth-function of elementary
propositions allows us to combinatorially define all the possible truth-
functions of a given set of bases, as in the standard truth-table method
(TLP 4.2-4.45). In order to account for determination of sense in a way
that takes into account the possible yes/no direction of a proposition,
Wittgenstein combines the picture theory and the truth-functions theory.
By this combination, Wittgenstein identifies the linguistic logical space
(the set of possible combinations of names into propositions) and the
logical space (the left part of the truth-table that includes all elementary
propositions).
Elementary propositions represent the possible states of affairs into
which objects can be combined as a result of their logical form (TLP
4.21, 4.3). The distribution of truth-values on the left part of the truth-
table represents the different ways the world could be depending on the
existence or non-existence of states of affairs. Each row of the left part
of the truth-table represents a possible world. The set of columns on the
right, the set of all truth-functions, represent all that can be said about
the world. The sense of a proposition is its agreement and disagreement
with possibilities of existence and non-existence of states of affairs (TLP
4.2). The sense of a proposition p is then determined by means of a divi-
sion of logical space into two sets of worlds, those in which p is true and
those in which it is false. The proposition p says that the world is one of
the worlds in which p is true; and ~p says that the world is one of the
worlds in which p is false (TLP 3.4, 4.022-4.023, 4.4).
4. Conclusion
Wittgenstein’s dialogue with Russell during the years 1911-1914 was

crucial for the development of the Tractarian doctrines and offers an ex-
traordinary framework in which to portray Wittgenstein’s moves as at-
tempts to overcome the difficulties that the evolving Russellian theories
encountered. (See Table 1 below for a summary of the moves.)
If my interpretation is correct, the problem indicated by Text 1 and
Text 1.2 stimulated Wittgenstein to develop the bipolar character of
propositions and to explain determination of sense as division of logical
space. Is there any textual evidence that Wittgenstein was stimulated by
Russell’s difficulties in the direction I have exposed above? His NL of-
fers immediate textual evidence that this was actually the case.11 For
11
The text in NL is very important as evidence of the fact that Wittgenstein was
motivated by Russell’s problems. Wittgenstein’s more important conversations
with Russell on his 1913 account of judgement took place in May 1913 and the
letter where he briefly objects to Russell’s theory was sent in June 1913 (Text 1).
Wittgenstein spent the summer in Hochreith and Bergen (September) where he
worked on Logic and reported to Russell that his work was going exceptionally
well. Finally, he went to Cambridge at the beginning of October and he dictated
the manuscript of NL to Russell. It is clear that it contains the work he had done
during that summer, work that was developed immediately after intense discus-
48 María Cerezo
immediately after the paragraph that includes Text 1.2, Wittgenstein ex-
plains how the form of a proposition symbolizes:
But the form of a proposition symbolises in the following way: Let us con-
sider symbols of the form “xRy”; to these correspond primarily pairs of ob-
jects, of which one has the name “x”, the other the name “y”. The x’s and
y’s stand in various relations to each other, among others the relation R
holds between some, but not between others. I now determine the sense of
“xRy” by laying down: when the facts behave in regard to “xRy” so that the
meaning of “x” stands in the relation R to the meaning of “y”, then I say
that the [the facts] are “of like sense” with the proposition “xRy”; other-
wise, “of opposite sense”; I correlate the facts to the symbol “xRy” by thus
dividing them into those of like sense and those of opposite sense. To this
correlation corresponds the correlation of name and meaning. Both are psy-
chological. Thus I understand the form “xRy” when I know that it discrimi-
nates the behaviour of x and y according as these stand in the relation R or
not. In this way I extract form all possible relations the relation, as, by a
name, I extract its meaning from among all possible things. (Wittgenstein,
NL: 104)
It is clear that this text does not give a complete account of the way in
which the Tractarian doctrines are combined to account for sense with
independence of truth. As we have seen, the picture theory and the truth-
functions theory are necessary for the project, since they provide some
important conceptual tools such as isomorphism, the depiction relation
and the idea of the sense of a proposition as its agreement and dis-
agreement with possibilities of existence and non-existence of states of
affairs. But three ideas are important in this text. First, the idea that the
way of symbolizing a proposition can be accounted for differently than
the way of symbolizing a name. Secondly, Wittgenstein shows the direc-
tion to take in resolving what Russell was attempting but unable to do,
namely, to explain how the determination of the meaning of constituents
can by itself determine the sense of a proposition.12 And, finally, by rela-
sions with Russell on his theory of judgement. See McGuinness (2005), Chapters
5 and 6, for further details.
12
See also McGuinness (1974).
ting bipolarity with sense (sense-2), he takes the first step, and a crucial
one, in the direction that he develops in the Tractatus.13
PROBLEMS MOVES / SOLUTIONS / NOVELTIES

1. Way of combination of con-
stituents
1.1. the way in which con- * Structure of proposition
stituents combine
1.2. there being actually * Factual character of propositional sign
combined
2. Falsehood * Recourse to form as (combinatorial) possibility
3. Assertion-truth * Factual character of propositional sign
* Bipolarity
* Sense-2
4. Yes-no direction * Bipolarity
* Sense-2
* Truth-functions theory
5. Right-constituents * Isomorphism
* Sense-1
6. Status of logical forms * Isomorphism
* Showing/saying distinction
Table 1. Summary of Russellian problems and Tractarian moves/novelties
References
Angelelli, Ignacio and María Cerezo (eds.), 2006. Studies on the History of Logic.
Berlin: Walter de Gruyter.
Candlish, Stewart 1996. The Unity of the Proposition and Russell’s Theories of
Judgment. In: Monk and Palmer (eds.), 103-135.
Carey, Rosalind 2007. Russell and Wittgenstein on the Nature of Judgement. Lon-
don Continuum.
Cerezo, María 2005. The Possibility of Language. Stanford: CSLI Publications.
13
I am aware that some of the Tractarian ideas presented here are simply sketched.
My only intention is to show how they arise as a consequence of Wittgenstein’s
struggle with Russell’s problems. For a more detailed interpretation of the Trac-
tatus that is behind some of the ideas of Section 3, see Cerezo (2005).
50 María Cerezo
Connelly, James 2011-2012. On ‘Props’, Wittgenstein’s June 1931 Letter and Rus-
sell’s ‘Paralysis’. Russell: The Journal of Bertrand Russell Studies 31, 141-166.
Griffin, Nicholas 1985. Russell’s Multiple-Relation Theory of Judgment. Philo-
sophical Studies 47, 213-247.
Griffin, Nicholas 1985-1986. Wittgenstein’s Criticism of Russell’s Theory of
Judgment. Russell: The Journal of Bertrand Russell Studies 5, 132-145.
Griffin, Nicholas (ed.) 1992. The Selected Letters of Bertrand Russell. London: Al-
len Lane The Penguin Press.
Hanks, Peter W. 2007. How Wittgenstein defeated Russell’s Multiple Relation
Theory of Judgment. Synthese 154, 121-146.
Hochberg, Herbert 1996. The Role of Subsistent Propositions and Logical Forms in
Russell’s 1913 Philosophical Logic and in the Russell-Wittgenstein Dispute. In:
I. Angelelli and M. Cerezo (eds.), 317-341.
Johnston, Colin 2007. The Unity of a Tractarian Fact. Synthese 156, 231-251.
Landini, Gregory 2007. Wittgenstein’s Apprenticeship with Russell. Cambridge:
Cambridge University Press.
McGuinness, Brian 1974. The Grundgedanke of the Tractatus. In: G. Vesey (ed.),
49-61.
McGuinness, Brian 2005. Young Ludwig: Wittgenstein’s Life, 1889-1921. Oxford:
McGuinness, Brian (ed.) 2008. Wittgenstein in Cambridge. Letters and Documents
1911-1951. Oxford: Blackwell Publishing.
Monk, Ray and Anthony Palmer (eds.), 1996. Bertrand Russell and the Origins of
Analytical Philosophy. Bristol: Thoemmes Press.
d’Ors, Angel and María Cerezo 1995. Tractatus 5.54-5.5423: Sobre los llamados
‘enunciados de creencia’ en el Tractatus Logico-Philosophicus de Ludwig
Wittgenstein [Tractatus 5.54-5.5423: On the so-called ‘belief statements’ in the
Tractatus Logico-philosophicus of Ludwig Wittgenstein]. Anuario Filosófico
28, 269-310.
Pincock, Chris 2008. Russell’s Last (and Best) Multiple-Relation Theory of Judg-
ment. Mind 117, 107-139.
Russell, Bertrand 1910. On the Nature of Truth and Falsehood. In: Philosophical
Essays. London: Longman, 147-159.
Russell, Bertrand 1937. The Principles of Mathematics, 2nd ed. London: Allen &
Unwin (1st ed.: 1903).
Russell, Bertrand 1984. The Collected Papers of Bertrand Russell, vol. 7, Theory of
Knowledge: The 1913 Manuscript. London: Allen & Unwin.
Stevens, Graham 2004. From Russell’s Paradox to the Theory of Judgment: Witt-
genstein and Russell on the Unity of the Proposition. Theoria 70, 28-60.
Stevens, Graham 2006. Russell’s Repsychologising of the Proposition. Synthese
151, 99-124.
Vesey, Godfrey (ed.) 1974. Understanding Wittgenstein. Ithaca, New York: Cornell
University Press.
Whitehead, Alfred North and Russell, Bertrand 1927. Principia Mathematica, 2nd
ed. (1st ed.: 1910). Cambridge: Cambridge University Press.
Wittgenstein, Ludwig 1913. Letter to B. Russell, June 1913. In: B. McGuiness (ed.),
p. 40.
Wittgenstein, Ludwig 1922. Tractatus Logico-Philosophicus. London: Routledge &
Kegan Paul.
Wittgenstein, Ludwig 1979a. Notes on Logic. In: L. Wittgenstein 1979b, 93-107.
Wittgenstein, Ludwig 1979b. Notebooks 1914-1916, 2nd ed. Oxford: Blackwell.
Paweł Grabarczyk
University of Łód
[email protected]
How to Talk (Precisely) about Visual

Perception? The Case of the Duck /
Rabbit
Abstract: In Remarks on the philosophy of psychology Wittgenstein uses ambigu-
ous illusions to investigate the problematic relation of perception and interpretation.
I use this problem as a starting point for developing a conceptual framework capa-
ble of expressing problems associated with visual perception in a precise manner. I
do this by discerning between subjective and objective meaning of the term “to see”
and by specifying the beliefs which are to be ascribed to the observer when we as-
sert that she sees a given object. The framework (detailed in section 2) is then used
to analyze the case of the duck/rabbit illusion. It shows that ambiguous illusions
present us with a specific skeptical challenge but that the challenge can be over-
come by empirical sciences. Along the way I explicate some of the common notions
associated with perception (“to look at”, “to have an impression of…”, “to react as
if one had an impression of…”, “to convince oneself that what one sees is…”).
Keywords: Wittgenstein, duck/rabbit, ambiguous illusions, optical illusions, per-

ception, impressions, belief ascription
1. Trouble with perception
It seems to be a paradox of sorts (or perhaps it’s just irony) that talking
about perception leads to so much confusion and so many misunder-
standings. After all, shouldn’t that which is given in perception be evi-
dent? Isn’t perception our model case of what “being evident” is? No-
where is this tension better expressed than in the works of later Wittgen-
stein. In Remarks on the philosophy of psychology (1980) he presents the
reader with a collection of visual puzzles the most famous being the
54 Paweł Grabarczyk
duck/rabbit picture.1 All of these illusions are there to examine the rela-
tion between perception and interpretation. What do we see directly and
what is interpreted or inferred from what we see? Is there such a thing as
pure, direct perception? Questions like this are, of course, closely related
to the question of relation between perception and language because in-
terpretation is first and foremost a linguistic activity. As such it is some-
thing Wittgenstein had been interested in long before he started to ana-
lyze ambiguous pictures. It is evident in one of his most famous quotes
from Tractatus – “What can be shown cannot be said” (1990: 79). One
way of interpreting the quote is to take it as stipulating that visual per-
ception contains a non-linguistic surplus. What illusions (ambiguous il-
lusions amongst them) show is that even if there are things you can only
show, things which escape successful description, it doesn’t mean that
these aspects of perception are pure. It may very well be the case that
our conceptual capacities always play an important role in perception.
What ambiguous pictures have in common is that they seem to show us
the backstage of our perception because we can clearly see that there is
some kind of decision-like process that underlies the seeing process.
Wittgenstein calls this process a “change of aspect” (1980: 8) but he is
clearly having difficulties with finding its exact nature. How is this pro-
cess different from normal illusions and normal, veridical perception?
Does it raise any new and important philosophical questions? What ex-
actly happens when we switch from the duck to the rabbit?
My aim in this article is not historical. I do not want to provide
an accurate interpretation of Wittgenstein but rather to pursue the above
problem and formulate it in a precise manner. I also want to stress that
although the problem itself lies between philosophy of perception and
philosophy of language, my paper leans heavily towards the latter. Let
me relieve the tension – I will not solve the duck/rabbit problem. I will
not solve it because, if I am right, it cannot be answered by philosophy
of language. But in order for the question to be answered by anyone it
has to be asked in a precise manner and this is something, I believe, I
can do. My aim is thus strictly conceptual – I want to create a framework
1
Originally introduced in Jastrow (1900).
How to Talk (Precisely) about Visual Perception? The Case of Duck/Rabbit 55
which will enable us to talk precisely about perception and restate the
duck/rabbit problem in a solvable manner.
Let me start by identifying two important reasons why talking about
perception often leads us astray. The first can be most easily introduced
by a simple dialog:
A: Yesterday I saw a red cup in the kitchen.

B: You couldn’t have seen a red cup. I don’t have one.
A: I think I know better what I saw!
It’s futile to ask who’s right here. A and B clearly use the word “to see”
with different meaning. Let’s call A’s meaning “the subjective meaning”
and B’s meaning “the objective meaning”. As we will see it is important
not to conflate these two meanings because they express two aspects of
perception which are irreducible to each other.2 The second problem can
be introduced via the following reasoning:
1. A dog sees a postman.

2. The postman is the best chess player in the city
3. Thus, the dog sees the best chess player in the city.
The argument is formally valid, but I doubt you accepted (3) without
hesitation. Why is that? It seems that (3) presupposes or at least suggests
that the dog knows that the postman is the best chess player in the city
which is rather absurd. Dogs just can’t have such sophisticated beliefs.
But the moment you realize this you also realize that accepting (1) is just
as risky as accepting (3) because having beliefs about postmen is proba-
bly no less cognitively advanced as having beliefs about chess players.
Does it mean that the dog didn’t see the postman after all? It may seem
that the natural way out of this puzzle is to say that the dog didn’t see the
postman as a postman. But what does it actually mean? Is it equivalent
to having beliefs about postmen or maybe having a concept of a post-
man? In fact – do we ever see postmen? Maybe the only thing you can
really see is a collection of color patches and shapes. It is easy to get lost
2
Valberg (1992) calls this split “the puzzle of experience”.
in these questions, so the second thing we have to be extremely careful

about is letting beliefs and concepts enter the picture unnoticed.
As I said, I believe that a solution of these problems will most likely
come to us from cognitive psychology, but it won’t happen if we don’t
formulate the questions in understandable and unequivocal manner. To
do that we have to create a conceptual framework powerful enough to
express the different possibilities that are taken into account in phi-
losophy of perception. As is evident from above considerations the
framework has to enable us to clearly distinguish between subjective and
objective meanings of “seeing” and clearly state which beliefs are to be
attributed to the observer. I construe such a framework in section 2. Sec-
tion 3 is devoted to some of its possible applications – the duck/rabbit
case amongst them.
2. A conceptual framework
Let’s start with the difference between subjective and objective meaning.
We shall focus on a simple model situation, where an observer O sees a
particular object A at a particular time:3
(S) O sees A at t1
What do we mean when we interpret the word “sees” in the objective

sense? I propose two conditions which have to be met if it is to be the
case:
W1: There exists an object x O looks at (at t1)

W2: x = A
There are a few things we have to clarify here. First of all, we have to
explain the term “looks at”. After all, substituting one common term for
another doesn’t help much. The term should be understood as follows:
3
“A” is to be understood as a constant.
An observer O looks at an object x iff the object x reflects light

which affects the receptors of O in a proper way.
I guess that the expression “in a proper way” may raise some eyebrows,
so let me clarify – it is intentionally unexplained, because the question of
what really is “proper” is not one a philosopher can answer. It is an em-
pirical question which, as far as I know, has already been quite exhaust-
ively addressed.
Let me elaborate on why these conditions look this way. First of all,
we couldn’t settle for a rather intuitive condition of vicinity (O sees A in
an objective sense if A is in a vicinity of O). It is quite obvious that the
object which is to be seen has to interact with the observer and not just
“be there”.4 But the interaction itself also has to be rather specific. To
start with, it has to be related to a given sense – if we want to diffe-
rentiate between seeing and hearing, we probably won’t be able to avoid
talking about the eyes, and not the ears. So maybe we could settle with
something like: O sees A in the objective sense if A interacts with eyes of
O? But it is easy to notice that even this isn’t enough. I could easily es-
tablish that a given object is cold by touching it with my eyes closed, but
it is far from the interaction we were thinking of.5 By talking of the
proper way the receptors register the light not only do we give credit
where it’s due (that is to empirical science) but we also leave the possi-
bility of talking about observers with photo receptors quite different than
eyes (for example robots or people using technologies like Brainport
– see Bach-y-Rita et al. 2003).
Note that even with these clarifications a conjunction of W1 and W2
does not give a definition of “seeing objectively”. It is meant only to be
a necessary condition but it is enough to differentiate it from “seeing in a
subjective sense” which we are now going to discuss.
Let’s get back to our situation S. What is meant by saying that O sees
A at t1 in a subjective sense? I propose to explicate it with the following
conditions:
4
See Grice (1961: 141) for some rather convincing examples
5
A similar argument can be found in Pitcher (1971).
W3: O is in an unconscious visual state S (at t1)

W4: S is a typical unconscious state for O and A
W5: O is in a conscious visual state V (at t1)
W6: V is a typical conscious state for O and S
First of all, note that I avoided the most natural way of expressing the
subjective meaning of “seeing” which is “having an impression of A”.
The reason I don’t want to talk about impressions is that they create in-
tensional context and that their identity conditions are rather mysterious.
Consider “an impression of a postman”. Is it the same impression as “the
impression of a man dressed as a postman”? Maybe, technically speak-
ing, there is no such thing as an “impression of a postman” and we
should talk only of simple impressions like color or shape impressions?
But then, consider “an impression of white” and “an impression of the
color of my shirt”. Granted that my shirt is white, do you have an im-
pression of the color of my shirt when you look at the margin of this
page? Could you have it without ever seeing my shirt? I don’t have a
good solution to these puzzles, and that is why I prefer to use W3-W6
instead of the notion of “impression”. Let me elaborate on these condi-
tions so we can learn that they may be treated as an extensional explica-
tion of the problematic notion of “impression”.
First of all, we have to discern between unconscious and conscious
stages of perception. After all, it is pretty well established that the three
dimensional picture we experience starts as a two-dimensional pattern.
The contours of objects I discern in a given scene have to be detected by
the edge detection cells, there is a blind spot in my vision which is filled
by my brain and so on. You may be curious about the specifics of this
stage – is it something that happens in the retina or does it perhaps span
to some parts of the visual cortex? We don’t have to go into these de-
tails. The only point is not to conflate conscious with unconscious stag-
es.
Note that neither W3 nor W5 says anything specific about the respec-
tive states. The only thing said is that the observer is in some internal
visual states. We learn more about these states in W4 and W5. The point
of these conditions is that they inform us that O has a disposition to be in
a given state. The key expression “being a typical (un)conscious state

for…” can be defined as follows:
A visual state X is a typical unconscious state for an observer Y and

an object Z iff there exists a moment tn when Y looked at Z and eve-
ry time Y looks at Z she is in an unconscious visual state X.6
A visual state X is a typical conscious state for an observer Y and an
unconscious visual state Z iff there exists a moment tn when Y was in
Z and every time Y is in Z she is in a conscious visual state X.
Note that what W3-W6 boil down to is much more intuitive than it may
seem at first glance. What these conditions try to express is that when we
describe someone as seeing an object in a subjective sense, what we re-
ally mean is that she is in some (unknown to us) internal state which she
normally is in when she sees this object in an objective sense. We do not
know the state, because it does not manifest itself to us in any way in
normal acts of communication.7 The states are thus relativized to the ob-
server – the state which is typical for observer O when she looks at A
may be quite different from the state you are in when you look at A. It is,
to use the famous Wittgesteinian expression, a beetle in a box (1958:
100). Similarly to conditions W1 and W2, conditions W3-W6 should be
understood as necessary conditions.
This way we can precisely differentiate between objective and sub-
jective meaning of “seeing”. Subjective seeing means that conditions
6
The reason this definition looks like that is that we have to block a well known
counterintuitive consequence of defining a dispositional notion via a material
implication. If we decided to use just the second conjunct, then considering that
there are things the observer has never seen and will never see – a dragon for ex-
ample – every state would have been “typical” for a dragon. The same mutatis
mutandis can be said about the second definition. Also, the term “looks at” is to
be understood in the technical sense defined earlier.
7
They can be registered in various nonstandard ways – for example we can use
PET scans to try to correlate them with different stimuli.
W5-W6 (but not necessarily W1-W4) hold. Objective seeing means that
conditions W1-W4 (but not necessarily W5-W6) hold.8
Once we have conditions W1-W6, what is left is to add beliefs to the
mix. In order to do that we have to differentiate between two senses of
“having a belief” – verbal and non-verbal (indicated by an asterisk).
O has a belief that p iff O has the disposition to assert a sentence “p”.
O has a belief* that p iff O’s behavior can be best explained by attri-
bution of a belief, that p.
The sense of having a belief* is very similar to Dennett’s “intentional

stance” (1989: 17). Although we can ascribe having a certain belief to
anything, only sometimes does this ascription really help us with our
predictions. This definition explicates the common expression: – “it be-
haves as if it believed that p”.
The good news is that introducing extra “belief conditions” will now
be extremely easy, because we will be using a simple sentence form
“O believes that Wn” or “O believes* that Wn”.9 The bad news is that
this triples the number of conditions. Fortunately, there is no point
in talking of certain combinations. For example, there is nothing like be-
lieving* that W3 – the whole nature of unconscious states boils down to
the fact that we do not register them (without advanced equipment) and
hence we never act as if we believed that we were in those unconscious
states.
Similarly, although we could speak about “believing that W3”, it is, in
fact, completely irrelevant to perception.10 The belief conditions we will
be using in the next section are:11
8
It may seem that the objective sense also demands W5-W6, but the blind seeing
phenomenon (mentioned in section 3) and the case described in Gazzaniga et al.
(1962) show that they are not necessary.
9
To shorten the number of conditions I sometimes use a conjunction: O believes
that Wm & Wn.
10
It is only possible since we can have third person knowledge of our own states.
11
I do not have the space here to elaborate on the reasons some of the possibilities
are missing from the list.
W7: O believes that W1 & W2

W8: O believes* that W1 & W2
W9: O believes that W5
After everything we said up to this point, it should be easy to understand

the conditions W7-W11. W7 means that the observer has the disposition
to affirm that she looks at A. W8 means that she behaves as if she looked
at A. W9 means that O has a disposition to affirm that she is in some
specific visual state, W10 means that she believes that she recognized
the state as being typical (in a defined sense). W11 may not be obvious
at first but it should be easier to grasp once we realize that it is supposed
to capture the common expression “reacts as if she had an impression…”
as opposed to “reacts as if she saw…”. It is well suited to express
(among other cases) our purely aesthetic or emotional judgments. For
example, when I look at a cinema screen and see a fast approaching
train, I might react to it because I like steam trains and thus I do react as
if I had an impression of a train but it does not mean that I reacted as if I
had seen a fast approaching train.
3. Framework applications
Let’s list all conditions in one place, because we shall be addressing

them quite often.
W1:. There exists an object x O looks at (at t1)

W2: x=A
W3: O is in an unconscious visual state S (at t1)
W4: S is typical for O and A.
W5: O is in a visual state V (at t1)
W6: V is typical for O and S.
W7: O believes that W1 & W2

Various combinations of conditions W1-W11 constitute the framework

we were looking for. To understand how it works consider two cases:
Case 1: W1-W11 obtain.

The observer looks at A, has proper visual states (conscious and un-
conscious), reacts accordingly (as if she saw A and as if she had the
impression of A) and believes that she sees A and has a visual state
she typically has when she looks at A. We can classify this case as a
model veridical perception.
Case 2: Only W1-W4 and W8 obtain.
The observer looks at A and reacts as if she saw this object, but there
is no conscious experience of seeing nor any belief about seeing the
object. This case corresponds to the phenomenon of blindsight.
Now, if you treat both cases as boundaries, what you get is a set of pos-
sible in-between cases. Using these combinations you can express vari-
ous philosophical positions or describe new interesting cases – real and
imaginary. For example, if you take Case 1 and subtract W1 and W2,
you end up with hallucination (let’s call it Case 3).12 If you take Case 2
and add condition W7, you obtain Block’s super-blindsight (let’s call it
Case 4):13 the observer does not have the conscious experience but not
only does she react as if she saw the object, she also learns to give prop-
er verbal reports of the object. It is a good place to remind us that the
framework I present here is purely conceptual. It is a set of linguistic
tools to speak about perception, not a theory of perception. The latter
should be provided by philosophy of mind or psychology. Because of
that we should not worry if the idea Block suggested is empirically pos-
sible or not. The only thing we are interested in is that it is coherent and
can be expressed via the combination of W1-W11. Generally speaking,
12
Hallucination would have been described differently if we wanted to embrace
disjunctivism. I use this framework to express different positions in philosophy
of mind in Grabarczyk (forthcoming).
13
Discussed in Block (1995).
there are two ways to approach the framework. You can analyze differ-
ent philosophical positions and try to express them within it or simply
combine the conditions and look for interesting cases. I discuss it in
more detail in Grabarczyk (forthcoming).
Before we can see how this framework can help us with Wittgen-
stein’s illusions, we have to see how it works for some of the more typi-
cal cases of illusion. Let’s use the following convention: we can trans-
form any condition Wn into an alternative condition Wn’, Wn’’ and so
on. We do that by replacing every instance of A with A’ (or A’’) where
A’ (A’’) is a name of a different object than A (or A and A’). Sub-
sequently we replace every instance of S and V with S’ and V’ (or S’’
and V’’), where S’ and V’ are names of states different than S and V (or
S, S’ and V, V’ accordingly).
Having established that, let us try to analyze the phenomenon of opti-
cal illusion. As we said earlier, philosophers often describe illusions as if
they were erroneous interpretations or inferences.14 If we wanted to take
their words for granted we would have to interpret the phenomenon of
illusion in a following way:
Case 5: W1-W6 obtain but instead of W7-W11, W7’-W11’ obtain.
What is meant by this is that the observer O looks at A and has all the
proper visual states (unconscious and conscious) but, for some reason
has a wrong set of beliefs (both verbal and non-verbal). At first it may
seem to fit the descriptions cited above, because it is natural to connect
the notions of inference or interpretation with beliefs. Even if we accept
that the act of interpretation or even an act of inference can start with
something different than a proposition, it seems obvious that it leads to a
proposition that is held by the interpreter. But upon further inspection
Case 5 seems to be more fitting for a description of delusion than illu-
sion. Fortunately we have at least two different possibilities of describ-
ing illusion using conditions W1-W11.
14
A notable example of this is Russell (1948: 167).
Case 6: W1, W2’, W3, W4, W4’, W5, W6, W6’, W7, W8, W9, W10,
W11 obtain.
Case 7: W1, W2’, W3’, W4, W4’, W5, W6, W6’, W7, W8, W9, W10,
W11 obtain.
In Case 6, the observer looks at something different than the object A,

but for some reason this different object produces in O the unconscious
state S (W3) which is the state O is disposed to be in when she sees A
(W4). Because of this O is in a wrong conscious state V (W5). It is
wrong for her, because she clearly has appropriate dispositions (W4,
W4’, W6,W6’) – O was perfectly capable of having a veridical percep-
tion if the error hadn’t happened. We have to assume that she has all of
these dispositions because this enables us to differentiate between her
having an illusion and her not being able to discern between A and A’.
Let’s call this type of illusion Lower level illusion.
Let’s now turn to Case 7. The only important difference here is that
the object A’ O looks at produces the right unconscious state S’, but for
some reason it is transformed to a wrong conscious state V. Let’s call
this type of illusion Higher level illusion.
It is higher level illusions which invite terms like “inference” or “in-
terpretation” into the visual discourse. And, as Wittgenstein (1980: 3)
points out, it is much more problematic than we may initially expect.
There is just no way for us to use these terms in their normal meaning.
Seeing is immediate and quick, while interpretation or inference is a
process which takes much more time. Postulating unconscious infe-
rences or interpretations also appears to be rather counterintuitive, but it
seems inevitable if we stick to these terms in this context. Fortunately, as
conditions W1-W11 show, we can avoid talking about inferences and
interpretations altogether and still differentiate between veridical percep-
tion and different types of illusion. You might object that it is a cheap
win, because we simply dodged the problem – conditions W1-W11 do
not tell us anything about the nature of the transition between uncons-
cious and conscious states (maybe it is an act of interpretation after all).
But the point is that it is an empirical question and as such shouldn’t be
pre-conceived by the definition of illusion. The framework I propose
gives us an opportunity to be neutral in this respect.
Unfortunately, as Wittgenstein anticipated, the duck/rabbit (and other

similar examples) introduce their own problems and lead to a unique
form of skepticism. With conditions W1-W11 we are in position to show
exactly why it is so.
One last time, we shall start with a set of two more familiar cases.
Case 8: W1, W2’, W3, W4, W5, W6, W6’, W7’, W8’, W9, W10,
W11 obtain.
Case 9: W1, W2’, W3’, W4’, W5, W6, W6’, W7’, W8’, W9, W10,
W11 obtain.
What Cases 8 and 9 represent are two types of conscious illusion (lower
and higher level). In Case 8 the observer looks at the object A’ (W1 and
W2’) and it produces in her the subconscious state S (W3) which she is
normally disposed to be in when she looks at A (W4). Because of this
she is in a state V (W5) which she should be in if the object is A. But,
contrary to this, she believes that what she looks at is actually A’ (W7’).
She acts as if she looked at A’ (W8’) although she believes that she is in
a visual state V (W9) and that it is something she normally is in when
she looks at A (W10) and cannot help but have associations and aesthet-
ical judgments about A (W11). Case 9 is analogous, the only difference
being that the error enters at a later stage (at the stage of W5).
Let me show it on a concrete example. Consider a rather typical case
of looking at a stick in the water. You see it as if it was bent (W3-W6)
and know that you see it as if it was bent (W9,W10). If you have any as-
sociations, let’s say that you associate it with a letter of an alphabet, you
will associate the stick with the letter “V” rather than “I” (W11). Fur-
thermore, you can differentiate between straight and bent sticks (W6,
W6’). But you realize that the stick is straight (W7’) and you act as if it
was straight (W8’) – for example, if you were to touch its submerged
end, you would probably know where to put the finger.
The most striking thing conscious illusions teach us is that not only do
we sometimes see things differently from what we believe them to be,
but also that we couldn’t see them veridically even if we tried. Exposing
the illusion does not break the spell. Consider the Ponzo illusion. Even if
you use a ruler and see for yourself that the lines have the same length
you cannot stop seeing one of them as if it was shorter. It leads to a con-
clusion that perception is somehow immune to belief. Let’s call it a rule
of belief independence. Note that this rule is a very important tool
against skepticism. Beliefs are somewhat arbitrary – they probably de-
pend on the language we use, the culture we live in, and some of our idi-
osyncrasies. If perception was dependent on them, we wouldn’t have
been able to position it as epistemically basic or pure as we often wanted
to do. This intuition is quite easily recognizable in history of philosophy,
from Locke’s simple/complex ideas division to the ideal of observational
sentences. It has been attacked and pronounced dead several times, but it
still returns in different forms, being very attractive for the non-skeptic.
And this is the main reason why the duck/rabbit example is so special.
Contrary to other conscious illusions, it seems to debunk the rule of be-
lief independence. The distinctive aspect of the duck/rabbit picture is our
ability to voluntarily switch between two ways we can see it – once as a
picture of a rabbit, once as a picture of a duck. If this switch is, as we
feel, something that we actually do, and not just something that random-
ly happens to us, then a question arises: how exactly do we do it? Note
that there are not that many options we can choose from. As should now
be obvious from earlier considerations, we can divide conditions W1-
W11 into three groups. The objective part (W1-W2), the subjective part
(W3-W6) and the belief part (W7-W11). We do not have the ability to
directly induce visual states (even the conscious states).
I can imagine a horse, but I cannot subjectively see the horse, just be-
cause I want to see it. Therefore we are not able to directly manipulate
with W3-W6. So, if the switch changes them, it can only happen indi-
rectly, as a result of voluntary change of something else. But then, if we
decide that this change is the result of the voluntary change in the belief
part (any of the conditions W7-W11), we negate the rule of belief inde-
pendence. Needless to say, it is everything that the skeptic wanted. If a
change of beliefs can lead to a change in our visual states, then perhaps
we could do the same trick in other, seemingly veridical cases? Maybe
everything is an ambiguous illusion waiting to be discovered?15
15
Something along these lines is suggested in Strawson (1974: 58).
Fortunately, there are at least two ways to describe the duck/rabbit

scenario without giving the skeptic the upper hand. Note that because of
the ambiguous nature of the duck/rabbit illusion we have to present the
cases as disjunctions.
Case 10: W1, W2, W3, W4, W5, W6, W6’, W6’’, W7, W8, W9’,
W10’, W11’or
W1, W2, W3, W4, W5, W6, W6’, W6’’, W7, W8, W9’’,
W10’’, W11’’
Let’s explain the idea behind Case 10. The observer looks at the picture
of the duck/rabbit and has visual states which she normally has when she
looks at the picture of the duck/rabbit. Apart from the disposition to see
the duck/rabbit picture she also has the disposition to see pictures of
ducks (W6’) and pictures of rabbits (W6’’). She believes that she looks
at the picture of duck/rabbit and acts accordingly (W7, W8). But (and
here comes the interesting part) she somehow manages to switch be-
tween beliefs that what she experiences is the rabbit visual state (W9’,
W10’) and the duck visual state (W9’’, W10’’). This switch does not re-
sult in producing the corresponding visual states, but it produces in her a
non-verbal belief W11’ (or W11’’ accordingly). It is the difference be-
tween these two states that gives the distinct feeling of aspect change.
What changes is our attitude to what we see, our associations (of some-
thing being similar to ducks or rabbits), and our aesthetic feelings to-
wards it (we may like rabbits but not ducks).16 Case 10 can be under-
stood as an explication of a common expression “she convinced herself
that she sees…”.
But there is one more way to preserve the rule of belief independence.
Remember that our aim, if we want to block the skeptical threat is to
avoid suggesting that a change in beliefs results in a change in visual
states. We achieved it in Case 10, because we suggested that the change
happens only in beliefs (verbal and non-verbal). But you may object that
it is not compatible with what we really experience, because we in fact
16
Or, as Schroeder (2010: 365) says, she may feel that the beak is simply too long.
do experience some change in visual state when we switch from the

duck to the rabbit. The solution is now evident. If any of the subjective
conditions are to be changed and the rule of belief independence is to be
preserved, the only place the switch can happen are the objective condi-
tions W1-W2. Let’s show this case and explain it in some detail.
Case 11: W1, W2’, W3’, W4’, W5’, W6, W6’, W6’’, W7, W8, W9’,
W10’, W11’or
W1, W2’’, W3’’, W4’’, W5’’, W6, W6’, W6’’, W7, W8,
W9’’, W10’’, W11’’
What is meant here is that the observer, in fact, changes the object she is
looking at. The rest of the conditions are simply the result of that. It is
worth to mention that we assume this in order to be able to make the
switch the observer has to have the disposition to be in visual states in-
duced by duck/rabbit pictures, duck pictures and rabbit pictures (W6,
W6’, W6’’) and that even though the switch is quite convincing, she still
believes that what she looks at is a duck/rabbit picture although in fact
she is looking at something else. How can it be possible? After all the
duck/rabbit picture does not change! One possible explanation is that in
order to change the visual state, the observer inadvertently looks at dif-
ferent parts of the picture.17 To see the duck she has to focus on the left
part, to see the rabbit she has to focus on the right part. If we assume that
the difference in focus is radical, then we could treat the two distinct
parts of the picture as different objects the observer looks at. The idea
that the switch of the aspect boils down to looking at different parts of
the picture has been suggested and tested by psychologists, but the re-
sults are unfortunately mixed (see Wimmer and Doherty 2007). There
are no fundamental reasons, however, to believe that the hypothesis will
not be conclusively tested in the future. The important fact is that we are
able to formulate the problem in a perfectly solvable way.
17
Some readers may be alarmed by the fact that parts of objects are now treated as
new objects. Are we entitled to change the ontology like that? I analyze the rela-
tion between perception and ontology in Grabarczyk (2013).
4. Summary
As we saw, talking about perception often leads to some confusion and

creates paradoxes. The main culprits of these confusions are: conflating
two different notions of the predicate “to see” (subjective and objective)
and the uncertainty as to which beliefs (if any) are ascribed to the ob-
server whenever she decides that she sees a given object. My aim was to
create a precise and unequivocal conceptual framework capable of ex-
pressing problems and solutions connected to the phenomenon of visual
perception. The framework consists of eleven conditions, combinations
of which give us different descriptions of acts of perception (veridical
and non-veridical). Using the framework we were able to reconstruct the
reasons why the duck/rabbit case can be thought of as leading to a spe-
cific type of skepticism and show how we could avoid the skeptical con-
clusions. Along the way, we replaced some of the unclear or problematic
common notions connected with perception by their explications: “to
look at”, “to have an impression of”, “to behave as if one believed
that…”, “to react as if one had an impression of…”, “to convince oneself
that what one sees is…”. If the proposed interpretation of the duck/rabbit
case is right, then whether it poses a skeptical threat Wittgenstein feared,
remains an open question. But it is now a question to which further em-
pirical studies can give a perfectly adequate answer.
References
Bach-y-Rita, Paul, Mitchell E. Tyler and Kurt A. Kaczmarek 2003. Seeing with the
Brain. International Journal of Human-Computer Interaction, 15: 285-296.
Block, Ned 1995. On a Confusion about a Function of Consciousness. Behavioral
and Brain Sciences 18 (2), 227-247.
Cottingham, John and Peter Hacker (eds.) 2010. Mind, Method and Morality: Es-
says in Honour of Anthony Kenny. New York: Oxford University Press.
Dennett, Daniel 1989. The Intentional Stance. Cambridge, Mass.: MIT Press.
Gazzaniga, Michael, Joseph Bogen and Roger Wolcott Sperry 1962. Some Func-
tional Effects Of Sectioning The Cerebral Commissures in Man. Proceedings of
the National Academy of Sciences 48: 1765-1769.
Grabarczyk, Paweł 2013. Do Animals See Objects? In: M. Miłkowski and K. Tal-
mont Kami ski (eds.), 86-102.
Grabarczyk, Paweł (forthcoming), Kłopoty z widzeniem. O filozoficznie istotnych

u yciach wyra enia „widzie [Trouble with seeing. On philosophically relevant
uses of the term „to see”].
Grice, Herbert Paul 1961. The Causal Theory of Perception, Proceedings of the Ar-
istotelian Society XXXV, 121-153.
Jastrow, Joseph 1900. Fact and Fable in Psychology. Oxford: Houghton, Mifflin.
McNamara, Danielle S. and Gregory, J. Trafton (eds.) 2007. Proceedings of the
29th Annual Cognitive Science Society. Austin, TX: Cognitive Science Society,
Miłkowski, Marcin and Konrad Talmont Kami ski (eds.) 2013. Regarding Mind,
Naturally. Newcastle: Cambridge Scholars Publishing.
Pitcher, George 1971. A Theory of Perception. Princeton, NJ: Princeton University
Press.
Russell, Bertrand 1948. Human Knowledge: Its Scope and Limits. New York: Si-
mon and Schuster.
Shroeder, Severin 2010. A Tale of Two Problems: Wittgenstein’s Discussion of
Aspect Perception. In: J. Cottingham and P. Hacker (eds.), 352-371.
Strawson, Peter 1974. Freedom and Resentment. London: Methuen.
Valberg, Jerry 1992. The Puzzle of Experience. Oxford: Oxford University Press.
Wimmer, Marina C. and Martin J. Doherty 2007. Investigating Children’s Eye-
Movements: Cause or Effect of Reversing Ambiguous Figures? In: D. S.
McNamara and J. G. Trafton (eds.), 1659-1664.
Wittgenstein, Ludwig 1958. Philosophical Investigations. Oxford: Basil Blackwell.
Wittgenstein, Ludwig 1980. Remarks on the Philosophy of Psychology Vol. 1. Ox-
ford: Basil Blackwell.
Wittgenstein, Ludwig 1990. Tractatus Logico-Philosophicus. London And New
York: Routledge.
Arkadiusz Gut
Catholic University of Lublin
[email protected]
Priority of Thought or Priority of

Language
Abstract: It has been indicated that Frege’s view concerning the thought-language
relationship contains inner tensions. They arise from the fact that in Frege’s writ-
ings we find two inconsistent statements, namely that the structure of sentence
serves as the structure of thought and that two structurally different sentences can
express one and the same thought. In the presented paper I start with Dummett’s
three partial claims leading to the conclusion that Frege supports the hypothesis of
the priority of language over thought. Next, I show that the three statements lead to
a general rule which, according to Dummett, says that for Frege and all the later
analytic philosophers the characteristics of a thought may be obtained through a
philosophical characteristics of a sentence (or more broadly – of language). Next, I
provide a set of arguments showing that that the view suggested by Dummett en-
countered, generally speaking, many lines of criticism. At same time, I develop a
suggestion that Frege’s project advocates rather the thesis that two sentences with
different predicative structure can express one and the same thought.
Key words: Frege, thought, meaning, language, logical form, analytical philosophy
0. Introduction
Starting from the works by Dummett (1981, 1991, 1993), Bell (1987,
1996), Currie (1985), Garavaso (1991), Burge (1990), Weiner (1996,
1997) or Mayer (1990), and ending up with the works by Bermúdez
(2003, 2001), Künne (2007), Textor (2009) and Kemmerling (2010), it
was pointed out that Frege’s view concerning the thought-language rela-
tionship contained inner tensions. They arise from the fact that in Fre-
ge’s writings two inconsistent statements are found, namely that the
structure of sentence serves as the structure of thought and that two
structurally different sentences can express one and the same thought.
72 Arkadiusz Gut
In numerous works by Dummett, the statement that the structure of

thought is isomorphic to the structure of sentence appears to be an essen-
tial part of the more general view according to which Frege advocated
the priority of language over thought. The constituents of that genera-
lized view in the form of the priority of language over thought are the
claims that for Frege and all later analytic philosophers, characteristics
of a thought may be obtained through philosophical characteristics of a
sentence, or that our thinking capacity is equated with our ability to use a
language, or that thoughts are nothing more but senses of sentences, or,
finally, that Fregean conception of a sense should be assimilated to a
modern notion of the conventional linguistic meaning.
The present text discusses a set of arguments showing that the sug-
gested view – according to which Frege advocated the priority of lan-
guage over thought – is open to substantial criticism. A key task is to
show that Frege’s project rather advances the thesis that two sentences
with different predicative structures can express one and the same
thought. Accordingly, it will be put forward that particular expression of
a thought in a sentence (i.e. a particular decomposition of some thought)
proves to be just one of many possible ways a thought can be split up.
This entails additionally, as Bell, Garavaso, and Kemmerling admit, that
the assumption according to which any thought has one clearly determi-
ned (or unique) inner structure, and that one of the alternative structures
of a thought may objectively be considered better than any other, ought
to be rejected. Following this argument, however, I still believe that de-
fending the thesis that two structurally different sentences can express
the same thought does not necessarily lead to the conclusion that
thoughts should be regarded as unstructured entities (the view favored
by Bell, Garavaso, and Kemmerling).
A key aspect in the discussion opened in this text is to place emphasis
on thought itself and on showing the deeper meaning of the statement
that one and the same thought can be split up (zerlegbar) in different
ways, which means that thought is decomposable by nature. Further-
more, a seldom made distinction between two pairs of sentences will be
presented, the first pair saying: “Line a is parallel to line b” and “The
direction of line a is identical with the direction of line b” and the second
stating: “Brutus killed Ceasar” and “Caesar was killed by Brutus”. The
Priority of Thought or Priority of Language 73
first pair will be considered as an accurate exemplification of the

statement that one and the same thought is decomposable in different
ways, whereas the second will just be treated as a hint and elucidation of
understanding that language has an extensive range of expressions which
allow people to express one and the same thought in different ways.
Moreover, it will be shown that clarifying the view according to which a
clear distinction between thoughts and sentence meanings (understood as
the conventional linguistic meaning) entails considering the identity of
thought in terms of intensionally understood equipollence which refers
to the so-called epistemic identity. Such line of arguments will serve to
maintain the stance that Frege espoused the view of the priority of
thought over language.
1. The legacy of Dummett’s approach: language priority over

thought
In order to sketch two competing approaches to Frege’s stance on the

issue of thought-language relationship, I begin the paper with Dum-
mett’s three partial claims that in a systematic manner organize the in-
terpretation of Frege’s position as someone who defends the stance rec-
ognizing priority of language over thought.
(1) […] a thought is the sense of a sentence; we can present a thought only
as the sense of a certain sentence. Thus presentation of the sense is a form
of understanding an expression. (Dummett 1981: 478)
(2) The concept of sense has been introduced as something objective and
common for all language users.
(3) A thought is presented during the presentation of the semantic properties
of a sentence; to speak about the structure of a thought is to speak about the
semantic relations occurring between parts of a sentence. (Dummett 1993:
7-8)
The three statements lead to a general rule which, according to Dum-

mett, says that for Frege and all the later analytic philosophers the char-
acteristics of a thought may be obtained through a philosophical charac-
teristics of a sentence (or more broadly: of language). In his work What
Do I Know when I Know a Language? Dummett precisely reveals the
74 Arkadiusz Gut
essence of the above rule. He emphasizes that philosophers before Frege,

interpreting the value of language, saw a certain kind of code in lan-
guage. Concepts are coded in words, while thoughts, being complexes of
concepts, are coded in sentences that, generally speaking, are supposed
to reflect the structure of the former ones. According to this view,
Dummett says, we need a language, since it so happens that we do not
have telepathic abilities and thoughts are not transferred directly to a lis-
tener. Communication is then understood as the use of a telephone: the
speaker codes his thoughts in a transmittable medium and the listener
decodes them. The proposition saying that there exists a possibility to
ascribe to people possession of thoughts and concepts that is indepen-
dent from possession of knowledge concerning language is the premise
of this vision of language.
According to Dummett, the whole analytic philosophical school is in a
way founded on the rejection of the above research perspective. In the
new perspective presented by Frege, Dummett thinks, the conception of
language as a code for thought is replaced by such an interpretation of
language that does not refer to the former understanding of thoughts and
concepts. Language is understood then not as a code but as a vehicle.
Thoughts are regarded to be embodied in linguistic vehicles hence we
grasp thoughts through understanding a language and analysis of
thoughts is carried on through language analysis. In consequence,
Dummett emphasizes, it is impossible to take off the language clothes
from the thought and study it without the clothes, and we find factual
reasons why philosophical studies of language are gaining value, becom-
ing the basis for all philosophical studies. In this context, it should be
stressed that thought by its nature may not exist without a vehicle and
that language is a vehicle whose working is most evident, being in this
way most susceptible to a systematic philosophical explanation. Accor-
ding to Dummett, this view is supported by two reasons (cf. Gut 2012).
Firstly, ascribing possession of a given thought to a subject in a defi-
nite time may only be justified by the presence of a background that can
be frequently broad, against which something serving as a tool for the
thought occurs. Hence the existence of language and its knowledge by
the subject are necessary, if we consider the issue of the possession of
thoughts by somebody.
And secondly, the structure of a sentence is the only tool that reflects
the structure of the thought. It is assumed that the structure of thought
inherits the structure of the sentence, mainly because language is like a
vehicle for thoughts, and thoughts, by their nature, cannot exist without
this vehicle.
With such an approach the basic understanding of what thought is
comes from deliberations about a sentence understood as a linguistic
meaning, which leads one directly to the statement that “we grasp
thoughts through understanding the sentences that express them”
(Bermúdez 2003: 15). Hence, when we pass on to the deliberations
about the inner composition of thought, its components, and the ways in
which thoughts establish inferential relations, our attention is directed to
the logical form of the sentences in which thoughts are expressed.
The Fregean approach to thought gives us a clear answer. Thoughts are

senses of sentences, and the structure of a thought is given by the logical
form of the sentence that expresses it – where the logical form of a sentence
is given by the appropriate logical regimentation of the sentence in Frege’s
concept-script. This logical form identifies the constituents of the thought,
that is, the senses of which it is composed. (Bermúdez 2003: 16)
Analogously, when attempting to explain generativity of thought, the

starting principle is to show that thoughts are the senses of sentences and
hence generativity of thought may be described as a derivative property
in relation to generativity of language. Then it is emphasized that:
[…] There is a further and more overreaching isomorphism in play, not

simply between a thought and the sentence expressing it, but between
thought as whole and language as whole. (Bermúdez 2003: 17)
All this serves to show that the model of thought suggested by Frege is
the model where thought should be treated as a language-dependent unit
that may be exclusively possessed by and ascribed to beings that have
language at their disposal. Let us also notice that if thoughts are senses
expressed by sentences, and the structure of a thought is given to us by
analysis of the structure of the sentence expressing it, then we receive a
clear epistemological indication about the way thoughts may be identi-
fied and ascribed to other subjects at all. There is no other way out than
76 Arkadiusz Gut
to accept that a formulation of a thought cannot be anything else than an

understanding of a sentence, and the undertaken understandings cannot
be anything else than a kind of transition between sentences (cf.
Bermúdez 2003: 19; Gut 2005).
The main motif of accepting such an interpretative profile is convin-
cingly connected, however, in Frege’s claim that sentence structure and
the structure of thought are isomorphic. As Noonan (2001: 31) stresses,
such an approach “led many to assimilate Frege’s conception of sense to
modern notion of conventional linguistic meaning”. Supporters of the
view connecting the category of thought with the category of linguistic
meaning most often refer to the following:
We can regard a sentence as a mapping of the thought: corresponding to the

whole-part relation of a thought and its parts we have, by and large, the
same relation for the sentence and its parts. (Frege [1919] 1979: 255)
2. Overview of alternative approach: relative priority of thought

over language
Generally speaking, the view presented here has been subjected to criti-
cism from various perspectives.
Firstly, it has been noticed that Dummett’s description does not re-
spect numerous source texts, in which Frege seems to unambiguously
suggest his status as that of a theorist of thought. A fragment of Begriffs-
schrift is often cited on this occasion, with the phrase about the task of
philosophy which is to overcome the rule of the word over human spirit
and to liberate the thought from the burden that is only the peculiarity of
the linguistic expression (cf. Frege [1879] 1997). Additionally, in Fre-
ge’s unpublished works statements can be found, where he says that “we
can be led by language to see things in the wrong perspective, and what
value it must therefore have to philosophy to free ourselves from the
domination of language” (Frege [1884] 1979: 67). According to Weiner
(1997: 11) “someone whose writings include frequent remarks about the
importance of freeing ourselves from the domination of language is not a
likely candidate for identification as an originator of the linguistic turn”
and as someone who defends the position of a theorist of linguistic

meaning.
Secondly, Dummett is accused of wrongly interpreting the very cate-
gory of sense (Sinn), treating it in a way analogous to what Church earli-
er did, namely, as linguistic meaning. If Frege is a theorist of the linguis-
tic meaning, it has to be accepted that for him the sense (Sinn) of an ex-
pression is only the linguistic meaning that is known to everybody who
has the knowledge of the language (Church 1943). Accordingly, the
sense is what we grasp only if we understand the language. The above
conclusion has been evaluated negatively by Burge. In his article Frege
on Sense and Linguistic Meaning (1990) he remarks that in analyses of
the Fregean category of sense one must not miss the general principle:
that the category of sense was established, among others, for explaining
what thought and proposition are. Burge’s argument, which is opposite
to Dummett’s, is based on the observations that:
(i) analysis of Frege’s remarks on names, occasional expressions, personal

pronouns – from which it is to follow that the sense (Sinn) of such expres-
sions as “Gustav Lauben”, “I”, “today”, “yesterday” cannot coincide with
their linguistic meaning;
(ii) showing that the so-called history of concepts (their historical nature) is
only either the history of our cognition of concepts, or of the meaning of
words; in other words, the problem is not that the effort aiming at under-
standing the meaning of a word should not be mistaken with forming a con-
cept, but it should exclusively be defined as a process in which our under-
standing or grasping a concept is formed. (See Burge 1990: 38-41, 1979:
398-432.)
Thirdly, it is stressed that in Frege’s texts we find a lot of explanations

with proofs that achieving knowledge about thought and its structure is
not a component of the knowledge that we achieve by analyzing the
ways to express thought and the way to form conventional linguistic
meanings. In this context the spirit of comments assumes the following
shapes:
(a) It is stressed “that no amount of investigation of the actual usage and

understanding of one’s language is, according to Frege, sure to reveal the
‘deeper structure’ and the nature of sense and thought. One may have to
78 Arkadiusz Gut
achieve genuine advances in non-linguistic knowledge before one can fully

master the sense and thoughts that one’s language expresses. This point
renders Frege’s view and modern methodologies for studying language
quite different. (Wettstein 1995: 137)
(b) It is also argued that although Frege stresses “That certain sentence get
correlated with certain thought contents is, he would surely agree, an arti-
fact of human institutional arrangements. Frege would insist, however, that
one’s thinking a certain thought content is no matter of human convention,
institutional arrangement, or anything of the like. It is a matter of one’s
mind grasping an objectively existing content. More important, then, at
least in a sense more important, than the fact that sentences express such
‘thoughts’ is the fact that by uttering sentences we assert them, we give
voice to the thought contents that we are thinking […] Frege, we might say,
puts forth a thought-driven picture of language. (Wettstein 1995: 138-9)
Fourthly, it is emphasized that the stance that grasps thought as some-

thing internally connected with linguistic expression is closer to Witt-
genstein’s, and not Frege’s ideas. It has repeatedly been emphasized that
when we juxtapose Wittgenstein’s and Frege’s positions we clearly no-
tice that only with the first of them it appears that thought is internally
integrated with language, that thought is a linguistically dependent enti-
ty, and that creating thought is understood as a process accompanying
speaking.
Fifthly, it is often stressed that Dummett too rashly reduced the thesis
about the priority of language over thought exclusively to the semantic-
analytical dimension, underestimating the ontological and the epistemo-
logical aspect. In this context, it is emphasized that Frege intuitively re-
spects the above distinction. It is shown in many important fragments
stemming from Sources of Knowledge and Logical Generality where we
read that although the awareness of a thought is always connected with
the awareness of a sentence this connection does not follow from the fact
that the thought and sentence are identical.
Sixthly, it is emphasized that through accepting the more epistemo-
logical interpretation profile we acquire a basis for a different treatment
of sense. This allows one to conceive the category of sense as the way
the referent is given (Art des Gegebenseins) or as the cognitive value
(Erkenntniswert) of the sentence. Supporters of the epistemological in-
terpretation argue that a proper answer to the question where the dif-
ferent cognitive values of the two sentences: “a = a” and “a = b” come

from, should be looked for in the fact that each of the thoughts expressed
in these sentences is justified in a different way. Thus the cognitive val-
ue of the sentence is determined by the ways of its justification (Carl
1994).
3. In quest of Fregean stance
Most researchers realize that the key argument that Frege uses to support
his theory of priority of thought is ultimately reduced to showing that in
the centre of Frege’s deliberations a thesis may be proposed about the
possibility of expressing the same thought by means of sentences with
different logical structure. In this context following examples from Fre-
ge’s works may be given:
Seeing that the same thought can be worded in different ways, we learn bet-
ter to distinguish the verbal husk from the kernel with which, in any given
language, it appears to be organically bound up. (Frege [1897] 1979: 142)
Therefore let us never forget that two different sentences can express the
same thought, that we are concerned with only that part of a sentence’s con-
tent which can be true or false. (Frege [1897] 1979: 143)
To be sure, we distinguish the sentence as the expression of a thought from
the thought itself. We know we can have various expressions for the same
thought. (Frege [1924] 1979: 269)
According to Burge (1999) and Bell (1996), it is implied in the above

quotations that two different sentences may – despite the difference be-
tween them – express precisely the same thought. Presenting the prob-
lem in this way seems to be at odds with the rule according to which
there is an isomorphic connection between a thought and a sentence. If
Frege, indeed, endorses the idea that one and the same thought can be
expressed by two structurally (syntactically) different sentences, then the
assertion that a sentence functions as a model for a thought could not be
taken as a pivotal (central) claim in Frege’s works. Hence, accepting the
rule that two different sentences may express the same thought one un-
dermines the logic of the argument leading to essentially connecting
both thinking and the very thought with language practice and with the
80 Arkadiusz Gut
sentence itself. If Frege’s stance is really in opposition to the stance ex-

pressed by Dummett and others, what needs to be found in Frege’s writ-
ings is a full justification for the following thesis:
Two structurally different sentences can express the same thought

(henceforth 2S-1T thesis – two sentences-one thought).
Constructing the argument purely historically, we could, on the one

hand, say that Frege’s view that a thought can be split up in many ways,
so crucial for the 2S-1T thesis, appears in his works for many times. In
numerous places do we read
I do not believe that for any judgable content there is only one way in which
it can be decomposed, or that one of those possible ways can always claim
objective pre-eminence. In the inequality 3>2 we can regard either 2 or 3 as
the subject. In the former case we have the concept <<smaller than 3>>, in
the later <<greater than 2>>. We can also regard <<3 and 2>> as complex
subject. As a predicate we then have the concept of the relation of greater to
the smaller. (Frege [1882] 1979: 101)
(ii) We must notice, however, that one and the same thought can be split up
(zerlegbar) in different ways. The word <<singular>> does not apply to
thought in itself but only with respect to a to particular way of splitting it
up. (Frege [1906] 1979: 202)
(iii) We can decompose the proposition <<8 = 2³>> into <<8>> and <<is
the third power of 2>> or into <<2>> and <<is something whose third pow-
er is 8>>, or into <<3>> and <<is something which, when the power of 2,
yields 8>>. (Frege [1903] 1967: 197)
(iv) But we must never forget that different sentences may express the same
thought.[…] This will be surprising only to somebody who fails to see that
a thought can be split up in many ways, so that now one thing, now another,
appears as subject or predicate. The thought itself does not yet determine
what is to be regarded as the subject. (Frege [1892] 1952: 49)
We could also quote a few basic exemplifications that Frege discusses,

which are supposed to prove that sentences with a different logical and
predicative form express one and the same thought. Here are the most
frequently given collections of examples (briefly defined in the paper as
[ ] examples):
Example A (1) Line a is parallel to line b.

(2) The direction of line a is identical with the direction of line b.
Example B (3) There is at least one square root of 4.
(4) The concept square root of 4 is realized.
(5) The number 4 has the property that there is something of
which it is the square.
Example C (6) There are 0 F’s.
(7) The number of F’s is 0.
(8) The concept F is not instantiated.
However, if we assume that in Frege’s works a full justification can be

found for the above proposition that at the same time is supposed to be a
good systematic argument for recognizing independence of thought from
language, we are eo ipso obliged to show that the 2S-1T thesis is more
about thought itself than about the specificity of language. In other
words, in the process of defending the 2S-1T thesis the following ele-
ments have to be articulated:
(a) a thought may be split up in different ways;

(b) sentences that are to express one and the same thought may be
sentences with a different logical form – that is sentences with a
different predicative structure;
(c) a new formulation of a thought, that is articulation of the same
thought in a new sentence, has informative character – which
means that the passage from the first to the second articulation
becomes a valuable broadening of our knowledge;
(d) new formulations of a thought should be treated as a method of
gaining new concepts, hence sentences expressing one and the
same thought may express different concepts.
4. Elucidation of core idea: one thought two sentences
In order to understand the proper sense and dimension of Frege’s argu-

mentation, to find an essential support for the 2S-1T thesis, one has to
first separate real representative examples (described above as [ ]
examples) that are to support the thesis from the so called exempli-
fications that only suggest the general direction of Frege’s position, but
82 Arkadiusz Gut
are not par excellence examples for the claim that two sentences with
different logical structures may express one and the same thought. Most
scholars, however, who support the thesis that distinct sentences can ex-
press the same thought (2S-1T thesis) do not draw an important distinc-
tion between [ ] examples and elucidative examples. Such a distinction
seems necessary in order to understand that the 2S-1T thesis is essential-
ly based on the idea that one and the same thought is decomposable in
different ways, but not only on the idea that language has an extensive
range of expressions which allow people to express one and the same
thought in different ways. Let us start with an analysis of the following
pairs of sentences that frequently appear in Frege’s writings:
(a) At Plataea the Greeks defeated the Persians.

(a’) At Pletaea the Persians were defeated by the Greeks.
(b) Brutus killed Ceasar.
(b’) Caesar was killed by Brutus.
In many cases, when he refers to sentences in active and passive voice,

Frege draws the reader’s attention to the difference between the means
of expression and the thought expressed by these pairs of sentences. At
the same time he stresses that “language has means of presenting now
one, now another, part of the thought as subject” [Frege [1892] 1952:
49). Prima facie one could think that analysis of pairs (a and a’) and (b
and b’) supplies the foundations on which the 2S-1T thesis is based.
However, it seems that such a speculation would undermine the sense of
the whole discussion, making its philosophical value trivial. In order to
understand it, we have to distinguish the pairs of sentences in the active
and passive voice from [ ] examples and show that the motivation for
introducing the 2S-1T thesis is much more profound than the one that
may be connected with analysis of pairs of sentences in the active and
passive voice.
Firstly, let us notice that quoting examples in the active and passive
voice should be exclusively associated with the willingness to present to
the reader the introductory idea of the possibility to express one and the
same thought in two sentences and of illustrating this it by examples that
can be understood by any natural language user. Secondly, when Frege
says that sentences in the above pairs express one and the same thought,
he at the same time admits that what matters here is the differences in
the means of expression. This fact – as he stresses in Begriffsschrift – is
connected with the resources we encounter in language. Language, he
goes on to say, is free to choose a means of expression. When he refers
to pairs of sentences in the active and passive voice, Frege says nothing
about the fact that a thought may be split up in different ways, but only
that “the sentence can be transformed by changing the verb from active
into passive and at the same time making the accusative into the subject”
(Frege [1918] 1977: 9). In his work entitled Logic Frege quotes the
following pair of sentences:
(c) M gave document A to N.

(c’) N received document A from M.
He emphasizes that sentence (c) is a transformation of sentence (c’). At

the same time he remarks that owing to this transformation “we learn not
a whit more or less from the second sentence than we do from the first”
(Frege [1897] 1979: 141). The stressed lack of “a smallest difference”
has to be associated with the fact that both sentences not only express
the same thought, but do it in the same way. Hence, the possibility to
formulate the sentences in the active and passive voice is connected with
grammatical rules that are present in natural language. Frege defines
these changes as stylistic and aesthetic ones. And he sees their
significance in making better communication possible.
The linguistic significance of the position of the subject in the word-order

lies in its marking the place where what one particularly wants to draw the
attention of the listener to it put. This can have the purpose, for example,
of indicating relation judgment and others, thereby facilitating the listener’s
grasp of all interconnections. Now all those features of language that result
only from interaction of speaker and listener […].(Frege [1879] 1997: 54)
Frege places this type of transformations in the grammar of natural

language. They are changes that today we would call purely grammatical
ones that are not connected with a new split-up of the thought. In other
words, he does not connect this type of transformations with broadening
84 Arkadiusz Gut
our knowledge about thought itself (Frege [1897] 1979: 153). We can
safely assume that the change from the active to passive voice may be
done purely mechanically, without considering the formulation of the
thought expressed by the sentence. This kind of transformations does not
have to be reflected in the language created for the needs of science, as it
is only connected with the structure of natural language itself. The trans-
formations are rather result of acquiring language competence. In the
work On the Foundations of Geometry, a distinction appears between
linguistic and logical decomposition of a sentence. With the use of the
above distinction we may assume that sentences in the active and
passive voice are characterized by different language structure that is
determined by grammatical rules (Frege [1879] 1997: 54). From the
point of view of logic this situation looks different – the two sentences
have the same logical structure. In an analysis of the instances given as
exemplifications of the 2S-1T thesis, Frege will repeatedly emphasize
that the sentences differ from each other not only (referring to the
distinction given above) linguistically, but also logically. This significant
difference is connected, as will be discussed later, with the fact that in
the case of [ ] examples we deal with thoughts that are split up in
different ways. The changes, that is the transitions from one sentence to
another, are rather connected with what is going on at the level of
thought. As will be shown, they cannot be made with the use of purely
grammatical resources of transformations.
Let us move on to another group of sentences that also have to be
distinguished from [ ] examples, i.e. from the proper exemplifications of
the 2S-1T thesis:
(d) This dog howled the whole night.

(d’) This cur howled the whole night.
Frege, by means of sentences (d) and (d’), highlights such functions of

language which are connected with mood (Stimmung), atmosphere
(Duft), illumination (Beleuchtung), colouring (Färbung).
What we called mood, atmosphere, illumination in poem, what is portrayed

by intonation and rhythm, does not belong to thought. (Frege [1918] 1977:
9)
But we must not forget that language does not simply express thought; it
also imparts a certain tone or colouring to them. And this can be different
even where the thought is the same. (Frege [1906] 1979: 193)
Quoting the deliberations about such questions as mood, smell, colour-

ing or illumination proves that natural language sentences, as Frege
indicates, over and above thought and assertion often contain a third
component not covered by assertion (Frege [1918] 1977: 9-10). Colour-
ing and illumination are very often used, Frege goes on, when we want
to act on feelings and mood of the hearer, or to arouse his imagination,
that is to direct him to something that is very often hard to formulate in
thought (Frege [1918] 1977: 8-9).
In our context, this can be read as a directive for making distinctions
of the type: content of a sentence and content of a thought. They have
their utmost importance in the debate about the relations between
thought and language. Referring to the issue of mood, etc., allows Frege
to show that not everything that sentences express should be considered
a component of the thought itself. Simply, the functions of language
exceed expressing just thoughts. The content of the sentence is broader;
it also comprises what can be called colouring or embellishing.
Decoding Frege’s intention one may also say that – according to him–
the means of expression existing in language, like adding the words
“already”, “still” (Frege [1918] 1977: 9) or substituting “dog” with
“cur”, have an external character in relation to the very thought (Frege
[1897] 1979: 141). And what is more important, introducing these means
expresses the speaker’s more subjective approach; these are the means
connected rather with the component of attitude than the component of
the content of thought.
However, Frege writes about the basic purpose of the deliberations on
this subject in his work entitled Logic, where we read:
We have seen that the series of sounds that compose a sentence is often not
sufficient for the complete expression of thought. If we wish to bring the
essence of a thought into as sharp a focus as possible, we ought not to over-
86 Arkadiusz Gut
look the fact that converse case is not uncommon, the case where a sentence
does more than express a thought and assert its truth. In many cases a sen-
tence is meant to have an effect on the ideas and feeling of the hearer as
well; and the more closely it approximates to the language of poetry, the
greater this effect is meant to be. (Frege [1897] 1979: 139)
According to this hint, we have to be sensitive, in Frege’s opinion, to the

complexity of language and to the multiplicity of its different functions,
which make the content of a sentence to include more than only the
thought it expresses. Hence, the ability of referring to the thought itself
is at the same time connected with the ability to separate, what Frege
calls, trappings of the very thought. Also it is important to stress that
changing some words or changing their order, although it does not
change anything in the thought itself, can significantly change the
presentations and emotions that appear in the hearer of the words. This is
connected with the fact that “sounds of language act as a sensory stimu-
lus” and influence associations that occur during the comprehension of
language messages. The point is that when two sentences produce two
different reflections or representations (Vorstellungen) which comprise
emotions, mood or attitude, then it should not be concluded on that basis
that the two sentences express two different thoughts.
In this sense, if we change the word “horse” into “steed”, we are
moving on the level of subjective states, something that should be linked
with our ability to imagine something that belongs to the content of the
sentence. (Frege [1897] 1979: 142) In order to defend the 2S-1T thesis,
it seems necessary to redirect this discussion to the level of conceptual
contents. In other words, when we change the word “dog” to “cur”, we
refer to our imagination or association net; and when we change the
parallelism of lines to identity of directions, it is hard to talk about a
recourse to our associations. Eo ipso, the valuable broadening of our
knowledge that, in Frege’s opinion, is given by [ ] examples should not
be linked with changes in attitude or illocutive power, but with changes
in the very contents of thought.
5. Elaborating the priority of thought approach
If we assume that up to this moment we have been able to show what the
2S-1T thesis should not be directly associated with, we now have to
show three further problems. We have to analyze pairs of sentences that
are considered proper exemplifications of the thesis about the possibility
of expressing one and the same thought by different sentences. Then we
have to show the way in which Frege justifies the thesis that sentences
differing in their logical structure and whose components express diffe-
rent concepts, may ultimately express one and the same thought. And
finally, the key problem is to give a criterion of identity of the thought
that, let us add, may be expressed in various ways. At each of these stag-
es we cannot escape the question concerning the crystallization of the
understanding of the sense by building opposition not only to presen-
tation and coloring (which have been mentioned above), but also to the
linguistic meaning. In order to cast more light on the issue we are now
dealing with, it is worth quoting two contexts in which Frege uses the
rule that the same thought may be expressed by two different sentences.
The first context comes from Foundations of Arithmetic:
The judgement <<line a is parallel to line b>>, or using symbols a || b can

be taken as an identity. If we do this, we obtain the concept of direction,
and say: <<the direction of line a is identical with the direction of line b>>
[…] We carve up the content in a way different from original way, and this
yields us a new concept. (Frege [1884] 1953: 75)
The second context is a fragment of the essay On Concept and Object in

which Frege seems to suggest that the thought concerning the square
root of 4 may be expressed in the following ways:
(3) There is at least one square root of 4.

(4) The concept square root of 4 is realized.
(5) The number 4 has the property that there is something of which
it is the square.
What we have here is a clear suggestion that sentences expressing the

same thought do not have to share the same logical structure. The possi-
88 Arkadiusz Gut
bility to express the same thought by sentences with different predicative

structures is supposed to confirm the rule that for the same thought there
exist many sentences differing from one another logically (cf. Garavaso
1991: 202; Beaney 1996: 134). If such a possibility is admitted, it may
not be said that parts of a thought are permanently ascribed to parts of a
sentence, or that expressing a thought in a sentence leads to distin-
guishing primary or the only components of the thought (Currie 1985).
Moreover, we see that differences between sentences that can express
the same thought may reach a much deeper level than the differences
between sentences listed in the comparisons given above [(a)-(a’), (b)-
(b’), (c)-(c’), (d)-(d’)]. In order to pass on, in Frege’s strategy, to the es-
sential motif, one has to start with the statement, in which the words are
uttered, saying that there are no obstacles for the situation that “now one,
now another, part of the thought could be presented as the subject” (Fre-
ge [1892] 1952: 49). There is a strong suggestion contained in it that the
thought in its “nature” makes it allowable “that one way of analyzing a
given thought should make it appear as a singular judgement; another, as
a particular judgement; and a third, as a universal judgement” (Frege
[1892] 1952: 49-50).
Let us notice that the equivalences Frege introduces:
(1) Line a is parallel to line b.

clearly show that at the foundations there is a deep conviction that the
sentences in the above pairs express the same thought – they have the
same conceptual content. Passing from (1) to (2) Frege clearly emphasi-
zes that he wants to express the same content (thought), but in a different
way. Let us also remember that if the above pairs of sentences did not
express the same thoughts, that is if they did not give the same set of in-
formation, the process of analysis underlying reduction of arithmetic to
logic would be utterly destroyed (Beaney 1996: 135-140; Currie 1985).
So the idea is that the same conceptual contents or judgeable contents
may be grasped in various ways. In this sense various ways of express-
ing the same thought are de facto consequence of different decomposi-
tions of the same thought. It is so, because the rule according to which
the same thought may be “split up” in various ways, is the assumption
underlying these propositions.
This will be surprising only to somebody who fails to see that a thought can
be split up in many ways, so that now one thing, now another, appears as
subject or predicate. (Frege [1892]1952: 49)
However, Frege connects the basic argument with the thought itself, stat-
ing that:
The thought itself does not yet determine what is to be regarded as the sub-
ject. (Frege [1892]1952: 49)
We can see a distinct move towards discussing the thought, the content
itself, that is subject to different decompositions, and only on this basis
is the proposition formulated that different sentences with different logi-
cal structures may express one and the same thought. We remember, that
when Frege says that sentences:
(1) Line a is parallel to line b.

express one and the same thought, he automatically adds that the transi-
tion from (1) to (2) consisted in splitting up the content (the thought) in a
different way, and this new split-up resulted in passing from speaking
about parallelism of the lines to speaking about equality of the direc-
tions. We know that using this example is subjected to showing a similar
relation between the sentences:
(9) There are as many F’s as G’s.

(10) The number of F’s is the same as the number of G’s.
When we compare the new split-up of the content, not only does a new
concept appear (respectively: direction and number), but also relations
change (respectively: we are not speaking about the relation between
concepts, but between objects). Hence splitting the contents goes “deep”,
90 Arkadiusz Gut
and the sentences that are supposed to express the same thought may be
distinguished from each other by their components and structures.
Before we pass on to the question of identity of thoughts, let us dis-
cuss two more things. Firstly, if a thought was connected with a sen-
tence, that is if we could talk about the composition of a thought only in
the derivative sense as far as the structure of the sentence was con-
cerned, or if the feature of generativity was inherited from generativity
of language means, then not only saying that different sentences express
one and the same thought would be excluded, but also the idea that sub-
ject and predicate are not pre-determined in the content of a thought
would not be the raison d’etre. And secondly, it is already at this stage
that we realize that the view according to which the above pairs of sen-
tences, despite significant differences, express one and the same thought
is only possible to accept when the thought is not numerically identified
with something we call conventional linguistic meaning. Transferring
these conclusions to the field of deliberations about the category of
sense, one has to say that (Fregean) sense may not be identified with the
term of linguistic meaning used at present, that is with something that is
created only by speakers of a language (Burge 1990).
Attention should be paid to one thing: a new decomposition of the
content is a way that leads to obtaining a new concept that appears due
to splitting the contents of the thought. This is a kind of such an intellec-
tual activity thanks to which the new concept is in a way brought to light
as a result of an analysis of the content of the thought that we express in
the first articulation. This is why, when talking about a new concept,
Frege keeps persuading the reader that we are dealing with the same
conceptual content. His reasoning is more or less such that in a new
split-up of the thought we do not extend beyond the contents of the giv-
en thought, but remaining in its scope we split it up in such a way that
allows a new articulation. To explain it visually in a way, it may be said
that what matters here is dividing the same field of, for example, a
square, in different ways, once by distinguishing four squares, and then
by distinguishing four triangles in the square (cf. Beaney 1996). Alt-
hough the division of the field of the square is different each time, we do
not exceed the limits of the square. This is what Frege means when in
paragraph 88 of Foundations of Arithmetic he writes:
If we represent the concepts (or their extensions) by figures or areas in

a plane, then the concept defined by a simple list of characteristics corre-
sponds to the area common to all the areas representing the defining charac-
teristics; it is enclosed by segments of their boundary lines. With a defini-
tion like this, therefore, what we do – in terms of our illustration – is to use
the lines already given in a new way for the purpose of demarcating an area.
Nothing essentially new, however, emerges in the process. But the more
fruitful type of definition is a matter of drawing boundary lines that were
not previously given at all. What we shall be able to infer from it, cannot be
inspected in advance; here, we are not simply taking out of the box again
what we have just put into it. (Frege [1884] 1953: 100)
At this stage of our deliberations, when we want to support the inter-

pretation suggesting independence of thought from language, it is worth
quoting Frege’s remark from Boole’s logical Calculus and the Concept-
script, where we read that the possibility of decomposing a thought into
the argument and the function supposes that the thought (judgable con-
tent) is internally complex (in sich gegliedert) (Frege [1880/81] 1979: 9-
12). Not only does this confirm the fact that no actual (unique) logical
form or actual composition of the parts is ascribed to a thought, but it
also emphasizes that a thought should be treated as something internally
complex. It is just a thesis about inner complexity of a thought that is
one of the arguments because of which Frege stresses the exceptional
value and advantage of the analysis with the use of the argument-
function pair over the analysis with the use of the subject-predicate pair.
The significant advantage of analysis by means of the argument-function
pair stems from the fact that in its course not only the argument and the
functional expression in the actual case are given, but the possibility of
an alternative analysis is offered as well. Frege writes about it in Be-
griffsschrift, presenting an example with hydrogen and carbon dioxide.
From the presented analysis we learn that in the course of splitting the
thought “hydrogen is lighter than carbon dioxide” into the argument
“hydrogen” and the function “to be lighter than carbon dioxide” we also
acquire the knowledge that the same content may be split up in a diffe-
rent way, namely, so that “carbon dioxide” will become the argument,
and “being heavier than hydrogen – the function. (cf. Frege, [1879]
1997; Beaney 1996: 238) So when formulating the content by means of
92 Arkadiusz Gut
the argument-function pair in one way, we see that it can be also for-
mulated in a different way. In this respect, an analysis by means of the
argument-function pair shows the truth of that a thought may be analy-
zed in various ways, i.e. for each judgeable content there is no one way
in which it can be split up. This fact reveals that thought is an entity
which is decomposable by nature. A primary characteristic of thought
constitutes its division into parts. Therefore, using a certain analogy re-
ferred to by Frege himself, thought – just as the area of a square – has
the potential for being split up in different ways (i.e. inside a square four
identical squares or four identical triangles may be distinguished).
How can one formulate conditions for identity of a thought in such a
situation? First, let us notice that Frege seems to realize that identity of
reference (Bedeutung) does not entail identity of the thought. “From
identity (Gleichheit) of reference (Bedeutung) there does not follow
identity of the thought” (Frege [1891] 1952: 30-32). In other words, a
definition of identity of the thought cannot be given exclusively on the
basis of identity of the meaning (reference) – since identity of the
thought is a stronger relationship than identity of the meaning (Bedeu-
tung). If two sentences express one and the same thought, then by defini-
tion they share the same logical value. However, if they only share the
same reference, we have no grounds for recognizing the two sentences
as expressing one and the same thought. Next, it has to be remembered
that when we speak about identity of the thought, we speak about identi-
ty of something that has to be characterized as cognitive content in the
first place, that is, what the subject formulates and then recognizes. Tak-
ing into consideration the logical side, the level of reference
(Bedeutung), is of utmost importance, albeit it is insufficient. Under such
circumstances, a fragment from the work A brief Survey of my logical
Doctrines seems to be a crucial one; Frege formulates the following cri-
terion there:
Now two sentences A and B can stand in such a relation that anyone who
recognizes the content of A as true must thereby also recognize the content
of B as true and, conversely , that anyone who accepts the content of B
must straightway accept that A. (Equipollence). It is here being assumed
that there is no difficulty in grasping the content of A and B. The sentences
need not be equivalent in all respects. For instance, one may have what we
may call a poetic aura, and this may be absent from the other. Such a poetic
aura will belong to the content of the sentence, but not to that which we ac-
cept as true or reject as false. I assume there is nothing in the content of ei-
ther of two equipollent sentences A and B [that is, not only logically equi-
pollent with respect to reference (Bedeutung), but equipollent in terms of
the content – A.G.] that would have to be immediately accepted as true by
anyone who had grasped it properly. The poetic aura then, or whatever else
distinguishes the content of A from that of B, does not belong to what is ac-
cepted as true; for if this were the case, then it could not be an immediate
consequence of anyone’s accepting the content of B that he should accept
that of A. For the assumption is that what distinguishes A and B is not con-
tained in B at all, nor is it something that anyone must recognize as true
straight off. […] So one has to separate off from the content of a sentence
the part that alone can be accepted as true or rejected as false. I call this part
the thought expressed by the sentence. It is the same in equipollent sentenc-
es of the kind given above. It is only with this part of the content that logic
is concerned. I call anything else that goes to make up the content of a sen-
tence the colouring of the thought. (Frege [1906] 1979: 197-198)
Quoting the above fragment is extremely important, as it is impossible to

grasp in the proper way the sense of the first sentence in which Frege
formulates the criterion according to which two sentences express one
and the same thought, without citing the conditions under which the cri-
terion is met. The conditions stated in the fragment above clearly imply
that Frege refers to the subject and its cognitive abilities. Then the atten-
tion is not focused on the logical value of the content itself but on the
way the content is conceived and comprehended by the subject. As a re-
sult, two sentences express the same thought only when someone who
accepts the content (thought) of sentence A as being true also accepts –
directly and with no obstructions – the content (thought) of sentence B
as being true. It is not the truthfulness itself – or, in other words, sharing
the same reference (Bedeutung) – that matters, but accepting something
as true, certainly preceded by grasping the content of the thought. A cri-
terion given in this way assumes that everyone who grasps the thought
expressed by sentence A and the thought expressed by sentence B, has to
accept the first of these sentences directly and with no obstructions, also
accept the second one. Here we see a clear reference to the subject’s
cognitive activity. It seems that going in this direction Frege realizes that
94 Arkadiusz Gut
without taking into consideration the subject, every other attempt to give
a definition of identity would be too coarse; it would aim at identity of
reference (Bedeutung). The question of truth – or in other words, refer-
ence – does not disappear. However, it is not examined in abstracto,
without taking into consideration the specificity of the very content and
the way the content is grasped. When we look at the above characteris-
tics considering identity of thought, we see that it says that everybody
who accepts content A as true, has to (immediately) accept content B.
Going further, two next things are crucial for this explanation. The first
one, that the object referred to by the subject is content (Sinn), and not
reference (Bedeutung). And the second one, that the content appears as
the object of oblique speech (oratio obliqua). For this reason exchanging
A for B cannot happen only on the basis of sharing the same logical val-
ue; if it were so, a subject accepting a proposition as true would have to
accept also all the other propositions that would be logical consequences
of the first proposition. If we want to avoid such a situation, it has to be
said rather that A and B express one and the same thought when and on-
ly when A and B are exchangeable (salva veritate) in oblique contexts
(oratio obliqua). In other words, if two sentences express the same
thought, the rule of exchangeability (salva veritate) has to apply to in-
tensional, and not only to extensional contexts. Two sentences A and B
have to be intensionally equivalent expressions. Knowing Frege’s expla-
nations about oblique speech and intensional contexts we know that in-
stead of the term “thought” we may use the term “sense”, and say that
two sentential expressions A and B express the same sense if they are
intensionally equivalent, that is, when they may be exchangeable in
oblique contexts. The criterion of identity outlined in this way distinctly
proves that thought/sense is a kind of stuff constituting the content of our
cognitive attitudes. Hence we may state that when Frege establishes
identity of thoughts, he means identity of cognitive content, something
that is contained in the subject’s cognitive horizon. Such an expression
of the notion of sense/thought that is fine-grained allows to clearly dis-
tinguish the content of a thought from the way the thought is expressed
(Künne 2007). When the criterion of thought identity is adopted, which
refers to intensionally captured equivalence, then, as Beaney (1996: 254-
255) argues, a thought from the thought as expressed may be distin-
guished, what finally makes it possible to show a wider dimension of

Fregaean stance for the priority of thought over language.
References
Beaney, Michael 1996. Frege Making Sense. London: Duckworth.

Beaney, Michael (ed.) 1997. The Frege Reader. Oxford: Blackwell Publishers.
Bell, David 1987. Thoughts. Notre Dame Journal of Formal Logic 28 (1), 36-49.
Bell, David and Neil Cooper (eds.) 1990. The Analytic Tradition. London: Black-
well Press.
Bell, David 1996. The Formation of Concepts and the Structure of Thoughts. Phi-
losophy and Phenomenological Research 56 (3), 583-596.
Bermúdez, José Luis 2001. Frege on Thoughts and Their Structure. Logical Analy-
sis and History of Philosophy 4, 87-105.
Bermúdez, José Luis 2003. Thinking without Words. Oxford: Oxford University
Press.
Burge, Tyler 1979. Sinning against Frege. Philosophical Review 88, 398-432.
Burge, Tyler. 1990. Frege on Sense and Linguistic Meaning. In: D. Bell and N.
Cooper (eds.), 30-60.
Carl, Wolfgang 1994. Frege’s Theory of Sense and Reference. Cambridge: Cam-
bridge University Press.
Church, Alonzo 1943. Carnap’s Introduction to Semantics. Philosophical Review 52
(3), 298-304.
Currie, Gregory 1985. The Analysis of Thoughts. Australian Journal of Philosophy
63, 283-298.
Dummett, Michael 1981. The Interpretation of Frege’s Philosophy. London:
Duckworth.
Dummett, Michael 1991. Frege and Other Philosophers. Oxford: Oxford Universi-
ty Press.
Dummett, Michael 1993. Origins of Analytical Philosophy. London: Duckworth.
Dummett, Michael 1993a. What Do I Know when I Know a Language? In: The
Seas of Language. Oxford: Clarendon Press, 94-105.
Beaney (ed.), 47-78.
Frege, Gottlob [1880] 1979. Boole’s logical Calculus and the Concept-script (trans-
lated by P. Long and R. White). In: H. Hermes et al. (eds.), 9-46.
Frege, Gottlob [1882] 1980. Letter to Marty 29.08.1882 (translated by H. Kaal). In:
G. Gabriel et al. (eds.), 99-102.
96 Arkadiusz Gut
Frege, Gottlob [1884] 1953. Foundations of Arithmetic. A logic-mathematical en-

quiry into the concept of number (translated by. J.L. Austin). Oxford: Blackwell
Press.
Frege, Gottlob [1891] 1952. Function and Concept (translated by P.T. Geach). In:
P.T. Geach and M. Black (eds.), 21-41.
Frege, Gottlob [1892] 1952. On Concept and Object (translated by P.T. Geach). In:
P.T. Geach and M. Black (eds.), 42-55.
Frege, Gottlob [1897] 1979. Logic (translated by P. Long and R. White). In: H.
Hermes et al. (eds.), 126-151.
Frege, Gottlob [1906] 1979. Introduction to Logic (translated by P. Long and R.
White). In: H. Hermes et al. (eds.), 185-196.
Frege, Gottlob [1906] 1979. A Brief Survey of My Logical Doctrines (translated by
P. Long and R. White). In: H. Hermes et al. (eds.), 197-202.
Frege, Gottlob [1918] 1977.Thoughts (translated by P.T. Geach and R.H.
Stoothoff). In: P.T. Geach (ed.), 1-30.
Frege, Gottlob [1923] 1979. Logical Generality (translated by P. Long and R.
White). In: H. Hermes et al. (eds.), 258-262.
Frege, Gottlob [1924] 1979. Sources of Knowledge of Mathematics and Natural
Sciences (translated by P. Long and R. White). In: H. Hermes et al. (eds.), 267-
274.
Gabriel, Gottfried, Hans Hermes, Friedrich Kambartel, Christian Thiel and Albert
Veraart (eds.) 1980. Gottlob Frege Philosophical and Mathematical Corre-
spondence. Chicago: The University of Chicago Press.
Geach, Peter T. and Max Black (eds.), 1952. Translations from Philosophical Writ-
ings of Gottlob Frege. Oxford: Oxford University Press.
Geach, Peter T. (ed.) 1977. Logical Investigations. Oxford: Blackwell Press.
Gut, Arkadiusz 2005. Gottlob Frege i problemy filozofii współczesnej [Gottlob Fre-
ge and the Problems of Contemporary Philosophy]. Lublin: Wydawnictwo
KUL.
Gut, Arkadiusz 2012. Gottlob Frege’s Thought in the Light of Inquiries Conducted
in Contemporary Philosophy. Studies in Logic and Theory of Knowledge 7,
125-162.
Hermes, Hans, Friedrich Kambartel and Friedrich Kaulbach (eds.) 1979. Gottlob
Frege. Posthumous Writings. Chicago: The University of Chicago Press.
Kemmerling, Andreas 2010. Thoughts without parts: Frege’s Doctrine. Grazer Phi-
losophische Studien 82, 165-188.
Künne, Wolfgang 2007. A Dilemma in Frege’s Philosophy of Thought and Langu-
age. Revista di estetica 34, 95-120.
Mayer, Verena. 1990. Zerlegung und Struktur von Gedanken. Grazer Philosophi-
sche Studien 37, 31-57.
Noonan, Harold 2001. Frege. A Critical Introduction. Cambridge: Polity Press.
Tait, William W. (ed.) 1996. Early Analytic Philosophy: Essays in Honor of Leon-
ard Linski. Chicago: Open Court.
Textor, Mark 2009. A Repair of Frege’s Theory of Thoughts. Synthese 167, 105-
123.
Weiner, Joan 1996. Has Frege a Philosophy of Language? In: W.W. Tait (ed.), 249-
271.
Weiner, Joan 1997. Frege’s Logic and the Theory of Meaning. Acta Analytica 18, 7-
22.
Wettstein, Howard 1995. Has Semantics Rested on a Mistake? And Other Essays.
Stanford: University Press.
Thomas J. Hughes
University of Durham
[email protected]
On the Ambiguity in Definite

Descriptions
Abstract: The following paper will defend the thesis that definite descriptions
should receive an attributive semantics only in those instances where they fall under
the scope of certain semantic operators. The ambiguity defended will therefore not
be the one noted by Donnellan (1966, 1968) and Kripke (1977), as it will not recog-
nise an ambiguity in one and the same proposition. Each and every proposition will
receive a single semantics for every occasion of use, and this semantics will be in-
formed by the grammatical configuration of the expressions. If the definite descrip-
tions falls under the scope of a particular semantic operator then it is attributive, if
not then it is referential.
Keywords: definite descriptions, demonstratives, reference, quantification, phases,

Russell
0. Introduction
The theory of descriptions advocated by Russell in his landmark paper

‘On Denoting’ (1905) argued that a proposition containing a definite de-
scription is to be assigned a quantificational semantics. In addition Rus-
sell stated that definite descriptions were not to be assigned a semantics
in isolation from a proposition. This theory was directly at odds with
Russell’s original position in the Principles of Mathematics (1903)
wherein he took definite descriptions to be referential and equated them
with constants. The position advocated in the following paper is that cer-
tain propositions containing definite descriptions are referential and oth-
ers quantificational, and that these facts are to be explained gramma-
tically. Additionally it will defend a modification of Russell’s position
that definite descriptions cannot be assigned a semantics in isolation
from a proposition, the modification will be that definite descriptions
100 Thomas J. Hughes
receive a semantics that is dependent on the larger grammatical structure

which is found within. The semantics of a definite description therefore
is dependent on a larger grammatical configuration, which will be ex-
plained below.
Keith Donnellan (1966, 1968) illustrated that in some instances defi-
nite descriptions are used referentially and in others, attributively. Criti-
cally Donnellan stated that the ambiguity occurs in one and the same
proposition and depends on the context of use. The sentence in (1) could
receive the paraphrase in (2a) or (2b) depending on the usage:
(1) Smith’s murderer is insane.

(2) a. Jones is insane.
b. Whoever murdered Smith is insane.
(2a) is the paraphrase of a referential use, whereas (2b) is a paraphrase of

an attributive use. It was suggested by Donnellan that this ambiguity was
one that occurred at the level of pragmatics rather than semantics (1966:
297). Kripke supported this and stated that the semantics of (1) should
be uniform across uses (1977: 262). The semantics of (1) should there-
fore be universally equated with the logical form in (3) irrespective of
how it is used on a particular occasion:
(3) ∃x( (x) & ∀y( (y) → y = x) & (x))
When using a description referentially the proposition retains the seman-

tics of (3), but through pragmatic effects akin to conversational implica-
tures (e.g. the speaker’s intention) it can be used to make reference to a
particular individual.
The problem with this treatment of the ambiguity is that by consider-
ing it as present in one and the same proposition a semantic difference
would be both unappealing and unintuitive. However, a further ambigu-
ity appears to rear its head in (4) and (5):
(4) The table is covered with books.

(5) The mother of every boy is proud of him.
On the Ambiguity in Definite Descriptions 101
In (4) we have a deictic use of the definite description, which if felici-

tous would be a referential use. Alternatively (5) is only felicitous if
used attributively, that is, it is felicitous only in the instance where we
state of all mothers that they are proud of their sons. (5) is interchange-
able with the paraphrase every boy’s mother is proud of him. The dis-
tinction between (4) and (5) is clear in that one is noticeably referential
and the other is noticeably attributive. It is the contention of this paper
that this ambiguity is the one present in definite descriptions, rather than
an ambiguity in one and the same proposition (I will continue to use the
term ambiguity even though, strictly speaking, ambiguity is by definition
a property of a single proposition).
1. Referential uses
One preliminary thing that needs to be assured is that (4) is indeed refer-
ential, and that it is referential in its semantics. The first point to flag up
is that (4) is deictic and is only felicitous if spoken in a context in which
a particular table is present. It is reasonable to suppose therefore that the
definite description is acting in a similar fashion to other deictic struc-
tures, such as demonstrative descriptions. This position is held in various
forms by at least Kaplan (1970), Devitt (1981), and Wettstein (1981,
1983). Kaplan defends the view that there is an identical semantic com-
ponent, ‘dthat’, acting to create the similarity between (6) and (7):
(6) Dthat [“the table”] is covered with books.

(7) Dthat [“that table”] is covered with books.
The identical element he terms ‘dthat’ acts as an ostensive device for

making reference, a device that is akin to pointing but expressed in lan-
guage. Kaplan states of this similarity “if pointing can be taken as a form
of describing, why not take describing as a form of pointing?” (1970:
223)
If there is a semantic ambiguity in definite descriptions then there are
instances when using a definite description will result in a singular prop-
osition and instances wherein it will result in a general proposition. The
instances where a definite description contains the ‘dthat’ element are

those where the proposition created is singular. For Kaplan the sense of
the demonstrative description and the definite description is not present
in the proposition itself but is instead used to identify the referent and
load that referent into the proposition. In Kaplan’s words “[i]nstead of
taking the sense of the description as subject of the proposition, we use
the sense only to fix the denotation which we then take directly as sub-
ject component of the proposition” (1970: 233). The referent, be it a
concrete referent or a discourse referent, is then loaded into the pro-
position, this is in contrast to the general propositions created in attribu-
tive uses where the sense is contained in the proposition rather than the
referent.
Devitt (1974, 1981, 2007, forthcoming) also defends the view that
there is a semantic ambiguity and that the referential use emerges in vir-
tue of its link to demonstratives, amongst other phrases (pronouns, prop-
er names). What Devitt says is that a referential description is such in
virtue of the fact that there is a ‘causal-perceptual’ link between the ex-
pression the F and an object (1974: 186, 1981: 515, 2007: 22, forth-
coming: 2). The explanation of this goes as follows “[t]here is a seman-
tic convention of using ‘the F’ to refer to x which exploits both a causal-
perceptual link between the speaker and x and the meaning of ‘F’ …
such a link accompanies all referential uses. It is then plausible to think
that it plays a semantic role, just as it does with a deictic pronoun or
demonstrative” (2007: 22). A proposition containing a definite descrip-
tion that makes use of this causal-perceptual link creates a singular
proposition, and is therefore semantically referential. Devitt’s position
again ties together definite descriptions with demonstratives, and states
that they have a common method of picking out their referents.
There is a clear intuition in both Kaplan’s and Devitt’s theories that
referential descriptions act in a way akin to demonstratives. The intuition
exists partially in virtue of the fact that the two can be used inter-
changeably in certain grammatical contexts without loss or alteration of
the proposition’s meaning. This fact is illustrated in (8) and (9) below:
(8) a. The table is covered with books.

b. That table is covered with books.
(9) a. Juan is such a bad driver the idiot will get himself killed.
b. Juan is such a bad driver that idiot will get himself killed.
Further examples can be given but (8), the deictic use, and (9), the epi-
thet use, illustrate the point at hand. However, whilst there are clear sim-
ilarities between the two types of expressions there are also clear differ-
ences. For example, it is impossible to use (10) and (11) interchangea-
bly, and this has something to do with the semantics of demonstratives:
(10) That is red.

(11) *The is red.
The fact is that demonstratives can appear without any noun phrase
complement whereas the definite article demands one. Before turning to
the reason why this distinction exists let us consider some further diffe-
rences where it is not possible to use the two types of expression inter-
changeably, consider the following pairs of sentences:
(12) a. Barcelona F.C. lost and the fans were furious.

b. Barcelona F.C. lost and those fans were furious.
(13) a. Juan crashed his car and the guy was furious.
b. Juan crashed his car and that guy was furious.
(14) a. I always get the bus.
b. I always get that bus.
The (a) sentences are not interchangeable with the (b) sentences in a way
similar to those pairs in (8) and (9). In each of (12-14) the (b) sentences
are interpreted deictically, that is relative to a referent that is salient in
the context. In (12) the definite description in (a) is felicitous primarily
in the context where we are talking about each and every Barcelona fan,
whereas in (b) the demonstrative description those fans picks out a sub-
section of the fans and would be felicitous only with a salient set of se-
lected fans. In (13) the definite description in (a) picks out ‘Juan’,
whereas in (b) the demonstrative is naturally interpreted as picking out
someone other than ‘Juan’. In (14) we see the same facts arise, (a) con-
tains an attributive description stating only a preferred mode of trans-
port, whereas in (b) the demonstrative description once again picks out a
salient particular bus.
The obvious conclusion to draw from the examples in (12-14) is that
demonstrative descriptions force a deictic reading in sentences where a
definite description does not. Demonstrative descriptions are therefore
more deictic than definite descriptions. These facts illustrate a difference
between the two that should be noticed when giving an account of the
semantics of such expressions, a fact that both Kaplan and Devitt appear
to overlook. The forced deictic readings for a demonstrative description
is born out of the fact that the demonstrative contains locational infor-
mation relative to the speaker/addressee that is absent in the definite ar-
ticle. Demonstratives can be seen as the morphological realization of the
equivalences in (15):
(15) That = [the + there]

This = [the + here]
The locational information gives us two distinct demonstratives, the dis-

tal demonstrative to pick out individuals distant from the speaker and the
proximal demonstrative that picks out individuals that are closer to the
speaker. The locational information is what causes these expressions to
be strongly deictic.
Interestingly, when a definite description and a demonstrative counter-
part are indeed interchangeable, the locational information of the demon-
strative is uninterpreted. In addition, it is always the distal demonstrative
rather than the proximal that is capable of being used. This fact becomes
more important when we consider that it is reflected cross-linguistically
in languages typically taken to lack determiners. Cyr (1993) notes that
Montagnais has begun to develop a definite article through the distal
demonstrative, and when the distal is used in the definite article manner
its locational information is not relevant to its interpretation. Huang
(1999) states that the same is true of spoken Mandarin. We can say that
the definite article is akin to a distal demonstrative except for the fact
that the article has been bleached of its locational information. These
facts imply that there is indeed a strong relationship between the definite
article and the distal demonstrative, yet there remain those differences
listed above.
2. The Determiner Phrase
The facts expressed above all have to do with the grammar of the deter-
miner phrase. It is within the determiner phrase that we see the simi-
larities and differences captured between the distal demonstrative and
the definite article. In addition an argument will be made to the tune that
the determiner phrase structures the referential capabilities of those ex-
pressions that fall in it, including the referential capabilities of definite
descriptions. It will be observed that definite descriptions fulfil the re-
quirement forced by the determiner phrase for being a referential expres-
sion, and that if no semantic operator alters the expressions interpreta-
tion then they receive a referential semantics.
Before turning to how we get the attributive readings of definite de-
scriptions it is important to secure their referential readings. Definite de-
scriptions fall grammatically within what is called the determiner phrase.
The determiner phrase contains a phrase head, the determiner position
(D), a complement, typically a noun phrase (NP), and a specifier posi-
tion (specD). The structure can be illustrated in the following tree:
(16)
The diagram in (16) depicts the grammatical structure of the determiner

phrase. The D position is that occupied by both the definite article and
demonstratives. When the D position is filled with a definite article the
noun phrase must also be filled, but when the D position is filled with a
demonstrative the noun phrase may remain empty.
Two further aspects of the diagram need to be explained, that is the
edge and interior dichotomy. Following Sheehan and Hinzen (2012), the
proposal here is that the edge of the determiner phrase is the locus of
reference. That is, the more the interpretation of an expression relies on
the interpretation of its edge, the more referential it is. The phrase struc-
ture of DPs gives us three logical possibilities for grammatical construc-
tion, the first are those instances where only the interior is filled, the
second are those instances where both the edge and the interior are
filled, and the final instances are those, like proper names, where only
the edge is filled. Sheehan and Hinzen neatly correlates these gramma-
tical facts with facts about forms of reference, they state that:
[W]here the phrase head D is radically underspecified, only the phrase inte-
rior is interpreted and reference is purely configured via the NP descrip-
tion… In definite specific readings, both the phrase edge and the phrase in-
terior appear to enter the interpretive process, and a reading that is inter-
mediate between a referential and a purely quantificational one is derived.
Finally, after substitution of the head of the phrase interior into the phrase
edge, only the edge can be interpreted, and no restriction enters the way in
which reference is configured, giving rigidity. (Sheehan and Hinzen 2012:
12)
The DP therefore creates a spectrum of referential possibilities merely in

virtue of possible valuations for the edge and interior.
The fact is that definite descriptions fill both the edge and interior of
the determiner phrase in order to be grammatical. In virtue of doing so
they can swing between referential and attributive readings. Demonstra-
tives do not require a noun phrase complement and as such are biased
towards the edge and consequently referential readings; however in in-
stances where the interior is filled they can indeed by attributive (see
King 2001 for examples). The determiner phrase creates a spectrum of
reference as illustrated below (with * standing for ungrammatical and ¬
for negation):
(17) a. Bare Nouns [DP *D [NP]]

b. Indefinite Descriptions [DP a *¬[NP]]
c. Definite Descriptions [DP the *¬[NP]]
d. Demonstratives [DP that ¬*¬[NP]]
e. Proper Names [DP John [*NP]]
f. Pronouns [DP I [*NP]]
The spectrum ranges from completely non-referential expressions as in

(a) to those that are rigidly referential (e-f). We can see that as expres-
sions become increasingly referential the ungrammatical star moves
from the edge to the interior, that is bare nouns are ungrammatical with
the edge filled, and proper names/pronouns are ungrammatical with the
interior filled. Definite descriptions fall in the middle.
The spectrum above illustrates how both the differences in (10-14)
and the similarities in (8-9) come about. The similarities exist in virtue
of the fact that both types of expression are felicitous with the edge and
interior filled, and as such are both capable of referential readings. The
condition on being referential is that the edge must be filled, both ex-
pressions satisfy that. The differences exist in virtue of the fact that the
demonstrative itself contains descriptive information morphologically
realised in a single word that allows it to be grammatical without the
need for a noun phrase. In this later sense the demonstratives interpreta-
tion is dependent solely on the edge, and it consequently forces a refer-
ential reading.
The determiner phrase illustrates how grammar informs the referential
status of the arguments of propositions. Definite descriptions fall be-
tween referential and attributive readings, which means that there needs
to be external pressures on the construction to establish which reading is
present in a particular use of a sentence. The determiner phrase itself is
not capable of telling us what reading is present in a particular use, for
that we need to look further into the sentence.
3. Referential and attributive readings
The proposal advocated here is that definite descriptions contain an am-

biguity, but that the ambiguity is present in grammatically contrasting
sentences, such as (4) and (5). The determiner phrase illustrates how def-
inite descriptions are capable of being in one instance referential and in
another quantificational but it remains to be seen how this valuation
comes about and what semantics we should assign.
Before turning to how definite descriptions are valued it is first neces-

sary to outline two possible semantic valuations for referential and attri-
butive uses. In situations where a description is used deictically to pick
out an individual in a clear context then it is reasonable to suppose that it
has a semantics similar to that of a demonstrative description. However,
the deictic role of a demonstrative is largely determined through the
(speaker/addressee) locational information that acts as a restriction on
the interpretation. Once this has been bleached and we are left with the
definite article there is no need to include this proximity information in
the semantics of the expression; however such expressions can still be
referential. Following Irene Heim (1982) we might take the semantics of
demonstratives and definite descriptions to be open formulae, restricted
by descriptive content, in addition to certain proximity information for
demonstratives. We can write the semantics in the following way:
(18) <x, , >, which reads ‘value x with the individual that satisfies
the descriptive restriction and the proximity restriction ’.
Applying this formalism to concrete sentences we can state the seman-

tics as follows:
(19) The manager of Barcelona F.C. is Spanish.

<x, manager of Barcelona F.C., unmarked> & S(x)
(20) That table is covered with books.
<x, table, marked distance from speaker> & C(x)
The above may well be a minimal semantics for such expressions but it
gets the important part of the interpretation correct. Rather than being
quantified over the variable is valued with an individual that satisfies
conditions and . A uniquely identifying description can be given the
semantics in (19) and this avoids equating all referential descriptions one
for one with demonstrative descriptions. For the attributive uses it will
be assumed that definite descriptions receive a Russellian semantics,
which is given by the formula in (3). Nothing in this theory depends on
the Russellian analysis being correct. So long as there is a contrasting
semantics for attributive uses the theory holds.
Turning to the process via which the semantics is assigned, the theory
put forward here is that a definite description is only attributive when it
falls under the scope of certain semantic operators. In all other instances
the definite description receives a referential semantics. The following
two example sentences will be used as motivating this theory, although it
is important to note that the examples are not exhaustive:
(21) I always get the bus.

(22) I want to be the mayor.
Both (21) and (22) contain definite descriptions that are attributive rather
than referential. There is no particular individual picked out by either
definite description. In (21) the definite description is used simply to in-
dicate a preferred mode of transport rather than a particular bus. In (22)
the definite description is not referring to a particular person but rather a
desired job. The fact that the definite descriptions receive an attributive
semantics is because they fall under the scope of a generic and intention-
al verb respectively. Those verbs force an attributive reading on the ob-
ject of the sentence.
Before we turn to a concrete example, the way in which grammar
and semantics interact must be explained. It is standard within the gene-
rative grammar literature to take grammatical structure to be built up in
computational chunks known as phases (Uriagereka 1999; Boeckx and
Grohmann 2007; Chomsky 2008; Gallego 2010). The grammatical struc-
ture is built from the bottom up and at certain key points it stops, when
computational load reaches its limit. At this point grammatical opera-
tions such as case marking, agreement, and movement take place, in ad-
dition it is at this point that grammatical structure is transferred to its in-
terfaces. The two main interfaces with grammar, which are taken to be a
sensory-motor interface and a conceptual-intentional interface, assign an
interpretation to the grammatical structure at the phase level.
The conceptual-intentional interface is where expressions receive a
semantic interpretation. The phases are typically taken to be the com-
plementizer phrase (C), the transitive verb phrase (v*), and the deter-
miner phrase (D) discussed above. At each phasal stage grammatical
structure is sent to receive a semantic interpretation. The phasal interior
is what is given a semantic interpretation at the point of transfer, and on-

ly the edge is observable to the grammatical component for further oper-
ations.
The idea presented here is that the semantics assigned to the deter-
miner phase is dependent upon the grammatical configuration surround-
ing it. If the definite description is in the object position then its semantic
value is assigned at the level of the transitive verb phase in virtue of be-
ing in its interior. If alternatively it is in the subject position then the se-
mantics is assigned at the level of the complementizer phase for the
same reason. The sentence in example (21) contains a definite descrip-
tion in object position and is given an attributive semantics, and this is
the result of the semantic operator contained in the generic verb. To
make this clear the process can be seen in the set of grammatical trees
below.
(23)
In (23) we see the building of the determiner phase and when this is
complete this triggers the transfer of the interior, the noun phrase com-
plement, which assigns a semantic interpretation to the descriptive con-
tent of the definite description. At this point we still do not have a refer-
ential or attributive semantics assigned, we merely have the semantics
assigned to the descriptive content. In (24) the next part of the grammat-
ical structure is built, at this point the grammatical engine can only work
with the phase edge. Once the verb has been merged into the grammati-
cal structure, the interior of the transitive verb phase is transferred to be
assigned a semantics, and this time the semantics assigned is either ref-
erential or attributive. The semantic type assigned will depend upon the
verb merged, with certain verbs containing semantic operators that force
an attributive reading, if on the other hand the verb lacks such an opera-
tor, a default referential reading is assigned.
(24)
In (24) we see the full theory in practice. Following the introduction of

the generic verb the semantic operator present in it causes the definite
description in object position to be interpreted attributively, receiving the
quantificational semantics.
4. Conclusion
The above theory is a stepping-stone towards understanding how defi-

nite descriptions come to contain an ambiguity between referential and
attributive readings. To begin with, it was argued, contra Donnellan
(1996, 1968) and Kripke (1977), that the ambiguity was one that occurs
only in different propositions, not in one and the same proposition. Ex-
amples (4) and (5) were given to illustrate the ambiguity. Next it was
argued that we can understand the referential uses as having a semantics
akin to that of demonstratives and that the differences between the two
occur only in respect to the demonstratives containing locational infor-
mation. Once the distal demonstrative is bleached of this it becomes a
definite article, as observed by Cyr (1993) and Huang (1999) cross lin-
guistically. These facts are captured neatly in the grammar of the deter-
miner phase. Finally an account was needed as to how a definite descrip-
tion comes to be valued as referential or attributive, which goes beyond
the grammar of the determiner phase in isolation. It was argued that def-
inite descriptions receive their semantic value in virtue of the phase that
they occupy the interior of. An example derivation was given of an at-
tributive interpretation, and it was seen that the attributive interpretation
is forced only in particular grammatical contexts; when the description
falls under the scope of a semantic operator that forces the reading.
Overall this is a programmatic paper and the analysis must be extended
to account for expressions like (5); however, it is clear that the grammat-
ical structure of propositions containing a description is key to how its
semantic value is assigned. Thus, the theory defends a semantic ambi-
guity in definite descriptions, whose origin is grammatical.
References
Boeckx, Cedric and Kleanthes K. Grohmann 2007. Remark: Putting Phases in Per-
spective. Syntax 10 (2), 204-222.
Chomsky, Noam 2008. On Phases. In: M. Kenstowicz (ed.), 1-52.
Cyr, Danielle 1993. Cross-Linguistic Quantification: Definite Articles vs
Demonstratives. Language Sciences 15 (3), 195-229.
Devitt, Michael 1974. Singular Terms. The Journal of Philosophy 71 (7), 183-205.
Devitt, Michael 1981. Donnellan’s Distinction. Midwest Studies in Philosophy 6
(1), 511-526.
Devitt, Michael 2007. Referential Descriptions and Conversational Implicatures.
European Journal of Analytic Philosophy 3 (2), 7-32.
Devitt, Michael forthcoming. Referential Descriptions: A Note on Bach. European
Journal of Analytic Philosophy.
Donnellan, Keith 1966. Reference and Definite Descriptions. The Philosophical
Review 75 (3), 281-304.
Donnellan, Keith 1968. Putting Humpty Dumpty Together Again. The Philosophi-
cal Review 77 (2), 203-215.
Epstein, Samuel and Norbert Hornstein 1999. Working Minimalism. Cambridge,
Mass.: The MIT Press.
Gallego, Ángel J. 2010. Phase Theory. Amsterdam & Philadelphia: John Benjamins
Publishing Company.
Heim, Irene. 1982. The Semantics of Definite and Indefinite Noun Phrases. PhD
dissertation, University of Massachusetts.
Huang, Shuanfan 1999. The Emergence of a Grammatical Category definite article
in spoken Chinese. Journal of Pragmatics 31, 77-94.
Kaplan, David 1970. Dthat. Syntax and Semantics 9, 221-243.
Kenstowicz, Michael (ed.) 2008. Ken Hale: A life in Language. Cambridge, Mass.:
The MIT Press.
King, Jeffrey C. 2001. Complex Demonstratives: A Quantificational Account. Cam-
bridge, Mass.: The MIT Press.
Kripke, Saul 1977. Speaker’s Reference and Semantic Reference. Midwest Studies
Russell, Bertrand 1903. The Principles of Mathematics. Cornwall: Routledge &
Kegan Paul.
Russell, Bertrand 1905. On Denoting. Mind 14 (56), 479-493.
Sheehan, Michelle and Wolfram Hinzen 2012. Moving Towards the Edge. Linguis-
tic Analysis 37 (3-4), 405-458.
Uriagereka, Juan 1999. Multiple Spell-Out. In S. Epstein and N. Hornstein (eds.),
251-282.
Wettstein, Howard K. 1981. Demonstrative Reference and Definite Descriptions.
Philosophical Studies 40, 241-257.
Wettstein, Howard K. 1983. The Semantic Significance of the Referential-
Attributive Distinction. Philosophical Studies 44, 187-196.
Carl Humphries
Ignatianum Academy, Cracow
[email protected]; [email protected]
Proceduralism and Ontologico-

Historical Understanding in the
Philosophy of Language
Abstract: Since Frege, analytical philosophers have mostly construed language
proceduralistically, treating reference and assertion as largely uniform procedures
for reidentifying entities and proposing states of affairs as true. Their conviction that
these procedures make sense as such typically presupposes a broadly Kantian intui-
tion that they reflect some larger self-validating normative sphere. This faces two
objections: that it misconstrues the role of ahistorical, ontologically significant
commitments/contexts, and that it ignores cases involving radically historical under-
standing. The first shows up inter alia in the Wittgensteinian claim that it fails to
leave room for a proper understanding of the contextual grounding of language use.
The second appears in the critique of historicistic forms of proceduralism. Each ob-
jection captures something, but it seems one cannot embrace both on pain of incon-
sistency, as they entail conflicting readings of the existential quantifier. A third po-
sition, explored by Heidegger and the ‘third’ Wittgenstein, rejects the disjunction
between ahistorical-ontological and historical-contingent forms of commitment and
context, thus avoiding having to choose between these two readings. This, however,
involves a quietistic stance regarding the distinguishability of ahistorical-onto-
logical and historical-contingent forms of commitment and context, which is some-
times counterintuitive. A possible alternative is sketched, involving the notion of
ontologico-historical understanding.
Keywords: language, reference, assertion, ontological commitment, historical

commitment, ontologico-historical understanding, proceduralism
1. Proceduralism and the philosophy of language
Proceduralism holds that what we do or think or know is partly validated

by how, in some more or less general sense, we go about doing or think-
116 Carl Humphries
ing or knowing it – that is to say, by what procedures we follow. Philo-

sophically, proceduralists typically claim that human thought or under-
standing exhibits, at least in certain respects, an intrinsically self-
validating character. They generally believe that there is an intelligible
sense in which the validity of such forms of thought or understanding
can be said to be a function not just of some particular content we enter-
tain, but also of some overall form that thought or understanding is re-
quired to possess. This form is only exhibited when the right procedures
are followed, therefore its presence testifies to the fact that those pro-
cedures have been adhered to. Taking under consideration even a broad
range of positions of this sort, it seems that most (if not all) kinds of pro-
ceduralism appeal in some way to a notion of rationality as a basis for
determining their preferred conception of what, precisely, should be in-
vested with this self-validating status.
Proceduralism in this broadly construed sense (as distinct from the
Rawlsian one) has, without doubt, played a significant role in modern
analytical philosophy of language. The competing claims about the na-
ture of linguistic reference and assertion put forward there point to an
ongoing dispute about which areas or aspects of human thought and un-
derstanding are to be construed proceduralistically, as possessing an in-
trinsically self-validating (and so potentially context-independent and
formalizable) character, and which not. The most salient example of this
is the tension between the essentially formal or propositional paradigm
of language that dominated early analytical philosophy (Frege, Russell,
the Tractatus), and the emphatically antiformalistic variant of contextu-
alism about meaning-related issues introduced by the later Wittgenstein.
Indeed, much of what has followed since in analytical philosophy of
language has aimed at achieving a satisfactory theoretical reconciliation
between these two overall approaches.
Within contemporary philosophy of language, proceduralism about
meaning and reference might be thought of as capturing, in a vague but
nevertheless fundamental sort of way, the underlying philosophical (vs.
psychological) significance of various putative anticontextualist commit-
ments. Introducing the concept of procedural validity as a positive term
here may serve to draw attention to wider philosophical concerns that
potentially lie behind and motivate the various positions in play – con-
Proceduralism and Ontologico-Historical Understanding in the Philosophy… 117
cerns which, given the technical orientation of such philosophy, have

tended to be passed over. But in that case why not also pose questions
about whether such positions are worthy of assent in these broader terms
too? Surely, it makes sense to wonder about the rightness of such posi-
tions, not exclusively as they pertain to more or less specific claims
about the context-dependence, or lack thereof, of certain philosophically
interesting features of language, but also as positions implicitly adopted
at a more general level – the level at which questions arise about the sig-
nificance of proceduralism. The upshot is that the introduction of the no-
tion of procedural validity into philosophical debates about language
constitutes an invitation to think about what is involved more generally,
where answers given at this level may then still impact on more specific
language-related disputes.
In that case, even if our primary concern is with digesting the signi-
ficance of the language-centred disputes in which figures such as Frege
and Wittgenstein have played a prominent role, it will make sense to ask
in a more general sort of way whether proceduralism is something we
should accept or reject. This article sets out to review the salient issues
involved in seeking an answer to this question, keeping in mind its im-
plications for language.
2. The varieties of proceduralism
Proceduralism typically holds that there is something about the workings

of the human mind, or about human practices or behaviour, that is self-
legitimating. The classic example is Kant’s conception of the human
mind as self-legislating with respect to the form and status of practical-
ethical and theoretical kinds of knowledge, in virtue of its possession of
reason. Indeed, within the history of Western philosophy, Kant played
the leading role in bringing about and helping to complete a shift from a
previously dominant conception of thought and knowledge as more or
less ontologically or ontologico-theologically grounded to one according
to which they are taken to be procedurally grounded in this kind of way.
This shift is generally interpreted as reflecting his epistemologically
‘critical’ turn, motivated by the aim of seeking to free human thought
118 Carl Humphries
from dogmatic tendencies within traditional metaphysics while neverthe-

less establishing, or preserving, the sort of understanding of how thought
and knowledge stand relative to the empirical sciences that would allow
us to think of human beings as freely rational agents – agents, that is,
who may then bear moral responsibility for their choices and actions,
and so on. In this respect, Kant is considered to have been motivated by
the broadly humanistic goal of harmonising theoretical knowledge (the
natural sciences) with practical-ethical knowledge, which he sought to
do by means of a unifying move that is tantamount to insisting on the
possibility of adopting a higher-level standpoint – one that purports to
take in both, but from a perspective that remains recognizably human in
its sense of the irreducible significance of both morality and rationality.
Of course, one of the great issues in philosophy of the last century or
so has been the question of how far this shift ought to be thought of as
calling for qualification or reversal. Is there, as many philosophers have
argued, a requirement for a viable post-critical philosophy, and if so,
does this lead back, at all, to certain pre-critical philosophical insights
(e.g. those of Aristotle)? How far would we then be justified in rejecting,
or at least distancing ourselves from, Kant’s humanism? The complexity
of these issues exceeds what can be addressed in one article, but the fact
remains that they are relevant to the standard problems dealt with by
philosophers working with language: our conceptions of language do
often reflect broader commitments, assumptions or intuitions about pro-
ceduralism. For example, the belief that one can pursue a philosophical
investigation of language without reference to such motivating com-
mitments at all is itself only tenable against the background of certain
assumptions about the self-standing character of language or linguistic
communication – assumptions that are often tantamount to a procedura-
listic understanding.
Historically, the Kantian shift mentioned above first manifests itself as
having implications for the philosophy of language with the ensuing rise
to prominence of the idea that the basic units of human thought,
knowledge and understanding are not the concepts denoted by individual
terms within propositions (i.e. those filling the places reserved for sub-
ject and predicate in such propositions), together with the structural pos-
sibilities opened up by these denotables via the formal relations of tradi-
tional term logic and their applications in syllogistic reasoning, but are
instead the judgements or thoughts expressed by propositions them-
selves. This idea acquired its paradigmatic modern formulation, of
course, in Frege’s context principle and, along with ideas about the ana-
lysability of linguistic sentences along lines derived from the logical
analysis of mathematical propositions, may be said to have initiated
modern analytical philosophy, at least in its early phase.
The logicism of early analytical philosophy amounts, then, to a ver-
sion of proceduralism centred on commitments regarding the nature of
propositional assertion: e.g. that everything stateable in natural language
can be expressed in mathematical logic (Frege, Russell), or that all state-
able facts are subject to strictly logical limits on how language pictures
the world (the Tractatus). The commitments involved are, in both cases,
proceduralistic, inasmuch as they imply general positions as to the na-
ture of our assertoric practices that would not make sense unless one
held the correspondingly general features they are concerned with to be
entailed by some overarching character that our assertings are required
to share, as a condition of their possessing the distinctive legitimacy they
have for us when regarded as acceptable instances of propositional as-
sertion. (In other words, that any such character should be maintained
over all conceivable variations in intralinguistic or extralinguistic con-
text consistent with the idea of a string of words counting as a purpor-
tedly fact-stating propositional assertion is simply a reflection of – is al-
ready implicit in – our prior intuitions or assumptions about the nature of
this particular kind of legitimacy.)
Indeed, similar commitments of a proceduralistic kind are implicit not
only in most subsequent philosophical characterizations of the nature of
linguistic meaning and reference, but also in those positions that seek to
understand these things against the background of some relatively gen-
eral conception of the communicative function of everyday language use.
Hence the claim of Strawson (1959), that all communication about par-
ticulars involves forms of speaker-hearer reidentification presupposing
uniquely identifying references to spatio-temporal continuants and their
spatio-temporal locations, amounts to a proceduralistic stance about ref-
erence qua one of the essential elements of communication (and some-
thing similar might be said of the more narrowly specific form of that
120 Carl Humphries
claim put forward by Evans (1982) in respect of demonstrative iden-

tification). Likewise, the various claims put forward about how we inter-
pret speaker intention in third-person terms on the basis of certain ra-
tionality-presupposing norms amount to variants or elements of a pro-
ceduralistic conception of the normative and rational character of lin-
guistic communication.1
In broad terms, the difference made to such theoretical construals of
language by their proponents’ adoption of a proceduralistic stance is that
an implication comes into play to the effect that we should think of as-
sertion and reference, and in some cases also language-use (or, where
such a term is appropriate, linguistic behaviour), as having some overall,
internally consistent character. In virtue of this character, some or all of
these can be understood as corresponding to, or being regulated by, pro-
cedures held to possess intrinsic and general validity where language is
concerned, and it is in some way thanks to this that these features can be
regarded as fulfilling some purpose – e.g. a practical goal of passing on
information – that is itself assumed to be generally in force.
Before considering critiques of proceduralism, let us note in passing a
couple of additional features that cannot be properly explored here. One
concerns the ultimate motivations identifiable in particular versions of
proceduralism when construed as an overarching philosophical position,
and how far these remain consistent with those of Kant’s version, with
1
Such a conception typically involves some general idea about the role of infe-
rentialistic forms of understanding, or about there being a universal practice in
place, either of ascribing, in a supposedly pragmatically rational sort of way,
maximally coherent and empirically grounded belief-sets to speakers, or of adhe-
ring to conversational maxims pertaining to the interpretation of utterances in the
light of reciprocal presumptions about participants’ commitment to rational co-
operation, or of interpreting such utterances inferentialistically for the sake of
holding speakers to account rationally for their discursive commitments. Such
norms are often seen as analogous to what is involved in scientific theory con-
struction, and where this is so it is worth pointing out that philosophical concep-
tions of the role played by such norms in that area normally amount to statements
of a methodological commitment to coherentism as a regulative epistemic ideal.
Such statements are proceduralistic, in that they reflect general epistemological
positions concerning the status of the methodological precepts of the sciences as
these relate to human knowledge generally.
its underlying humanism. The other concerns the extent to which proce-
duralism should, or should not, be understood as a commitment about
matters that are essentially ahistorical, and so incompatible with any
form of historicism.
Regarding the first, Kant’s humanism involves the idea that by affirm-
ing a priori commitments of a minimally necessary kind, some positive
constraints can be imposed, within epistemology, on the scope of both
potentially dogmatic forms of rationalism and potentially ultra-sceptical
forms of empiricism. Such constraints, as we noted, imply an overview
of relations between contrasting forms of human knowledge and under-
standing, and with this a basis for reconciling them into a unified totali-
ty. For Kant the relations at issue are those holding between ethico-
practical and theoretical-scientific forms of knowledge, an overall char-
acterization that might also take in Hegel and Marx. For neo-Kantian
humanists like Cassirer, they are between the ‘symbolic forms’ mani-
fested in human culture and those manifested in the empirical and formal
sciences.2 For Frege and Husserl, they are between logical constraints on
thought and its experiential correlates (in the sense of phenomenologi-
cally disclosed ‘intending’) on the one hand, and psychological factors
on the other.
Turning to the second issue, we should note that most forms of proce-
duralism in Western philosophy are ahistorical in that they represent a
theoretical take on supposedly invariant aspects of the human condition,
construed at one and the same time in epistemological and ethico-
practical, or in some cases cultural-symbolic, terms. This also holds for
more specifically political and/or ethical manifestations of procedu-
ralism that invoke a formal-procedural ideal of how human beings
should conduct their dealings with one another (e.g. contractualism). But
sometimes proceduralism is historical. This is so where the idea of an
ultimate appeal to procedural validity derives its justification specifically
from its perceived capacity to furnish a response to certain problems that
emerge in the light of a prior commitment to historicism. Such forms of
proceduralism are historicistically historical: they presuppose histori-
cism. That is the case for the dialectics of Hegelianism and Marxism,
2
See Cassirer (1955/1957).
122 Carl Humphries
with their respective proceduralistic commitments to determinate nega-

tion. It is also true of a wide swathe of twentieth-century post-Marxist,
post-Nietzschean critical theorising about culture and society, whose
central point of reference seems to be the thought that there are singu-
larities of personal and/or collective experience whose overriding impor-
tance in the contemporary socio-political context requires a special form
of reflective acknowledgement – one that can only be achieved theoreti-
cally, utilising interpretative procedures specifically oriented towards
resisting the incorporation and perpetuation of ideologically implicated
claims.
The distinction between ahistorical and historical versions of proce-
duralism matters because these invite quite different forms of critique,
whose relationship to each other then raises further problems. These, as
we shall see, also bear on linguistic reference and assertion, but have
largely been ignored.3
3. Critiques of proceduralism
Critiques of proceduralism can be grouped according to what sort of

proceduralism they criticize. Firstly, there are critiques that claim that
ahistorical proceduralism cannot capture the real nature of the commit-
ments that frame our empirically oriented understanding and knowledge
of how things are.4 As such framing commitments are typically associa-
ted with notions of ‘ontological’ commitment (under one or other of the
various meanings that the term ‘ontological’ has assumed in modern phi-
losophy), it seems reasonable to say, loosely, that these critiques are ‘on-
tologically motivated’, and that together they furnish options for what
3
Proceduralism sometimes presents itself as both ahistorical and historicistically
historical, as when the same form of validity is viewed both as a historical at-
tainment and as a priori transcendental ideal. Both critiques are then potentially
relevant. Even if this were not so, the points brought to light by each critique
would need to be squared with those of the other one.
4
Our use of the word ‘frame’ here is meant to reflect its etymological connection
with notions of ‘making possible’, ‘projecting from’, and ‘helping forward’ – as
reflected in the common Indo-European root of the English terms ‘frame’ and
‘from’.
we shall hereon refer to as the ontological critique of proceduralism.

Secondly, there is the critique that claims that historicistically historical
proceduralism cannot capture the real nature of our commitments about
things that are historical in the sense that they are understood by us to be
historically contingent, and which then furnishes a historical critique of
proceduralism. Thirdly, and finally, there are critiques that argue that
both kinds of proceduralism are wrong, because both of them presup-
pose the drawing of a fundamentally disjunctive distinction between
ahistorical and ontological concerns, commitments, and forms of unders-
tanding on the one hand, and historically contingent ones on the other,
where the drawing of such a distinction at a fundamental level is itself
targeted for critique. This yields one or more critiques of a position we
shall call ‘ontologico-historical disjunctivism’. These may be motivated
by quietism about the very possibility of making a basic-level distinction
between ontologically framing and historically contingent forms of con-
cern, commitment and understanding. We shall call this position ‘onto-
logico-historical quietism’.
3.1. The ontological critique of ahistorical proceduralism
The basic charge levelled against ahistorical proceduralism pertains to

our framing commitments – commitments we take ourselves to have be-
cause of their role as preconditions for the possibility of certain kinds of
practical or theoretical knowledge and understanding that we more or
less take for granted, so that they are presupposed by the latter. Proce-
duralism – it is claimed – fails to differentiate sufficiently between ele-
ments or aspects of language or thought as these specifically pertain to
such framing commitments on the one hand, and as they specifically per-
tain to what is framed by those framing commitments on the other. More
specifically, it fails to differentiate between these in respect of what it
says or implies about linguistic meaning, reference, assertion, etc. The
critique comes in three versions, which may be respectively character-
ized as neo-Aristotelian, early-Heideggerian (Being and Time), and later-
124 Carl Humphries
Wittgensteinian (Philosophical Investigations I).5 We shall focus here on

just those versions of these critiques that carry the most direct implica-
tions for language.
The neo-Aristotelian critique may itself be arrived at from several di-
rections, all of which involve some rehabilitation of metaphysical essen-
tialism. It may begin from a realism about phenomenologically disclosed
essences, where this, in turn, may, as in some forms of neo-scholastic
philosophy, be theologically inspired, or may correspond to a strictly
phenomenological stance – the sort of stance that, historically, as with
Ingarden, sought to distance itself from Husserl’s embrace of transcen-
dental idealism as a basis for phenomenology. Alternatively, it may be
arrived at (as has been more typically the case in the context of recent
analytical philosophy) on the basis of what is thought to be implied by
some more or less universal conceptual presuppositions of ordinary lan-
guage (and the everyday understanding of the world that these reflect):
here a neo-Aristotelian form of metaphysical analysis makes an appear-
ance, and is used to explicate everyday conceptions of thinghood (arte-
facts, living things, etc.), causal powers, dependency relations, disposi-
tional capacities, etc.6 Taken to its logical conclusion, this may suggest
advantages for term logic over Fregean mathematical logic when it
comes to respecting the validity of inferences between general kind
terms – inferences of the sought that, according to realist forms of essen-
tialism, are needed to express our understanding of essences themselves.
Fregean mathematical logic emerges as implicitly anti-essentialist and so
by no means neutral with respect to ontological matters. In its place, a
return to an ontologically grounded conception of logic may be advocat-
ed – the latter disclosed via phenomenologico-categorial intuitions per-
taining to essences, universal kinds, etc. This would vindicate the
longstanding claim of neo-scholastics (e.g. Veatch 1969) that the recur-
sivity of Fregean mathematical logic misrepresents the nature of essen-
tialist commitments in language (and the inferences based on them) by
implying a generalized possibility of logical ascent as regards contexts
5
See Heidegger (1922/1962) and Wittgenstein (1953).
6
See Wiggins (1980), Lowe (1998), and for a recent overview, Tahko (2012).
of predication.7 It also lends force to the idea that has come to promi-
nence in some recent analytic philosophy, according to which essential-
ism and logical modality are held to be divergent in respect of what they
imply for basic modal categories like necessity and possibility.8
The early-Heideggerian critique takes a different line, first arguing
that acts of assertion only make sense when their apophantic (showing)
character has a point, and then claiming (in a phenomenologically moti-
vated way) that this depends on when it makes sense for things to be
shown by us to one another, inasmuch as they have not already shown
themselves. What then follows is that assertibility must be relativized to
some more basic context than any mere context of formal assertibility.
For the early Heidegger, that more basic one is furnished by a herme-
neutically ontological phenomenology of practical enworldment and
practical-existential temporality (Being and Time).
The later-Wittgensteinian critique offers a parallel approach, inasmuch
as what Heidegger does in relativizing assertibility to practical-exis-
tential context, Wittgenstein does (in the Investigations) in relativizing
conceptual reference-determining meaning to practice-constituted con-
texts. According to the latter, the intelligibility of the ‘grammaticality’ of
human concept-use depends on how linguistic act-types function within
‘language games’, while language games, in turn, derive their point from
their role in helping to sustain practices. The upshot is that universally
applicable logical constraints on propositional sense (viz. the Tractatus)
are to be replaced by ‘grammatical’ ones internal to practices, where
these latter reflect extra-linguistic variations between practice-consti-
tuted forms of life. A final step in the analysis is the recognition that such
forms of life express human values, concerns and purposes, construed by
Wittgenstein in a pluralistically piecemeal way whose implications for a
range of issues of philosophical concern (e.g. relativism) remain contro-
versial.
7
This runs parallel to the Wittgensteinian idea that our intuitions about ethical
value simply leave no room for logical ascent to second-order metaethical char-
acterizations of ethicality itself.
8
See Fine (1994) and Vetter (2011).
126 Carl Humphries
Summing up, we may note that each of these three critiques proposes
that we should replace a relatively far-reaching but abstract form of pro-
ceduralism about thought and language, manifested in presumptions
about the overall analysability of language as a formally logical system,
with something else. For the neo-Aristotelian it must make way for the
possibility of entertaining essentialist modal commitments, construed as
constitutive of the essences of certain kinds of thing, and not as reflect-
ing any corresponding feature of language, thought or the human mind
per se. For early Heidegger a formal paradigm of intelligibility is to be
replaced by the practical intelligibility internal to the basic ‘situation’ of
Dasein’s existing temporally in its world. For the later Wittgenstein it is
to be replaced by the ‘grammatical’ contrast between making and lack-
ing sense, as this specifically pertains to our participation in practices
that reflect human purposes and concerns that may themselves vary from
one situating context to another. In each case the replacement is a con-
text of intelligibility that purports to be less abstract, or less far-reaching
(as regards inclusiveness), but still self-validating. Such critiques, rather
than entailing a total rejection of what is involved in ahistorical procedu-
ralism, suggest scope-restrictions on its applicability to human affairs.9
3.2. The historical critique of historicistically historical

proceduralism
Historicistically historical proceduralism, as was noted earlier, presup-

poses historicism, since it proposes a form of understanding – in this
case that of an overarching historical dialectic – that claims to validate
itself proceduralistically on the basis of the general premise that reality-
as-a-whole should be construed in constitutively historical terms – that is
to say, in terms that might be said to be radically historical (i.e. histori-
cal ‘through and through’). This means we are dealing with something
whose intelligibility is held to be entirely independent of ahistorical
9
The deeper issue here, of what proceduralistic and non-proceduralistic concep-
tions of self-validatingness might be said to have in common, is something I will
have to explore on another occasion. It concerns the ultimate form of alignment
of order and value in reality – a theme that goes back to Plato’s Symposium.
framing commitments of a substantive (i.e. ontological) rather than

merely procedural (formal-logical) kind.
The most straightforward critique of such a position, construed as an
instance of proceduralism, argues that it is internally problematic, be-
cause either the historicism in question is not, after all, radically histori-
cal in the sense just outlined (in that an ontological commitment to a his-
torical telos’s being in play is required for it to make sense), or it is so,
but presupposes a commitment to a fatalistic determinism about the fu-
ture that is held to be intuitively implausible. This stems from the idea
that historical intelligibility, properly construed, is always constituted ex
post factum. If that is so, then historicism (as a general thesis) presup-
poses anticipatory knowledge of the end of everything, and such
knowledge can only be grounded in either the grasp of an ontologically
(and ahistorically) intelligible historical telos or in a commitment to fa-
talism.10
The idea that historical intelligibility is always constituted ex post fac-
tum certainly makes sense if we think that what is historically intelligible
consists in a series of occurrences that bring us to a point where they
themselves stand revealed as having terminated in an outcome. That out-
come then retrospectively confers an ex post factum significance upon
them, in virtue of which they constitute a ‘self-standing’ historical epi-
sode – as with, for example, the events leading up to some unforeseen
but overridingly significant development. Such an episode will be made
up of changes in a diachronically unfolding modal scenario whose end-
state is the unavoidability of the outcome that has occurred.11 As such,
10
A further significant critical point is that such thinking prevents us from seeing
that historical understanding can also be ‘radically historical’ in the absence of
historicism, inasmuch as it figures as such as part of our ordinary common-sense
historical understanding. (taking the latter to be distinct from the essentially in-
terpretative discipline of ‘history’ construed as a form of scholarly research into
areas of the past with which one is not directly acquainted by virtue of having
witnessed them oneself).
11
For formal analysis of this, see von Wright (1984a). The central idea is that
something is understood historically when we understand how it came to be un-
avoidable through a process of successive eliminations of alternative possibilities
brought about by antecedent events. The intuition is that what matters about a
128 Carl Humphries
they can only be grasped after they have happened and a final point of
‘historical closure’ has been reached. In the absence of divinatory pre-
monitions about how the future will look from an ultimate final vantage
point, it is hard to see how this could ever be possible with respect to
‘reality-as-a-whole’.
3.3. The critique of ontologico-historical disjunctivism as a

proceduralistic premise
The historical critique of historicistic proceduralism opens up a per-

spective on historical intelligibility that is significant for other reasons
too. These are connected with the thought that even in the absence of
historicism, human understanding may be radically historical: that is, it
may be so if or when it concerns something whose only or ultimate sig-
nificance is historical, in that its significance is necessarily revealed– if
it is so at all – ex post factum, without reference to any substantive fram-
ing commitments. (This, as we shall see later, when combined with ele-
ments of the ontological critique of ahistorical proceduralism, opens up a
further perspective on general and language-related issues related to pro-
ceduralism.) If one attaches importance to such instances of non-his-
toricistic yet radically historical intelligibility, then the historical critique
(of historicistically historical proceduralism) will also have critical im-
plications for ahistorical proceduralism. To grasp this, we must first un-
derstand why instances of this sort of historical intelligibility might mat-
ter to us.
Such cases may be construed in two ways, each with potentially dif-
ferent implications. They can be construed either in essentially first-
person-recollection-based terms, or in essentially third-person-based
terms. In the former case, radical ex post factum historical intelligibility
will be constituted from the recollective standpoint of individual or col-
historical understanding is that it charts, and by doing so already divulges a sig-

nificance in, the particular form taken by this process of ‘progressive enfate-
ment’. Our curiosity about counterfactual scenarios then makes sense as an incli-
nation to try to discern the provenance of once contingent events and states of af-
fairs.
lective subjects engaged in the first instance in reflective contemplation

of their past – of the sort most obviously typified in Western cultures by
religious and/or poetic forms of introspection. In the latter, it will reflect
third-person-based construals of the intrinsic significance of completed
historical episodes – episodes pertaining in the first instance to the lives
of others, or to one’s own just insofar as it is conceivable as standing on
a par with these.
Central to the idea of first-person-based radically historical intelligi-
bility is the notion that acts of recollection (and rituals of remembrance)
play a constitutive role in relation to reflection on the meaning of a rec-
ollected past: a ‘subject’ contemplates the ex post factum significance
that past events have when viewed from the perspective of the present,
where this implies, in turn, that the relevant episodes of historical devel-
opment themselves really only count as such in virtue of the ex post fac-
tum significance they have when viewed from the standpoint of recol-
lection in the present. However, Wittgensteinian considerations pertain-
ing to the public-private distinction make any attempt to spell out the
wider implications of this philosophically problematic: there is no agree-
ment about how we might separate out structural features of public lan-
guage-use (and any structures of understanding these express) from pri-
vate introspection. Perhaps, then, we should look to third-person-based
radically historical intelligibility to furnish any implications relevant to a
further critique of proceduralism.
To get a sense of why the latter might be considered important, we
must first invoke some sources of a kind that tend not to figure even in
the remote theoretical background as this normally pertains to procedu-
ralism. There is, within contemporary philosophising about ethical mat-
ters, a strand that might best be described as committed to a form of non-
ontological personalism, whose distinctive claim is that our fundamental
conception of a person is constituted not in terms of a mere continuous
form of self-identity, or even some richer notion (such as Aristotle’s) of
how a human being comes to be thought of as subsisting through chang-
es. Instead, it is held to be constituted as a reflection of what is implied
in respect of it by our ethical responses to some series of developments
whose significance, given our response (e.g. compassion, admiration,
etc.), is such as to require that we then predicate of it that it corresponds
130 Carl Humphries
to the life-history, and thus also the existence presupposed by this, of a

being construed as the bearer of ethical personhood.
This is certainly not an easy idea to grasp, but it might be taken to be
the essence of the thought expressed in Wittgenstein’s oft-quoted re-
mark: “My attitude towards him is an attitude towards a soul. I am not of
the opinion that he has a soul” (Philosophical Investigations II. iv).12
What we respond to in such cases, and what is then invested with signif-
icance because of our having responded to it, is what we would then typ-
ically go on to think of as having already transpired in the way of things’
being historically significant in such a life, where this commits us to
conceiving of such a life as historically constituted.13 If, moreover, the
life defined in terms of our responses happens to already be over, then
the form of personhood entailed by our responses will be exhaustive of
personhood as it pertains to that life (qua life-history). Viewing such a
life as the life of this or that person will just mean viewing it with refer-
ence to something itself understood in strictly non-ontological, radically
historical terms – as historical ‘through and through’.14
The interesting implications of this emerge when we consider what
follows for language from this form of antiproceduralism, alongside
what is entailed by the ahistorical critiques mentioned earlier.
To begin with, let us note that prior to any scope-restrictions stem-
ming from the ahistorical (ontological) critiques mentioned earlier, ahis-
torical proceduralism in the philosophy of language (Frege, Russell,
Tractatus) assumes that we can unify all kinds of commitment about the
‘existence’ or ‘reality’ of logical objects (i.e. occurring events, existing
12
See Wittgenstein (1953: 178). For a groundbreaking attempt at an exposition of
this line of thinking, that takes as its point of departure Wittgenstein’s remark,
see Cockburn (1990).
13
The most obvious and pronounced instances of this mode of receptiveness are to
be seen in how we respond to the lives of others when we encounter those lives
as historically terminated episodes – something that provides the motivation for
biography as a genre.
14
One could construct an analogous response-based conception of historically ter-
minated communities – one applicable, for instance, to the way of life of Central
European Jews prior to the Holocaust, or, indeed, to any community whose tradi-
tional form of life can be said to have been destroyed by the advent of modernity.
things, etc.) under a single minimally thick notion of so-called ‘existen-

tial commitment’, typically expressed by using the ‘existential quanti-
fier’ that is symbolized as ∃x and standardly paraphrased by means of
informal phrases like ‘there is’. This form of commitment, it must be
noted, signals no more than the availability of something within some
domain of discourse to serve as an argument satisfying some propo-
sitional function.
Each of the ahistorical (ontological) critiques surveyed before implies
that this will be problematic in some way or other. For neo-Aristote-
lianism, it wrongly encourages us to think of all inferences as running
between propositions resting on existential commitments of this mini-
mally thick sort (to the exclusion of syllogistic inferences involving kind
terms, grounded in an essentialist construal of the referents of these). For
Heideggerians, it fails to differentiate between what it means for mere
entities to ‘be’ or ‘exist’, and what it means for reflective beings to do
so, where these are construed as instantiating temporalized Being (to-
gether with its distinctive phenomenology of practical engagements,
where this then also contextualizes our grasp of entities themselves). For
the later Wittgenstein (of the Investigations), it fosters a false conception
of ‘existence’ as involving some sort of underlying context-invariable
mode of commitment – one that will be wrongly thought of as invariant
across different practice-constituted forms of life, language games, etc.
The most straightforward way of capturing what these three positions
have in common is to note that in all of them the restrictions imposed on
procedurality engender a requirement to differentiate expressively not
only between commitments pertaining to actualities on the one hand and
to hypothetical possibilities on the other (as in the generalized modal on-
tological discourse of actualism and possible worlds theory), but also
commitments pertaining to constitutive possibilities (‘potentialities’) on
the one hand and to merely hypothetical ones on the other.
Embracing both of these distinctions entails ontologico-historical dis-
junctivism. This is because the three ahistorical (ontological) critiques
suggest that an existence-commitment with respect to some object of
reference will, at least on some occasions, imply a commitment to cert-
ain (non-hypothetical) possibilities being taken as constitutive of the
framing context that is a precondition of the intelligibility of the exis-
132 Carl Humphries
tence-commitment itself (and of any ‘Russellian’ reference that implies

it). Hence, consistency requires that the notion of existential commit-
ment we take as implicit in such cases be thick enough to also capture
the particular sort of existential commitments invoked when we seek to
express that kind of implication. The significant feature here is that if the
common minimum of commitment expressed by the standard use of the
existential quantifier (∃x) is itself interpreted as thick enough to leave
room for this (in the sense that it allows for the existence of ‘potentia-
lities’ in the sense of unactualized but constitutive possibilities), then it
will conflict with what we want to say about radically historical pheno-
mena. This is because a commitment to the radical historicality of some
phenomenon just is, as we have seen, a commitment to there being no
such implied framing commitments in play as preconditions for the intel-
ligibility of that phenomenon, and hence no context in which the exis-
tence of such ‘potentialities’ could intelligibly be posited. The upshot is
that embracing both critiques of proceduralism would require us to have
in play, at one and the same time, two distinct interpretations of the con-
cept of existential commitment: one that expressly leaves room for such
implications, and one that expressly does not. One might then be in-
clined to think that this places ahistorical and historical dimensions of
our structures of understanding and commitment in a disjunctive rela-
tionship with one another, at least with respect to how they are mani-
fested through language at the level of our readings of the significance of
the existential quantifier. That, in turn, would threaten to preclude any
possibility of reaching a unified overview of the relative scope and im-
portance of ahistorical and historical forms of understanding of the
world – the sort of unifying standpoint that might furnish a basis for a
humanism comparable in its significance to Kant’s. At any rate, it now
looks as if the unity of commitment expressed by a proceduralistic under-
standing, rather than being replaced by an alternative understanding that
is more thickly defined, more restrictive, but somehow still capable of
expressing the logico-linguistic basis for an overview, has been lost.
This suggests that if we really seek such a unity, but also accept the va-
lidity of both critiques, we must look beyond the disjunctivistic conflict
between ontological and historical forms of commitment that these cri-
tiques themselves together imply.
This is where the third form of critique, which precisely takes as its
target the perceived disjunction between ontological and historical forms
of understanding apparently brought to light by these differences per-
taining to existential commitment, comes properly into view. This criti-
que, at least in its versions to date, invokes a kind of quietism about the
very distinction between ontological and historical forms of under-
standing as a basis for dissolving the disjunction in terms not tantamount
to a return to proceduralism. Such a position is to be found in the later
Wittgenstein, but also in the later Heidegger, for whom (much as with
his earlier critique of proceduralism) the phenomenological conditions
bearing on the intelligibility of the apophantic character of assertion call
for a more basic context than that of formal linguistic assertibility.15
In this case of Wittgenstein, such a position emerges as a response to
the concerns explored in On Certainty.16 In that work, as part of a re-
sponse to the issue of scepticism raised by Moore, he suggests that his
account of the context-dependency of linguistic meaning (as given in the
Investigations) can be understood as also, in effect, redefining the Kant-
ian idea of an analytic/synthetic epistemological boundary pertaining to
15
For Heidegger, such a context is to be furnished by a poetics of Being as both
self-revealing and self-concealing, conceived so that it leaves no room for a
foundational distinction to be made between ontologically and historically
grounded forms of commitment, since it involves quietism about explanatory
grounding per se. He seeks to draw attention to the existence of a longstanding
tradition in Western thought of assuming that one must always choose between
justifying something either by rationalizing it in the sense of showing how it is
required as part of a general structure of possibility (cf. Kant’s a priori know-
ledge), or by treating it as explainable in terms of either historical factors or other
causally explanatory conceptions that are tantamount to regarding it as fun-
damentally contingent. His point is that this implicitly presupposes the more
basic idea that there is some sort of general requirement for grounding of any sort
in play – an idea that, quoting the mystical poet German-Silesian poet Angelus
Silesius, he rejects. In life, Heidegger argues, when confronted with simple phe-
nomena such as a living thing (a rose), we do not invariably ask why it is there at
all, so the question of what sort of answer this ‘why’ would then demand (i.e. an
ontological or a historical one) need not necessarily arise. See Heidegger (1991:
41-49).
16
See the whole of Wittgenstein (1969).
134 Carl Humphries
judgements or commitments, so that it now demarcates logico-gramma-

tical commitments presupposed by our practices as quasi-analytically
necessarily true from empirically accountable commitments that are con-
tingently true or false, but whose conceptual constituents presuppose
these same practices as making those conceptual items intelligible. The
consequence he draws is that for any given assertoric context of referen-
tial concept-use, there will be a practice-constituted form of life which
requires that certain statements count as expressing bedrock presupposi-
tional commitments, while others count as expressing empirically de-
terminable truth claims, while the demarcatory boundary responsible for
determining which statements fall into each category will shift from one
practice-constituted form of life to another. What sets the boundary, ac-
cording to Wittgenstein, is just the sheer (ungrounded) fact of our being
committed to this or that practice by our form of life – and thus the sheer
(ungrounded) fact of our being involved in that form of life. As he says
in Investigations, II.xi: “What has to be accepted, the given, is – so one
could say – forms of life”.17 Of course, if this involvement is unproblem-
atically ungrounded, there can be no meaningful question to answer
about whether the grounding involved is properly ontological or histori-
cal.
This move by Heidegger and Wittgenstein would seem to resolve the
issue of how to accommodate ahistorical and historical critiques of pro-
ceduralism without relinquishing the unifying goal that motivated proce-
duralism itself. But it is not without problems. We may wonder whether
quietism of this sort really is the only way to achieve such an outcome.
Could there not be other ways that might be valid, say, at least for some
instances or areas of human concern and commitment – and, by exten-
sion, for related context-sensitive aspects of language use?
4. Ontologico-historical understanding and truth
Although there is insufficient space to properly address the matter here, I

shall briefly point the reader in the direction of a potentially fruitful posi-
tive alternative to quietism, which I shall call ontologico-historical un-
17
See Wittgenstein (1953: 226).
derstanding. The basic thought here is that there could be structures of

intelligibility and significance at work in human affairs that involve a
clear distinction between ahistorically grounded (ontological) elements
on the one hand, and historical grounded (contingent) elements on the
other, where these are nevertheless mutually dependent and conjoined,
and mutually irreducible.
The example that springs to mind here is that of intergenerational fam-
ily relationships. If we reflect on what is involved in these, then it does
seem to make intuitive sense to say both that the historically construed
life-history of a human being qua progenitor (parent, etc.) is rendered
ontologically significant by its grounding implications for their descend-
ants’ (children’s) life-possibilities (as a kind of family legacy), and that
the ontologically construed life-possibilities of a human being qua de-
scendent (child, etc.) are rendered historically significant by their prove-
nance as this pertains to their progenitor’s (parents’) life-history (as a
kind of family inheritance). If that is right, then what we observe in such
a case is a complex relationship-structure involving referential depend-
ency relations running in both directions between that which is con-
strued ahistorically (ontologically) on the one hand, and that which is
construed historically (as contingent) on the other. As a ‘legacy’ (of one
person to another) the referential dependency-relation runs ‘forwards’
from the historical to the ontological, while as an ‘inheritance’ (of one
person from another), it runs ‘backwards’ from the ahistorical (ontologi-
cal) to the historical (as contingent). Such an analysis might, moreover,
be widened to take in some other kinds of relationship involving similar
role-constituting legacy-inheritance structures, such as non-family-based
intergenerational relationships between individuals or between commu-
nities.
This may shed new light on why certain referring expressions figure
in linguistic usage in the way that they do. These are ones that we take as
serving to identify integral unities of placehood or objecthood of a sort
that would lack any such unified character were it not for the contexts
furnished by structures of ontologico-historical understanding of the sort
just mentioned. Consider, for instance, the understanding invoked when
the term ‘home’ is construed as referring to the spatio-temporal and ob-
jectual correlate of a familial or communal legacy-inheritance structure:
136 Carl Humphries
in such cases, the term invokes both an entity comprehensible as defined

in strictly historical terms (as the family home, the home of one’s child-
hood, one’s hometown, etc.), and one comprehensible as defined strictly
with reference to its constitutive practical potential as an artefact (i.e.
something built and/or chosen for the sake of its capacity to provide
shelter, and so on).
From the point of view of the philosophy of language, such cases are
inherently complicated: they require us to admit assertoric hybrids, in
the sense of claims or statements exhibiting features of two quite differ-
ent modes of assertion at one and the same time. On the one hand, they
resemble those whose constituent references imply existential com-
mitments involving a more than minimally thick notion of existence, re-
flecting ontological concerns. But, on the other, they also resemble those
whose constituent references imply existential commitments involving
no more than a minimally thick notion of existence, reflecting an abso-
lute bare minimum of ahistorical concern that can only then be construed
in proceduralistically formal terms (cf. Adorno 2001: 174-179). Similar-
ly, if there are facts whose obtaining is specifically tied to just those do-
mains of human concern that require such conjoined forms of assertion
to be in operation when we make claims about them, then the statements
made true by these facts’ obtaining could be said to be ‘ontologico-
historically true’, and thus to instantiate something perhaps deserving of
investigation in its own right: ‘ontologico-historical truth’.
References
Adorno, Theodor W. 2001. Kant’s Critique of Pure Reason. Translated by R. Liv-

ingstone. Cambridge: Polity.
Cassirer, Ernst 1955/1957. The Philosophy of Symbolic Forms. Vol.1-3. New Ha-
ven: Yale University Press.
Cockburn, David 1990. Other Human Beings. New York: St. Martin’s Press.
Evans, Gareth 1982. The Varieties of Reference. Edited by John McDowell. Oxford:
Fine, Kit 1994. Essence and Modality. Philosophical Perspectives 8, 1-16.
Heidegger, Martin 1922/1962. Being and Time. Translated by J. Macquarrie and E.
Robinson. London: SCM.
Heidegger, Martin 1991. The Principle of Reason. Translated by R. Nelly. Indian-

apolis: Indiana University Press.
Lowe, Jonathan 1998. The Possibility of Metaphysics: Substance, Identity and Time.
Oxford: Oxford University Press.
Strawson, Peter Frederick 1959. Individuals. London: Methuen.
Tahko, Tuomas (ed.) 2012. Contemporary Aristotelian Metaphysics. Cambridge:
Veatch, Henry 1969. Two Logics. Evanston: Northwestern University Press.
Vetter, Barbara 2011. Recent Work: Modality without Possible Worlds. Analysis 71
(4), 742-754.
Wiggins, David 1980. Sameness and Substance. Oxford: Blackwell Publishers.
Wittgenstein, Ludwig 1953. Philosophical Investigations. Oxford: Blackwell Pub-
lishers.
Wittgenstein, Ludwig 1969. On Certainty. Oxford: Blackwell Publishers.
von Wright, Georg Henrik 1984a. Diachronic and Synchronic Modality. In: von
Wright (1984b), 96-103.
von Wright, Georg Henrik 1984b. Truth, Knowledge, and Modality. Oxford:
Blackwell.
Gary Kemp
University of Glasgow
[email protected]
Quine’s Criticisms of Semantics

Abstract: Quine had a number of criticisms of the idea of natural language seman-
tics. Without going into them all, I try to separate them into the interlinguistic and
intralinguistic, suggesting that semantics can survive the interlinguistic criticisms,
but that some of the more piecemeal intralinguistic criticisms remain.
Keywords: Semantics, Quine, reference, indeterminacy, underdetermination of the-

ory, Chomsky, Davidson, linguistic disposition, inconsistency.
0. Introduction
Let us then recognize that the semantical study of language is worth pursu-
ing with all the scruples of the natural scientist. We must study language as
a system of dispositions to verbal behaviour, and not just surface listlessly
to the Sargasso Sea of mentalism. (Quine [1975b]: 252)
After Frege, Russell, and the early Wittgenstein, came Carnap, who at-
tempted to analyse meaning in terms of the concepts of intension, exten-
sion, and lexical structure (in Carnap 1942, 1956). Other famous figures
contributing to this growing development in the 1950’s through the early
1970’s include Church, Montague, Kaplan, and David Lewis. Quine’s
was a dissenting voice. Until the 1960 publication of Word and Object,
Quine was intimately involved with and critical of the work of Carnap
(and a few others). Afterwards, although he seldom criticised philoso-
phers directly, and according to his autobiography, was very selective in
what philosophers he read, many of his implicit criticisms of philoso-
phers were well known and indeed are still discussed – in particular, the
central dogma discussed in ‘Two Dogmas of Empiricism’ of 1951 of an
unjustified appeal to analyticity and meaning in epistemology; and the
attempt in Word and Object to show that the notion of the meaning of a
140 Gary Kemp
sentence, a proposition, is indeterminate in its empirical application. The

earlier paper accused the philosopher of being insupportably dogmatic;
the later book made the positive case for the negation of that dogma.
Quine’s interest in semantics was in the end subservient to his episte-
mological agenda. His overarching aim was to identify the real presup-
positions, and to sketch the main lines, of a naturalistic and scientific ac-
count of the whole of human knowledge. Most of Word and Object, and
the subsequent monographs The Roots of Reference, Pursuit of Truth,
and From Stimulus to Science, were devoted to that constructive topic
(see Hylton 2007). Indeed Quine’s doctrine concerning truth and refer-
ence implies that semantics cannot be fundamental, as a so to speak key-
stone, as it was for Hume, Carnap, Ayer, and Dummett in their different
ways. It is an illusion, for Quine, to think that philosophy could first de-
cide a criterion for sense, and then proceed to particular problems such
as the external world or induction (Kemp 2010: 289-290, 2012: 57-61).
Quine’s message – naturalism as he understands it – is the Neurathean
one that there is no such stable starting point; one has to accept science
more or less as we have it (which is not to say that scientists are above
reproach).
But that is, or would be, accepted by many practicing semanticists. It
seems to leave open the question of whether we can regard semantics as
a science, no higher but no lower than other special sciences. Perhaps
semantics is just not at present a hard science, unlike say Chemistry;
maybe it is just a young science. And since 1960 attention has shifted
from the explication of meaning in ideal languages as in Carnap to the
detailed analysis of meaning in natural language. The discipline of natu-
ral language semantics has arisen, with branches in philosophy, psycho-
logy and of course linguistics, comprehending a bewildering variety of
approaches and paradigms: teleosemantics, conceptual role semantics,
game-theoretic semantics, dynamic semantics, intention-based seman-
tics, expressivist semantics, inferentialism, and more. It would be pre-
sumptuous to pretend to have boiled down a Quinean challenge that di-
rectly threatens all of it. Nevertheless, I’m not at all confident that Quin-
ean worries about semantics have quite been answered.
Thus, in this paper, after introducing what I think are the primary
Quinean objections to the reliance on meaning in philosophy, I shall try
Quine’s Criticism of Semantics 141
to articulate some of Quine’s criticisms to semantics in this less ambi-

tious sense.1
Quine’s actual attitude towards semantics has at times seemed more
positive than his most famous negative doctrines suggest. In particular,
in the 1970’s Quine explicitly supported Davidson’s programme of con-
ceiving the semantics of natural language in terms of Tarski’s truth-
apparatus (as in Quine 1974: 65). But that was in the initial heyday of
Davidson, when it seemed that his approach was a technically souped-up
version of Quine’s idea of radical translation, one that like Quine’s did
not require for the interpretation of a speaker that one has to assume the
concepts of meaning or analyticity (see Quine [1999]: 159-165). Gradu-
ally, certain irrevocable differences came to light – over truth and refer-
ence (see Kemp 2012 for the whole story).2 Yet, as will become evident,
I will make use of a key move in Davidson’s approach, one that is con-
sistent with Quine’s views.
1. External objections
Quine was suspicious about the use of the concept of meaning in serious
science from very early – not of meaningfulness (roughly, the concept
which counts out ‘Colourless green ideas sleep furiously’) but the mean-
ings of sentences, which in turn implies the objectivity of synonymy. In
1946 he wrote an unpublished piece called ‘On the Notion of an Ana-
lytic Statement’, in 1949 ‘Animadversions on the Concept of Meaning’,
and of course in 1951 the famous ‘Two Dogmas’ article (first delivered
in 1950). The two main conclusions of the latter are (1) that no attempt
to explain or reduce the notion of analyticity can fail to presuppose other
meaning-theoretic notions; and (2) that no such notion is needed to give
an account of knowledge – confirmational or epistemic holism can do
1
Space prevents me from considering the Quine’s objections to a science of inten-
tion, of propositional attitudes de re.
2
Lepore and Ludwig (2005) have argued strenuously that not only was the con-
cept of meaning assumed by Davidson as an intensional concept, his programme
seems hopeless without it; I don’t agree with the first bit; see Kemp (2012: 83-
86).
142 Gary Kemp
the job. The suspect notion of analyticity is a full-blooded one: it covers

all the inferences or truths which empiricists ascribed to the meanings of
words, not just a few trivial ones such as No bachelor is married. Much
later, in the Roots of Reference and elsewhere, Quine recognised certain
truths that can plausibly be thought to be accepted in the learning of cer-
tain words. This accounts for the bachelor example’s being trivial (but
somewhat rare). He thus can recognise such examples without granting
that analyticity is capable of anything like the role that Carnap and Ayer
envisaged, as the cement needed to hold our entire conceptual scheme
together.
But Quine’s overriding interest in ‘Two Dogmas’ concerns the fun-
damentals of epistemology, not semantics per se. One could accept epis-
temic holism combined with semantic atomism, and that the notion of
analytic truth is seldom exemplified, but still hang on to the idea of sen-
tential meaning – it is just that as a matter of fact, synonymies and ana-
lytic truths are thin on the ground. Nothing Quine says in that article
completely undermines the idea of sentential meaning or proposition,
and in particular there is nothing in the article to disqualify those notions
from being empirically well-founded.
Chapter Two of Word and Object constitutes a direct frontal assault
on the notion of sentential meaning. There is an occasional tendency to
dismiss it as the ravings of an old-fashioned behaviourist in the style of
Skinner. But Quine restricts his behaviourism to semantics – and it is not
that the meaning of a sentence or term can be identified with the corre-
sponding behaviour, and nor is it that semantic expressions themselves
reduce to behaviour. It is that the only facts relevant to meaning are be-
havioural facts – where these comprise not just actual behaviour but be-
havioural dispositions. It is roughly a supervenience thesis (but only
roughly, since it is not the case that semantic facts are determined by be-
havioural facts, as we shall see in a moment). This is harder to deny; it is
why Quine says such things as ‘one may or may not be a behaviourist,
but in linguistics one has no choice’ (Quine [1987]: 341). True, a Quine-
an must hold that the basis for the decision is ultimately pragmatic. Still,
if two creatures have the same linguistic behavioural dispositions – in
every circumstance they would say the same thing – then for Quine it is
pointless to suppose there might be a semantic difference between them.
Ascriptions of meaning, whatever else they are, are in principle veri-

fiable; meaning, if anything, is what is communicated, is what can be
held in common by different speakers and handed down from teacher to
pupil (see Quine 1969: 81).
Doubts about this take two forms: First, one might think that for ex-
ample colour-inverts are possible but not behaviourally detectable. To
which the reply is just the wooden one that any such differences simply
would not show up in the meanings of such words as ‘red’ (Quine
1975b: 245). Second, one might point to those ineliminable uses of first-
person pronouns as show up in de se statements of propositional attitude.
And again, the correct reply is the wooden one that the facts communi-
cated by such uses of first-person pronouns, and indeed those communi-
cated by the uses of indexicals generally, may in some sense be other-
wise incommunicable, but rules for the use of indexicals are necessarily
public, can be taught, and so on.3 The very first sentence of Word and
Object is that “Language is a social art” (Quine 1960: ix).
Brute dispositions are not only mysterious but would violate Quine’s
extensionalism. He thinks of linguistic dispositions as ultimately neural,
and the relation of disposition to fact as exemplified by the solubility of
salt to its underlying chemical basis (Quine 1960: §46). The expression
of the disposition of salt to dissolve in certain circumstances is the
non-extensional idea that if a sample were placed in water in those cir-
cumstances, it would dissolve. According to Quine, the counterfactual
conditional can in turn be replaced in a regimented science by certain
extensional, universally quantified conditionals, dealing with molecular
structure, explaining the counterfactual. We don’t know the neural corre-
lates for linguistic dispositions, but so long as we remain confident that
they’re there, we can go on speaking of the dispositions.
Thus to Word and Object. The thesis that translation is indeterminate
is really two claims:
3
A third response is to cite the idea of someone who always lies. In that case, one
should posit a tacit negation in front of everything he or she says. A different
case is the person who lies as often as not, where there is no empirical means of
telling which is which; I think Quine would simply bite the bullet, contending
that such behaviour is not meaningful. See Quine (1975b: 252).
144 Gary Kemp
(1) Holophrastic Indeterminacy, which at this stage was that a native

sentence could be correctly translatable in terms of one empirically ade-
quate scheme as P, but according to another empirically adequate
scheme, as not-P (thus one must relativise ascriptions of truth to the
translation scheme chosen). Especially in cases where the native opinion
doesn’t choose between the sentence and its negation, and where our
own opinion doesn’t choose between the respective translations, the idea
is that there is nothing to choose between them; there is no factual dif-
ference. What’s significant is the mere conceivability of it, where what is
meant is that the possibility is not incoherent, not inconsistent; in that
sense one could translate a speaker in various ways, even if no one
would (see Quine [1999]: 159).
Why would one believe it? For Quine, sentences can be sorted into the
observational and the non-observational or ‘theoretical’. Observation
sentences, holophrastically considered, admit of unique translation – in
the sense that observation circumstances of the source language sentence
can be equated with those of a sentence of the target language (I’ll skip
the distinction between observation sentences and non-observational oc-
casion sentences). Holophrastic indeterminacy sets in with the non-
observational part of language, whose translation depends on slicing up
sentences into sub-sentential expressions. In Word and Object, Quine
merely thought it more or less obvious, once one pays close attention to
the plight of a radical translator, one who has nothing to go on but native
speech-behaviour.
Subsequently one main argument emerged, one that proceeds from the
alleged fact of the general underdetermination of theory. The thesis is
accepted by many philosophers of science: it is that if our theory – our
entire theory, or system of beliefs minus its observational component– is
A, there is in principle another theory B which is incompatible with A
but is empirically equivalent to it.4 Empirical equivalence is just same-
ness of entailed observation statements (the question of whether this
4
In fact it is difficult to see how empirically equivalent theories can be incompati-
ble that doesn’t collapse it into practical incompatibility. If logical, just rewrite
one theory to eliminate logical incompatibility between A and B. See Quine
(1975a: 239-241), but also see Quine [1992: §§41-42].
equates to justificatory sameness, sameness of warrant, is a thorny issue

– surely not, if theoretical virtues such as simplicity count for warrant;
I’ll just assume for simplicity’s sake that A and B are also equivalent
with respect to these other virtues).5 Thus, if the situation actually arose
within science, the choice would be real: we would have a factual deci-
sion as to what to count as our ‘ultimate parameter’, our theory of na-
ture, for one must always presuppose in the background, if vaguely,
some theory of nature. The relation of indeterminacy to underdetermina-
tion is that the former is ‘parallel but additional’ (Quine 1969: 303; also
see 1969: 302-304, [1970b]: 209-210). For a radical translator – who ac-
cepts A as his theory of nature – the decision whether to translate the na-
tive as holding A or as holding B would not make a difference with re-
spect to the native’s behaviour, so would not correspond to any fact, so
is indeterminate. The moral is not that no translation is correct, but that
either choice is correct.6
Admittedly, the idea of underdetermination, even in principle, de-
pends on broad and manifestly speculative considerations.7 Nevertheless,
if one accepts it, then it is difficult to see how the argument can be re-
sisted. The argument does depend on the contrast between ‘inside’ a the-
ory or language, and ‘outside’ one, as in the case of the radical transla-
tor. In his earlier reply to Chomsky, Quine says that the very notion of
‘matter of fact’ is an internal notion, a notion which only makes sense
within or inside a theory. Does this mean, absurdly, that English-physics
differs from Polish-physics? Strictly speaking yes, but in a looser sense
such a distinction is matter of degree, not all or nothing. English and
5
Actually theories don’t imply observation sentences directly, but observation
categoricals, like ‘If smoke, fire’; see Quine (1980: 27).
6
It is sometimes supposed – not by Quine – that epistemic holism, together with
Quine’s behaviourism, entails holophrastic indeterminacy. The trouble is that ho-
lism is insufficient for underdetermination, which seems necessary for indeter-
minacy. For holism, rightly considered, is lacking the necessary broadness of
scope; it is ‘unrealistic to extend a Duhemian holism to the whole of science, tak-
ing all science to be the unit that is responsible to observation’ (Quine [1975a]:
229-230).
7
Indeed in Word and Object, the indeterminacy of translation was taken to how
the underdetermination of theory, not the other way round.
146 Gary Kemp
Polish have the advantage of longstanding proximity, communication,

and practices of inter-translation; one seldom if ever calls into doubt the
adequacy of the Polish translation of such a sentence as ‘All bodies have
mass’. All such matters as the truth-value of the Polish sentence or the
English sentence can for all practical purposes be assumed to be ques-
tions arising within a theory; no harm will arise. Only the case of radical
translation reveals the true situation.
(2) The second type of semantic indeterminacy is the Inscrutability of

Reference – the indeterminacy of reference of terms, what famously
Quine called for a while ‘ontological relativity’. It’s that certain differ-
ences in the reference schemes for individual terms do not affect the
truth-values of sentences, thus make no behavioural difference, hence
there is nothing factual to choose between them. This was what the fa-
mous ‘Gavagai’ example was about, with ‘rabbit’, ‘rabbit-stage’ and
‘undetached rabbit-part’. But a few years after the publication of Word
and Object Quine saw the complexities of the ‘Gavagai’ example were
strictly irrelevant to his claim, and came to favour a simple argument for
inscrutability by permutation (see Quine 1969: 55-58, and 1981: 19-20).
Suppose we had a complete truth-theoretic interpretation of a language.
By a proxy-function Quine means a 1-1 function defined over the entire
domain that is not the identity function. Applying such a function in the
metalanguage to the envisaged interpretation, along with corresponding
re-interpretation of the predicates – so that for example a sentence of the
form ‘Fa’ becomes ‘The proxy of a is the proxy of an F’– will not affect
the truth-value of any sentence. So we could translate the natives as talk-
ing about the proxies of rabbits – say cosmic complements of rabbits
(the whole universe minus the rabbit), unit sets of rabbits, or even num-
bers – rather than rabbits, and the new scheme will predict their speech-
behaviour as well as the original scheme did. So reference is inscrutable;
again, meaning that there is no such thing as the right choice amongst
indefinitely many equivalent schemes.
The common complaint that the idea is psychologically implausible
fails to reckon with Quine’s behaviourism. Quine does not think of ref-
erential claims as saying anything about the psychologies of language
users (see Quine 1995: 75). Similarly to Frege, claims about reference
are justified entirely by their effects on the truth-values of statements,

which for Quine are the most that show up in linguistic behaviour or
dispositions. Whatever goes in one’s brain may be as ‘rabbity as ever’
without constituting an ‘empirical bar to the re-interpretation’ as Quine
puts it [1992: 34]; indeed one cannot suppose that it does without buying
into a private language, a gratuitous, free-spinning wheel (imagine one’s
interpreting ordinary utterances proxy-fashion, then saying ‘rabbit’ when
one means privately rabbit-proxy – such would be an inner ceremony
that amounts to nothing). Any clash with intuition can be explained in
terms of the use of language: to learn the word ‘rabbit’ and the rest is to
learn a manner of speaking. Saying ‘cosmic complement of a rabbit’,
and correspondingly for the rest, would come to exactly the same thing
so far as reference is concerned, but no one would understand you, at
least not at first. Understanding words is partly a matter of familiarity
with them.
These arguments – against the determinate application of meaning and
reference – concern entire languages, the whole of their semantics, from
the outside. If these arguments cannot be refuted, then I would propose
two things in mitigation
The first appertains to the claim of holophrastic indeterminacy. The
way to cope is to hold that the notion of the meaning of sentence should
not be expunged from serious science, but that it must be understood
within the context of semantics for particular languages. By this I don’t
just mean that it needs to be relativised, that, since an expression may
occur in different languages with different meanings, the expression ‘x
means y’ requires at least another argument place for the specification of
a language. I refer to the idea that the meaning of a sentence cannot be
explained in general, and perhaps even that it makes no definite sense in
general – but the semanticist’s activity of formulating an account of the
semantics of a particular language may be allowed to proceed, so long as
he proceeds ‘immanently’, and does not explicitly use the general notion
of meaning. In other words, we go with Davidson, whose theories of
meaning don’t actually employ the notion of meaning (whether they do
so implicitly is a harder question).8 The situation is analogous with re-
8
But for dissent, see Lepore and Ludwig (2005).
148 Gary Kemp
spect to Tarski’s definitions of truth. Truth, on his scheme, is explicitly

and eliminatively defined for particular (artificial) languages, but no ex-
plicit and general recipe is provided for definition of truth for other lan-
guages, artificial or natural.9 At most, his definitions provide a model for
emulation. And in both cases – Tarskian truth and Davidsonian meaning
– the key semantical concept is satisfaction (others may use ‘x denotes y’
or ‘x is the semantic value of y’), an extensional concept that is used on-
ly within the semantics of a particular language, not as a general con-
cept. Semantics is thus strictly speaking relative to a particular scheme, a
particular interpretation or T-theory of the language in question, but one
is free to speak breezily in terms of the general notions, so long as they
are understood as partial.
The second appertains to the second indeterminacy claim, the inscru-
tability of reference. What it shows is not that the notion of reference
should be expunged by serious science. When musing on ontology, the
conclusion that Quine draws from inscrutability is that reference is con-
cerned, strictly speaking, not with word-thing relations, but with struc-
ture. To get reference right, it is sufficient to get structure right. But so
long as semantics is conceived as a special science – not as a fundamen-
tal science – nothing prevents the semanticist from choosing one of the
infinitely many adequate schemes of reference. The point may be simply
forgotten about; he may blithely get on with ‘Fido’ referring to Fido.
2. Internal objections
I now shift to less sweeping, less famous Quinean points that appertain
directly to immanent semantics, semantics as a special science.
Quine’s naturalism has as a key tenet a ‘doctrine of gradualism’. By
this he means that although it is possible to ‘limn the structure of reality’
in a stark language containing only mathematical and logical vocabulary,
and physical predicates – where physical objects are just regions of
space-time – he does not think of the softer sciences, the humanistic sci-
ences, as being reducible to physics. In that sense he agrees with Da-
9
Except in languages of infinite order, in which case the truth-predicate can be at
most axiomatized rather than defined.
vidson: he accepts anomalous monism. But only in its negative aspect:

he differs from Davidson in that he takes the fact that desire-belief psy-
chology, sociology, political science, economics, and so on, as not being
free of human interests, and in point of fact resistant to assimilation to
physics, as reason to rank them lower on the scale of objectivity, to call
them Grade B. These sciences are comparatively vague, loose-jointed,
and sometimes ill-defined in their exact purpose; the supervenience rela-
tion mentioned earlier that he envisages is at most a loose one. But – and
this is a point that is frequently missed – to call them Grade B is not to
dismiss them: the ways of dividing up reality which they embody are
enormously complex – having human beings at their centre – and, espe-
cially for purposes of prediction, they are the best we can do, at least for
the foreseeable future. So Quine is not committed to the idea that be-
cause they resist tidy assimilation to physics, semantics – to take the
central case – is therefore nonsense, consigned to the rubbish, along with
homeopathy and astrology. This is an example of the doctrine of gradu-
alism: gradualism on the score of objectivity. It’s just that semantics
cannot be a fully objective and exact science.
It stands to reason, then, Quine doesn’t have some sweeping objection
to semantics-even-as-admittedly-grade-B.
Consider now Quine’s response to Chomsky in ‘Methodological Re-
flections on Current Linguistic Theory’ (Quine [1970a]: 215-227).10 The
subject is not meaning as such but grammar, in the deep sense that
Chomsky was pushing. Quine points out that Chomsky gets him wrong
in supposing that Quinean verbal dispositions are measured by the prob-
ability of an unprompted, out-of-the-blue utterance of a sentence (Chom-
sky [1969]: 57-58); no, they are measured by the disposition to assent to
a queried sentence in a given state of sensory stimulation (where assent,
despite persistent readings to the contrary, is a narrowly behavioural
matter – unlike Davidson’s holding-true, which imports intentional mat-
ter into the situation of the interpreter or translator; see Quine [1975b:
252].
The meat of the essay voices a constant theme of Quine’s philosophy,
namely that the postulation of mental structures is not only empirically
10
Thanks to John Collins for guidance is this section.
150 Gary Kemp
unjustified, it carries an illusion of explanation, an appeal to dormative

virtues or a just-so story. Linguistic competence cannot be explained by
one’s being explicitly guided by an articulate and exhaustive set of rules.
The most that can be said is that one’s speech can be described as fitting
such rules. Of course Chomsky denies this dilemma; he speaks of ‘tacit
knowledge’, or ones ‘cognizing’ such a set of rules, yet for Quine it re-
mains entirely obscure how one can be implicitly guided by, covertly or
tacitly guided by a theory (which one is unable to formulate or recog-
nise).
But Quine does posit innate structure, as Chomsky recognises [1969:
54-55, 63-64]. Unlike the Lockean tabula rasa, Quine’s infant begins up
to its ears in innate dispositions – without them, learning would be im-
possible, as Quine cheerfully recognises (especially in 1974: 12-32; but
also in 1960: 83-85). And I shall assume that Quine’s objections to
Chomsky’s programme are insufficient to derail it. I shall think of such a
theory as describing, in an abstract way, functions realized by the brain.
The mind does have its innate shape, and as many have pointed out, to
theorise about it, about the character or nature of ‘tacit knowledge’, can
just be regarded by way of an analogy of principles of human hearing, of
the human visual system, or even of say digestion. Let us ignore the im-
precision of the analogy. The argument remains that without Chomsky’s
model or something like it of generative grammar or of grammatical in-
variances, certain facts – the astonishing rapidity of language-learning of
children in view of the poverty of the stimulus, the seeming fact that the
generation of well-formed syntax requires operations on underlying
trees, not just sentences or strings themselves, and others – are simply
inexplicable. Such are the sort of fine-grained dispositions that
Chomskian grammar aims to explain.11
Now the details of Chomskian grammar are, thank goodness, not my
concern. My concern is semantics. We’ve seen that even if Quine’s ab-
stract arguments against meaning are accepted, then there is still room
for an immanent, language-specific semantics. We’ve also seen that for
Quine, an account of reference is ultimately an account of structure, the
structure of an individual’s linguistic dispositions. What I propose, then,
11
For a recent round-up of these considerations, see Berwick et al. (2011).
is that a (short-armed) functional theory of language should be regarded

as a substantive model of Quinean linguistic dispositions. Such a theory
is the closest one may come to a language-transcendent linguistic theory,
but such a theory delivers virtually all that is needed. Let me expand on
this.
First, the linguistic interplay between human creatures, and between
them and their environment, is so complex that it is almost inevitable
that we think of certain phenomena as ones of ‘meaning’ or ‘reference’,
as a matter of word-thing relationships; but nevertheless the habit is just
a matter of convenience, if not a measure of our laziness or ignorance, a
place-holder for future work. Consider Quine’s response to Davidson’s
attempt to get him to give up his theory of stimulus meaning – a theory
of linguistic dispositions where the inputs are stimulations of the nerves
and the outputs are the disposition to assent or dissent from sentences
queried – in favour of Davidson’s own theory which takes as basic the
referents ‘out there’.12 The difficulty that besets Quine’s view, but not
Davidson’s, is that one cannot assume sameness of structure in our indi-
vidual neurologies such as the precise layout of photoreceptor cells in
our respective retinas; you have your neural patterns of stimulation, I
have mine. Quine’s reply is first to describe a puzzle: How can it be that
we manage to have more or less the same dispositions with respect to the
environment, when what is going in each of us may be quite different?
The explanation he gives for what he calls the ‘pre-established intersub-
jective harmony’ of standards of perceptual similarity appeals to natural
selection: the standards of perceptual similarity are defined only for the
individual, but the more our ancestors’ standards correlated with the en-
vironment, the more they tended to procreate successfully. Since we do
inhabit similar environments, we all come to it neurologically equipped
to solve the same sorts of problems, despite our anatomical differences.
Thus, to a comparatively unreflective translator, it will seem that a basic,
unchangeable fact is that the natives refer to rabbits and so on; but all
12
These passages (Quine 1995: 20-21, 1996: 473-476, 1999: 160-61, 2000:
408-410) have often been taken as Quine’s giving in to Davidson; I don’t think
so. See Kemp (2012: 124-143).
152 Gary Kemp
that’s going on is certain neural events which make it possible to speak

in that unreflective way.
This leads to the second point. The envisaged functional theory is an
internalist theory of human language, and, by extension, of human cog-
nition in general. Some dispositions – innate or acquired – might even fit
into the slots required for concepts. The presence of such innate disposi-
tions or concepts might for all we know be extensive, as people from
Meno to Chomsky and Fodor have argued. But the subject matter as
Quine sees it would still be neurological; maybe not Fodor’s doorknobs
(Fodor 1998), but the neural correlates that make it possible to have a
word for doorknobs. That is the most that can be made of concepts, for
Quine. If I may assume a functional theory of concepts for the moment,
the concept of a doorknob, in its purely recognitional office, has as its
inputs the exquisitely diverse range of glimpses and feels, conceived
strictly neurologically, of what are in fact doorknobs, together with a
few illusory cases. There is no reason to call the concept that of door-
knob, except that it is by far the most convenient.
Which leads to the third point. If the ultimate subject matter of a theo-
ry of language is the mechanism underlying linguistic dispositions, then
the story will be unaffected by differences in the creature’s distal envi-
ronment that make no proximate difference, that don’t leave a mark in
its hardware. Putnam’s Oscar on Earth in 1750 will have the same dispo-
sitions as his twin-earth doppelganger. If our intuition is that Oscar now
means H2O and not XYZ, then that of course shows up in a disposition;
if we say the claim is about Oscar in 1750, with a perfect doppelganger
on Twin Earth, then the claim is deniable. Something similar can be said
to a worry about indexicality generally; when Oscar and twin-Oscar say
‘My mother is lovely’, surely they refer to different women. And of
course: linguistic dispositions do not settle reference and truth-
conditions; for that, we need context. But once we know the story of a
creature’s linguistic dispositions, we can plug him in to a context and get
truth-conditions; it isn’t a case of something else magically being gener-
ated.
Fourth, our theory is a theory of human language, so it will have lim-
its. Creatures from alien worlds might be translatable, without the theory
of human language being applicable to them. Perhaps they cannot be de-
scribed as having ‘concepts’ in common with us. But although transla-

tion in that case might well be slow work, there is no reason to rule it out
as impossible. Once we see our way around the idea that translation re-
quires sameness of meanings, the same propositions, the position with
respect to the aliens will not be one of missing the essential thing they
have in mind, a meaning. The idea that if one communicates with an al-
ien, then the same theory of language must apply to both can be dis-
missed as an illusion.
And this leads to the fifth point. Whereas we say ordinarily that mean-
ing is what we understand, that we understand by virtue of our grasp of
meaning, or some such thing, Quine proposes that understanding can be
viewed directly in behavioural terms. He thinks of a self-conscious lexi-
cographer as a sort of coach, or a lion-tamer:
Understanding, behaviourally viewed, is thus a statistical effect: it resides

in multiplicities. The nucleus is the word, and mass is made up the count-
less sentences in which the word occurs… Lexicography has no need of
synonymy […] and it has no need of a sharp distinction between under-
standing and misunderstanding either. The lexicographer’s job is to impro-
ve his reader’s understanding of expressions, but he gets on with that with-
out drawing a boundary. He does what he can, within a limited compass, to
adjust the reader’s verbal behaviour to that of the community as a whole, or
of some preferred quarter of it. The adjustment is a matter of degree, and a
vague one: a matter of fluency and effectiveness of dialogue. (Quine
[1979]: 300)
By ‘lexicographer’ Quine means someone compiling dictionaries. The

lexicographer’s art is to coach the user in the use of further words.
Sometimes he gives an explicit definition, but other times gives only suf-
ficient or necessary conditions, or gives examples of proper usage in
characteristic sentences, or describes the things the word denotes; his
overriding aim is to instil fluency in the use of lexemes by hook or
crook. The lexicographer does not overstep his bounds if he finds it nec-
essary to speak of extralinguistic fact – dictionaries are not sharply dis-
tinguished from encyclopaedias, as Quine puts it. One thinks of Wittgen-
stein’s point of the entanglement of language in forms of life.
154 Gary Kemp
3. Residual Quinean doubts
To repeat: The anti-analyticity points do not spell the end of meaning.

The indeterminacy of translation – the radical, holophrastic version–
makes trouble for a general or interlinguistic theory of meaning, but does
not rule out the construction of specific or immanent ones, in the way
that, for example, Davidson described. The inscrutability of reference
shows that, in principle, an account of reference is just an account of the
structure of a constellation of linguistic dispositions, dealing in the final
analysis with matters of neurology. It can in turn be theorised about at a
more general and abstract level by means of a functional theory. The re-
sult is perhaps an emaciated semantic theory, but it is all the same rec-
ognisably semantics. It is not just syntax. Indeed the theorist can keep to
saying ‘rabbit’ refers to rabbits and the like; there is pragmatically noth-
ing wrong there.
But Quine has further doubts about the science of semantics, having to
do with the lack of ordinary determinacy. In a little known piece called
‘Cognitive Meaning’ [1979: 287-302] Quine, assuming an intralinguistic
standpoint, argues that synonymy requires an objective distinction be-
tween a variant of de dicto occurrences which he calls de voce such as
‘Tom does not know that Hesperus is the Evening Star’, and normal de
dicto occurrences such as ‘The Sumerians did not know that Hesperus is
the Morning Star’. Intuitively, Tom is semantically ignorant, the Sume-
rians factually so. But ‘[w]hen we try to find a criterion for that cut-off
point, we sense the quality of myth that invests it’ [1979: 297]. Objec-
tively there is at most a partial ordering of terms by degree of synonymy:
one term is more cognitively synonymous to another than it is to a third
if the set of occurrences where it can supplant the third salva veritate is a
proper subset of cases in which it can supplant the other (the idea made
its first appearance in ‘The Problem of Meaning in Linguistics’ Quine
[1951b]: 47-64).
Perhaps at most this will only lay bare one’s antecedent commitment
to the reality of semantics. But I think there is a more powerful strain in
Quine that runs against semantics: that actual linguistic performance is
slipshod in comparison with any semantic models (see Quine [1970a]:
219f, [1975b]: 244f); that the models idealize too much – unlike say me-
teorology, where the actual phenomena are more complex than our mod-
els, semantic models tend to fill in areas that are ontologically grey. Two
related points.
First, not all the vagueness of ordinary language can be captured by
the dominant theories of vagueness (Quine 1961: 1ff). Degrees of satis-
faction, fuzzy sets, supervaluation, context-sensitive theories, all assume
there is a fact of the matter concerning the precise fuzziness etc., or that
in a particular communicative situation there is some exact thing to be
modelled. And that is at any rate powerless for what has been called
conceptual vagueness. In fact, much ordinary language is vague in this
respect to some degree; circumstances occur in which the law of exclud-
ed middle breaks down, where the problem is not like the way the prin-
ciple breaks down in cases like ‘is bald’ – it’s not a case that might be
circumvented, for example, by replacing a vague monadic predicate with
relational predicate. Wittgenstein dramatized it at PI §80:
I say: “There is a chair”. What if I go up to it, meaning to fetch it, and it

suddenly disappears from sight? – So it wasn't a chair, but some kind of il-
lusion”. – But in a few moments we see it again and are able to touch it and
so on. – So the chair was there after all and its disappearance was some
kind of illusion” – But suppose that after a time it disappears again – or
seems to disappear. What are we to say now? Have you rules ready for such
cases – rules saying whether one may use the word “chair” to include this
kind of thing? But do you miss them when we use the word “chair”; and
are we to say that we do not really attach any meaning to this word, because
we are not equipped with rules for every possible application of it? (Witt-
genstein 1953: §80)
The notion of a chair is as solid as an ordinary notion can be, but in de-
scribable circumstances one finds the concept slipping through one’s
fingers. Not that this is something that ever happens; the point is that
concepts need only be as sharp as the world demands in practice (one
can say ‘stand roughly there’). ‘It is idle to brook definitions against im-
plausible contingencies’, wrote Quine [1992: 21]. And in other cases –
art, the self, indigenous – this sort of thing is not rare. Normal usage of
such terms is variable, vacillating, and sometimes uncertain or confused
if pressed. It seems inevitable that any semantic model will idealise,
running well beyond the behavioural data, filling in and rounding off.
156 Gary Kemp
Second is an issue that is roughly the converse of the foregoing. As

pointed out earlier, Quine recognised that certain sentences can plausibly
be thought to be accepted or rejected in the learning of certain words.
This accounts for the felt analyticity of examples such as the bachelor
(Quine 1991: 395-397), as well as the idea that elementary logic, and
systems of words such as those describing kinship, are analytic. Thus if
we accept that certain statements are, as we say, constitutive of the con-
cepts they involve, then there is a clear sense in which ordinary language
is contradictory.
Charles Chihara once described a certain league of clubs which sub-
scribe to a rule that no secretary of a club in the league is eligible to join
the same club (Chihara 1979). The secretaries of such clubs decide to
form their own club – SecLib, thus declaring a rule:
Df: For every x, x is eligible to join SecLib iff there is a club of

which x is a secretary and which x is not eligible to join.
Ms. Fineline is hired as the club’s secretary; she happens not to belong
to any other club:
Ms. Fineline is secretary of SecLib.

Ms. Fineline is not secretary of any other club.
Is Ms. Fineline eligible to join SecLib? Contradiction either way. The

moral: (1) is pragmatically consistent – it lands one in no trouble– so
long as situations such as that represented by the conjunction of (2) and
(3) don’t occur.
This is ‘more’ paradoxical than Russell’s Barber. The statement
‘There is a barber who shaves all and only those who do not shave them-
selves’ is plainly false; whereas (1) seems a good definition – and in a
loose and popular sense is a definition except for awkward cases like
Ms. Fineline. General rules may be perfectly all right yet have excep-
tions. No word is defined in this case, but Chihara, and I believe Quine,
think that this is what is going on with the Liar Paradox, which arguably
does trade on unrestricted T-sentences taken as constitutive of the word
‘true’: it’s a good rule that with the introduction of the word via the
T-schema, an arbitrary statement is equivalent to its own truth-predi-

cation, even though in weird reflexive cases the rule runs into trouble
(see Quine 1961: 5-9). Similarly with certain other reflexive paradoxes,
and sorities paradoxes (‘If n is small, so is n+1’): what one learns in ac-
quiring the words is susceptible to paradox.
Quine’s position is not that further analysis will always reveal that
there is no antimony, or that we should abandon such habits with words.
Rather, if such problems do arise, then, provided we do require an exact
and consistent description of the relevant part of reality, we should
abandon ordinary language and certain beliefs for regimented theory – as
for example replacing the unrestricted T-schema with a Tarskian-style
definition of truth, intuitive notions of set with axiomatic set theories, or
sorites-culprits such as ‘big’ with the corresponding relative terms
(Quine 1960: 53). Such is the way of scientific progress.
What then of semantics? Something like the above indicates that a
consistent account of ordinary reference is not in the end possible– that a
register of all a creature’s linguistic dispositions is all that can strictly be
required. One can reply that the solution is the adoption of paracon-
sistent logic or some variant, but Quine would agree with Strawson on at
least this point, that “ordinary language has no exact logic” (Strawson
1950: 344).
References
Berwick, Robert C., Paul Pietroski, Beracah Yankama and Noam Chomsky 2011.
Poverty of the Stimulus Revisited. Cognitive Science: A Multidisciplinary
Journal 35, 1207-1242.
Carnap, Rudolf 1942. Introduction to Semantics. Cambridge, Mass.: Harvard Uni-
versity Press.
Carnap, Rudolf 1947/1957. Meaning and Necessity. Chicago: University of Chicago
Press.
Chihara, Charles 1979. The Semantic Paradoxes: A Diagnostic Investigation. The
Philosophical Review LXXXVIII (4), 590-618.
Chomsky, Noam 1969. Quine’s Empirical Assumptions. In: D. Davidson and J.
Hintikka (eds.), 53-68.
Davidson, Donald and Jaako Hintikka (eds.) 1969. Words and Objections: Essays
on the Work of W. V. Quine. Dordrecht: Reidel Publishing.
158 Gary Kemp
Fodor, Jerry 1998. Concepts. Oxford: Oxford University Press.

Hylton, Peter 2007. Quine. Oxford: Routledge Press.
Kemp, Gary 2010. Quine: The Challenge of Naturalism (On Quine, by Peter
Hylton). The European Journal of Philosophy 17 (2), 283-95.
Kemp, Gary 2012. Quine Versus Davidson: Truth, Reference and Meaning. Oxford:
Lepore, Ernie and Kirk Ludwig 2005. Donald Davidson: Meaning, Truth, Lan-
guage and Reality. Oxford: Oxford University Press.
Orenstein, Alex and Petr Kotatko (eds.) 2000. Knowledge, Language and Logic:
Questions for Quine. Vol. 210 of the Boston Studies in the Philosophy of Sci-
ence. Dordecht: Kluwer Academic Publishers.
Quine, Willard V. O. [1946]. On the Notion of an Analytic Statement. In: Quine
2008a, 24-35.
Quine, Willard V. O. [1949]. Animadversions on the Notion of Meaning. In: Quine
2008a: 152-156.
Quine, Willard V. O. [1951a]. Two Dogmas of Empiricism. In: Quine 1961, 20-46.
Quine, Willard V. O. [1951b]. The Problem of Meaning in Linguistics. In: Quine
1961, 47-64.
Quine, Willard V. O. [1954]. The Scope and Language of Science. In: Quine 1976,
228-245.
Quine, Willard V. O. 1960. Word and Object. Cambridge, Mass.: MIT Press.
Quine, Willard V. O. 1961. From a Logical Point of View, second edition. Cam-
bridge, Mass.: Harvard University Press.
Quine, Willard V. O. [1963]. Necessary Truth. In: Quine 1976, 68-76.
Quine, Willard V. O. 1969. Ontological Relativity and Other Essays. New York:
Columbia University Press.
Quine, Willard V. O. [1969]. Reply to Chomsky. In. D. Davidson and J. Hintikka
(eds.), 302-311.
Quine, Willard V. O. [1970a]. Methodological Reflections of Current Linguistic
Theory. In: Quine 2008a: 215-227.
Quine, Willard V. O. [1970b]. On the Reasons for Indeterminacy of Translation. In:
Quine 2008a, 209-214.
Quine, Willard V. O. 1974. The Roots of Reference. La Salle, Ill.: Open Court.
Quine, Willard V. O. [1975a]. On Empirically Equivalence Theories of the World.
In: Quine 2008a: 228-243.
Quine, Willard V. O. [1975b]. Mind and Verbal Dispositions. In: Quine 2008a, 244-
256.
Quine, Willard V. O. 1976. Ways of Paradox, revised edition. Cambridge, Mass.:
Harvard University Press.
Quine, Willard V. O. [1977]. Facts of the Matter. In: Quine 2008a: 271-286.
Quine, Willard V. O. [1979]. Cognitive Meaning. In: Quine 2008a: 287-302.
Quine, Willard V. O. 1981. Theories and Things. Cambridge, Mass.: Harvard Uni-
versity Press.
Quine, Willard V. O. [1987]. Indeterminacy of Translation Again. In: Quine 2008a,
341-346.
Quine, Willard V. O. 1992. Pursuit of Truth, second edition. Cambridge, Mass.:
Quine, Willard V. O. [1992]. Structure and Nature. In: Quine 2008a, 401-406.
Quine, Willard V. O.1995. From Stimulus to Science. Cambridge, Mass.: Harvard
University Press.
Quine, Willard V. O. [1996]. Progress on Two Fronts. In: Quine 2008a, 473-477.
Quine, Willard V. O. [1999]. Where Do We Disagree? In: Quine 2008b, 159-165.
Quine, Willard V. O. 2000. Response to George. In: A. Orenstein and P. Kotatko
(eds.), 408-410.
Quine, Willard V. O. [2001]. Confessions of a Confirmed Extensionalist. In: Quine
2008a: 498-506.
Quine, Willard V. O. 2008a. Confessions of a Confirmed Extensionalist and Other
Essays (edited by Dagfin Føllesdal and Douglas Boynton Quine). Cambridge,
Mass.: Harvard University Press.
Quine, Willard V. O. 2008b. Quine in Dialogue (edited by Dagfin Føllesdal and
Douglas Boynton Quine). Cambridge, Mass.: Harvard University Press.
Strawson, Peter 1950. On Referring. Mind 59/235, 320-344.
Wittgenstein, Ludwig 1953. Philosophical Investigations (translated by G. E. M.
Anscombe). Oxford: Basil Blackwell.
Siu-Fan Lee
Hong Kong Baptist University
[email protected]
Who Wants to Be a Russellian about

Names?
Abstract: Russell had two theories of names and one theory of description. Logi-
cally proper names are Millian names, which have only denotation but no connota-
tion. Ordinary names are not genuine names but disguised definite descriptions sub-
ject to quantificational analyses. Only by asserting that ordinary names are definite
descriptions could Russell motivate his theory of description to solve three prob-
lems for Millian names, namely, Frege’s puzzle, empty reference and negative exis-
tentials. Critics usually discuss Russell’s theories of names and his theory of de-
scription separately. This paper takes a new perspective and presents a dilemma for
the overall project, arguing that it is hard to be a Russellian about names coherently.
The central issue is whether contextualisation is semantic or pragmatic in nature, an
issue very much alive in contemporary debates. This paper traces Russell’s ambigu-
ity on this matter back to his conception of the roles of knowledge by acquaintance
and knowledge by description in naming.
Keywords: Russell, Kripke, Donnellan, Sainsbury, theory of description, theory of

names, semantic-pragmatic distinction
1. Two theories of names and a theory of description
Russell has two theories of names. One applies to what he calls ‘logically
proper’ names, and is based on the realist theory of meaning and the princi-
ple of acquaintance. The other applies to what he calls ‘ordinary proper’
names, for example, ‘Aristotle’, ‘Troy’, ‘Margaret Thatcher’, and is the
theory that these are ‘truncated or telescoped’ descriptions (1918: 243; cf.
1912: 29). (Sainsbury 1979: 57)
Like Mill (1872), Russell regarded a name as having no other function

but to name. Thus, a name is a simple symbol not composed of any in-
dependently meaningful parts. Since a symbol stands for the thing it de-
162 Siu-Fan Lee
notes, a simple symbol can have no other meaning apart from its denota-
tion.
‘Scott’ taken as a name has a meaning all by itself. It stands for a certain
person, and there it is. (Russell 1918: 253)
Names and objects are associated through knowledge by acquaintance. It

is so because under the influence of Cartesian epistemology, acquain-
tance guarantees a direct linkage between object and the mind (Russell
1912: 25). Knowledge by acquaintance becomes the foundation of
knowledge (1912: 26). It also provides the foundation for naming: to
name a thing is for the agent to stand in a naming relation with the object
by acquaintance. Since the agent is in direct contact with the object ac-
quainted, no intermediary is required and the function of a name lies
solely in its contribution of the object as a constituent of the proposition.
A name like this is object-dependent and its utterance guarantees the ex-
istence of the object referred. This is what a genuine name is in its logi-
cal proper sense. Only indexicals and demonstratives, such as ‘this’,
‘that’ and ‘I’, represent knowledge by acquaintance and hence only they
are Russellian genuine logically proper names.
Ordinary names, like ‘Aristotle’, ‘Margaret Thatcher’, are not Russel-
lian names because it is possible for a person to utter an ordinary name
without knowing the object by acquaintance, thus failing to fulfil what
Evans (1982: 43) called ‘Russell’s criterion’, stated as follows:
Whenever the grammatical subject of a proposition can be supposed not to

exist without rendering the proposition meaningless, it is plain that the
grammatical subject is not a proper name, i.e. not a name directly represent-
ing some object. (Whitehead and Russell 1927: 66)
Russell claimed that ordinary names are truncated descriptions and the
attribution is made only on the level of individuals’ thought:
Common words, even proper names, are usually really descriptions. That is
to say, the thought in the mind of a person using a proper name correctly
can generally only be expressed explicitly if we replace the proper name by
a description. Moreover, the description required to express the thought will
vary for different people, or for the same person at different times. The only
Who Wants to Be a Russellian about Names? 163
thing constant (so long as the name is rightly used) is the object to which
the name applies. (Russell 1912: 29-30, emphasis added)
The quotation, together with others such as Russell (1918: 201) and
(1959: 168-169), shows explicitly that Russell took idiolect meaning as
primary. Thus an ordinary name is only synonymous with a definite de-
scription in an individual speaker’s idiolect. It does not entail, however,
that a name is synonymous with a definite description in a public lan-
guage. Such synonymy is necessary for solving the puzzles about names
because they are general puzzles about the public meaning or the seman-
tic function of names. Sainsbury argued for a way to defend public syno-
nymy and this will be explained in section 2.1
How would a definite description be analysed then? Russell (1905)
provided a theory of description independently of his name theories
which claims that descriptions are subject to quantificational analyses.
Descriptions do not provide constituents to propositions but produce
truth or falsity by ranging over a domain, namely, the set of things that
serve as possible values for its variable in a logical system. A definite
description is analysed as follows:
If ‘C’ is a denoting phrase, say ‘the term having the property F’, then ‘C has
the property φ’ means ‘one and only one term has the property F, and that
one has the property φ’. (Russell 1905: 490)
For example, ‘The present King of France is bald’ is true if and only if
(i) There is at least one x which is a present King of France.

(ii) There is at most one x which is a present King of France.
(iii) Whoever is a present King of France is bald.
Let us call the first claim an existential claim, the second a uniqueness
claim, and the third an attributive claim.
1
Russell himself never used the word ‘synonymous’, but simply that common
words ‘are’ usually descriptions. There are of course multiple possible meanings
of the verb to be. However in the context, it seems clear that ‘are’ means ‘have
the same meaning’ or ‘play the same semantic function as’ in Russell’s claim.
This is the reading adopted in this paper.
164 Siu-Fan Lee
Russell justified his theory of description by utility. The theory solves

the puzzle about identity statements, the law of excluded middle con-
cerning empty expressions, and the existence of non-entities or negative
existential statements. Since ordinary names are definite descriptions ac-
cording to Russell, he could (and did) transfer the solutions to corre-
sponding puzzles regarding names in avoidance of positing any extra
semantic notions such as Frege’s sense (1892).
This is an admirable project. But how well does it work? So far critics
usually discuss Russell’s theory of ordinary names and his theory of de-
scription separately. This paper attempts a new perspective to give an
overview of the two sides. I argue that Russell can reply to problems of
either side but only at the expense of sacrificing the cogency of the oth-
er. The upshot is that either way Russell cannot make a coherent account
of both ordinary names and definite descriptions.
I present the argument in the form of a dilemma as follows.
1. According to Russell, an ordinary name is synonymous with a defi-
nite description.
2. According to Russell, a sentence composed of a definite description
is analysed semantically to entail an existential claim, a uniqueness
claim, and an attributive claim.
3. To solve the puzzles about names, Russell must hold that an ordi-
nary name is synonymous with a definite description in a public
language.
4. An ordinary name is synonymous with a definite description in a
public language after contextualisation.
5. Contextualisation is either semantic or not semantic.
6. If the contextualisation process in question is semantic, then an or-
dinary name is synonymous with a definite description in a public
language. This avoids Kripke’s challenge to Russell’s theory of
names (to be explained in section 2 below). However, because syn-
onymy is a symmetrical relation, a definite description is synony-
mous with an ordinary name. Given that an ordinary name does not
have to satisfy an attributive claim in order to refer, a definite de-
scription thus does not have to satisfy an attributive claim. There-
fore, (2) is false.
7. If the contextualisation process in question is not semantic, then

even if a person associates a definite description with a name in her
thought, such a definite description is not synonymous with the or-
dinary name in a public language. So a definite description does not
have referential uses apart from attributive ones. This avoids Don-
nellan’s challenge to Russell’s theory of description (to be ex-
plained in section 3). However, it also entails that (3) is false.
8. By (5)-(7), either (2) is false or (3) is false.
9. Russell’s project requires that both (2) and (3) are true.
10. Therefore, Russell’s project is inconsistent.
Let us examine various premises in details below, in particular the two

horns of the dilemma as specified at (6) and (7).
2. The first horn of the dilemma: a reply to Kripke’s modal

argument
Kripke (1980) argues that an ordinary name cannot be synonymous with

a definite description because they have different modal profiles:
If ‘Aristotle’ meant the man who taught Alexander the Great, then saying
‘Aristotle was a teacher of Alexander the Great’ would be a mere tautology.
But surely it isn’t; it expresses the fact that Aristotle taught Alexander the
Great, something we could discover to be false. So, being the teacher of Al-
exander the Great cannot be part of [the sense of] the name. (Kripke 1980:
30)
According to Kripke, names are rigid designators which pick up the

same reference in all possible worlds in which that referent exists. How-
ever, descriptions are not rigid designators because they denote different
objects in different contexts of evaluation. Hence, names cannot be syn-
onymous with definite descriptions and descriptions cannot give the
meaning of a name. The argument seems strong and is accepted by many
philosophers (cf. Soames 1998).
Sainsbury (1993, reprinted in 2002) defended Russell. He believed
that Russell had relativised names as definite descriptions on the level of
166 Siu-Fan Lee
individual’s thought and it was possible to make ordinary names and de-
scriptions synonymous in a public language by reference to such con-
text.
…Russell’s theory of the meaning (in more or less Kripke’s sense) of ordi-
nary proper names is about the same as Kripke’s: they are rigid designators
whose meaning is their bearer. (Sainsbury 2002: 87)
How can this be the case? We have already noted above that Russell
(1912, 1918) identifies an ordinary name with a definite description on
the level of an individual speaker’s idiolect. Sainsbury characterised this
process as follows:
For every name in the relevant category, every speaker, and every occasion
of the name’s use, there is a description such that in order to make explicit
the thought in the mind of the speaker on that occasion, one must use a sen-
tence in which the name is replaced by the description. (Sainsbury 2002:
86)
Sainsbury argued that the meaning in an individual speaker’s idiolect

can turn into a public meaning via a process akin to a Gricean theory of
meaning.2 First, a speaker S formulates a communicative practice with
S’s audience that by uttering a name N, S intends to convey information
about X, which uniquely satisfied a description the F with which N is
associated in S’s thought. Sainsbury calls this the descriptive theory of
communication, abbreviated as DTC (2002: 91).
The next step makes the object specified above the public reference of
a name. X becomes “the public reference of a name N in a group G if
and only if whenever a member of G uses N, then he and any audience
meet the conditions set by DTC and X satisfies the quantification [speci-
2
Sainsbury did not claim an inspiration after Grice. I see however that his formu-
lation resembles the Gricean approach of meaning as expounded in ‘Utterer’s
Meaning, Sentence-Meaning, and Word-Meaning’ (reprinted in Grice 1989).
Sainsbury emphasised the role of communication because Russell focused on
that as well. Sainsbury claimed: “Although Russell takes idiolect meaning as
primary, he thinks that one can make derivative sense of the successful use of a
name in communication. It is not a matter of thought-sharing, but of reference-
sharing” (2002: 90).
fied in DTC]; that is X is the F” (Sainsbury 2002: 92). This is the de-
scriptive theory of public reference of names, abbreviated as DTR.
The final procedure is to specify a descriptive theory of truth-condi-
tions of utterances containing names from this public reference. This
theory provides formulas of the following form: an utterance of a sen-
tence ‘…N…’, where X is the public referent of N, is true if and only if X
satisfies ‘…ζ…’ (Sainsbury 2002: 93).
Note that in the process, unlike a descriptive phrase, the variable X is
bound to the assignment obtained in the utterance occasion in the idio-
lect stage and is kept fixed by such assignment ever since. In other
words, N is ‘assigned an object before the truth condition is evaluated’
(Sainsbury 2003: 93, my emphasis). Since the assignments of reference
in all subsequent steps are bound by this assignment as well, ‘…N…’ is
expressing the same truth condition of a singular proposition as if N is a
rigid designator.3 Idiosyncrasy is therefore not a problem for Russell.
If Sainsbury was right, then an ordinary name is synonymous in a
public language with a definite description once a contextualisation pro-
cess described above takes place in the initiation stage of a name-use,
ensuring that an ordinary name is a rigid designator. This is a semantic
process because the value produced in the contextualisation process con-
tributes to and is kept fixed all through the definition of the public refer-
ence of a name among users of that name in a community. Note also that
a name so formed under contextualisation is de jure rigid, the type of
rigidity Kripke required of names. A de jure rigid designator designates
the same object in every possible world in which it exists by stipulation.
A de facto rigid designator is rigid because indeed one and only one ob-
ject in every possible world where the object exists fulfils the descrip-
tion.4 A contextualised definite description is de jure rigid. It is thus rea-
sonable to believe that even Kripke would not reject this kind of rigidity.
3
More details are found in Sainsbury (2002: 90-95), especially its footnote 10.
4
Kripke distinguished de jure and de facto rigidity. De jure rigidity obtains ‘where
the reference of a designator is stipulated to be a single object’. In contrast, de
facto rigidity appears “where a description “the x such that Fx” happens to use a
predicate “F” that in each possible worlds is true of one and the same unique ob-
ject (e.g., “the smallest prime” rigidly designates the number two)” (Kripke
1980: 21n). Sainsbury proposed to interpret the meaning of ‘the F’ as bound to a
168 Siu-Fan Lee
The solution sounds perfect, but nothing comes for free really. I argue
that while Sainsbury’s account could successfully defend Russell from
Kripke’s attack on names, it also renders Russell’s theory of description
untenable. The argument is as follows. Synonymy is a symmetrical and
reflexive relation. If a name is synonymous with a definite description,
then conversely, a definite description is also synonymous with a name.
Therefore, it is possible that a definite description would have at least
one possible reading that is always associated with a particular object
prior to any analysis, quantificational or otherwise. Suppose a speaker S
has mistakenly thought that an object named by N satisfies the property
of the F thus linking the name N and the definite description ‘the F’ to-
gether, and all members of the linguistic community G follow S or defer
their judgement to people who follow S, then the mistake of associating
the F with N could go all the way into defining the public reference of N.
5
The result is that ‘the F’ would have a referential use which denotes
something that does not satisfy the property F. This undermines Rus-
sell’s theory of description, especially the attributive claim. No wonder
Sainsbury (2004) himself disagreed with Russell’s theory of description
but supported a referential theory which analyses ‘the F’ as ‘that F’.
3. The second horn of the dilemma: a reply to Donnellan’s argument
Strawson (1950) and Donnellan (1966) criticised Russell’s theory of de-

scription. Strawson argued against the existential and uniqueness claims
speaker’s thought. This procedure is governed by convention. Hence, I claim that

Russell’s contextualised description is de jure rigid.
5
Knobe (2007: 83) reported that in a survey by Machery, Mallon, Nichols, and
Stich, the majority of East Asian non-philosophy subjects who were given Krip-
ke’s classic stories of Gödel and Schmidt (1980: 83-84) offered the response that
the word ‘Gödel’ referred to the person who really discovered the incompleteness
of arithmetic. I think this response can be explained by Sainsbury’s contextuali-
sation. If a name is associated with a definite description from the start, it is psy-
chologically possible that such association would strike the speakers as an essen-
tial feature of the name. Thus even when the object does not satisfy the descrip-
tion indeed, people might still retain the impression that the name is supposed to
designate something that does.
while Donnellan argued against the attributive claim. The common idea
is that definite descriptions can sometimes be used as referring expres-
sions. Donnellan argued, for example, that ‘Smith’s murderer is insane’
can be interpreted attributively as anyone who murdered Smith is insane
when it is uttered in the crime scene against the violated body of Smith.
It can also be interpreted referentially to denote a particular person
whom was believed to be Smith’s murderer in a court by the audience
when they saw that person behaving weirdly even though he may not
have been the actual murderer of Smith. Other examples were raised and
discussed in the literature subsequently.
Kripke (1977) argued that Donnellan confused the semantic and
pragmatic uses of an expression, or the semantic reference and the
speaker’s (pragmatic) reference in these cases. Donnellan’s criticisms
are misguided because similar problems could occur to other expression
types such as proper names. For example, suppose two persons saw
Smith raking leaves in the distance and mistook him as Jones. Their
conversation ‘What is Jones doing there?’ ‘Raking leaves’ is indeed true
of Smith, not of Jones. This does not imply however that Smith thereby
changed his name or that the name ‘Jones’ has two semantic functions
denoting Smith and Jones respectively.
Kripke argued that if the ambiguity was semantic, then it could be dis-
ambiguated by stipulation. For example, the word ‘bank’ is ambiguous
in English and it would be disambiguated if there is a convention that
accepts only one use but reject the other, or restricts the word for one
meaning but creates another word to signal the other. Donnellan’s ambi-
guity is not of this kind however. People miscall an object because they
have different beliefs about the object concerned. This misuse of lan-
guage will persist even if a convention is set to limit an expression type
to one use but not the other.
I reckon that Donnellan’s ambiguity is more complex, so Donnellan
and Kripke might have just talked past each other.6 However, I shall not
6
I developed this argument in my dissertation (unpublished) using Kamp’s (1981)
and Heim’s (1982) discussion of anaphoric expressions, represented as dynamic
changes of references in Kamp’s Discourse Representation Structure (DRS), of
which an introduction is found in Kadmon (2001). I believe that not all quanti-
fiers are used in English as anaphors but it is part of the English grammar that
170 Siu-Fan Lee
develop this line of argument in this paper because I want to address an-
other issue. Let us suppose for the moment that Kripke was right; does it
imply that a Russellian approach to names would welcome Kripke’s res-
cue? I think not. A Russellian about names would hesitate to embrace
Kripke’s argument because it would make it difficult, if not impossible,
to defend the thesis that ordinary names are definite descriptions. Let me
explain.
As discussed above, a plausible line to synonymise an ordinary name
and a definite description starts at the level of an individual’s thought.
Russell pursued this line and Sainsbury developed it into a possible pro-
gramme generating public meanings. Kripke (1977) would deny such
possibility because according to Kripke, for example, whether the court
believes that Jones is Smith’s murderer at a certain point of time does
not have any bearing on the semantic reference of the phrase; ‘Smith’s
murderer’ always refers, and only refers, to the real murderer regardless
of the utterer’s intention, knowledge and beliefs. ‘Smith’s murderer’ is
but one example of a definite description. So the claim must be that in
general anything people mistake as the denotation of a definite descrip-
tion in a context of utterance is irrelevant to the semantic function of it.
However, if this is so, for the sake of consistency, we must also accept
that whatever people take (rather than mistake) as the denotation of a
definite description in a context of utterance is irrelevant to the semantic
function because the same procedure applies. A person simply may not
know whether or not his belief is true indeed at the time he holds it.
Hence, if we were to deny the role of knowledge in determining the
denotation of a definite description in order to avoid determination based
on mistaken beliefs, we must also deny the same procedure when true
beliefs are at stake. The upshot is that Russell would not be able to syn-
onymise an ordinary name with a definite description in any case be-
definite descriptions are used like pronouns. Since anaphoric expressions are typ-
ically referring expressions, it is also part of the English grammar that definite
descriptions are capable of being used referentially. If this is correct, Donnellan’s
ambiguity is not simply a matter about a speaker’s contingent belief of things,
but a matter about the mechanism for communicators to encode and decode a
name with its reference, which brings us back to the interface between semantics
and pragmatics and question whether contextualisation can be semantic.
cause an ordinary name can only pick up the pragmatic speaker’s refe-
rence of a definite description but never its semantic reference. Sains-
bury’s programme of public meaning would not be able to kick start. In-
deed, it would seem impossible to have any programme trying to link
people’s thought (Russell 1912, 1918, 1959 as quoted in section 1) with
the semantic meaning of their utterance. No wonder Kripke (1980: 96-
97) himself would prefer a causal theory of names which replaces the so-
called Frege-Russell descriptive theory and consider any fulfilment of
descriptive beliefs in naming as trivial (Kripke 1980: 88n).
In sum, although it is convenient to sweep any unwanted semantic as-
signment swiftly under the carpet of pragmatic alterations, once we
adopt the strategy, consistency would demand that we sweep away any
assignment possibly desirable for Russell’s purposes. This is not a result
that a Russellian would want in order to keep Russell’s theories coher-
ent.
4. Contextualisation: Semantic or pragmatic?
The central issue behind this dilemma is the role of contextualisation in

meaning. Sainsbury endorsed contextualisation as a part of the meaning-
giving process, thus it is possible to establish a synonymous relation be-
tween two expressions which are only linked together contingently in
some people’s minds as long as other members of the linguistic commu-
nity use the association in the same way. Kripke (1977) implies a diffe-
rent picture according to which the semantic function of a linguistic ex-
pression consists purely in some logical mechanisms operating inde-
pendently of people’s use. If names and descriptions have different
mechanisms to pick up things, then even if they happen to yield the same
reference, it is a coincidence that they do and so they still do not have
the same meaning. Contextualisation, under this view, can only be prag-
matic but not semantic.
172 Siu-Fan Lee
There is no dispute over whether contextualisation can be pragmatic;

the issue is whether contextualisation can be semantic, which is still a
much debated issue.7 How would Russell look at this matter?
Russell might not be aware of the contextualisation problem and he
surely was not worried about the synonymous relation. He claimed that
as long as the object referred is constant, communication will succeed
because the proposition expressed is the same:8
But so long as this remains constant, the particular description involved

usually makes no difference to the truth or falsehood of the proposition in
which the name appears. (Russell 1912: 30)
However, the fact that we can get two readings from Russell supporting
either his theory of ordinary names or his theory of description shows
there must be some ambiguity and it is a significant matter. I argue that
the root of this ambiguity lies in the types of knowledge Russell conside-
red as relevant to and required for a semantic process, namely know-
ledge by acquaintance or knowledge by description. Given that know-
ledge by acquaintance and knowledge by description are quite different
in nature, Russell’s free use of these two types of knowledge turns out to
be the cause of subsequent confusion.
Russell illustrated three ways to use a name as follows:
Assuming that there is such a thing as direct acquaintance with oneself,

Bismarck himself might have used his name directly to designate the par-
ticular person with whom he was acquainted… But if a person who knew
Bismarck made a judgment about him, the case is different. What this per-
son was acquainted with were certain sense-data which is connected (right-
ly, we will suppose) with Bismarck’s body. His body, as a physical object
7
For example, there is the debate between contextualism and invariantism. Con-
textualists (Recanati 2004; Travis 2006) argue that meanings, at least of par-
ticular types, vary with context. Invariantists (Cappelen and Lepore 2005) argue
that sentences have shared type-contents and we can find a ‘minimal’ invariant
meaning (overlapping content) to each word, thus disallowing contextual ad-
justments except in the cases of demonstratives and indexicals. Borg (2004) is a
minimalist. Between the two extremes lie many possible intermediate positions,
including Stanley, Bach, Szabo and others.
8
See also Sainsbury (2002: 85-101).
and still more his mind, were only known as the body and the mind con-
nected with these sense-data. That is, they were known by description…
When we, who did not know Bismarck, make a judgement about him, the
description in our minds will probably be some more or less vague mass of
historical knowledge… Thus it would seem that, in some way or other, a
description known to be applicable to a particular must involve some refer-
ence to a particular with which we are acquainted… (Russell 1912: 30)
There is no doubt that Russell required a name to be initiated by some

knowledge by acquaintance. Since the initiation of a name is obviously a
semantic process, if not the whole of it,9 knowledge by acquaintance
must have some semantic role in a name use. Acquaintance is a causal
event. So Russell must hold a certain causal theory of names. However,
exactly what kind of a causal theory would Russell hold? Would it be a
pure causal theory such that knowledge by acquaintance is necessary and
sufficient for determining the reference of a name? Or, would he rather
hold one in which descriptive beliefs also play some semantic, viz. refer-
ence-determining, role? In other words, would knowledge by description
play some semantic role in the use of a name? This is a significant ques-
tion for Russell particularly because he also claimed that a name is a def-
inite description.
Contextualisation typically involves restricting a domain or reference
according to speaker’s beliefs at a particular temporal-spatial location.
So we can assume that if knowledge by description plays some semantic
role, contextualisation is possibly semantic.
Let us examine the possible answers in turn. There are two possible
interpretations to the claim that names are initiated by knowledge by ac-
quaintance. A strong interpretation requires that a name is initiated sole-
ly through knowledge by acquaintance; a moderate interpretation re-
9
Indeed a Millian theorist of names would hold that initiation is the only semantic
process relevant to a name. Kripke defined the Millian theory of names as hold-
ing that: “The linguistic function of a proper name is completely exhausted by
the fact that it names its bearer” (Kripke 1979: 383). A name can be linked to its
bearer by ostensive definition in perceptual contact. Hence, a Millian could hold
that the only semantically relevant process of a name is the tagging of a name on-
to an object. All other processes may be regarded as pragmatics or matters of
deference among members of a linguistic community.
174 Siu-Fan Lee
quires that initiation involves both types of knowledge. The strong inter-
pretation can be further divided into two strands: S(i) states that changes
in transmission play no semantic role, so a semantic analysis of names
only involves knowledge by acquaintance; S(ii) holds that transmission
is part of the semantic processes of a name use and knowledge by de-
scription may play some regulatory role in this stage. What implication
does each of these three readings have on the nature of contextualisation
and which interpretation would Russell embrace?
S(i) recognises contextualisation as only a pragmatic process because
knowledge by description plays no semantic role in the whole process,
from initiation to transmission. Descriptive beliefs about an object in a
speaker’s mind would not affect the reference determination of the name
no matter how dominant the beliefs have become or whether the beliefs
are indeed true. This reading entails a pure causal theory of naming,
which fits squarely with the Millian view of names.10
However, this view also forms the first horn of Russell’s dilemma as
stated above. Names and definite descriptions are not possibly synony-
mous because the meaning of a name is solely determined by knowledge
by acquaintance. All three puzzles about Millian names would return
and Russell would not be able to solve them. Empty names can hardly be
initiated under this view because naturally no knowledge by acquain-
tance is possible when there is no object to be acquainted. Negative exis-
tential statements would be meaningless as the subject terms are empty.
Frege’s puzzle would also reappear because an object can be acquainted
in many ways, generating knowledge by acquaintance not recognisable
as belonging to the same object.
Furthermore, since subsequent users of a name would not have
knowledge by acquaintance of the designated object, the transmission of
a name-use would depend only on deference. It might be hard for an
agent to use a name without being able to identify what it refers. There-
10
By a ‘pure’ causal theory, I mean a doctrine that does not recognise any semantic
role played by descriptive beliefs. Kroon (1987) distinguished three types of
causal theories of names: causalism, causal neutralism and causal descriptivism.
A pure causal theory is roughly the same as Kroon’s causalism.
fore, a name-use practice of this type would be short, limited in scope of

circulation, and easily be mistaken.11
To avoid these problems, Russell is likely to take the remaining read-
ings which both acknowledge some semantic role of knowledge by de-
scription.
The moderate interpretation states that the reference of the name is de-
termined both by knowledge by acquaintance and knowledge by descrip-
tion. Suppose I name a particular table I saw in a particular library at a
particular time yesterday P. ‘P’ names the table I saw (or, the sense-data
I acquainted, according to Russell); it also names the table I believed I
saw at that particular time and space (i.e. a description linking the sense-
data with the object). All subsequent users of the name have to follow
my practice if they are to use the same name. ‘P was antique’ and ‘the
table I saw yesterday in the library was antique’ would say the same
thing because the two expressions pick out the very same object back in
the naming situation. Contextualisation so described is semantic, not
merely pragmatic.
This interpretation fits well with Russell’s description of the case
when someone who knows Bismarck makes a judgement about him.
11
If only deference were in use and speakers definitely have no knowledge by de-
scription, then people would be using the name without any idea of what it refers.
Yet people use language to express their thoughts, so there must be knowledge
shared and to be communicated among speakers when they use a name. A certain
body of knowledge must be associated with the use of a name and motivate peo-
ple to use it in certain ways, typically to refer to the object that satisfies the asso-
ciated knowledge. I thus doubt that people would have the interest to keep using
it if absolutely no knowledge by description is involved. Besides, even if they
keep the practice, it is easy to make mistakes just like when people are playing
Chinese whispers. Yet in reality, ‘Aristotle’ designates the unique person even
after more than 2000 years and no one alive has knowledge by acquaintance with
the designated person anymore. So it seems that knowledge by description must
somehow be involved and be used to regulate the usage of a name. It is thus not a
trivial matter to explain the mechanism in detail. Kripke (1980) distinguished
reference-fixing from meaning-giving functions. I think however that this dis-
tinction is not helpful to understand Russell’s case because Russell did hold the
Millian view of names that the meaning of a name is to refer. So Russell was not
trying to explain meaning separately and independently from reference-fixing.
176 Siu-Fan Lee
Baptism typically involves someone (other than the person or object be-
ing named) using a name on some object he or she acquaints, so baptism
can easily be described as belonging to this category.
The moderate reading is intuitively plausible, too. Sainsbury (2005:
106-107) argued that baptism involves both object-related intentions and
descriptive intentions. Suppose during a ceremony a priest baptised a
boy and a girl who are twins, the priest uttered a girl’s name to the boy
and a boy’s name to the girl because they looked so alike. Our intuition
is that the boy would not thereby get a girl’s name and the girl a boy’s
regardless of what causally happened in the ceremony. The fact that we
would consider the case an error shows exactly that baptism does in-
volve, and is regulated by, descriptive intentions rather than being de-
termined by acquaintance alone.
There is a catch for the moderate interpretation however and this con-
stitutes precisely the other source of Russell’s incoherence. The reading
implies that reference as determined by knowledge by acquaintance and
by knowledge by description may diverge. If baptism includes know-
ledge by description, then an initiator of a name could attach a mistaken
description to the object from the start and it is not easy to distinguish
whether subsequent mistaken judgements using the name are mistakes in
language or in judgement. Suppose in a maternity ward a nurse mis-
placed a baby before it was named, a mother then named the baby in her
arms having a mistaken descriptive belief that it was her baby.12 Suppose
the mistake was never recognised and corrected, such false description
would associate with the baby in all subsequent uses of the name. ‘My
baby is beautiful’ uttered by the mother looking at the baby in her arms
would be false of her biological child but true of the baby in acquain-
tance. A similar situation happens for the name, too. Suppose the mother
gave the name ‘Jack’ in the scenario above. Is the baby in acquaintance
Jack, or is the mother’s biological baby Jack? Neither is a better answer
than the other because the moderate interpretation acknowledges know-
12
This example is modified from a similar one in Sainsbury (2005: 119-120). In
Sainsbury’s example, the mistake was corrected and he used his example to show
that a referent is not forever. My example however is to show another point.
ledge by acquaintance and knowledge by description as both semantic.

Therefore we cannot disregard either as illegitimate or derivative.
This brings about a situation akin to Donnellan’s examples. ‘Smith’s
murderer is insane’ may mean whoever murdered Smith, just like ‘my
baby’ uttered by the mother could mean her biological child. Similarly,
‘my baby’ may refer to the baby the mother is holding just like ‘Smith’s
murderer is insane’ may mean in some contexts the person believed to
be the murderer by the utterer. If contextualisation is possibly semantic,
then Donnellan’s challenge to Russell is likely to stay.
S(ii) holds that knowledge by description can be taken into the seman-
tic account of a name in the transmission process even if not in the initia-
tion process. Evans’ (1973) causal theory of names belongs to this cate-
gory. It takes informational links rather than baptismal ones as the de-
terminant of reference. For example, ‘Madagascar’ was originally the
name of a part of the African continent. Marco Polo mistook the natives’
descriptions about that place and thought they were descriptions of the
island. This usage became more dominant and now ‘Madagascar’ is the
name of the island, not the continent.13
S(ii) allows semantic contextualisation, so Donnellan’s challenge
would apply similarly to this reading. That is, reference as determined
by knowledge by acquaintance and by knowledge by description may
diverge and there is no overriding semantic reason to determine which is
the one and only correct reference. Indeed, Evans’s causal theory per-
mits reference change; correctness of reference also depends on con-
tingent factors such as dominance of usage and authority.14
13
There are disputes over how the phenomenon is explained. Evans (1973) consi-
dered it a case of reference shifting. Sainsbury (2005) considered that Marco Po-
lo had unwittingly initiated a new name. Either admits the role of descriptive be-
liefs in reference determination and so this matter does not influence my overall
argument. Evans admits the role of descriptive beliefs in the transmission; Sains-
bury in the initiation. Sainsbury allowed an initiation to be unintentional to the
baptiser also because there are obvious cases such as a name originated through a
slip of the tongue. So he believed that the existence of a name is defined by us-
ers’ subsequent practice, not the baptiser’s initiation.
14
Evans identified dominance as a factor in determining reference. He wrote, “I
think we can say that in general a speaker intends to refer to the item that is he
178 Siu-Fan Lee
In sum, I argue that Russell would endorse some form of a causal the-
ory because naming is initiated somehow through knowledge by ac-
quaintance. If knowledge by acquaintance is the only semantically rele-
vant knowledge, then contextualisation would be merely pragmatic. This
could happen under a certain understanding of the strong interpretation;
but all puzzles of names would return and this move would defeat Rus-
sell’s cause. To avoid this, Russell is likely to allow knowledge by de-
scription to play some semantic role in the initiation (the moderate inter-
pretation) or the transmission processes (the second version of the strong
interpretation). It is thus reasonable to believe that Russell would regard
contextualisation as possibly semantic. In any case, any interpretation
would lead to the dilemma mentioned in the first half of this essay. So it
is still hard for Russell to maintain coherence.
5. Conclusion
To conclude, this paper investigates some inherent tensions in Russell’s

project about names and definite descriptions. It argues that Russell
could save his theory of ordinary names from Kripke (1980) if he en-
dorsed contextualised description as a semantic tool for reference deter-
mination of a name. Alternatively, he could save his theory of de-
scription from Donnellan (1966) if contextualisation is merely a tool for
pragmatic reference. Russell’s project will be compromised either way.
dominant source of his associated body of information” (Evans 1973: 202). In

turn, who and what define dominance can be a matter of power and authority.
For example, in the Madagascar case should Europe not be the dominant power
of trade and military expansions in modern times, or should Marco Polo’s infor-
mation not be incorporated in the map-making process, it is imaginable that
‘Madagascar’ could revert to be the name of a part of the African continent as
used by the natives rather than that of the island when Marco Polo’s error was
discovered. That this recovery did not happen shows authority does play a role.
In another study, Dickie (2011) compared the Madagascar case with the case of
Geoffrey Chaucer, which once also involved a large body of misinformation. Yet
in the latter case, the errors are acknowledged and recovered without anyone
considered that reference shifting of the name is involved.
I argue that the root of Russell’s problem lies in his ambivalence over
how knowledge by acquaintance and knowledge by description are re-
lated in fulfilling the semantic function of a name. If Russell allowed on-
ly knowledge by acquaintance in initiating a name, then synonymy
would not be established and the Millian puzzles would not be solved.
So it is likely that he would allow knowledge by description to play
some semantic roles instead, either in the initiation or the transmission
processes. A certain kind of causal descriptivism would thus provide a
coherent reading of Russell’s literature about names. However, such a
move would also entail departure from Russell’s theory of description.
Either way has costs to pay, it thus seems hard to be a Russellian about
names coherently.
References
Borg, Emma 2004. Minimal Semantics. Oxford: Oxford University Press.

Cappelen, Herman and Ernie Lepore 2005. Insensitive Semantics. Oxford: Black-
well.
Dickie, Imogen 2011. How Proper Names Refer. Proceedings of the Aristotelian
Society 111 (1), 43-78.
Donnellan, Keith 1966. Reference and Definite Descriptions. Philosophical Review
77, 203-215.
Evans, Gareth 1973. The Causal Theory of Names. Proceedings of the Aristotelian
Society Supplement 47: 187-208.
Evans, Gareth 1982. The Varieties of Reference (edited by John McDowell). Ox-
ford: Clarendon Press.
Frege, Gottlob 1892. On Sense and Reference. In: A.W. Moore (ed.), 23-42.
Groenendijk, Jeroen, Theo Janssen and Martin Stokhof (eds.) 1984. Truth, Interpre-
tation and Information. GRASS 2. Dordrecht: Foris.
Grice, Paul 1989. Studies in the Way of Words. Cambridge: Harvard University
Press.
Heim, Irene 1982. The Semantics of Definite and Indefinite Noun Phrases. Ph.D.
dissertation. University of Massachusetts, Amherst.
Kadmon, Nirit 2001. Formal Pragmatics. Oxford: Blackwell.
Kamp, Hans 1981. A Theory of Truth and Semantic Representation. In: J. Gro-
enendijk et al. (eds.), 1-41.
Knobe, Joshua 2007. Experimental Philosophy. Philosophy Compass 2/1, 81-92.
180 Siu-Fan Lee
Kripke, Saul 1977. Speaker’s Reference and Semantic Reference. Midwest Studies
Kripke, Saul 1979. A Puzzle about Belief. In: A. P. Martinich (ed.), 382-410.
Kripke, Saul 1980. Naming and Necessity. Cambridge: Harvard University Press.
Kroon, Fred 1987. Causal Descriptivism. Australasian Journal of Philosophy 65, 1-
17.
Lee, Siu-Fan 2009. Names and Empty Names. Ph.D. dissertation. King’s College
London.
Marsh, Robert C. (ed.) 1956. Logic and Knowledge. London: George Allen & Un-
win.
Martinich, A. P. (ed.) 1996. The Philosophy of Language. 3rd ed. Oxford: Oxford
University Press.
Mill, John Stuart 1872. A System of Logic. Collected Works of John Stuart Mill,
Vol. VII. Toronto: University of Toronto Press, 1974.
Moore, Adrian W. (ed.) 1993. Meaning and Reference. Oxford: Oxford University
Press.
Recanati, Francois 2004. Literal Meaning. Cambridge: Cambridge University Press.
Reimer, Marga and Anne Bezuidenhout (ed.) 2004. Descriptions and Beyond. Ox-
ford: Oxford University Press.
Russell, Bertrand 1905. On Denoting. Mind 14, 479-493.
Russell, Bertrand 1912. The Problems of Philosophy. Oxford: Oxford University
Press.
Russel, Bertrand 1918. The Philosophy of Logical Atomism. In: Robert C. Marsh
(ed.), 175-281.
Sainsbury, R. Mark 1979. Russell. London: Routledge.
Sainsbury, R. Mark 1993. Russell on Names and Communication. In: M. Sainsbury
2002, 85-101.
Sainsbury, R. Mark 2002. Departing from Frege: Essays in the Philosophy of Lan-
guage. London: Routledge.
Sainsbury, R. Mark 2004. Referring Descriptions. In: M. Reimer and A. Bezuiden-
hout (ed.), 369-389.
Sainsbury, R. Mark 2005. Reference Without Referents. Oxford: Clarendon Press.
Soames, Scott 1998. The Modal Argument: Wide Scope and Rigidified Descrip-
tions. Nous 32 (1), 1-22.
Strawson, Peter 1950. On Referring. Mind 59, 320-344.
Travis, Charles 2006. Insensitive Semantics. Mind and Language 21, 39-49.
Whitehead, Alfred North and Bertrand Russell 1927. Principia Mathematica. vol.1,
2nd ed. Cambridge: Cambridge University Press.
Gabriele M. Mras
University of Vienna
[email protected]
Bradley, Russell, and the Structure of

Thought
Abstract: Russell’s multiple theory of judgement is commonly regarded as a fail-
ure. This is so because any attempt to appeal to some entity as that in virtue of
which an ascription of properties is true, inevitably invokes a regress. But Russell
could have known this all along. He was familiar with this objection from Bradley’s
work and he used a “regress argument” himself while he still was philosophizing in
the tradition of British idealism.
The reason why Russell, despite his own insistence that no ‘third thing’ could unite
the items of a sentence, ends up with a view that makes the appeal to such a thing
necessary has to do with what is commonly held against him: Russell’s particular
way to rely on the notion of structure led him astray as a critique of Bradley.
Keywords: theory of judgement, Russell, Bradley, Frege, unity of the proposition,

structure, relations, first and second order functions
1. Bradley and the unity of a thought
“Bradley won” – says Davidson in Truth and Predication; which is a

remarkable thing to say. Why did he win? According to Davidson, be-
cause Bradley’s former pupil failed by a criterion that Britain’s most fa-
mous idealist rightly declared crucial for being able to account for
thought.
The philosopher in contrast to whom Bradley is assessed here so fa-
vourably is one of the founders of analytic philosophy: Bertrand Russell.
His theory of judgement is said to miss the conditions of the possibility
of thought by failing to comprehend “thought’s peculiar unity” (Da-
vidson 2005: 99ff).
I would want to agree with this judgement. The expectations, howev-
er, that I will now put forward Davidson’s verdict as a diagnosis, as has
182 Gabriele M. Mras
been done recently (paradigmatically Hanks 2007, 2012), will not be sat-
isfied. What is at issue between Russell, and in a way not only Bradley
but also Wittgenstein insisted against him on a non-reductive ‘top down’
approach, is not grasped in its depth by insisting on what is conception-
ally necessary for being a part, e.g. what is not sufficient for being what
a part is a part of.
One should not forget: Russell himself emphasized many times, as
here in his “Reply to Mr. Bradley” that “no enumeration” of “the con-
stituents” “will reconstitute” a unity, since “any such enumeration gives
us a plurality, not a unity” (Russell 1910: 373).
If one acknowledges this point about decomposing and rebuilding,
then aiming at a unity while defending ‘strict pluralism’ seems to involve
no inconsistency. Being an advocate of both, according to Russell, is in
this sense not only possible – it is required. A picture of thought that pre-
sents thought as not decomposable would not be the right medium to al-
low inferential relations to be represented.
A Bradleyan account of the conditions of there being a unity of
thought seems, indeed, to be deficient in this respect. Bradley wanted to
interpret our saying something of the form A stands in relation to B as
“pointing” to “a unity”, “a substantial totality beyond relations” (Bradley
1946: 141). His answer to “Can we … have a plurality of independent
reals […]?” is consequently “No” in an ontological sense (Bradley 1946:
124). This, however, gives the form of a judgement a telos that runs
counter to our ordinary understanding of the conditions met, if a judge-
ment fulfills its purpose.
Russell focuses exactly on this mismatch. Attributing it to a confusion
concerning the relation of the pairs true or false and necessary or possi-
ble, he declares:
I am inclined to think that a large part of my disagreement with Mr. Bradley

turns on a disagreement as to the notion of ‘necessity’. I do not myself ad-
mit necessity and possibility as fundamental notions: it appears to me that
fundamentally truths are merely true in fact […]. deducibility is […] a fact,
i.e. it has no modal property of necessity not possessed by the facts which it
relates. (Russell 1910: 374)
Mr. Bradley’s contention that such notions as ‘a man’ contain an element of
negation is one which I cannot admit. […] It is of course undeniable that ‘a
Bradley, Russell, and the Structure of Thought 183
man’ implies the denial of more than one man but it does not follow that
this denial is part of its content. Such a view involves the assumption […]
that all inference is essentially analytic […]. […] This view appears to me
to be erroneous […]. (Russell 1910: 377)
As often as he repeats how opaque it is what is to be understood by the

famous claim concerning the un-reality of relations – “it is very difficult
to state this doctrine in any … precise meaning” (Russell 1966: 139) –,
Bradley’s doctrine for him is clearly connected with the idea of truth be-
ing grounded in “a substantial totality” that bears with it the idea that the
opposite of what is said by a judgement is unconceivable/unthinkable.
A “substantial totality without division” (Bradley 1946: 128) could
because of that, for Russell, not possibly be condition for thought. It
simply fails to make intelligible that what is claimed to be so might not
be as it is claimed.
So looking at what Russell says about wholes and parts in its historical
context ought to put the problem “that exercised the fathers of analytic
philosophy following a distinguishable tradition traceable to Plato’s dia-
logues” (García-Carpintero 2010: 279; see also Picardi 2008) into a dif-
ferent light.
Since it is not the case that any unity will do in order to account for
the possibility of thought – not the one suggested by Bradley, at least not
for analytic philosophers – it is not the case either that the right rela-
tion(ship) between ‘parts’ becomes any clearer, if one is reminded that
parts are parts only in respect to a unity of which they are such.
For philosophers in the Hegelian tradition it is the copula that serves
as hint that the relation of ‘parts’ is not to be understood as mere combi-
nation. Given that the judgement’s unity is claimed to be possible in vir-
tue of a third thing x, to which one cannot appeal but by two distinct
terms, what the copula shows is, that what a thing is, can be expressed
by a judgment only.
To Russell this approach seemed to ground truth in something object/
substance-like and to dissolve this object/substance as possible reference
of a judgement. What if the object assumed to be appearing as subject
and predicate was judged not to be as it seems? Russell takes this ques-
tion to support his view that the “form” of a judgment has to match the
conditions of our ordinary understanding of what is, respectively is not

so, if a judgement is true, respectively is false, and so further to
acknowledge an ‘ontology’ that is not Bradley’s.
As it is not the case that emphasizing that ‘parts’ must be determined
in respect to the whole (thought) is not what Russell would want to deny,
it is also not the case that Russell would have wanted to ignore that the
role of the ‘parts’ in question has to do with “the capacity to be true or
false” (see Gaskin 2008; King 2007; Hanks 2007).
There is no question that Russell would have happily embraced:
[…] what is judged must be capable of being true or false, and a disunified
collection of objects, properties and relations […] lacks that capacity. […]
the answer to the question ‘What is judged?’ is the sort of thing that is true
or false. (Hanks 2012: 39f; see also Davidson 2005: 99ff)
The whole point of Russell’s understanding of analysis is that what is

judged must be capable of being true or false, i.e. “truths are merely true
in fact” as he insists against Bradley.
It was in the course of rethinking what has to be presupposed for
judgements to be true that Russell became convinced that a regress à la
Bradley’s does not have the argumentative force assumed. It does not
prove that “non-intrinsic external relations” could be no part of a sen-
tence, thought or a fact. Why? Not (only) because “A is left to B”, “A >
B”, “A lives north of the Danube Canal” as assertions do not state some-
thing that is constitutive of the items talked about. And, not because A is
something different from the relation it bears to B. This so far could be
taken to point right into the direction of Bradley’s approach; as it did for
Russell for a long time (see Griffin 2013 and Hylton 2008). What con-
vinced him that, even if there is no thing that constitutes/makes “> B” to
be true, the relation “ > ” functions (as point of correspondence are con-
siderations that have to do with falsehood).
For Russell, that something is false requires understanding what it is
that is denied to be true. If it does sound plausible to say that what is
false could well have been, one could think of some A’s, B’s, and C’s
and D’s as connected by the range of possibilities that is true of them.
The original idea now is that this denial of what could have been repre-
sents a part of the ‘situation’ to which sentences like “A is left to B”, “A
> B”, “A lives north of the Danube Canal” are thought to correspond.
If Russell ends up with an account of the unity of a thought by enume-
rating the terms used to express it, he does face the difficulty Davidson
highlights (Davidson 2005). But not because he would not appreciate
what he or other philosophers regard as important.
García-Carpintero is therefore right to write: To say “a sentence has
the capacity to say something true or false” is helping oneself “without
further ado” to that “which is precisely what we wanted to understand in
the first place” (García-Carpintero 2010: 285). More has to be said about
a part’s contribution to the truth of an assertive sentence, if one wants to
take up a side in the discussion about the unity of the proposition.
2. The structure of thought
Had Russell stayed with Frege’s notion of a concept expression as in-

complete he would have found an answer to the question how the parts
of a sentence contribute to its truth and would not have defended a term
theory of judgement which is widely believed to fail for a Bradleyan or
other kind of regress.
But the few passages normally cited by philosophers time and again
can hardly support the view that Frege wanted to “solve […] the prob-
lem of the unity of sentences and the thoughts they express” (Dummett
2001: 10). Frege nowhere mentions Bradley, his regress, or a unity prob-
lem as such. Granted that one does not have to be concerned with a
question like “What is it for a thought to be about an object?” in order to
be able to solve problems connected with it, there is still reason to doubt
that the development of a modern conception of a predicate as expres-
sion with (empty) places (Davidson 2005: 132) is a contribution to the
topic “What is the structure of a thought?” (Dummett 1993: 128f).
Dummett is convinced that the quantifier-variable notation is the an-
swer to “What is the structure of a thought?”:
[…] Frege’s invention of the quantifier-variable notation yielded him sever-

al fundamental insights. […] the concept of concept expressions as incom-
plete solved the problem of the unity of sentences and the thoughts they ex-
press. (No glue is needed to make the part of the sentence adhere to one an-
other.) The concept-expression or relation-expression is of its nature inca-

pable of standing alone […]. (Dummett 2001: 10)
Indeed, the quantifier-variable notation seems to be the appropriate can-

didate for “a structure of thought” because it represents the requirement
that what is said about one thing includes or entails a comparison. If this
were a necessary condition for a thought to be about an object, then all
sentences – as expressions of thought – would, in a way, have to exhibit
this comparison. I think Frege might have agreed to this as desideratum
but not as an answer to “What is the structure of a thought?”.
Frege introduced the idea of expressions with empty places in a con-
text (Frege 1879: §9) where even the example used shows that he is in-
terested in preserving truth in an inferential relation. The example in
which truth is preserved is: If oxygen is lighter than carbon dioxide, and
if Carbon dioxide is lighter than C, then oxygen is lighter than that in
relation to which carbon dioxide is lighter. The decomposition of the
judgeable content that is for Frege most adequate involves the famous
function/argument division:
[…] an expression is […] is split up into a constant part … and a symbol,

imagined as replaceable by others, that stand for the object related by the
relations. I call one part the function, the other an argument. (Frege 1879:
§9)
To stay with his example “Oxygen is lighter than carbon dioxide”– de-
pending on the way it is split up – “… is lighter than carbon dioxide” or
“Oxygen is lighter than …” – leaves different parts as exchangeable. As
one part of a sentence is regarded as a function-expression that takes dif-
ferent expressions as argument, depending on what is put in the argu-
ment place will change the whole expression. But if it is the case that
both, “Oxygen is lighter than carbon dioxide” and “Carbon dioxide is
lighter than C” are true, the substitution of “carbon dioxide” through C
in the first sentence will give a different expression, yet something that
in respect to truth – is the same. As this, for Frege, is of the greatest sig-
nificance, he regards concept-expressions as leaving ‘room’ for other
parts, i.e. as expressions which not until they are completed ‘express’
whose part they really are.
It is, however, not the case that every whole expression that another
expression is part of is an expression of something that is either true or
false.
The (functional) expression “… is lighter than carbon dioxide”, when
completed, yields a sentence. The functional expression “… + 3”, when
completed by “5”, does not. It gives an expression that stands for a num-
ber, here 8. So a function as expressed by the “+” sign cannot be regard-
ed as having the same function as the expression “is lighter”, which is
the result of deleting “carbon dioxide” from “… is lighter than carbon
dioxide”. Expressions that take other expressions as (their) parts do not,
if they are not sentences, have parts which depend on each other in a
way that the parts which are parts of expressions that are sentences do
depend. So they could not be regarded as equivalent in structure.
This means that there being empty places ‘inside’ an expression does
not indicate that what is left is a predicate or a concept expression.
As we already saw, if structure means dependency, then it applies to
too many thing. Turning on the idea of structure serves therefore rather
as demarcation: thought cannot be represented as what is an undividable
whole.
That Dummett cites Frege’s invention of the quantifier-variable nota-
tion as paradigm of “the” form of thought has nonetheless some support
in passages from “Funktion und Begriff” where Frege shows us what
made it necessary for him to extend the notion of a function. Only if a
particular instance of an expression of a relation like “+” is enriched by
“=” as in “… + 3 = 8” the result is something that is true or false. So on-
ly under this condition a function is a concept that matches arguments
onto truth values.
The extension of the notion of a function not only tries to do justice to
a fact that was presupposed in §9 of the Begriffsschrift. Is also presents
more clearly how one should think of an aspect Frege addresses already
in the Begriffsschrift in the §§9, 11, 12 which follow the function-
argument distinction. It is for finding the form of ‘generality’ that a sys-
tematic account of the conditions under which expressions are ex-
changeable is in Frege’s concern (Frege 1879: §9). The generality that is
presupposed in the notion of function that takes more than one argument
does, however, not mean that a sentence like “Edinburgh is north of

London”, “3 >1” is an expression of generality.
As in the expression of generality the objects are undetermined “is
north of London” as part of “that there is something that is north of Lon-
don” is not the same as “is north of London” ascribed to an object A.
The “left-over” part therefore does not show by its brackets that gener-
ality is a presupposition in order to understand that an object falls under
a concept. It therefore appears to be futile to try to identify as form “the”
relation of dependency in virtue of which expressions are to be com-
bined/completed.
In a way, this is exactly what Dummett emphasizes when he speaks of
a concept expression as being “of its nature” “incapable of standing
alone”. In Frege’s famous words:
I think of a concept as having arisen by decomposition from a judgeable

content. […] I do not believe that concept formation can precede judgement
because that would presuppose the independent existence of concepts […].
(Frege 1882: 101)
That one must not “presuppose the independent existence of concepts”

because no conditions could prove the applicability of concepts, goes
together, so Frege, with regarding the opposite as a view, which claims
to be possible what is not.1 But since a conception of a concept as unsa-
turated is one that is “of its nature” incapable to get hold of these con-
cepts as concepts, it is not only superfluous to look for conditions that
represent the “relation” a concept has to an object. It is, in fact, a mis-
take.
If one honors Frege’s distinction of first and second order function/
concept one cannot claim that the quantifier-variable notation is the form
in virtue of which thought is about an object.
A judgement of the form “a is F” for Frege is to be seen as fundamen-
tal and not reducible to judgements whose form expresses that a concept
is satisfied in n objects. A singular predicative sentence is not to be ana-
1
The relevant passage in Frege’s “Der Gedanke” might appear to use a Bradleyan
regress as an argument. But it doesn’t. It is close, however, to what is said at the
beginning of Kant’s Transcendental Deduction; see also Glock (2010).
lyzed as saying anything about a concept. On the contrary, “… is wise”

is a ‘part’ of an expression, but not in the sense that the expression ‘∃x’
could be used to express what “Socrates is wise” states. One could put
that point by saying that when one talks about the ‘part’ of an expression
‘∃x’ that takes something indeterminate as part, one talks not about the
concept that as function maps arguments onto truth values.
Claims about how to arrive at constant or invariable expressions by
decomposition do therefore not fit well with the assumption of a struc-
ture: The brackets ‘inside’ an expression do not indicate that what is left
is a concept expression, nor that there is only one possible decomposi-
tion or that anyone must be prior to another
My point is here not that there are many structures for being able to
claim “the” one “common form of thought” (see also Kemp 2011,
Waismann 1965: 304ff). In a Fregean framework with its distinction of
first and second functions or concepts it makes no sense to appeal to a
concept in isolation. To divide the part of the expression of a thought to
get hold of a concept would be to focus on mere signs or symbols by
themselves as being relevant for questions of truth.
Dummett’s claim that what “makes the parts of the sentence adhere to
one another” is “the capacity of predicate logic … to analyze the struc-
ture of … sentences” (Dummett 2001: 17, italics mine) which could be
isolated as structure of thought is not what Frege would or could have
wanted to defend.
3. Russell and the quantificational form
From what has been discussed so far the so-called ‘unity of the proposi-
tion’ seems not to be an original question. It is rather a criterion whose
fulfillment, however, does not settle whether an approach can or cannot
account for how we are able to think about an object.
This, I believe, is the source of the recent reservations about the poten-
tial fruitfulness of the whole Bradley–Russell–Wittgenstein ‘debate’
(García-Carpintero 2010; Kemp 2011). This critical attitude towards the
rapidly growing number of publications about the “thought’s unity”
brings out something very important. How far can philosophy disengage
itself from the common understanding of a judgement? This appears to

be what really lies at the heart of the “unity of the proposition” problem.
Russell seemed to have thought that by staying close to ‘grammar’ one
could accomplish the task once pursued by a “table of judgement”. He
obviously saw himself confronted with the problem of how to ascribe
something as a property to an object whose identity lies in nothing but a
true ascription. Frege’s idea of functions as concepts struck him as being
useful in finding a solution to this challenge. But since Russell regarded
it as unnecessary or erroneous to distinguish two levels of speech in the
way Frege secured their “intranslatability”, he sought to understand a
judgement of the form “A is F” in a way that integrates what is for Frege
to say something about a concept being satisfied in n objects. If first or-
der concepts as incomplete expressions are necessarily missed whenever
one attempts to refer to them – the part crucial in these expressions
would be lost – then this is proof for their having been constructed in a
wrong way. Russell finds support for this criticism:
Frege wishes to have the empty places where the argument is to be inserted
indicated in some way; thus he says that in 2x3 + x the function is 2 ( )3 +
( ). But here his requirement that two empty places be filled by the same
letter cannot be indicated; […]. (Russell 1903: 509)
But Russell did not assume that for a relation to be is to be abstracted

from particulars. Russell's realism requires one to understand universals
– which predicates and relations are according to him – as being actually
independent of the descriptions that identify objects:
[…] the relation ‘north of’ does not seem to exist in the same sense in which
Edinburgh and London exist. If we ask ‘Where and when does this relation
exist?’ the answer must be ‘Nowhere and nowhen’. (Russell 1912: 55f)
If the analysis of “a is F” or “aRb” tries to respect this “nowhere”, then,

however, two criteria of a judgement appear to be true: firstly, the par-
ticular objects; and secondly, the relation that possibly holds between
them. It is exactly the quantificational form regarded as that in virtue of
which judgement is to be understood that makes it impossible to judge
something of objects as being true. This form does not settle the question
whether what is said hypothetically is so/is true of a.
And if there is nothing but this form and “a”, then “a is B” and “a is
non B” together with “a” are supposed to answer the question what cor-
responds to “Edinburgh is north of London”, if true. Not surprisingly,
this view of truth conditions makes it possible to raise the question
again, how one can understand what is stated by “A is in Relation to B”
if what is stated needs one to have understood “A is in Relation to B”.
This being so, the moral I would want to draw is, nonetheless, not to
adopt the following stance:
Where Gaskin appeals to a regress he claims not to be vicious to his ac-

count for propositional unity, King appeals to a very small circle he claims
to be virtuous. (García-Carpintero 2010: 288)
References
Amoretti, María Cristina and Nicla Vassallo (eds.) 2008. Knowledge, Language and
Interpretation, Frankfurt: Ontos.
Bradley, Francis Herbert 1897. Appearance and Reality. A Metaphysical Essay. Ox-
ford: Clarendon Press 1946.
Cottingham, John and Peter M. S. Hacker (eds.) 2010. Mind, Method, and Morality
Essays in Honour of Anthony Kenny. Oxford: Oxford University Press.
Davidson, Donald 2005. Truth and Predication. Cambridge, Mass.: Harvard Uni-
versity Press.
Dummett, Michael 1993. Origins of Analytic Philosophy. Cambridge: Harvard Uni-
versity Press.
Dummett, Michael 2001. Gottlob Frege (1848-1925). In: A. P. Martinich and D.
Sosa, (eds.), 6-20.
Frege, Gottlob [1879] 1960. Begriffsschrift (translated by P. Geach) In: P. Geach
and M. Black (eds.), 1-20.
Frege, Gottlob [1882] 1980. Letter to Marty. In: Philosophical and Mathematical
Correspondence. Chicago: University of Chicago Press, 99-102.
García-Carpintero, Manuel 2010. Gaskin’s Ideal Unity. Dialectica 64 (2), 279-288.
Gaskin, Richard 2008. The Unity of the Proposition. Oxford and New York: Oxford
University Press.
Geach, Peter and Martin Black (eds.) 1960. Translations from the Philosophical
Writings of Gottlob Frege (trans. by P. Geach), Oxford: Basil Blackwell.
Glock, Hans Johann 2010. Concepts: Between the Subjective and the Objective. In:
J. Cottingham and P. M. S. Hacker (eds.), 306-329.
Griffin, Nicholas 2013. Russell’s Philosophical Background. In: N. Griffin (ed.),
The Cambridge Companion to Russell. Cambridge, New York: Cambridge
University Press, 84-107.
Hanks, Peter 2007. How Wittgenstein defeated Russell’s multiple relation theory of
judgement. Synthese 154, 121-146.
Hanks, Peter 2012. Early Wittgenstein On Judgement. In: J. L. Zalabardo (ed.),
Wittgenstein’s Early Philosophy, Oxford: Oxford University Press, 37-63.
Hylton, Peter 2008. Selected Essays on Russell’s Philosophy. Oxford: Clarendon
Press.
Kemp, Gary 2011. The Unity of the Proposition in the Later Wittgenstein. Concep-
tus. Zeitschrift für Philosophie 97, 7-28.
King, Jeffrey C. 2007. The Nature and Structure of Content. Oxford: Oxford Uni-
versity Press.
Martinich, A.P. and David Sosa (eds.) 2001. A Companion to Analytic Philosophy.
Oxford: Blackwell.
Picardi, Eva 2008. Frege and Davidson on Predication. In: M. C. Amoretti and N.
Vassallo (eds.), 49-79.
Russell, Bertrand 1903. The Principles of Mathematics. Cambridge: Cambridge
University Press.
Russell, Bertrand 1910. Some Explanations in Reply to Mr. Bradley. Mind 19 (75),
373-378.
Russell, Bertrand 1910/1966. Monistic Theory of Truth. In: Philosophical Essays.
London: Allan & Unwin.
Russell, Bertrand 1912. The Problems of Philosophy. London: Hutchinson Home
University Library.
Russell, Bertrand 1986. The Philosophy of Logical Atomism and Other Essays
1914-19. Edited by J. Slater. London: Routledge.
Waismann, Friedrich 1965. The Principles of Linguistic Philosophy. London:
MacMillan.
Zalabardo, José L. (ed.) 2012. Wittgenstein’s Early Philosophy. Oxford: Oxford
University Press.
Jaroslav Peregrin
Academy of Sciences of the Czech Republic, Prague
& University of Hradec Králové
[email protected]
Logic and the Pursuit of Meaning*

Abstract: The ‘linguistic turn’ of philosophy of the twentieth century led to the
overestimation of the role of logic in the process understanding of meaning and in
the consequent ‘dissolution’ of traditional philosophical problems. This is not to say
that logic, in this respect, would be useless – on the contrary, it is very important;
but we must understand that the role it can sensibly play is the Wittgensteinian role
of helping us build simplified models of natural language (with all possibilities and
limitations models have), not the Carnapian role of reducing meanings, without a
remainder, to logico-mathematical constructs. In this paper, I try to throw some new
light on this situation in terms of distinguishing two perspectives that may be as-
sumed to look at an expression: the expression-as-object perspective (looking at the
relation between an expression and its meaning as a contingent, a posteriori matter)
and the expression-as-medium perspective (looking at this relation as something
necessary or a priori).
Keywords: logic, meaning, linguistic turn, formal semantics
We want to establish an order in our knowledge of the use of

language: an order with a particular end in view; one out of
many possible orders; not the order.
Wittgenstein, Philosophical Investigations, §132
*
An original version of this paper was written some fifteen years ago, as a by-
product of my work on the book Doing words with worlds (Peregrin 1995). A
Czech version appeared in the collection of my essays Logika a jazyk (Peregrin
2003). The work on the present version was supported by the research grant No.
13-21076S of the Czech Science Foundation.
194 Jaroslav Peregrin
0. Introduction
The linguistic turn that occurred in the minds of various philosophers

during the first half of the twentieth century has led to the conclusion
that to resolve the traditional philosophical problems means to dissolve
them by means of the logical analysis of the language in which they are
formulated. The spread of this insight, which presented something truly
novel, is probably the most significant event in the history of twentieth-
century philosophy; at the same time, however, it is the source of pro-
fuse misunderstandings and misinterpretations.
Some of the proponents and followers of the linguistic turn have come
to the conclusion that this turn amounts to the ultimate word on philoso-
phy, meaning the end of philosophy in the traditional sense and the rise
of a new kind of scientifico-philosophical thinking shaped by Cartesian
rigor. Formal logic, viewed as the means of uncovering of the ‘true
structure’ of language and consequently of the ‘true structure’ of the
world, has moved to the center stage of philosophy.
In this paper, I would like to show that however great the significance
of the linguistic turn and of the employment of logical means in philo-
sophical analysis, such expectations were – and are – unwarranted.
I would like to show that the import of logic for philosophy is neither
that it lets us get hold of meanings in an explicit way, nor that it shows
us the ‘real structure’ of the world, a structure otherwise obscured by
language; rather, its value is that it provides what can be called per-
spicuous representations. It offers us vantage points from which we can
comprehend the vast variety of language, and consequently of the world
that we cope with by means of language, allowing us to better under-
stand their nature.
I would like also to indicate in which way awareness of the limitations
of logical analysis is what distinguishes philosophers like Frege, Witt-
genstein or Quine from Tarski, Carnap and many of the subsequent ana-
lytic philosophers and theoreticians of language.
Logic and the Pursuit of Meaning 195
1. The linguistic turn of philosophy
The linguistic turn, as Rorty (1967: 3) puts it, is based on “the view that
philosophical problems are problems which may be solved (or dis-
solved) either by reforming language, or by understanding more about
the language we presently use”. The idea behind this is that insofar as all
the mysterious entities with which philosophy comes to deal, entities
like matter, justice, knowledge, consciousness, evil, etc., are meanings of
some words (in particular of the words matter, justice, knowledge, con-
sciousness, evil, etc.), the only thing a philosopher must do is analyze
and understand meanings of words. And this brings his business from
the strange and shadowy realms where such entities are supposed to be
found back to the all too well known public arena in which we play our
language games. Replacing a question what is (an) X? by what is the
meaning of ‘X’?, the move that Quine (1960: 271) later called the ‘se-
mantic ascent’, we seem to be able to reduce many quite obscure or en-
igmatic questions to ones that can be answered by straightforward and
down-to-earth observations on how we use our language.
Russell, Carnap and other exponents of the linguistic turn pointed out
that the problem with language is that expressions which appear to stand
for an object may well not do so. In his path-breaking paper ‘On Denot-
ing’, Russell (1905) showed, by the freshly discovered art of logical
analysis, that, despite appearances, expressions such as someone, every-
one or the king of France are not names; and Carnap (1931, 1934) strug-
gled to demonstrate how that kind of logical analysis can be used to elu-
cidate the nature of ‘names’ such as God, being or nothingness and in
this way reveal the emptiness of many classical philosophical problems.
Such considerations resulted in the conclusion that the surface or appar-
ent structure of natural language is not the structure which is relevant for
the semantics of language, that the relevant structure is hidden, and that
the task of the philosopher is to bring it to light.
This then led to the view that natural language is only an imperfect
embodiment of an ideal structure which can be disclosed by an analysis;
and logic was promoted as the general tool for this kind of analysis. In
this way logic launched its triumphant campaign in the realm of philo-
sophy, conquering or exterminating its parts one after another.
2. The formalistic turn of logic
At approximately the same time at which the linguistic turn was finding
expression in the writings of Russell, Wittgenstein, Carnap and others,
another important event, closely connected with it, took place as well. This
was the birth of formal logic in the strict sense.
To avoid misunderstanding, let me stress the difference between what
I call formal logic and logic that I dub merely symbolic.1 Both formal
and symbolic logic are based on the substitution of symbols for natural
language statements and expressions; however, whereas within the
merely symbolic approach symbols are employed solely for the purpose
of regimentation (in Quine’s sense), i.e., of suppressing those aspects of
natural language expressions which are considered irrelevant for the
analysis of consequence, within the truly formal approach the resulting
systems of symbols – logical calculi – are taken to be abstract algebraic
structures. Aristotle used letters to represent unspecific terms; hence he
could be considered an early symbolic logician. Frege and Russell were
symbolic logicians par excellence; but neither of them was a formal lo-
gician.2
It was Hilbert who, for the first time, viewed logic as a strictly formal
matter; however, a tendency towards such a conception of logic is clear-
ly recognizable already in the writings of the logical school of Boole and
Schröder. For Frege, a symbolic formula represents a definite statement,
a definite ‘thought’. There are situations in which it may be reasonable
to disregard the particular statement a formula represents; but there is no
way to detach the latter from the former completely. For Hilbert, on the
other hand, a formula is first and foremost an abstract object, an object
which we are free to interpret in various alternative ways.
The nature of the difference between Fregean symbolic and Hilbertian
formal logic becomes clear when we consider the controversy between
1
These terms have been applied to logic in very various ways. See Dutilh Novaes
(2011) for an overview.
2
For both Frege and Russell, symbols were, as Tichý (1988: ii) puts it, “not the
subject matter of their theorizing but a mere shorthand facilitating discussion of
extra-linguistic entities”.
the two logicians about the nature of axioms and implicit definitions3.
For Frege, as for the Ancients, an axiom is a statement the refutation of
which is beyond the scope of human imagination; therefore there can
hardly be a discussion on whether something is or is not an axiom. For
Hilbert, on the other hand, an axiom is a statement which differs from
other statements only in that we choose it as foundation; we are free to
choose axioms according to our liking.
It was the formal approach to logical calculi which allowed logicians
to develop metalogic and model theory, to prove theorems about logical
calculi. The work of Löwenheim, Skolem, Gödel, Tarski and others who
entered the vast new world of ‘liberated signs’ elevated logic to a new
paradigm. Tarski’s model theory then presented the next step in the
takeover of philosophy by logic: after the logical analysis of language as
pursued by Frege, Russell and Carnap eliminated the old metaphysics,
model theory slowly moved in to fill the gap. A volume of selected pa-
pers on model-theoretic semantics of one of the most influential twen-
tieth-century theoreticians of meaning, Montague (1974), simply bears
the title Formal Philosophy.
3. Correspondence
The linguistic turn requires us to look at an expression as a mere type of

sound or string of letters and check whether there is a meaning attached to
it. Similarly for the formalistic turn of logic: to be able to consider a
system of logical formulas as something self-contained, something that
can be interpreted in various alternative ways, we must regard formulas as
not having a priori meanings, but rather as mere strings of letters.
At first sight, there might seem to be nothing easier than consider an
expression of a language as an object deprived of any meaning. After all,
meaningful expression is a amalgamation of such meaningless expres-
sion with its meaning, a combination that does not hold together by it-
self, but is held together solely by the powers of the human mind; so it
seems that to consider an expression as meaningless is easy, for it mere-
ly relieves us of the mental effort of holding it together with its meaning.
3
See also Peregrin (2000).
However, I am convinced that looking at meaningful expressions in

this way is basically misguided, they are not complexes formed by at-
taching meanings to sounds; they are more adequately seen as meanings
embodied in sounds. An expression serves as a mere way of presenting
its meaning; normally we do not perceive an expression as such, but ra-
ther look ‘through it’ at its meaning. An expressions and its meaning are
inseparably connected; they are, as de Saussure (1931) put it, two sides
of a single sheet of paper. Thus, the connection between an expression
and its meaning can be said to be a priori, not to be found in the world,
but rather being constitutive of our grasp of the world. And the linguistic
turn requires intentional suppression of this normal perception.
Once we realize that the Saussurean two sides of a sheet of paper pic-
ture the composition of an expression and its meaning much more ade-
quately than the amalgamation of an expression with a meaning, we can
see that it might take a nontrivial effort to disengage an expression from
its meaning, while what is effortless is to see it as an embodiment of the
meaning – to see it as meaningless need not be any easier than to see,
say, a human ear as a mere chunk of meat, rather than an ear.4
In sum, both the linguistic turn of philosophy and the formalistic turn
of logic require us to view the relation between an expression and its
meaning as a contingent, a posteriori, fact. However, we cannot adopt
this view universally without undermining our ability to use language
and engage in argumentation. If we held the relation between an expres-
sion and its meaning to be always contingent, then no statement could be
true necessarily, in every possible world (and hence we would not be
able to articulate any universally valid argument) – for even for a state-
ment that means (in the actual world) something which is true in all pos-
sible worlds there would be possible worlds in which the statement
would be false, namely each possible world in which the statement
4
As Dutilh Novaes (2012), Chapter 6, duly points out, ‘de-semantification’ is a
cognitive mechanism which is very non-trivial and which is crucial from the
viewpoint of the deployment of the methods of modern formal logic. In particular
“by countering our automatic (or default) tendency towards semantic activation,
de-semantification allows for the deployment of reasoning strategies other than
our default strategies, thus enhancing the ‘mind-altering’ effect of reasoning with
formalisms”.
would mean something false (in that world).5 To say that the relation be-
tween a statement and its meaning is contingent is to say that there are
possible worlds in which the statement means something false; and for
the sake of rational argumentation we need statements which have mean-
ings independent of possible worlds (which are, so to say, ‘about’ possi-
ble worlds).6
Hence we need both the perspective which allows us to look at the re-
lation between an expression and its meaning as something contingent (a
posteriori, ‘within the world’) and the one which allows us to look at
this relation as something necessary (a priori, ‘about the world’). Let us
call the former the expression-as-object perspective and the latter the
expression-as-medium perspective. The expression-as-object perspective
is the perspective of a foreigner trying to figure out how to translate our
expressions into those of his own language (or that of a linguist inten-
tionally reflecting upon our usage of language); the expression-as-
medium perspective is that of our fellow speakers chatting away without
any awareness of their use of language. As philosophers we need both
perspectives, and, moreover, we need to go back and forth between
them. The need to switch between the two perspectives is quite obvious
when we try to state explicitly what a given expression means. Let us
consider the statement (1), or its Tarskian variant (2).
(1) “Snow is white” means that snow is white

(2) “Snow is white” is true if, and only if, snow is white
Such articulations of the correspondence between language and the

world, which are at the heart of the foundation of the Tarskian corres-
5
Contingency of meaning thus makes for a double-dependence of the link of a
statement to its truth value on possible worlds: not only that a proposition can be
true in some possible worlds and false in others, but also that a statement can
mean different propositions in different possible worlds. (This has come to be
explicitly reflected by the so-called ‘two-dimensional semantics’; see Stalnaker
2001.)
6
This is to say that possible worlds cannot be used to explain language, because
they themselves make sense only on the background of a language. See Peregrin
(1995).
pondence theory of truth, have initiated a broad and still continuing dis-
cussion.7 The central issue in this discussion is the status of sentences
articulating correspondence: is (2) a necessary or a contingent truth? If it
is necessary, then the correspondence theory manages to state the truth
conditions of a contingent statement without telling us anything factual,
which seems absurd. If, on the other hand, it is contingent, then how is it
possible that we directly see its truth?
If we consider the sentence snow is white as a priori equipped with its
meaning (i.e., if we use the expression-as-medium perspective), then to
say either (1) or (2) is to utter a truism. If, on the other hand, we were to
look at snow is white as a string of letters whose meaning (if any) is a
matter of empirical investigation (hence adopting the expression-as-
object perspective), we would make the intelligibility of (1) and (2) itself
an empirical issue. In other words, understanding this sentence presup-
poses knowledge of its truth. What we need to do is use the expression-
as-object perspective for the first occurrence of snow is white in (1) or
(2) and the expression-as-medium view for the second; only then are we
able to see the statement as a nontrivial piece of information, on a par
with “Schnee ist weiß” is true if and only if snow is white. (This switch
of perspective is, of course, what the apostrophes are employed to ef-
fect.)
This example illustrates that it is only through the ability to treat
meanings as detachable and to switch between the expression-as-
medium and the expression-as-object view (comparable with switching
between perceiving a window and looking through it) that we can make
sense of correspondence. More generally, it is this ability that underlies
both the linguistic turn of philosophy and the formalistic turn of logic.
The ability to view logical formulas both as self-contained objects and as
mere ways of pointing to their meanings is what makes it meaningful to
consider alternative interpretations of formulas. It is this ability which
made possible the development of genuine formal logic and model theo-
ry. And it is the same ability, applied to expressions of natural language,
that makes it possible to understand truth as correspondence and to com-
plete the ‘semantic ascent’. However, the art of playing hide-and-seek
7
See, e.g., Leitgeb (2007) and the literature quoted there.
with meanings can be deceptive: we may delude ourselves into thinking

that we have gained everything when in fact we have lost everything be-
cause we have lost firm ground beneath our feet.
4. The two faces of language
It may be helpful to use the spatial metaphor and to speak about ‘inside
language’ and ‘outside language’. To be inside means to use language as
the medium of grasping the world; to be outside means to perceive lan-
guage as a thing among other things of our world. To be inside is to take
an expression as inseparably and unquestionably connected with its
meaning, while to be outside is to perceive the connection between an
expression and its meaning as an empirical fact. If we are inside a house
and perceive the sky through a hole in the house’s roof, then it makes no
sense to ask whether we really do perceive the piece of sky we do;
whereas if we are outside the house, then the question whether an inside
observer can perceive this or that piece of the sky is meaningful and
nontrivial. The perspective from inside is the expression-as-medium per-
spective; whereas that from outside is the expression-as-object perspec-
tive.
What makes language capable of constituting an ‘inside’, which we
can ‘enter’? As I have explained in greater detail elsewhere,8 it is I think
the fact that language is, essentially, a complicated system of rules that
have come to interlock in a robust, but delicate way to delimit the ‘space
of meaningfulness’, in which we can take up meanings. In fact, the kind
of ‘Janus-faced’ characteristic of language is merely the most sophisti-
cated version of a property of everything that is norm-driven (and thus
rational). Any norms are bound to be outgrowths of human communities
and viewed as such they appear as contingent products of factual historic
developments; but we, as rational beings, are characterized by the ability
to obey norms; in other words, to assume the viewpoint from which they
appear to us as necessary.
If we are inside English, then we perceive what statements like (1) and
(2) say as a priori; if we are outside, we perceive it as a posteriori.
8
See especially Peregrin (2012, 2010).
Hence we may draw the conclusion that to be inside a language prevents

us from seeing the language in the unprejudiced way, and that therefore
we should try to stay outside every language. However, this seems to be
simply impossible for a human being, and it is surely impossible for a
theorizing human being. There is no necessity in adopting a particular
language, but it is necessary to adopt some language, and adopting a
language means to approve the necessity of its necessary statements. As
Wittgenstein (1956: II, §30) puts it, “the must corresponds to a track
which I lay down in language”. If we do not speak German, then finding
out that the sentence “Schnee ist weiß” means that snow is white is like
finding out that, say, Hamburg is a port; if, on the other hand, it is Ger-
man that is the language we use to cope with the world, then we cannot
find out anything of this kind, because the knowledge of it underlies the
very possibility of ‘finding out’.
Moreover, the replacement of the study of the mind and of the world
by the study of language which underlies the linguistic turn is meaning-
ful only due to the fact that language acts as our universal means of cop-
ing with the world, that it is a medium. In other words, the linguistic turn
makes sense only when related to the language we are inside of.9
To sum up: our language is a Janus-faced being; it may be ‘in the
world’ (when we are outside it), or it may be ‘about the world’, i.e.,
‘transcendent to the world’ (when we are inside it). We can move in and
9
In fact, the problem of the two faces of language is nothing new; it is only the
modern reincarnation of the much more traditional problem of the ambiguity of
subjectivity. The subject, the ego, can be considered either as a thing on a par
with other things of the world (‘psychological’ subject) or as something that is
transcendent to the world, that is, in Wittgenstein's words, not a part of the world,
but rather its boundary (‘transcendental’ subject). If we want, as Husserl did, to
use an analysis of the subject as a step toward the analysis of the world, we must
consider the subject in the latter sense, as a transcendental ego; we do not need
subjectivity as Seelenleben, but rather subjectivity as “Geltungsgrund aller ob-
jectiven Geltungen und Gründe” (Husserl 1977: 27). The linguistic turn then
means only the replacement of the subject by language. The opposition between
the psychological and transcendental subject reappears as the opposition between
the notion of language-as-object (‘grammatical language’) and the notion of lan-
guage-as-medium (‘transcendental language’), only the latter being able to un-
derlie ontological considerations.
out; but we cannot be both in and out in the same time. However, to real-
ize all of this means to question the philosophical significance of the cor-
respondence theory, and of the idea of linguistic turn as resting on this
theory. The point is that to make the theory of correspondence nontrivi-
al, we need to be outside the language in question; but to make the theo-
ry into a path-breaking piece of philosophy we would have to be inside
it. If we are inside, then the theory of correspondence is trivial, whereas
if we are outside, then it is one of the numerous hypotheses of natural
science to be tested by field scientists. What we can get hold of and thus
use to articulate correspondence is the language-as-object; but the notion
of correspondence is philosophically significant only when related to
language-as-medium.
In contrast to Tarski, Wittgenstein was clear about this predicament
from the beginning. Like Tarski, the Wittgenstein of the Tractatus was
convinced that correspondence was the key concept, but unlike Tarski he
immediately realized its essential deceptiveness. He clearly saw that if
we understand language in terms of picturing reality, then we question
all necessary statements (tautologies and contradictions), because these
are not pictures. Thus he was led to the seemingly counterintuitive con-
clusion that the statements of philosophy cannot be in fact meaningful –
the reason is that although truth may be indeed considered reducible to
correspondence, no theory of correspondence that would imply the re-
duction can be consistently articulated.
5. The two kinds of logic
The primary aim of logic is to summarize basic instances of consequence,

basic patterns of our reasoning used in arguments and proofs. Thus, logic
is inseparably linked to natural language – the medium of expression in
which arguments and proofs are originally formulated. The use of symbo-
lic and formal devices within logic arises from recognition of the fact that
such patterns are easier to summarize if we do not take natural language at
face value but reconstruct it instead as a strictly rule-based grammatical
system. This leads us to the concept of a formal calculus, a calculus
consisting of a formal grammar determining the class of well-formed
expressions, plus axioms and rules of inference determining the relation of

consequence and hence providing the needed criterion of validity of
proofs.
Formal logic suspends the relationship between natural language and
its formal reconstructions in order to permit the undisturbed analysis of
properties of formal calculi. However, once formal calculi began to be
studied independently of their relationship to natural language, they
slowly came to be seen as languages of their own – not as reconstruc-
tions of natural language, but rather as alternatives. Taken in this way
they turned out to be substantially incomplete: whereas it is essential for
natural language expressions that they be linked to their extralinguistic
denotations, expressions of the formal calculi lack such links. This was
the point at which Tarski entered the scene: his model theory appeared to
provide precisely what was needed, namely extralinguistic entities to
which expressions of formal calculi could be linked. Thus the paralle-
lism between natural language and languages of formal logic seemed to
be complete; and scholars like Montague began to deny any real diffe-
rence between the two kinds of languages.
Notice that the original aim of logic is compatible with the language-
as-medium perspective. We need not speak about language, we only
need to replace natural language statements and arguments by their for-
mal regimentations which allow us to ignore all irrelevant idiosyncrasies
and so to see the relevant patterns. Thus, it allows us to capture the unity
of sense within the multiplicity of surface forms and to account for the
infinite class of valid instances of consequence by finite means.10
However, if interpreted formal calculi are seen as alternatives to natu-
ral language rather than as its regimentation, logical analysis might be
seen not as a schematization of natural language sentences, but rather as
a way of making their meanings explicit by furnishing them with model-
theoretical interpretations.11 The problem of explicating meaning has
come to be understood as the problem of finding a model theory ade-
10
The intention to use symbolic means precisely to this effect has been clearly for-
mulated in the introduction of Frege’s Begriffsschrift. See Frege (1879: v).
11
This institutes an important ambiguity of the term ‘interpretation’; see Peregrin
(1994).
quate for natural language. Many theoreticians have embraced so-called

‘representational semantics’, claiming that we must first develop ade-
quate set-theoretic representations of what the world is like and what it
could be like, and only then study the relations of sentences to these rep-
resentations.12 However, to develop an explicit semantics means to step
outside natural language and hence to demote it to a mere object among
other objects of our world.
From the vantage point of the basic aim of logic this whole approach
is disputable. We can, of course, consider a formal calculus as a self-
contained whole, study various relations between its formulas, and talk
about some of these relations as relations of ‘consequence’; but doing so
means doing algebra, not logic in the genuine sense of the word. Alge-
braic theories resulting from the autonomous study of logical calculi are
respectable as such, and provide useful tools to the logician; however,
they are not as yet logic; similarly as the theory of differential equations,
surely indispensable for a physicist, is not as yet physics.
Axiomatic systems were introduced to characterize and explicate the
pre-theoretical notion of consequence; their basic aim was to characte-
rize the infinite number of instances of consequence by finite means (by
reconstructing them as potentially inferable by means of a finite number
of inference rules from a finite numbers of axioms), i.e., to deliver a cri-
terion of consequencehood. Model theory is merely another such method
of characterization (and it is in fact questionable as a method, in that it
does not restrict itself to finite means and hence need not provide a real
criterion). Thus the formal completeness of a logical calculus does not
prove its axiomatization to be ‘right’ (i.e., to adequately capture the con-
sequence relation as it is ‘directly’ presented model-theoretically), ra-
ther, it shows that two alternative formal characterizations of conse-
quence, the axiomatic and the model-theoretic one, coincide (thus cor-
roborating the – essentially formally unprovable – claim that they both
adequately capture the pre-theoretical notion of consequence).
12
Thus Etchemendy (1990). In fact, this means a return to metaphysics, although in
a set-theoretical disguise. Brilliant samples of systems of such a set-theoretical
metaphysics can be found in Cresswell (1973) or in Barwise and Perry (1983).
We have distinguished between two notions of language, the notion of

language as an object among other objects of our world, and the notion
of language as a medium of presentation of the whole world. Given the
basic dependence of logic on language, we can draw a similar distinction
for logic: a logical calculus can either be taken as a mere object within
our world, or it can be understood as a regimentation of language in its
transcendental capacity. This is tantamount to the distinction between
logic as calculus and logic as language introduced by Heijenoort (1967).
There is little doubt that our medium of reasoning is not a language
‘within the world’, but rather a language ‘about the world’; i.e., that it is
the medium view of language that must form the ultimate basis of logic.
Calculi of formal logic, if they are not to be understood simply as alge-
braic structures on a par with groups, rings or vector spaces, must be
seen as ‘regimentations’ of our language in its transcendental capacity.
Thus, if a logician proposes that “we simply put the logic which we are
studying into one compartment, and the logic we are using to study it in
another”,13 then he is stepping on thin ice, because unless the logic we
study is the same as the one we use, it is in fact no logic at all in the au-
thentic sense of the word.
Neurath’s classic metaphor seems to be particularly apt here: we can-
not step out of the boat of our language, we have to rebuild it while stay-
ing aboard. We can make logical calculi to capture and to explicate im-
portant points of the way we use language, but we cannot throw away
our language and put a calculus in its place.
6. The elusiveness of semantics
So is there any way at all to make sense of the linguistic turn and use logic
for philosophical purposes? Do we not, as soon as we begin to speak about
language, eo ipso adopt the language-as-object perspective and hence do
‘mere’ linguistics? And are we not doing ‘mere’ mathematics as soon as
we set out to do model theory?
An answer to this question is indicated in Wittgenstein’s Tractatus:
one can create a picture, articulating correspondence in such a way as to
13
Kleene (1967: 2-3).
feature language-as-object in the role of language-as-medium, and hope

that the reader will get it and yet not take the picture literally. This is
why Witthenstein says Tractatus offers no learnable truths, but rather a
kind of ladder to be thrown away once the reader has used it to climb
higher.
Hence the difference between linguistics and model theory, on the one
hand, and a philosophical account of correspondence, on the other, is not
that the former speak about language-as-object and the latter about lan-
guage-as-medium – whenever we speak about language, we eo ipso
speak about language-as-object. Language-as-medium can be used, but
not fully spoken about. The difference is that in the framework of lin-
guistic or logical discourse we take speech about language literally,
whereas in the framework of philosophical discourse we take it as a met-
aphor, as a picture. In doing Wittgensteinian philosophy we may make
use of language-as-object to the extent that it can serve as a vehicle of
metaphor; but we must avoid taking the metaphor literally, mistaking
language-as-object for language-as-medium. We should devise a theory
which permanently reminds us of its metaphorical character.
Let us return to (2). The sentence Snow is white is true if and only if
snow is white. We may be tempted to say that it is true due to the fact
that ‘out there in the world’ or possibly in a model structure which is
considered to offer a faithful representation of the world the entity snow
instantiates the property of being white, or that snow is an element of the
set of white things, or that there exists a fact of the coincidence of snow
and whiteness. However, all of this is rather tricky: we can either con-
sider snow is white as a mere string of letters (from the object perspec-
tive, i.e., from outside English), and then it is in itself neither true nor
false; or we can take for granted that it has its usual meaning (using the
medium perspective, i.e., staying inside English), and then (2) turns out
to be self-evident.14 To say that the entity snow has the property of being
14
If I say that the entity denoted by ‘snow’ instantiates the property denoted by ‘is
white’, then I speak about English and I hence treat of English from outside.
However, that such a statements really requires the perspective from outside
means that it says something more than every statement that could be made from
inside, especially that it says something over and above the statement that snow
is white. This is just the case when I insist that the fact of snow’s instantiation of
white is not an explanation for the truth of the sentence snow is white; it
is only its cumbersome paraphrase.15 It is, in fact, as Rorty (1989: 7) puts
it, like explaining why opium makes you sleepy by talking about its
dormitive power.
If we realize that our language is the ‘universal’ (the illuminating
German word unhintergehbar unfortunately has no exact English equi-
valent) medium, then we must conclude that its semantics is in a certain
sense fixed. Moreover, we must conclude that this semantics is essen-
tially elusive – to be able to grasp it we would have to step outside lan-
guage, and this is essentially impossible. “There is no outside”; as Witt-
genstein (1953: §103) puts it, “outside you cannot breathe”.
By providing a model-theoretical interpretation for a formal calculus
or for a natural language we offer a new perspective which may help us
perceive patterns and regularities which would remain hidden to our
eyes otherwise; however, it is inadequate to see this act as the act of go-
ing from the words to what the words are about.
7. Formal logic as ‘perspicuous representation’
The exclusive acceptance of the logic-as-calculus notion prevalent now,

and the mistaking of this notion for the notion of logic-as-language, has
led many philosophers to misguided conclusions. However, there is also a
more or less continuous tradition exhibiting awareness of the limitations of
this notion in philosophical contexts. As was shown especially by
Hintikka, the notion of logic-as-language has been central not only for
Frege, but also for some of the most outstanding analytic philosophers of
this century, especially Wittgenstein and Quine.16
whiteness is a fact independent of, and casually determining the fact of the truth
of snow is white.
15
Some paraphrases of such kind, if carried out systematically, may have a purpo-
se, namely helping us see a relevant structure of language; however, this has little
to do with the language-world relationship and with the question of what makes
sentences true.
16
See Hintikka (1984: 27-49); Hintikka & Hintikka (1986); Hintikka (1990).
The employment of formal logic for philosophical purposes is justi-

fied only to the extent that it helps capture language in its transcendental
capacity. In other words, formal logic, and especially model theory, is
not as yet philosophy; it is a device which can be utilized (correctly or
incorrectly) by philosophers. Moreover, there is no rule for correct us-
age. This was clearly recognized by Frege, Wittgenstein and Quine,17 but
largely ignored by Tarski, Carnap, Montague and many other philo-
sophers and semanticists.
The purpose of formalization is to help us see certain aspects of lan-
guage and its functioning clearer, to achieve what we might call, after
Wittgenstein (1953: §122), an übersichtliche Darstellung (‘perspicuous
representation’) It is justified to the extent, and only to the extent that it
fulfils this function; and it must be constantly evaluated from this point
of view. Stekeler-Weithofer has described the situation as follows:
With the development of the functio-logical semantics one constructs a

(mathematical) ‘object of comparison’, a logico-mathematical ‘picture’ or
‘model’, and compares certain aspects or regularities of our common lan-
guage usage, especially of our usual talk about the meanings of linguistic
expressions and of normal judgements of correctness (adequacy) and hence
truth of statements, with aspects and regularities in the picture. Such a com-
parison can yield certain keys to understanding how language ‘functions’
and it can help us formulate explicit rules of meaning sensibly, … One
should never forget, however, that what is at stake are constructed pictures,
perspectives which can be varied, and not descriptions adequate in general,
nor generally approvable criteria governing correct speech and argumenta-
tion. (Stekeler-Weithofer 1986: 141-142)
Russell, Carnap and other scholars were convinced that the structure of
language, although something quite definite, is hidden inside language
or behind it, and that we need logical analysis to bring this structure to
light. In this view, doing logical analysis can be compared to opening the
lid of a complicated machine, thereby revealing the machine’s inner
17
It is instructive to see how Frege understands the role of formal logic in his Be-
griffsschrift. For him, his concept script is like microscope: it is a tool excellent
for some purposes (namely for the purposes of science demanding extraordinary
acuity and differentiation), but useless for others.
workings. This metaphor is misguided, however: there is nothing about

language that is hidden and can only be made visible by opening a lid.
Language is accessible to us in all its aspects; our problem is to compre-
hend it – to command, as Wittgenstein (1953: §122ff.) puts it, a clear
view of it. If language is to be seen as a machine, then it is a machine
with all its wheels and gears in full view. Thus, the use of logical formu-
las to analyze language is more felicitously compared to drawing up a
scheme to facilitate comprehension of the operating principle of an en-
gine that is itself fully accessible to inspection but too complicated to be
understood. No logical calculus is the scheme that would guarantee un-
derstanding language; it is at most one of many possible schemes that
may contribute to it.18
8. Conclusion
The linguistic turn is based on the fact that whatever we can speak about
is the meaning of an expression of our language and that ontology is thus
in a sense reducible to semantics. Model theory, as developed within the
framework of modern formal logic offered means for the explicit captur-
ing of semantics; hence it is tempting to promote model-theoretical se-
mantics as ontology.
However, this might be really misguiding. If we look at our language
‘from inside’ and if we understand logic ‘as language’, then model theo-
ry can be at most one of the formal ways of summarizing ways of using
language; and as such it cannot be an explanation over and above being
a summarization and making language more comprehensible. On the
other hand, if we look at language from outside and if we pursue logic as
calculus, then there is no immediate philosophical relevance of model
theory; model theory is simply a part of mathematics and model-theo-
retical semantics is a part of empirical linguistics. Such enterprises can
be considered philosophically relevant only as metaphors; metaphors
which may (and do) help us see how is our language related to the
world, which are nevertheless no direct theories thereof.
18
For a further elaboration on these themes see Peregrin (1995).
References
Barrett, Robert B. and Roger F. Gibson (eds.) 1990. Perspectives on Quine. Oxford:
Blackwell.
Barwise, Jon and John Perry 1983. Situations and Attitudes. Cambridge, Mass.:
MIT Press.
Carnap, Rudolf 1931. Überwindung der Metaphysik durch logische Analyse der
Sprache. Erkenntnis 2, 219-241.
Carnap, Rudolf 1934. Logische Syntax der Sprache. Vienna: Springer.
Cresswell, Max John 1973. Logic and Languages. London: Meuthen.
Dutilh Novaes, Catarina 2011. The Different Ways in which Logic is (said to be)
Formal. History and Philosophy of Logic 32, 303-332.
Dutilh Novaes, Catarina 2012. Formal Languages from a Cognitive Perspective.
Cambridge: Cambridge University Press.
Etchemendy, John 1990. The Concept of Logical Consequence. Cambridge, Mass.:
Frege, Gottlob 1879. Begriffsschrift. Halle: Nebert.
Heijenoort, Jean van 1967. Logic as Calculus and Logic and Language. Synthèse
17, 324-330.
Hintikka, Jaakko 1984. A Hundred Years Later: the Rise and Fall of Frege’s Influ-
ence in Language Theory. Synthèse 59, 27-49.
Hintikka, Jaakko 1990. Quine as a Member of the Tradition of the Universality of
Language. In: R.B. Barrett and R.F. Gibson (eds.), 159-175.
Hintikka, Merril Bristow and Jaakko Hintikka 1986. Investigating Wittgenstein.
Oxford: Blackwell.
Husserl, Edmund 1977. Cartesianische Meditationen. Hamburg: Meiner.
Kleene, Stephen Cole 1967. Mathematical Logic. New York: John Wiley.
Leitgeb, Hannes 2007. What Theories of Truth Should Be Like (but Cannot Be).
Philosophy Compass 2, 276-290.
Montague, Richard 1974. Formal Philosophy: Selected Papers of R.Montague.
New Haven: Yale University Press.
Peregrin, Jaroslav 1994. Interpreting Formal Logic. Erkenntnis 40, 5–20.
Peregrin, Jaroslav 1995. Doing Worlds with Words. Dordrecht: Kluwer.
Peregrin, Jaroslav 2000. The ‘Natural’ and the ‘Formal’. Journal of Philosophical
Logic 29, 75-101.
Peregrin, Jaroslav 2003. Filosofie a jazyk. Triton: Praha.
Peregrin, Jaroslav 2010. The Enigma of Rules. International Journal of Philosophi-
cal Studies 18, 377-394.
Peregrin, Jaroslav 2012. Semantics without Meaning? In: R. Schantz (ed.), 479-502.
Quine, Willard Van Orman 1960. Word and Object. Cambridge, Mass.: MIT Press.
Rorty, Richard 1967. The Linguistic Turn. Chicago and London: The University of
Chicago Press.
Rorty, Richard 1989. Contingency, Irony and Solidarity. Cambridge: Cambridge
University Press.
Saussure, Ferdinand de 1931. Cours de linguistique générale. Paris: Payot. English
translation Course in General Linguistics, Philosophical Library, New York,
1959.
Schantz, Richard (ed.) 2012. Prospects of Meaning. Berlin: de Gruyter.
Stalnaker, Robert 2001. On Considering a Possible World as Actual. Proceedings of
the Aristotelian Society Supplem, 141-156.
Stekeler-Weithofer, Pirmin 1986. Grundprobleme der Logik. Berlin: de Gruyter.
Tichý, Pavel 1988. The Foundations of Frege’s Logic. Berlin: de Gruyter.
Wittgenstein, Ludwig 1953. Philosophische Untersuchungen. Oxford: Blackwell.
English translation Philosophical Investigations, Blackwell, Oxford, 1953.
Wittgenstein, Ludwig 1956. Bemerkungen über die Grundlagen der Mathematik.
Oxford: Blackwell. English translation Remarks on the Foundations of Mathe-
matics, Blackwell, Oxford, 1956.
Ulrich Reichard
University of Durham
[email protected]
Objects, Concepts, Unity1

Abstract: The paradox of the concept horse has often been taken to be devastating
for Frege’s ontological distinction between objects and concepts. I argue that if we
consider how the concept-object distinction is supposed to account for the unity of
linguistic meaning, it transpires that the paradox is in fact not paradoxical.
Keywords: Frege, paradox of the concept horse, predication, ontology, structure,

relational properties, unity of the proposition
1. The paradox of the concept horse
The most fundamental distinction in Frege’s ontology is that between

concepts and objects. The difference between them is that concepts are
‘incomplete’ or ‘unsaturated’ entities, whereas objects are ‘complete’ or
‘saturated’; that is, concepts are functions, in the simplest case taking
objects as arguments to return objects as values. Yet, Frege’s way of
drawing this distinction gives rise to the famous ‘paradox of the concept
horse’: “the concept horse is not a concept” (Frege 1892: 42). Whereas
this follows straightforwardly from Frege’s theory, it has often been tak-
en to be a serious problem for Frege. Dummett (1973: 212), for example,
argues that, if the paradox is not resolvable in some way or other, it con-
stitutes “a reductio ad absurdum of Frege’s logical doctrines”. Further-
more, according to Soames (2010: 21), the paradox shows the ‘self-
refuting’ character of Frege’s philosophy. Also Lowe (2006: 84) argues
1
For very helpful comments on an earlier draft of this paper, I am grateful to Da-
vid Kirkby, Andrew Woodard, Alex Malt and Wolfram Hinzen. I also thank the
participants of PhiLang 2013 in Lodz and the participants of the 2013 MLM
workshop in Durham for discussions on the topic of this paper, in particular John
Collins, Jonathan Lowe, and Jan Westerhoff.
214 Urlich Reichard
that the paradox ‘vitiates’ the concept-object distinction, and Davidson

(2005, chapter 6) maintains that it shows that Frege has not solved the
problem of the unity of meaning (which Davidson calls ‘the problem of
predication’). Davidson’s judgement is particularly problematic for Fre-
ge, given that one of Frege’s central motivations for drawing a distinc-
tion between concepts and objects in the first place was to ensure the
unity of the meanings of sentences, which Frege called ‘thoughts’: Frege
(1892: 54) argues that without a distinction analogous to his distinction
between concepts and objects, it remains unexplained how “all parts of a
thought […] hold together”.
In the following, I argue that if we consider how an ontological dis-
tinction like the object-concept distinction could provide an answer to
questions concerning the unity of meaning, it transpires that the paradox
of the concept horse is in fact not paradoxical, as Frege (1892) also ar-
gued himself.
2. A relational ontology
The problem of the unity of meaning consists in the following puzzle:2

sentences have meaning, but words on their own also have meaning. The
meanings of the words of a sentence are systematically related to the
meaning of the sentence. But the sentence is not just the sum of the
meanings of the words: it exhibits a certain unity that the sum of the
meanings of its constituents misses. For example, the meaning of the
sentence John sits is a proposition (or ‘thought’ in Frege’s terminology).
In contrast to the meaning of a list of two words, such as John, Mary, the
unity of the proposition exceeds that of an ordered set. The clearest sign
for this unity is that the proposition is evaluable for truth and falsity
whereas the meaning of the list is not. Thus, we face the question of how
the meanings of the words combine to make up a proposition. It might
be tempting to think that the problem can be resolved by referring to the
2
Versions of this problem have recently been discussed from a systematic point of
view by Davidson (2005), Gaskin (2008), King (2009; 2013), Soames (2010),
and Collins (2011). See also the metaphysical literature on Bradley’s Regress
summarized in Maurin (2012).
Objects, Concepts, Unity 215
ingredients: a sentence requires (at least) a noun and a verb, two nouns
like John and Mary are not sufficient to get a sentence. Though, having
the right ingredients is only a necessary, not a sufficient condition for
unity (cf. Gaskin 2008; Schnieder 2004). Exchange Mary for sits in the
list John, Mary – the result John, sits is still a list of words and its mean-
ing not a proposition evaluable for truth and falsity.
As we shall see below, the unity of linguistic meaning faces a further
complication which originates in the nature of natural language. To sep-
arate this complication from the unity issue, let us first consider a sim-
pler case: the unity of a complex spatio-temporal object like an arch.
What is it, we may ask, that unites all the stones which the arch is made
from, such that together they make up an arch, not a house, a pile of
stones or a set of stones scattered around the world? The answer to this
question is probably that the stones make up an arch, rather than any
other object or no unified object at all, because they are spatially ar-
ranged in a particular way. In other words, the stones make up an arch,
because they are all part of a particular spatial structure. Needless to say,
the structure is not itself a further stone that, if added to the other stones,
unites them in the required way. To assume otherwise would, it seems,
mean making a category mistake and would also lead straightforwardly
into Bradley’s Regress, as Russell (1903, chapter 4) had to note.
Nonetheless, the spatial structure does not only unite the stones in the
appropriate way, it also determines some relational properties of the
stones. Relational properties are properties which an object has only in
virtue of standing in a relation to some other object. For example, being
a grandfather is a relational property, since in order to be a grandfather,
there has to be a child that happens to be the offspring of one of your
children. Similarly, the stone in the centre of the arch (the one with the
rose) is the keystone – not because of its form or mass or any other in-
trinsic property, but in virtue of playing a particular role in the arch. If
this stone was part of the foundation of the arch, or if it was lying around
somewhere else, it would fail to be a keystone. Thus, we could say that it
is the keystone which unites the arch, instead of saying that it is the spa-
tial structure which does so; for, given that the keystone is only a key-
stone when it plays this particular role in a certain structure, these two
claims are equivalent: if there is a keystone, there has to be the structure
216 Urlich Reichard
of an arch; if there is the structure of an arch, there has to be a keystone

(special circumstances aside).
Keystones, then, are not only stones: they are keystones only in virtue
of being part of a structure – we could say, a keystone is a stone with
gaps for other stones. In this respect, keystones are quite similar to Fre-
gean concepts: concepts are also said to contain ‘gaps’ for arguments
and are therefore ‘incomplete’ or ‘unsaturated’. The structural nature of
Fregean concepts is relatively explicit in the following quote:
Instead of putting a judgement together out of an individual thing as subject

and an already previously formed concept as predicate, we do the opposite
and arrive at a concept by having the judgeable content fall into pieces [zer-
fallen]. However, for it to be possible to fall into pieces, the expression of
the judgeable content has to be structured. […] Yet, it does not follow that
the ideas of these properties and relations are formed independently of the
objects [to which they apply] […]. For this reason, their expressions never
stand on their own in the Begriffsschrift; rather, they always occur in com-
binations which express judgeable contents. (Frege 1880-1881: 18-19)
Concepts, then, originate in complete thoughts (‘judgeable contents’) by

abstracting over some parts of the thought. What remains, the concept,
retains the structure of the complete thought; concepts therefore are not
independent of the thoughts and ‘never stand on their own’. As Diamond
(1979) suggests, Frege’s (1884: x) context principle, “never [to] ask for
the meaning of a word in isolation, but only in the context of a sen-
tence”, may well be motivated by the insight that the meaning of a word
depends on the function it plays in a sentence. Also Frege’s (1891a: 96;
1892-1895) explicit distinction between concepts and their extension
(and the more general distinction between functions and their value
ranges (Frege 1891b: 32; 1892-1895: 132) seem to be motivated by the
structural nature of concepts (functions). Something is only a concept if
it plays a certain role in a thought; yet, the extension of a concept, the set
of objects falling under it, is not sensitive to the structure of the thought.
If concepts are thus structural entities, it is clear how they can be used
to account for the unity of thoughts. Without there being a complete
thought, there is no concept, in the same way as there is no keystone un-
less there is an arch. Therefore, concepts guarantee the unity of thoughts

and can thus be said to ‘hold’ the parts of the thought together.
The structural, or relational, nature of the distinction between concepts
and objects has to be separated from the question whether an entity is a
concept or object necessarily. In the case of the arch, it is clear that the
stone which functions as keystone in a particular arch could have been
used in a different function in the same or another arch, or in a house, or
a pile of stones. Keystones, thus, are keystones contingently, not neces-
sarily. Frege suggests at some places that the same is true of concepts:
when he writes that “it is a mere illusion to suppose that a concept can
be made an object without altering it” (Frege 1884: x), he seems to im-
ply that, in principle, concepts can be made objects. Whether something
is a concept or object, then, is a contingent matter. Also in a later article,
Frege (1892: 46) writes that, before it can be made the referent of a sub-
ject, “the concept […] must first be converted into an object”, which
again seems to commit Frege to contingent ontological categories. How-
ever, at this place, Frege adds that, “speaking more precisely, [the con-
cept has to be] represented by an object”, which relativizes Frege’s
commitment to the contingent nature of the concept-object distinction.
Nonetheless, whether or not the distinction is a contingent one, what
matters in respect to the unity question is that it is relational. If entities
possess their ontological categories necessarily, they fail to exist when
they do not play the role that constitutes the particular ontological cate-
gory. The category itself may nonetheless be relational.
Frege’s distinction between objects and concepts is closely related to a
distinction in (logical) syntax. Frege writes: “The concept (as I under-
stand it) is predicative”; and he adds in a footnote: “It is in fact the refe-
rent of a grammatical predicate”, before proceeding: “On the other hand,
a name of an object, a proper name, is quite incapable of being used as a
grammatical predicate” (Frege 1892: 43). The referents of singular terms
are, therefore, always objects and the referents of predicates are always
concepts – in this way, syntax decides over the ontological category of
the referents of the terms (cf. Ricketts 1986: 66). As Collins (2011: 37)
concludes, “for Frege, logic or semantics has priority over any meta-
physical conception of objects and properties”.
218 Urlich Reichard
Furthermore, Frege’s logical syntax is very close to the grammar of

natural language. For example, consider Frege’s definition of functions,
that at first sight seems to be unrelated to grammatical distinctions:
Suppose that a simple or complex symbol occurs in one or more places in

an expression […]. If we imagine this symbol as replaceable by another (the
same one each time) at one or more of its occurrences, then the part of the
expression that shows itself invariant under such replacement is called the
function; and the replaceable part, the argument of the function. (Frege
1879: 13)
This definition allows for grammatical singular terms to act as functions.

We could, for example abstract over sits in the sentence John sits and
thereby receive a predicate John ( ). Yet, following natural language
grammar, Frege never considers this as a possibility. As Gaskin (2008:
180) concludes, “it is a mark of Frege’s mature thought […] that the ob-
ject-concept dichotomy is not treated by him as a purely logical distinc-
tion […], but is assumed to line up neatly with traditional grammatical
categories”. Frege’s most fundamental ontological distinction, therefore,
is ultimately based on the grammatical distinction between referential
and predicative expressions.
The grammatical origin of Frege’s object-concept distinction is in line
with the structural nature of the respective grammatical categories, since
the grammatical distinction between referential and predicative expres-
sions is also determined relationally (cf. Hinzen and Sheehan 2013):
Whether radium, for example, acts as a name or as a predicate depends
on the context of the sentence. In (1), radium is a mass term and is thus
used predicatively, whereas in (2) radium is used to refer to a kind (cf.
Longobardi 1994):
(1) Radium was found in this lake.

(2) Radium was discovered by Madame Curie.
Similarly, as Frege (1892: 50; cf. 1882) observes, Vienna can be used to
refer to an object as in (3), but it can also be used as a predicate as in (4):
(3) Vienna is a beautiful city.

(4) Trieste is no Vienna.
Frege (1892: 50) concludes from this that “language often uses the same
word now as a proper name, now as a concept-word” and warns us not
to be ‘deceived’ by this fact.3
Given the relational nature of the grammatical distinction between
referential and predicative expressions, the unity of linguistic meaning
can already be guaranteed on the level of grammar: if something is only
a grammatical predicate if it stands in the right grammatical configura-
tion, something is only a grammatical predicate if it is part of a sentence
(or rather part of a clause). And sentences, as opposed to mere lists of
words, have the unitary meanings we seek to account for. Hence, so far,
the analogy between the arch and the linguistic case holds: what the spa-
tial structure is for the arch, the grammatical structure is for the sen-
tence; and what the stones are for the arch, the constituents (that is,
words and phrases) are for the sentence.
3. Lexicalization of structural properties
The case of linguistic meaning is in at least one way more complex than
that of the arch. As noted, the structure of the arch cannot be used as a
stone in another building. Also, given the structural understanding of
keystone discussed above, when the keystone of the arch is taken out of
the arch and made part of a house, it stops being a keystone, as it is not
part of the structure anymore that makes it a keystone. However, treating
structure as a lexical constituent is something language can do. We can
talk about the grammatical structure of a sentence or the predicate of a
sentence as in (6) and (7). I will call this process ‘lexicalization of struc-
tural categories’.
(5) John sits.

(6) The grammatical structure of (5) is simple.
3
One could insist that different lexical items underlie the different readings of
these expressions. However, even if so, this does not change the relational nature
of the distinction (cf. the discussion on necessity above).
220 Urlich Reichard
(7) The predicate of (5) is sit.
Is the grammatical structure of (5) grammatical structure or a lexical

constituent? It seems it is both: it picks out the structure of (5), but fea-
tures in (6) as a constituent. Given that constituents are not structure, the
grammatical structure of (5) is both structure and not structure. Yet, this
is not a contradiction, since it is structure and not structure in different
respects: Grammatically, it is not structure in respect to (6), but the lexi-
cal content renders it structure in respect to (5). The same is true of (7):
the lexical content of the predicate of (5) renders it a predicate in respect
to (5), but it is grammatically not the predicate (but the subject) of (7).
Again, this is not a contradiction, as the predicate of (5) is a predicate
and not a predicate in different respects.
However, although lexicalized predicates are predicates in a lexical
sense, there is one important aspect of predicativity which they lack:
they cannot account for the unity of meaning anymore. Whereas we can
say that sits in (5) is the predicate of the sentence and thus guarantees the
unity of the meaning of (5), the predicate of (5) cannot be taken to play
this role. Correspondingly, (8) cannot be a sentence and does not have a
proposition as meaning:
(8) John, the predicate of (5)
Given that one of Frege’s concerns was to account for the unity of mean-
ing, lexical predicates, then, are not really predicates, as they don’t guar-
antee unity. Therefore, he has reason to only count grammatical predi-
cates as predicates. And in that case, the predicate of (5) in (7) will not
count as a predicate. Hence, it is true that that the predicate of (5) is not
a predicate, in the relevant sense of predicate.
Given Frege’s (implicit) assumption that the structure of thoughts mir-
rors that of natural language, the case of the concept horse is analogous
to the grammatical case (Frege 1892: 46, n. 2). Language can not only
pick out (which involves lexicalization of) grammatical structure, but all
kinds of structure and treat it, not as structure, but as a constituent. Thus,
we can refer to concepts and say something about them, disregarding the
thought they are a part of.
(9) John is a horse.

(10) The concept horse is instantiated.
In (9), for example, horse is a concept structurally. It therefore guaran-

tees the unity of the thought. However, only the lexical content renders
‘the concept horse’ a concept in (10), not the structure of (10). Structu-
rally it is a constituent like John in (9). Treated in this way, the concept
inevitably loses one aspect of its structural nature: it cannot guarantee
unity anymore. However, given that for Frege the ability to guarantee
unity is central for concepts, lexical concepts are not proper concepts.
Hence, the concept horse is not a concept.
4. Conclusion
In sum, when considered in light of Frege’s urge to account for the unity
of linguistic meaning, concepts have to be understood in a structural
way; that is, whether something is a concept or not is determined relatio-
nally, by the role it plays within a thought. It is an ‘awkwardness’ (Frege
1892: 46) of natural language that we can refer not only to things but al-
so to structures, thus treating them as things, rather than structures in dif-
ferent thoughts.4 Yet, when treated in this way, they inevitably lose their
ability to account for unity. The same is true of concepts: language pro-
vides the possibility to refer to a concept. Yet, thereby, the concept loses
its structural aspect and thus cannot account for the unity of the thought
anymore. If this ability is taken to be essential for concepts, concepts
thus referred to are not concepts. This, it seems, is fully consistent. It is,
therefore, not ‘a reductio ad absurdum of Frege’s logical doctrines’, a
‘vitiation’ of the concept-object distinction, or evidence of the ‘self-
refuting’ character of Frege’s philosophy. And, most importantly, it is an
account – even if perhaps not an explanation – of the unity of linguistic
meaning.
4
Perhaps it is needless to add that the ‘awkwardness’ of language which gives rise
to the purported paradox of the concept horse in the first place is an aspect of
what increases the expressive power of natural language greatly – and is there-
fore (perhaps pace Frege) not to be legislated away.
222 Urlich Reichard
References
Collins, John 2011. The Unity of Lingustic Meaning. Oxford: Oxford University
Press.
Davidson, Donald 2005. Truth and Predication. Cambridge, Mass.: Belknap Press
of Harvard University.
Diamond, Cora 1979. Frege and Nonsense. In: The Realist Spirit: Wittgenstein, Phi-
losophy, and the Mind, 73-93. Cambridge, Mass.: MIT Press (1991).
Dummett, Michael 1973. Frege: Philosophy of Language. London: Duckworth.
Frege, Gottlob 1879. Begriffsschrift: A Formalized Language of Pure Thought
Modelled upon the Language of Arithmetic. In: Peter T. Geach and Max Black
(eds.), 1-20.
Frege, Gottlob 1880-1881. Booles rechnende Logik und die Begriffsschrift. In:
Hans Hermes et al. (eds.), 9-52.
Frege, Gottlob 1882. Frege an Marty 29.08.1882. In: Gottfried Gabriel et al. (eds.),
163-65.
Frege, Gottlob 1884. The Foundations of Arithmetic: A Logico-Mathematical En-
quiry into the Concept of Number. Translated by J. L. Austin. Oxford: Basil
Blackwell (1950).
Frege, Gottlob 1891a. Frege an Husserl 24.5.1891. In: G. Gabriel et al (eds.), 94-98.
Frege, Gottlob 1891b. Function and Concept. In: Peter T. Geach and Max Black
(eds.), 21-41.
Frege, Gottlob 1892. On Concept and Object. In: Peter T. Geach and Max Black
(eds.), 42-55.
Frege, Gottlob 1892-1895. Ausführungen über Sinn und Bedeutung. In: Hans
Hermes et al. (eds.), 128-36.
Gabriel, Gottfried, Hans Hermes, Friedrich Kambartel, Christian Thiel and Albert
Veraart (eds.) 1976. Gottlob Frege: Wissenschaftlicher Briefwechsel. Hamburg:
Meiner.
Gaskin, Richard 2008. The Unity of the Proposition. Oxford: Oxford University
Press.
Geach Peter T. and Max Black (eds.) 1952. Translations from the Philosophical
Writings of Gottlob Frege, Oxford: Basil Blackwell.
Haaparanta, Leila and Jaakko Hintikka (eds.) 1986. Frege Synthesized: Essays on
the Philosophical and Foundational Work of Gottlob Frege. Dordrecht: Reidel.
Hermes, Hans, Friedrich Kambartel and Friedrich Kaulbach (eds.) 1969. Gottlob
Frege: Nachgelassene Schriften. Hamburg: Meiner.
Hinzen, Wolfram and Michelle Sheehan 2014. The Philosophy of Universal Gram-
mar. Oxford: Oxford University Press (forthcoming).
Hochberg, Herbert and Kevin Mulligan (eds.) 2004. Relations and Predicates.
Frankfurt a. M.: Ontos Verlag.
King, Jeffrey C. 2009. Questions of Unity. Proceedings of the Aristotelian Society

109, 257-77.
King, Jeffrey C. 2013. Propositional Unity: What’s the Problem, Who Has It and
Who Solves It? Philosophical Studies (forthcoming).
Longobardi, Giuseppe 1994. Reference and Proper Names. A Theory of
N-Movement in Syntax and Logical Form. Linguistic Inquiry 25, 609-65.
Lowe, E. Jonathan 2006. The Four-Category Ontology. A Metaphysical Foundation
for Natural Science. New York: Clarendon Press.
Maurin, Anna-Sofia 2012. Bradley’s Regress. Philosophy Compass 7, 794-807.
Ricketts, Thomas G. 1986. Objectivity and Objecthood: Frege’s Metaphysics of
Judgement. In: Leila Haaparanta and Jaakko Hintikka (eds.), 65-95.
Russell, Bertrand 1903. Principles of Mathematics. London: Routledge (2010).
Schnieder, Benjamin 2004. Once More: Bradleyan Regresses. In: Herbert Hochberg
and Kevin Mulligan (eds.), 219-56.
Soames, Scott 2010. What Is Meaning? Princeton: Princeton University Press.
Piotr Stalmaszczyk
University of Łód
[email protected]
The Legacy of Frege and the Linguistic

Theory of Predication
Abstract: This paper discusses possible approaches to linguistic predication inspi-
red by the philosophy of language, distinguishing ‘Aristotelian’ (or concatentaive)
predication and ‘Fregean’ (or functional) predication. It also investigates the rele-
vance of Fregean semantics for contemporary linguistics, in particular generative
grammar. Though Fregean semantics is not concerned with natural language catego-
ries, Frege’s line of reasoning may be applied to analyzing predication understood
as a strictly grammatical relation. The paper also offers a preliminary classification
of predication types into thematic, structural and propositional.
Keywords: Frege, predication, generative grammar, predicate, function, copula
0. Introduction
The present paper discusses the legacy of Gottlob Frege in contemporary

linguistics, focusing on the influence of Fregean semantics upon natural
language analyses, especially within the generative paradigm.1 The ma-
jor claim is that some of the ‘Fregean tools’ (in the sense of Pietroski
2004, see below) may prove useful in analyzing such traditional linguis-
tic notions as predication.
For the purpose of this study predication is considered as a strictly
grammatical operation: both a semantic relation and an appropriate
structural configuration enabling this relation to occur (i.e. a method of
constructing sentences).
Section 1 briefly discusses the legacy of Frege, concentrating on his
critique of natural language, the ideas behind conceptual notation (the
1
This paper is a substantially revised and extended version of Stalmaszczyk
(2006, 2010).
Begriffsschrift), and his ideas on functions and arguments; section 2 in-

troduces the notion of predication; section 3 provides a preliminary clas-
sification of predication types into ‘Aristotelian’, or concatenative predi-
cation, and ‘Fregean’, or functional predication; section 4 focuses on
predication in generative grammar; and finally section 5 proposes a ten-
tative semantics for structural predication.
1. The legacy of Gottlob Frege
Paul Pietroski has recently remarked that Frege “bequeathed to us some

tools – originally designed for the study of logic and arithmetic – that
can be used in constructing theories of meaning for natural languages”
(Pietroski 2004: 29-30). The ‘Fregean tools’ still prove useful in analy-
zing not only the fundamental issues of sense and reference, the seman-
tics of proper names, but also, as will be shown below, such traditional
linguistic notions as predication. Frege’s “philosophy of language […]
remains intensely vital today. Not since medieval times has the connec-
tion between logic and language been so close” (Mendelsohn 2005:
xviii), and his “contributions to philosophy of language are so numerous
and so fundamental that it is difficult to imagine the field without them”
(Heck and May 2006: 3). Scott Soames adds that Fregean systems “were
the starting points for the stunning development of mathematical logic in
the twentieth century, and for the use of logical ideas and techniques in
the study of natural languages” (Soames 2010: 7), and finally, Michael
Potter points that:
One of the most important contributions made by Frege was to place langu-
age at the center of philosophical, and in particular metaphysical, inquiry by
recognizing its importance as a route to the structure of our thinking about
the world. […] This approach seeks to discuss metaphysical questions about
the structure of the world by means of a discussion of the structure of the
language in which we represent the world. (Potter 2013: 853)
Michael Dummett (1993) points to Frege’s early work as seminal for the
development of the linguistic turn in philosophy. The relevant fragment,
which introduces the context principle, is to be found in the Introduction
The Legacy of Frege and the Linguistic Theory of Predication 227
to Die Grundlagen der Arithmetik, where Frege formulates his three

fundamental principles, claiming in the second one “never to ask for the
meaning of a word in isolation but only in the context of a proposition”
(Frege [1884]: xxii); further on in §62 of Grundlagen he stresses that “it
is only in the context of a proposition that words have any meaning”
(Frege [1884]: 73).2
Dummett (1993: 3) observes that the context principle is “formulated
as one governing an enquiry into language rather than into modes of
thought”, and the thesis is “that it is only in the context of a sentence that
a word has meaning”, with sentences and words being linguistic units.
Though Carl (1994) claims that the context principle does “not belong to
a particular semantic theory but to an epistemological theory that draws
a fundamental distinction between thinking and having representation”
(Carl 1994: 43-44), it is clear that any epistemological enquiry is to be
grounded in appropriate linguistic investigation, which is the very es-
sence of the linguistic turn in philosophy, initiated by Frege.
Before discussing Fregean inspirations and implications for the theory
of predication, it is necessary to introduce briefly his achievements im-
portant for philosophy of language and linguistics, especially his work
on conceptual notation, and functions and arguments.
1.1. Conceptual notation
Frege was one of the first modern logicians to notice the need for a for-
malized conceptual notation, and devoted to this problem one of his first
major works, Begriffsschrift, eine der aritmetischen nachgebildete For-
melsprache des reinen Denkens (‘Conceptual Notation. A Formula Lan-
guage of Pure Thought, Modeled on Arithmetic’, 1879, henceforth CN).3
In the Preface he claimed that:
2
In a recent translation, Dale Jacquette renders these fragments as “The meaning
of a word must be inquired after in propositional context, not in isolation”, and
“Only in the context of a proposition do words refer to something” (Frege
[1884]/2007: 17; 66).
3
Unless otherwise noted, all page references following the abbreviated (or full)
titles are to the English translations of Frege’s texts collected in Beaney, ed.
If it is a task of philosophy to break the power of words over the human

mind, by uncovering illusions that through the use of language often almost
unavoidably arise concerning the relations of concepts, by freeing thought
from the taint of ordinary linguistic means of expression, then my Begriffs-
schrift, further developed for these purposes, can become a useful tool for
philosophers. (CN, 50-51)
It is clear from the above quotation that Frege did not consider spoken
language as a sufficiently precise instrument for logic. He pointed to the
need for creating a language made up of signs, the concept-script (or ‘id-
eography’), clear of any double meaning; he also claimed that “[t]he
main task of the logician is to free himself from language and to simplify
it” (Letters to Husserl, 1906, 303). Frege’s words were later echoed by
his student, Rudolf Carnap, who in the forward to The Logical Syntax of
Language (1937) argued that “the aim of logical syntax is to provide a
system of concepts, a language, by the help of which the results of logi-
cal analysis will be exactly formulable” (Carnap 1937: xiii). 4
Michael Losonsky claims that as result of concentrating on language
viewed (or even constructed) as a formal system, Frege “filtered out
what might be thought as the human dimension of language, namely its
psychological properties” (Losonsky 2006: 148). This is a natural (and
obvious) consequence of his anti-psychologism, as introduced already in
the Grundlagen, in the first fundamental principle: “always to separate
sharply the psychological from the logical, the subjective from the ob-
jective” (Frege [1884]: xxii), and repeated in several other places, for
instance in Logic:
In logic we must reject all distinctions that are made from a purely psy-
chological point of view. What is referred to as a deepening of logic by
psychology in nothing but a falsification of it by psychology. (Logic, 243)
(1997) and listed in the References. The German term Begriffsschrift has been
translated also as ‘concept script’ or ‘ideography’.
4
Further on Carnap remarks that “Philosophy is to be replaced by the logic of sci-
ence – that is to say, by the logical analysis of the concepts and sentences of sci-
ence, for the logic of science is nothing other than the logical syntax of the logic
of science” (Carnap 1937: xiii, italics in the original).
Additionally, as observed by Collins (2011: 37), “for Frege, logic or se-

mantics has priority over any metaphysical conception of objects and
properties”. In this study I do not investigate that issue any further, con-
centrating on the ‘Fregean tools’ as applied to analysing grammatical
structures.
In a somewhat different methodological context, Jan Łukasiewicz elu-
cidated the idea of formalism and observed that every “scientific truth, in
order to be perceived and verified, must be put into an external form in-
telligible to everybody”, and this aim “can be reached only by means of
a precise language built up of stable, visually perceptible signs. Such a
language is indispensable for any science” (Łukasiewicz 1957: 15).
Whereas Frege stresses that “Logic should be the judge of languages”
(Letters to Husserl, 1906, 303), Łukasiewicz observes that “Modern
formal logic gives the utmost attention to precision of language. What is
called formalism is the consequence of this tendency” (Łukasiewicz
1957: 15-16).
Elsewhere Frege states that “[i]nstead of following grammar blindly,
the logician ought rather to see his task as that of freeing us from the fet-
ters of language” (Logic, 244).5 As observed by Carl (1994: 54), this
“struggle against language and grammar” directs the logician’s concern
to the issue of the thought expressed by a sentence, and aims at discover-
ing the “logical kernel”.
Consequently, Frege distinguishes between “judgement” (Urteil) and
“judgeable content” (beurteilbarer Inhalt – the possible content of jud-
gement, the thought expressed by a sentence, or the ‘complex of ideas’).
In the conceptual notation, the former is represented by “|—A”, whereas
the latter by “—A”. Frege explains the convention in the following way:
A judgement will always be expressed by means of the symbol |— which

stands to the left of the symbol or complex of symbols which gives the con-
tent of the judgement. (CN §2, 52)
The horizontal stroke, from which the symbol |— is formed, binds the sym-
bols that follow it into a whole, and assertion, which is expressed by means
of the vertical stroke at the left end of the horizontal, relates to this whole.
5
For a critical discussion of Frege’s views on the ‘defects of language’, see Han-
fling (2000: 153-163).
The horizontal stroke may be called the content stroke, the vertical the
judgement stroke. The content stroke serves generally to relate any symbol
to the whole formed by the symbols that follow the stroke. What follows
the content stroke must always have a judgeable content. (CN §2, 53)
In other words, the content stroke serves to distinguish the formation of

a judgeable content from the act of judging. This sign indicates asser-
tion, it acknowledges the truth of a thought:
We […] need another special sign to be able to assert something as true. For
this purpose I place before the name of the truth-value the sign ‘|—’, so
that, for example, in ‘|— 22 = 4’ it is asserted that the square of 2 is 4. I dis-
tinguish judgement from thought in such a way that by judgement I under-
stand the acknowledgement of the truth of a thought. (GGA, 215)
Frege’s notation has met with criticism from other logicians and phi-
losophers; already Wittgenstein commented that: “Frege’s ‘judgement
stroke’ ‘|—’ is logically quite meaningless” (TLP 4.442).6 On the other
hand, it is possible to consider this sign as a metalogical symbol.
Frege is concerned in CN with splitting up the content of a judgment,
which he accomplishes by introducing the distinction between argument
and function. In CN §9, a constant component which represents the tota-
lity of the relations is called a function, and the symbol which is regard-
ed as replaceable by others and which denotes the object which stands in
these relations is the function’s argument. Furthermore, Frege observes
that the distinction between function and argument “has nothing to do
with the conceptual content, but only with our way of grasping it” (CN
§9, 66) and “for us, the different ways in which the same conceptual
content can be taken as a function of this or that argument has no im-
portance so long as function and argument are fully determinate” (CN
§9, 68). In other words: “the way we distinguish between argument and
function is not fixed by the conceptual content of a sentence” (Carl
1994: 62). I return to the consequences of this claim below; here it needs
to be added that in Frege’s system the grammatical categories of subject
6
See Dudman (1970) for a critical overview of the issue. For more recent re-
evaluation of Frege’s judgment stroke, see Smith (2000) and Green (2002). See
also the chapter on assertion in Dummett (1981).
and predicate have no significance: “I believe that the replacement of the

concepts subject and predicate by argument and function will prove it-
self in the long run” (CN, Preface, 51), and “[a] distinction between sub-
ject and predicate finds no place in my representation of judgment” (CN,
§3, 53). This is a major issue which recurs throughout Frege’s writings,
e.g. in one of the letters to Husserl he stresses that “[w]e should either
tidy up logic by throwing out subject and predicate or else restrict these
concepts to the relation of one object’s falling under a concept (sub-
sumption)” (Letters to Husserl, 1906, 303), and in Logic he concludes
that “from all this we can see that the grammatical categories of subject
and predicate can have no significance for logic” (Logic, 242).
Very interestingly, superficially similar calls for abandoning the tradi-
tional notion of predication can be found in Otto Jespersen’s Analytic
Syntax (1937), one of the first modern (linguistic) attempts at forma-
lizing the grammar of natural language. Jespersen’s motivation, how-
ever, is very different from Frege’s; in a sense, it reverses the logician’s
argumentation:
It would probably be best in linguistics to avoid the word predication alto-

gether on account of its traditional connexion with logical theories. In
grammar, we should not of course forget our logic, but steer clear of every-
thing that may hamper our comprehension of language as it is actually used;
this is why I have coined the new term nexus with its exclusive application
to grammar. (Jespersen 1937: 120)
Jespersen’s bipartite and relational approach to predication shows affini-

ties with traditional grammar and classical semantics; however, it is also
to a certain degree an interesting antecedent of the more recent genera-
tive inquiries.
1.2. Functions and arguments
The notion of a function was further investigated and developed in

Funktion und Begriff (‘Function and Concept’, 1891, henceforth FC),
and Grundgesetze der Arithmetik, Volume I, (1893, henceforth GGA),
where Frege introduced a strict differentiation between functions and
functional expressions. These studies show the fundamental shift in Fre-

gean semantics: from the unary semantics of conceptual content (as de-
veloped in Begriffsschrift) to the two-tiered semantics, where the notion
of conceptual content splits into sense and reference.7
This article focuses on one issue only – the treatment of grammatical
predicates – an issue of rather marginal importance for the main line of
Frege’s inquiry, but of considerable interest for linguistic research. In
Fregean semantics, indicative sentences are analyzed similarly to analyt-
ic expressions and mathematical formulae, hence a grammatical predi-
cate (i.e. a subject-predicate expression) is treated as a type of function
expression denoting a function, and as such has certain properties com-
mon to all functions. The function splits into two parts: the sign of the
argument and the expression of the function. The latter contains an emp-
ty place, it is incomplete or unsaturated: “the argument does not belong
with a function, but goes together with the function to make up a com-
plete whole; for a function by itself must be called incomplete, in need
of supplementation, or unsaturated” (FC, 133). Elsewhere, Frege ex-
plains that “The expression of a function is in need of completion, un-
saturated” (GGA, 211). Saturation is achieved through insertion of an
argument into the empty place. The argument “only serves to complete
the function that in itself is unsaturated” (GGA, 212). Importantly, the
same requirements hold for grammatical predicates, consider the follow-
ing examples:
(1) Caesar conquered Gaul.

(2) Berlin is a capital city.
In (1) the predicate conquered Gaul is incomplete and takes a name (ar-
gument) – Caesar – to saturate it, similarly in (2) the predicate is a capi-
tal city requires a completing argument – Berlin. The argument (object-
name) is complete in itself. The expression …conquered Gaul is a func-
tional expression, which designates a concept, i.e. conqueror of Gaul.
7
More precisely, conceptual content splits into three different notions: sense, ref-
erence and extension; this tripartite distinction, however, applies only to predi-
cates, whereas in names and sentences Frege identifies reference with extension,
cf. the discussion in Penco (2003).
Only after saturating the functional expression with a “proper name, or

an expression that replaces a proper name, does a complete sense ap-
pear” (FC, 139).
In Fregean logic, “concepts are functions”,8 and therefore they take
arguments and have truth-values, where the value of a function for an
argument is “the result of completing the function with the argument”
(FC, 134). The following examples illustrate this issue:
(3) a. Caesar conquered Gaul. (conqueror of Gaul = Caesar)

b. Frege conquered Gaul. (conqueror of Gaul = Frege)
c. 2 conquered Gaul. (conqueror of Gaul = 2)
As observed by Thiel (1968: 46), all examples in (3) are senseful since
Frege’s expansion of the concept of function allows any object to serve
as argument, but only for (3a) is the truth-function true. A grammatical
predicate may be polyadic and take more than one argument, as in (4):9
(4) Berlin is the capital of the German Empire.
Here, the expression is the capital of is a two-place predicate, or relation

(i.e. a function of two arguments), requiring two arguments (names) for
saturation. According to Frege, functions with two arguments “are dou-
bly in need of completion in that a function with one argument is effect-
ed. Only by a further completion do we reach an object, and this is then
called the value of the function for the two arguments” (GGA, 214).
Frege was primarily concerned with deriving arithmetic from logic;
however, the idea of function saturation applies also to grammatical
8
Cf. the following definition: “a concept is a function whose value is always a
truth-value” (FC, 139). A concept, just like a function, is unsaturated “in that it
requires something to fall under it; hence it cannot exist on its own” (Letter to
Marty, 1882, 81). On the analogy between concepts and functions, and concepts
and relations, see Dummett (1981: 255-257).
9
The same is true for sentence (1), with conquer being a dyadic predicate.
predicates. Adopting the convention used by Higginbotham (1990), the

open places in the predicate are marked with numerals:10
(5) Berlin [is the capital of (1,2)] the German Empire.
In the following examples, from Higginbotham (1990), the expressions

headed by one of the major syntactic categories of N(oun), V(erb),
A(djective), and P(reposition) are understood as n-place predicates. The
open places associated with the predicates are marked with numerals,
their appropriate arguments are indexed, and the underlined constituents
belong to the categories noted:
(6) a. Mary considers them1 [N fools (1)]

b. Mary1 [V persuaded (1,2,3)] me2 of something3
c. John1 left the room [AP proud of Mary (1)]
d. John1 is [P in (1,2)] the garden2
e. The1 [N’ conviction that the earth is in danger (1)] is widespread.
A sentence can contain no open positions, since no true or false state-

ment can be made with an open sentence. As noted by Higginbotham
(1990), the open positions in the various words and phrases that make up
a sentence must be eliminated, or discharged, as defined by the rules of
compositional semantics.
2. Predication
Research in philosophy of language, logic, and linguistics makes ample,

both explicit and implicit, use of the concept of predication.11 In tradi-
10
This convention of representing argument places with numerals was already used
by Quine and Davidson, see also Higginbotham (1985).
11
In this paper I focus on predication as a grammatical operation. I do not consider
an approach to predication advocated by, for example, V. H. Dudman, who
claimed that “a theory of predication for a language is meant to explain what can
be said in it, and how” (Dudman 1985: 43). For some recent (critical) discussion
on the relation between predication and purported mental processes, see Peregrin
(2011).
tional grammar, predication is the relation between the subject and the
predicate. In logic, predication is the attributing of characteristics to a
subject to produce a meaningful statement combining verbal and nomi-
nal elements. This understanding stems from Aristotelian logic, where
the term (though not explicitly used by the philosopher) might be de-
fined as “saying something about something that there is”.12
In more recent logical inquiries, this classical definition is echoed by
the ‘thing-property relation’ (e.g. in Reichenbach 1947). Quine (1960)
treats predication as the basic combination in which general and singular
terms find their contrasting roles; he also considers it to be one of the
mechanisms which joins occasion sentences. This idea is close to Lo-
renzen’s (1968) ‘basic statements’ (Grundaussagen), the simplest struc-
tures of a language that are composed of a subject and a predicate.
Strawson (1971) stresses that predication is an assessment for truth-
value of the predicate with respect to the topic, and according to Link
(1998) it is the basic tool for making judgments about the world. Simi-
larly Krifka (1998) claims that predication establishes a relation of a
specified type between a number of parameters, or semantic arguments.
For example, sentences with intransitive verbs establish a relation that
holds of the subject for some event, and sentences with transitive verbs
establish a relation between the subject, the object, and some event.
In Davidson’s approach to verb semantics predication can be specified
as a relation between a verb and one of its semantic entailments (Da-
vidson 1967), or as a combinatory relation which makes it possible to
join a property and an argument of the appropriate semantic type to form
a formula whose truth or falsity is established according to whether the
property holds of the entity denoted by the argument or not. Elsewhere
Davidson related predication to the problem of ‘the unity of proposition’
and claimed that “[…] if we do not understand predication, we do not
understand how any sentence works, nor can we account for the struc-
12
This definition may be inferred from Aristotle’s concept of a proposition, under-
stood as a “statement, with meaning as to the presence of something in a subject,
or its absence, in the present, past, or future, according to the division of time”
(On Interpretation, 17a23).
ture of the simplest thought that is expressible in language” (Davidson

2005: 77).13
Early generative grammar, in the Chomskyan tradition, paid very lim-
ited attention to the notion; later studies, e.g. Williams (1980) and Roth-
stein (1985, 1992), viewed predication as a primitive syntactic relation.
Whereas Williams argued for an indexing approach and almost equaled
predication with semantic role assignment, Rothstein claimed that the
semantic and syntactic concepts of predication are distinct, and the rela-
tion which holds between predicates and subjects at S-structure can be
defined in purely syntactic terms. Higginbotham (1987) understands
predication as a formal binary relation on points of phrase markers, and
Bowers (1993) postulates the existence of a functional category respon-
sible for implementing the relation.14
3. Types of predication
Frege has been commonly credited with proposing a bipartite analysis of

expressions into a functor and its argument(s). His approach to predica-
tion, understood here as a primitive logical relation, may be thus termed
functional: a function has to be saturated by an argument.15 Fregean ap-
proach contrasts with the concatenative approach to predication, rooted
13
For a recent critical discussion of this claim, see Peregrin (2011). This paper does
not discuss the issue of the unity of the proposition, for some recent studies in-
vestigating, from different angles, predication and the unity of the proposition,
see Gaskin (2008), Collins (2011), and Peregrin (2011).
14
For a background discussion on the notion of predication, and its importance for
linguistics and philosophy of language, see Rothstein (1985), Lenci (1998), and
Stalmaszczyk (1999).
15
Fregean predication should be analyzed in connection with his theory of truth
and semantic relations. However, this paper focuses only on aspects relevant for
a linguistic theory of predication. The mutual relations between the notions of
truth, existence, identity and predication are discussed in Klement (2002) and
Mendelsohn (2005). For a different discussion of Fregean predication, in the con-
text of the unity of the proposition, see Gaskin (2008) and Collins (2011); see al-
so Oliver (2010) on Frege’s account of predicates.
in the Aristotelian tradition. Importantly, both these approaches find ap-

plication in modern theories of grammar.16
3.1. Aristotelian predication
In Aristotelian semantics, predication is the relation constituted by two

elements, the subject with the predicate, with tense specification being a
third, concatenating, element, cf. (7):17
(7) Aristotelian predication:

Proposition Subject∩Tense∩Predicate
Whereas predication in grammatical theory is concerned with linguistic

items, Aristotle was concerned with relations between entities in the on-
tology, it is therefore necessary to distinguish between reference to lin-
guistic items, i.e. linguistic predication, and reference to items in the on-
tology, i.e. ontological predication:18
(8) Linguistic predication:

A predicate (a linguistic item) is linguistically predicated of its
subject (a grammatical item).
(9) Ontological predication:
A predicable (a metaphysical item) is ontologically predicated
of its subject (an item in the ontology).
16
For a re-analysis of the Fregean approach in contemporary generative grammar,
see Rothstein (1985), Eide and Åfarli (1999). A formal approach to Fregean se-
mantics is presented by Chierchia (1985) and Bowers (1993). Some implications
of Fregean semantics for categorial grammar are discussed by Wiggins (1984).
For a discussion of functionality and predication, see Klement (2002: 28-32).
17
Cf. the Aristotelian definition of simple proposition quoted in note 12, above.
This is not to claim, however, that Aristotelian predication is limited to structural
configurations, on the contrary, it has deep ontological grounding. As observed
by Moravcsik (1967: 82), Aristotle “takes predication to be showing the ontolo-
gical dependence of the entity denoted by the predicate on the entity denoted by
the subject”.
18
Lewis (1991: 4) refers to the latter as metaphysical predication.
Two relations which are variants of metaphysical/ontological predication

are introduced by Aristotle in the Categories:19
(10) X is SAID OF Y
i. Socrates is a man
ii. If ‘Socrates is a man’ is true, then man is SAID OF Socrates.
(11) X is IN Y
i. Socrates is pale
ii. If ‘Socrates is pale’ is true, then pallor is IN Socrates.
(10) and (11) exemplify the relations holding between items in the on-
tology: between a metaphysical subject, Socrates, and a predicable, man
or pallor. ‘SAID OF a subject’ is a relation of ontological classification,
‘IN a subject’ is a relation of ontological dependence. Furthermore, (10i)
tells us something fundamental about what kind of thing Socrates is, it is
therefore an example of essential predication. On the other hand, (11i)
tells us something that happens to be the case, it is an example of acci-
dental predication. In (10ii) and (11ii) the linguistic predication is relat-
ed to ontological predication, however, Aristotle is concerned primarily
with giving the metaphysical configurations that underlie sentences (10i)
and (11i), and, as pointed out by Lewis (1991:55), the philosopher is si-
lent on how the two kinds of predication are related. As observed by
Lewis (1991: 4, n.4), the relation between linguistic predication and
metaphysical predication is not bi-directional: the subject of a linguistic
predication can be either a linguistic item or an entity in the ontology,
however the subject of a metaphysical predication will always be an on-
tological item, and not a linguistic one.
A similar point is made by Mesquita (2012), who has recently stressed
the necessity to distinguish two levels in Aristotelian thought: “the onto-
logical level, where we speak of predicates as something that pertains to
things; and the logical level, where we speak of predicates as something
that is said of things” (Mesquita 2012: 20). In the former case the predi-
cate is an entity ‘utterly extra-logical and extra-linguistic’, in the latter
the predicate is a term, a part of a sentence, hence a linguistic item.
19
Cf. the full discussion in Lewis (1991: 53-63).
3.2. Fregean predication
In Fregean semantics the application of a function to an argument is not

a mere juxtaposition of the two elements. The function combines with
the argument into a self-contained whole due to the fact that it contains a
logical gap (the place-holder, or argument-place) which needs filling. As
concluded by Frege in Über Begriff und Gegenstand (‘On Concept and
Object’, 1892, henceforth CO): “not all the parts of a thought can be
complete; at least one must be unsaturated or predicative; otherwise they
would not hold together” (CO, 193). Tichý (1988: 27) comments that
“the function latches on its argument, sticking to it as if through a suc-
tion effect”. Thus, in Fregean semantics, predication is a relation in
which an argument saturates an open position in the function, cf. the
simplified formula (12):
(12) Fregean predication:

Proposition [Function (1, …, n)]∩Argument(1, …, n)
It needs to be stressed at this point that Fregean semantics is not concer-

ned with natural language predicates. His line of reasoning, however,
may be applied to analyzing predication as a grammatical relation. For-
mula (12) aims at capturing Tichý’s observation that “the function latch-
es on its argument”, furthermore, the “suction effect” is attributed to the
presence of the open position(s) – ‘(1, …, n)’. I use the term “Propo-
sition” in (12) as a generalized term for Frege’s “act of judgement” and
“assertion”.20 As observed by Stevens:
Rather than analyzing the proposition into a series of elements (subject,

predicate, copula), Frege construes the predicative part of the proposition as
a function which is essentially incomplete or ‘unsaturated’. (Stevens 2003:
224)
20
Cf. the distinction introduced by Frege in Thought (329):
(1) The grasp of a thought – thinking,
(2) The acknowledgement of the truth of a thought – the act of judgment,
(3) The manifestation of this judgment – assertion.
In other words, there is a need for a predicative constituent in a propo-

sition to bind the propositional content into a ‘fully-fledged unit’, cf.
Stevens (2003: 230-231).
In On Concept and Object Frege invokes the concept of unsatura-
tedness, or incompleteness, in order to elucidate the distinction between
singular terms and general terms. In Fregean semantics, a singular term
is a name of an object or a definite description. It is a saturated expres-
sion that can be used in the subject position of a subject-predicate sen-
tence. A general term is a name of a concept. It is an unsaturated expres-
sion, realized by a common noun, adjective or verb, which may be used
in the predicate place of a subject-predicate sentence, e.g.:
(13) a. Socrates is a man.

b. Socrates is wise.
c. The Greek philosopher sleeps.
The distinction between singular and general terms has interesting con-
sequences for the status of the copula. In sentences (13a) and (13b) the
singular term is linked to the general term by the copula is. Frege very
carefully distinguishes here between two different uses of the word is.21
Consider the following sentences, based on Frege’s examples:
(14) a. He is Alexander the Great.

b. It is the number four.
c. It is the planet Venus.
d. It is green.
e. It is a mammal.
In the first three instances, is is the “is of identity”, used “like the
‘equals’ sign in arithmetic, to express an equation” (CO, 183). In the last
two examples, it serves as a copula, “a mere verbal sign of predication”
21
Obviously, this distinction has an ancient tradition, cf. Aristotelian essential vs.
accidental predication; for a detailed discussion, see Lewis (1991). Frege’s origi-
nality, however, lies in showing the different underlying patterns of names and
predicates. See Wiggins (1984) and Collins (2011) for a recent discussion of this
issue.
(CO, 182). In other words, “something falls under a concept, and the
grammatical predicate stands for this concept” (CO, 183). One more ex-
ample helps to clarify the distinction:
(15) a. The Morning Star is Venus.

b. The Morning Star is a planet.
In (15a) we have two proper names, “the Morning Star” and “Venus”,
for the same object. In this sentence, the word is forms an essential part
of the predicate, it carries full predicative force. The predicate is formed
by is together with the name. This is the relation of equation, which in-
volves two arguments. In (15b), on the other hand, we have one proper
name “the Morning Star”, and one predicate, the concept-word “planet”.
Here the word is is just the copula, the “verbal sign of predication”. In
this instance, we have the relation of an object’s falling under a concept
(i.e. subsumption). Note that Frege distinguishes here between two dif-
ferent relations: that of one object falling under a concept (subsumption),
and that of one concept being subordinated to another (subordination).
Only in the first case, can we talk about the subject-predicate relation. In
contrast to equation, the relation of subsumption is irreversible.
We may add labeled brackets to (15), in order to explicitly illustrate
the distinction made by Frege:
(16) a. [Name The Morning Star] [Predicate [is] [Name Venus]]

b. [Name The Morning Star] [Copula is] [Predicate a planet]
In yet another formulation, it may be claimed that the sense of the copula
“is essentially a schematic truth rule that links names and predicates in
terms of their respective contribution to the determination of the truth
conditions of the host structure” (Collins 2011: 78). Following insights
from Frege, and under direct influence of Wiggins, Collins (2011: 78)
specifies this ‘schematic truth rule’ in the following way:22
22
Wiggins (1984: 318) stresses the fact that the copula does not stand for a relation,
it “does not need to do so, in order to contribute to the sense of the sentence”.
Where n is a name and P is a Predicate, True [n + copula + P] iff the

reference of n falls under the reference of P.
Collins also comments on the relation between the copula and jud-
gement: “The copula, as the basis of unity, is not a component of judge-
ment but signifies the unity of the judgement as constituted by the tran-
scendental unity of the judger” (Collins 2011: 14, n. 6).23
The distinction between singular terms and general terms features ex-
plicitly in yet another definition of predication. According to Quine
(1960), predication is the basic combination in which general and sin-
gular terms find their contrasting roles: “Predication joins a general term
and a singular term to form a sentence that is true or false according as
the general term is true or false of the object, if any, to which the singu-
lar term refers” (Quine 1960: 96). Since in Quine’s approach to predica-
tion the focus is on proposition formation, this instance of the relation
may be termed propositional predication.
The logical approach to functions might be insightful for the gramma-
tical analysis of predicates. In this paper, I am predominantly concerned
with the nature of the relation holding between two types of “linguistic
devices”: those which have an identifying function (Frege’s arguments),
and those which have a predicating function (Frege’s functions).
I firmly believe that the same line of argumentation applies to the
structural organization of syntactic arguments and predicates. Sentence
(4), repeated below, was analyzed as involving a two-place predicate
(17b):
(17) a. Berlin is the capital of the German Empire. (= (4))

b. Berlin [is the capital of (1,2)] the German Empire. (= (5))
At the same time, however, in this sentence the entire expression is the
capital of the German Empire is a one-place predicate saturated by the
sentential subject:
23
It needs to be added that Collins makes this remark in the context of his investi-
gation of the problem of the unity of linguistic meaning, and in this particular
remark he also comments on Kant’s theory of judgment.
(18) Berlin [is the capital of the German Empire (1)]
Frege himself introduces this two-step approach, cf. the following com-
ment from his Notes for Ludwig Darmstaedter:
The sentence ‘The capital of Sweden is situated at the mouth of Lake Mä-
lar’ can be split up into a part in need of completion and the saturated part
‘the capital of Sweden’. This can further be split up into the part ‘the capital
of’, which stands in need of completion, and the saturated part ‘Sweden’
(Notes for Ludwig Darmstaedter, 364).
Comparison of (17b) with (18) points to the necessity of distinguishing

between two types of predication: polyadic (i.e. ‘grammatical’ in Frege’s
terminology) and monadic (i.e. structural, to be discussed below) e.g.:
(19) a. Berlinα [is the capital of] the German Empireβ

(polyadic predication)
b. Berlinα [is the capital of the German Empire]
(monadic predication)
I will assume here that polyadic predication is a relation which occurs

between a predicate and its argument(s), and is a consequence of the
predicate’s semantic properties.
4. Predication in generative grammar
In generative grammar (e.g. Chomsky 1981, 1982), polyadic predication

is associated with semantic interpretation, and as such falls under the
scope of Theta Theory, one of the modules in the Government and Bind-
ing model of generative grammar. For this reason, it may be referred to
as thematic predication. The appropriate context, together with a brief
description, is provided below:
(20) Thematic predication:

i. Argument1 [Predicate (1,2,…)] Argument2 …
ii. Thematic predication deals with semantic interpretation of

arguments and thematic role assignment.
Monadic predication, on the other hand, involves a one-place predicate

and a referring term functioning as its unique argument (e.g. structural
subject), and can be thus termed structural predication:24
(21) Structural predication:

i. Argument [Predicate (1)]
ii. Structural predication deals with configurational relations be-
tween nodes described/defined on phrase markers.
In the Government and Binding model of generative grammar, the obli-

gatory presence of the closing argument (subject) in a subject-predicate
structure follows from the second clause of the Extended Projection
Principle (EPP). The Projection Principle formed in Chomsky (1981:
29) states that the subcategorization properties of each lexical item must
be represented categorially at each syntactic level, the EPP adds a sec-
ond requirement: that clauses have subjects, cf. Chomsky (1982: 10).25
Later, Chomsky (1986) suggested that this part of the EPP might be de-
rived from the theory of predication, as developed in Williams (1980)
and Rothstein (1985), and observed that the EPP “is a particular way of
expressing the general principle that all functions must be saturated”
(Chomsky 1986: 116). Chomsky explicitly referred to Frege, and ob-
served that a maximal projection (e.g. VP or AP) may be regarded as a
syntactic function that is “unsaturated if not provided with a subject of
which it is predicated” (Chomsky 1986: 116).
There is one more crucial property of structural predication not cap-
tured by the above descriptions. As observed already by Aristotle, predi-
cation is constituted by three elements: the subject, the predicate, and the
tense element, cf. (7) above. Also Frege, though working in a completely
24
I understand monadic predication similarly to Rothstein (1992: 153), who defines
it as the relation holding between the subject of a sentence and the “remainder”.
25
More recently, Åfarli and Eide (2000) proposed that the EPP is the effect of a
proposition-forming operation of natural language, induced by a predication op-
erator, and Åfarli (2005) shows the effects of semantic saturation.
different tradition, referred to the copula as the “verbal sign of predica-

tion” (see below), possibly implying that predication is a relation involv-
ing three, rather than two, elements. It is also possible to regard the ‘con-
tent stroke’, introduced by Frege in CN §2, as a third element involved
in predication.26
In accordance with the above observations, it is assumed here that
structural predication involves two terms – the subject argument and the
predicate function – and additionally an operator of predication, which
means that the schema and definition in (21) are inadequate, and require
the following reformulation:27
(22) Structural predication:

i. Argument [Operator [Predicate (1)]]
ii. Structural predication deals with configurational relations
between nodes described/defined on phrase markers trigge-
red by the operator of predication.
In other words, I make here two basic claims about structural predica-
tion:28
(23) I. Structural predication is monadic, in the sense that the predi-

cate takes only one argument (predication subject);
II. Structural predication is tripartite, in the sense that an appro-
priate operator is required to trigger predication.
The following table summarizes the correspondences that relate the ap-
propriate notional elements in the definitions of structural predication
introduced in this study: (i) shows the general non-theoretic pattern, (ii)
26
A similar observation has been made by Åfarli and Eide (2000: 47, n. 5). See
also Smith (2000) on the meaning and function of the judgment stroke in Frege’s
logic.
27
In the semantics developed by Chierchia (1985) and Bowers (1993), the predica-
tion operator is a function that takes the property element to form a propositional
function, which in turn takes an entity to form a proposition; for further details,
see Eide and Åfarli (1999) and Åfarli (2005).
28
For a discussion of these claims, see Stalmaszczyk (1999).
its Fregean functional equivalents, (iii) gives the relational definitions of

grammatical functions in Standard Theory, (iv) provides the X-bar theo-
retic category neutral elements, and (v) the elements from predication
theory grounded in more recent generative grammar (where Pr is a sepa-
rate functional category with its own projections Pr’, PrP):
(i) Subject Operator Predicate

(ii) Argument sign of predication Function
(iii) Subject-of: Predicate-of:
[NP, S] [PredP, S]
(iv) [Spec, XP] X [X’, XP]
(v) [Spec, PrP] Pr [Pr’, PrP]
Table 1. Elements of structural predication
A further layer of corresponding notions will be added after presenting

the semantics of predication; first, however, I discuss the predicational
domain and the issue of syntactic evidence for structural predication.
5. The semantics for structural predication
This section provides some informal semantic interpretation for structu-

ral predication and the elements involved in its constructing. The indis-
pensability of this move follows directly from Bouchard’s (1995: 22)
Principle of Full Interpretation:
(24) Principle of Full Interpretation:

Every (morpho-)syntactic formative of a sentence must have a
corresponding element in the semantic representation.
Every formative of a semantic representation must be identified
by a (morpho-)syntactic element in the sentence, which is asso-
ciated with that representation.
I understand the “syntactic elements” mentioned in the Principle very

broadly, including all the abstract elements used in defining structural
predication. Below I propose a restricted version of semantics for struc-
tural predication.29 I assume that there are three primitives: proposition

(P), entity (e), and property (R). Additionally, I regard the predicate as a
function from an entity to a proposition (a propositional function), and
the predication operator as a function from a property to a predicate.30
The basic correspondences are illustrated below:
(25) PrPP
NPe Pr’<e,P>
Pr<R,<e,P>> VPR
As far as the property (i.e. VP in syntactic terms) is concerned, it is as-

sumed here that it is of compositional nature: it is composed of a func-
tion from an entity (or property) to a property predicate:
(26) VPR
NPe V’<e,R>
V<e,<e,R>> NPe/XPR
In other words, the verb is defined as a function from an entity (or prop-
erty) to a property predicate, and consequently V’ is interpreted as a
property predicate, i.e. a function from an entity to a property.
Table 2 shows correspondences between the syntactic and semantic
elements, with (i) the syntactic symbols, (ii) the semantic elements (also
with less formal terminology), and (iii) the semantic symbols:31
29
For a general background to this type of semantics, stemming from the work of
Kazimierz Ajdukiewicz, see Bowers (1993), Chierchia (1985), Eide and Åfarli
(1997), Tałasiewicz (2010).
30
Note that these notions are introduced operationally, rather than ontologically.
For the methodological background behind this distinction, see Lorenzen (1968).
31
This table complements Table 1, which provided the elements of structural predi-
cation within different approaches.
(i) syntax (ii) semantics (iii) semantic symbol

V function from an enti- e<e,R>
ty to a property predicate
V’ function from an enti- <e,R>
ty to a property
VP property R
Pr function from a prop- R<e,P>
erty to a predicate
(predication operator)
Pr’ function from an enti- <e,P>
ty to a proposition (pred-
icate)
PrP proposition P
NP entity e
Table 2. Semantic elements in structural predication
Table 3 shows once again the relevant patterns of correspondences, this

time explicitly specifying the properties of appropriate functions, i.e.
their arguments and values:
syntactic sym- semantic

Function Argument Value
bol symbol
Operator of Property Predicate Pr <R, <e, P>>
predication
Predicate Entity Proposition Pr’ <e, P>
Verb Entity Property V <e, R>
Table 3. Functions, arguments and values
The full semantic derivation, along the lines suggested above, of a simp-
le transitive sentence (27) proceeds in the following way, with indexing
added for explanatory purposes only (and disregarding the issues in-
volved in appropriate tense derivation):
(27) Caesar conquered Gaul.

(28) 1. V = e<e, R> = {e, conquered e}
2. V’ = <e, R> = {Gauli, conquered ei}
3. VP = R = {conquered Gaul}
4. Pr = R<e, P> = {(conquered Gaul)i, e P}

5. Pr’ = <e, P> = {Caesari, ei conquered Gaul}
6. PrP = P = {Caesar conquered Gaul}
The above derivation resembles and reflects to some extent the process
of Numeration (in the sense of Chomsky 1995: 225), and therefore it
might be referred to as ‘Semantic Numeration’. Together, the two pro-
cesses provide grammatical and predicational domains of the derived
structure. Within the grammatical domain the grammatical functions are
determined and grammatical properties like modality, tense and aspect
are realized. The predicational domain provides the input to semantic
interpretation. At the same time the Principle of Full Interpretation (24)
is fulfilled, since there is full correspondence between the (morpho-)
syntactic formatives of a sentence and the elements in the semantic rep-
resentation. This integrated approach to predication clearly demonstrates
that the different aspects of predication are related.
6. Conclusion
The aim of this paper was to demonstrate that Frege’s approach to func-
tions has consequences for the treatment of predicates and predication in
the theory of grammar. I have tentatively postulated the existence of
thematic, structural and propositional predication, every instance of the
relation being linked to Frege’s logical and philosophical inquiries.
References
Åfarli, Tor 2005. Predication in Syntax. Research in Language 3, 67-89.

Åfarli, Tor and Kristin Eide 2000. Subject Requirement and Predication. Nordic
Journal of Linguistics 23, 27-48.
Aristotle [1941]. De Interpretatione (On Interpretation). Translated by E. M.
Edghill. In: The Basic Works of Aristotle. Edited by R. McKeon. New York:
Random House, 38-61.
Beaney, Michael 1997. Introduction. In: M. Beaney (ed.), 1-47.
Bouchard, Denis 1995. The Semantics of Syntax. A Minimalist Approach to Gram-

mar. Chicago and London: The University of Chicago Press.
Bowers, John 1993. The Syntax of Predication. Linguistic Inquiry 24, 591-656.
Carl, Wolfgang 1994. Frege’s Theory of Sense and Reference. Its Origins and
Scope. Cambridge: Cambridge University Press.
Carnap, Rudolf 1937. The Logical Syntax of Language. London: Routledge and
Kegan Paul.
Chierchia, Gennaro 1985. Formal Semantics and the Grammar of Predication. Lin-
guistic Inquiry 16, 417-443.
Chomsky, Noam 1981. Lectures on Government and Binding. Dordrecht: Foris.
Chomsky, Noam 1982. Some Concepts and Consequences of the Theory of Gov-
ernment and Binding. Cambridge, Mass.: MIT Press.
Chomsky, Noam 1986. Knowledge of Language. Its Nature, Origin, and Use. New
York: Praeger.
Chomsky, Noam 1995. The Minimalist Program. Cambridge, Mass.: MIT Press.
Collins, John 2011. The Unity of Linguistic Meaning. Oxford: Oxford University
Press.
Davidson, Donald 1967. The Logical Form of Action Sentences. [Reprinted in
Donald Davidson, 1980. Essays on Actions and Events, Oxford: Clarendon
Press, 105-122.]
Davidson, Donald 2005. Truth and Predication. Cambridge, Mass. and London:
Belknap Press of Harvard University Press.
Dudman, Victor Howard 1970. Frege’s Judgment-stroke. The Philosophical Quar-
terly 20 (79), 150-161.
Dudman, Victor Howard 1985. Towards a Theory of Predication for English. Aus-
tralian Journal of Linguistics 5, 143-196.
Dummett, Michael 1981. Frege: Philosophy of Language (Second edition). Lon-
don: Duckworth.
Dummett, Michael 1993. Origins of Analytical Philosophy. London: Duckworth.
Eide, Kristin and Tor Åfarli, 1999. The Syntactic Disguises of the Predication
Operator. Studia Linguistica 53, 155-181.
Beaney (ed.), 47-78.
Frege, Gottlob [1882] 1997. Letter to Marty, 29.8.1882 (translated by H. Kaal). In:
M. Beaney (ed.), 79-83.
Frege, Gottlob [1884] 1960. The Foundations of Arithmetic (translated by J. L. Aus-
tin). Second revised edition. New York: Harper & Brothers.
Frege, Gottlob [1884] 2007. The Foundations of Arithmetic (translated by D.
Jacquette). New York: Pearson Education Inc.
Frege, Gottlob [1891] 1997. Function and Concept (translated by P. Geach). In: M.
Beaney (ed.), 130-148.
Frege, Gottlob [1892] 1997. On Concept and Object (translated by P. Geach). In:
M. Beaney (ed.), 181-193.
Frege, Gottlob [1893] 1997. Grundgesetze der Arithmetik, Vol. I (translated by M.
Beaney). In: M. Beaney (ed.), 194-223.
Frege, Gottlob [1897] 1997. Logic (translated by P. Long and R. White). In: M.
Beaney (ed.), 227-250.
Frege, Gottlob [1906] 1997. Letters to Husserl, 1906 (translated by H. Kaal). In: M.
Beaney (ed.), 301-307.
Frege, Gottlob [1918] 1997. Thought (translated by P. Geach and R. H. Stoothoff).
In: M. Beaney (ed.), 325-345.
Frege, Gottlob [1919] 1997. Notes for Ludwig Darmstaedter (translated by P. Long
and R. White). In: M. Beaney (ed.), 362-367.
Gaskin, Richard 2008. The Unity of the Proposition. Oxford and New York: Oxford
University Press.
Green, Mitchell S. 2002. The Inferential Significance of Frege’s Assertion Sign.
Facta Philosophica 4, 201-229.
Hanfilng, Oswald 2000. Philosophy and Ordinary Language. The Bent and Genius
of our Tongue. London and New York: Routledge.
Heck, Richard and Robert May 2006. Frege’s Contribution to Philosophy of Lan-
guage. In: E. Lepore and B. Smith (eds.), 3-39.
Higginbotham, James 1985. On Semantics. Linguistic Inquiry 16, 547-593.
Higginbotham, James 1987. Indefiniteness and Predication. In: Eric R. and A. ter
Meulen (eds.), 43-70.
Higginbotham, James 1990. Frege and Grammar. Unpublished MS, Oxford: Oxford
University.
Jespersen, Otto 1937. Analytic Syntax. Helsingor. Reprinted New York: Holt, Rine-
hart and Winston, 1969.
Klement, Kevin C. 2002. Frege and the Logic of Sense and Reference. London and
New York: Routledge.
Krifka, Manfred 1998. The Origins of Telicity. In: S. Rothstein (ed.), 197-235.
Lenci, Alessandro 1998. The Structure of Predication. Synthese 114, 233-276.
Lepore, Ernest and Barry C. Smith (eds.), 2006. The Oxford Handbook of Philoso-
phy of Language. Oxford: Oxford University Press.
Lewis, Frank A. 1991. Substance and Predication in Aristotle. Cambridge: Cam-
Link, Godehard 1998. Algebraic Semantics in Language and Philosophy. CSLI
Lecture Notes No. 74, Stanford: Center for the Study of Language and Infor-
mation.
Lorenzen, Paul 1968. Methodisches Denken. Frankfurt am Main: Suhrkamp Verlag.
Losonsky, Michael 2006. Linguistic Turns in Modern Philosophy. Cambridge:
Łukasiewicz, Jan 1957. Aristotle’s Syllogistic from the Standpoint of Modern For-
mal Logic (second enlarged edition). Oxford: Clarendon Press.
Mendelsohn, Richard 2005. The Philosophy of Gottlob Frege. Cambridge: Cam-
Mesquita, António Pedro 2012. Types of Predication in Aristotle (Posterior Analyt-
ics I 22). Journal of Ancient Philosophy 6 (2), 1-27. Retrieved from
http://revistas.usp.br/filosofiaantiga/article/view/46736.
Moravcsik, Julius M. E. 1967. Aristotle on Predication. Philosophical Review 76,
80-96.
Oliver, Alex 2010. What is a Predicate? In: M. Potter and T. Ricketts (eds.), 118-
148.
Penco, Carlo 2003. Two Theses, Two Senses. History and Philosophy of Logic 24,
87-109.
Peregrin, Jaroslav 2011. There is no such Thing as Predication. Conceptus 97, 29-
51.
Pietroski, Paul M. 2004. Events and Semantic Architecture. Oxford: Oxford Univer-
sity Press.
Potter, Michael 2013. Frege, Russell, and Wittgenstein. In: G. Russell and D. Graff
Fara (eds.), 852-859.
Potter, Michael and Tom Ricketts (eds.) 2010. The Cambridge Companion to Fre-
ge. Cambridge: Cambridge University Press.
Quine, William V. O. 1960. Word and Object. Cambridge, Mass.: MIT Press.
Reichenbach, Hans 1947. Elements of Symbolic Logic. Berkeley: University of Cali-
fornia Press. Reprinted 1966, New York: The Free Press.
Reuland, Eric and Alice ter Meulen, (eds.) 1987. The Representation of
(In)definitness. Cambridge, Mass.: MIT Press.
Rothstein, Susan D. 1985. The Syntactic Forms of Predication. Bloomington: Indi-
ana University Linguistics Club.
Rothstein, Susan D. 1992. Predication and the Structure of Clauses. Belgian Journal
of Linguistics 7, 153-169.
Rothstein, Susan D. (ed.) 1998. Events and Grammar. Dordrecht: Kluwer Academic
Publishers.
Russell, Gillian and Delia Graff Fara (eds.) 2013. The Routledge Companion to Phi-
losophy of Language. New York and London: Routledge.
Smith, Nicholas 2000. Frege’s Judgement Stroke. Australasian Journal of Philoso-
phy 78, 153-178.
Soames, Scott 2010. Philosophy of Language. Princeton and Oxford: Princeton
University Press.
Stalmaszczyk, Piotr 1999. Structural Predication in Generative Grammar. Łód :
Wydawnictwo Uniwersytetu Łódzkiego.
Stalmaszczyk, Piotr 2006. Fregean Predication: Between Logic and Linguistics.

Research in Language 4, 77-90.
Stalmaszczyk, Piotr 2010. Gottlob Frege, Philosophy of Language and Predication.
In: P. Stalmaszczyk (ed.), 295-314.
Stalmaszczyk, Piotr (ed.) 2010. Philosophy of Language and Linguistics. Volume 1:
The Formal Turn. Frankfurt am Main: Ontos Verlag.
Stevens, Graham 2003. The Truth and Nothing but the Truth, yet Never the Whole
Truth: Frege, Russell and the Analysis of Unities. History and Philosophy of
Logic 24, 221-240.
Strawson, Peter F. 1971. Logico-Linguistic Papers. London: Methuen.
Sullivan, Peter 2010. Dummett’s Frege. In: In: M. Potter and T. Ricketts (eds.), 86-
117.
Tałasiewicz, Mieszko 2010. Philosophy of Syntax. Foundational Topics. Dordrecht:
Springer.
Thiel, Christian 1968. Sense and Reference in Frege’s Logic. Dordrecht: D. Reidel
Publishing Company.
Tichý, Pavel 1988. The Foundations of Frege’s Logic. Berlin and New York: Wal-
ter de Gruyter.
Wiggins, David 1984. The Sense and Reference of Predicates: A Running Repair to
Frege’s Doctrine and a Plea for the Copula. The Philosophical Quarterly 34,
311-328.
Williams, Edwin 1980. Predication. Linguistic Inquiry 11, 203-238.
Wittgenstein, Ludwig [1922] 1995. Tractatus Logico-Philosophicus (translated by
D. F. Pears and B. F. McGuiness). London and New York: Routledge.
Piotr K. Szałek
Catholic University of Lublin
[email protected]
Russell, Wittgenstein, and the Notion of

False Propositions
Abstract: The paper deals with one of the major puzzles for the Wittgensteinian
picture theory of language: if sense of propositions is determined by picturing the
possible state of affairs, how then can false propositions have sense if their corre-
sponding state of affairs does not exist? In order to answer the question, the paper
reconstructs the main structure of the argument for the theory in the context of its
relation to Bertrand Russell’s view on judgments. The paper argues that the Witt-
gensteinian solution to the problem of the false propositions relies on the notion of
the intrinsic symbolic (syntactic) structure of the proposition in virtue of the ana-
logy to the pictorial representation and possible configuration of its components.
Keywords: Wittgenstein, Russell, Frege, semantics, picture theory of language,

propositions, falsehood
0. Introduction
Ludwig Wittgenstein in his Tractatus Logico-Philosophicus aimed to

explicate the nature of language and its relation to the world. In more
technical terms, this work could be regarded as an attempt to clarify the
nature of the proposition of logic and of the logical constants, by means
of reference to the notion of a representation in general. This task culmi-
nated in the so-called picture theory of language, which in terms of syn-
tactical symbolism, described propositions as picture mirroring reality.
In this paper, I consider the major difficulty that directs this theory
and its possible solution, namely, the problem of the meaning of the
false propositions: if the sense of propositions is determined by picturing
the possible state of affairs, how then can false propositions have sense
if their corresponding state of affairs does not exist?
256 Piotr K. Szałek
In order to accomplish this task, I reconstruct the main structure of the

argument offered by Wittgenstein in his Tractatus (1922/1974), Note-
books 1914-1916 (1961), and Notes on Logic (1961). The paper is or-
dered in the following way. In the first part, I introduce and explain the
idea of picturing of the representation in general. In the second, I
demonstrate the adaptation of this idea to the theory of propositions. In
the third, I clarify the problem of falsehood in the context of the theory
of propositions. Finally, in fourth part, I answer the original question on
the basis of my prior analysis. I argue that the Wittgensteinian solution
to the problem of false propositions relies on the notion of intrinsic sym-
bolic (syntactic) structure of the proposition in virtue of the analogy to
pictorial representation and possible configuration of its components.
1. The idea of picturing
In Tractatus Wittgenstein offered clarification of the nature of the propo-

sition by an analogy to the pictures. He tried to adapt the framework of a
general theory of representation to the explanation of the structure and
functions of the proposition in language (see Wittgenstein 1922/1974: 6-
19). Following Kenny, we can state that in the case of any representation
there are two things one should consider: (i) what it is a representation
of, and (ii) whether it represents what it represents accurately or inaccu-
rately.
The distinction between these two features of representation corresponds to

the distinction, concerning proposition, between what the proposition
means and whether what it means is true or false – between sense and truth
value. (Kenny 1973/1975: 54)
Both these elements are present in Wittgenstein’s considerations of

propositions.
According to Wittgenstein, the nature of propositional representation
is an application of the general account of any kind of representation.
Any picture, model, or representation represents something in virtue of
isomorphic relation between it and the reality depicted. A picture must
be composite, that is must be composed of a multiplicity of elements that
Russell, Wittgenstein, and the Notion of False Propositions 257
are by ‘proxy’ or stand for the elements of the situation, which it depicts.
To obtain this, a picture possesses both structure and form. The structure
is conventionally determined by the manner of the depiction. In other
words, the structure of a picture represents the relationship between the
constituent elements of the picture. This relation is in itself a certain sit-
uation or fact, hence Wittgenstein wants to interpret a picture as a fact
(Wittgenstein 1922/1974: 14-15). The possibility of the actual structure
of pictures is the pictorial form (Form der Abbildung) (Wittgenstein
1922/1974: 16-17) – for instance, the three-dimensionality of a diorama,
the two-dimensionality of a painting, or the linear order of a musical
score. Additionally, the different representations of the same state of af-
fairs must share the same logical form, that is, the same (logico-mathe-
matical) multiplicity of separate elements and (conventionally deter-
mined) possibilities of combinations of these objects (Wittgenstein 1922/
1974: 14-15, 42-43).
The bare minimum of a pictorial (respectively representational) form
is therefore the logical form: whatever has a pictorial form also pos-
sesses logical form. “Any representation shares with what it represents
its logico-pictorial form” (Hacker 1972/1986: 59). In other words, the
elements of the picture are correlated with the elements of the situation
or state of affairs that it depicts in virtue of this pictorial relationship. In
order to be representational, the structure of a picture (an arrangement of
its elements) must represent a corresponding situation, namely, a cor-
responding possible arrangement of the objects depicted. Thus pictorial
relationship is isomorphic. What it actually represents in a picture “is not
the complex of objects of which it consists, but the fact that they are ar-
ranged as they are” (Hacker 1972/1986: 59). As Wittgenstein puts it, on-
ly a fact can represent a situation, that is, another fact.1
To summarize the most important features of the idea of picturing, we
can emphasise the following: Firstly, every picture must have a logical
1
See in this respect Hacker’s remark: “A model is true if things are in reality as it
represents them as being. To know whether it is true or false, however, it must be
compared with reality, i.e. verified in experience. One cannot, merely by looking
at a model, discern whether things are actually thus or not. The model only de-
termines that things either are so or are not so, and in that sense, only a logical
space, not a place” (Hacker 1972/1986: 59).
form in common with the pictured reality – the logical form is a part of
every pictorial form of any picture. Secondly, every picture represents a
possible state of affairs that determines its sense: the picture is true if its
sense agrees with reality, and is false if it disagrees (see Wittgenstein
1922/1974: 16-19).
2. The propositional representation
Wittgenstein applied the above general theory of a picture to the account

of the proposition. That is, a propositional representation is explicated in
its terms. A proposition then is described as “a picture of reality” that
corresponds to reality by depicting it (Wittgenstein 1922/1974: 38-41).
A proposition as a picture shares the essential features of a picture in
general. Namely, (i) a proposition must be the composite (of a function
and arguments in that case); (ii) the elements of a proposition stand for
the elements of the depicted situation or state of affairs; (iii) a proposi-
tion has a form, that is, combinatorial possibilities of its elements or-
dered by the rules of logical syntax; (iv) a proposition has structure, i.e. a
determined connection between its elements (Hacker 1972/1986: 60;
Glock 1996: 301).
Moreover, in analogy to the structure of a picture, the structure of a
proposition has to be isomorphically related to the logical structure of
what it portrays. Therefore, there must be a strict correlation between
configurations of the represented objects and configurations of the simp-
le signs in a propositional sign. As noted by Hacker,
[w]hat represents in the propositional-sign is the fact that its constituent ex-
pressions are arranged as they are, given the conventional rules of syntax
and given an ‘interpretation’ of the constituent names. (Hacker 1972/1986:
60)
The crucial point for our purpose is that a proposition is described as bi-
polar: it could be true or false. The truth of a proposition is characterized
as agreement between a proposition and a fact. In order to recognize if a
proposition is true or false, it has to be compared with the world. In
some sense then, it must be verified.
The above account of a proposition consists in the basic facet of the

picture theory, and could be most directly adapted only to the so-called
elementary propositions. Ordinary propositions are not like this. The
natural language contains de facto, amongst the others, complex con-
cept-words (or concept-terms), names of complex, and vagueness. Nev-
ertheless, according to Wittgenstein, this ordinary language is intrin-
sically in a proper order. In principle, thanks to the logical analysis, it is
possible to reveal the proper order, or, in other words, to discriminate the
elementary propositions that satisfy this order.
What is then the order revealed by elementary propositions? The ele-
mentary propositions:
(i) contain no logical constants, i.e. such words as “not”, or “and”,

or “all” (they do not have reference in reality, as they are not
names);
(ii) are logically independent of each other;
(iii) consist of names;
(iv) these names represent simple objects in the world; assert (truly
or falsely) the existence of a state of affairs.
(Wittgenstein 1922/1974: 10-11, 22-23, 58-59, 76-77; Keyt 1966: 378).

The elementary propositions are constituents of our natural language.
This language is made of them in virtue of a truth-functional combina-
tion.
3. The problem of falsehood
The theory of proposition as a picture directs us to the problem of false

propositions. How is it possible that a false proposition has sense if its
corresponding state of affairs does not exist? This problem is an instan-
tiation of the more general, old problem of the intentional reference of
falsehood or false proposition, i.e., of the intentional reference of ele-
ments of a language to non-existent objects of the world asserted by this
language. A proposition is true if and only if it depicts how things are in
the world, that is, by corresponding to a relevant fact. However, when a
proposition is false, it still remains meaningful, although no fact corre-

sponds to it.
In order to understand the Wittgensteinian solution to the problem, it
is helpful to recall some connected considerations of Russell and Frege.
Wittgenstein formulated his theory of propositions as pictures in reaction
to the failure of the Russellian theories of judgement, and in the discus-
sion with the Fregean theory of sense and reference (see Pears 1987:
115-152; Glock 1996: 298-299; Ricketts 1996: 64-80).
Russell’s early theory (until 1907), the so-called Single Object Theory
of Judgement (or Dual-Relation Theory of Judgement) met and fell foul
of the problem of falsehood. Namely, in the case of the, S’ propositional
attitude, where it is S’ belief that there is some relation between a and b,
say aRb. This relationship cannot be described, as suggested by Rus-
sell’s theory, as a dual relation between subject S (i.e. his/her state of
mind) and its object (i.e. a proposition that expresses this state). In other
words, if for a belief to be true, it is to be related to a fact, then for a be-
lief to be false it is to relate to an objective falsehood (see Russell
1912/1991: 70-73). This consequence engaged theory in the ontological
commitment to the assertion of the objective falsehood a la Alexius
Meinong, i.e., assertion of the different degree of existence, including
existence of non-existent objects (see Meinong 1904/1960: 76-117),
what was the object of criticism of Russell himself (see Russell 1905 and
1907).
To overcome this difficulty, Russell proposed his new Multiple-
Relation Theory of Judgement (sustained and developed by him in 1910-
1913, and partly rejected later because of the criticism of Wittgenstein).
According to this proposal, S is “acquainted with” or related to the sepa-
rate propositional constituents: a, R, and b, not to the proposition as a
whole however. In other words, every judgement is a relation of S’ mind
to multiple objects, including a relation to this relation. Schematically,
with J for judgement, it could be presented as follows J(a, R, b, S).
However, Wittgenstein noted that this theory would allow S to judge
nonsense: it would be no longer granted that these constituents are
joined in a meaningful way (see Wittgenstein 1961: 95). In reaction to
this observation, Russell modified his Multiple-Relation theory by add-
ing the condition that S is acquainted not only with constituents of the
proposition, but also with its universal logical form, say x y, which
stands for a general fact.
Wittgenstein insisted, however, that this modification is still not
enough to save the theory as the satisfying description of a proposition.
He charged the Russellian theory with inconsistency. According to him,
there are two contradictory points in the theory: on one side, judgement
is described as a fact (general complex), and at the same time, on the
other, it is supposed to be a simple object of acquaintance (Wittgenstein
1961: 100-101).
The first alternative creates a third-man regress: it explains why a, R and b

can combine to form certain facts (aRb, bRa) but not others (RRb, abR) by
reference to a further fact. The second simply adds a further constituent to
the proposition, without ensuring that its constituents to the proposition, in-
cluding the additional one, are combined in a licit way. (Glock 1996: 299)
The dissatisfaction with Russell’s theories leads Wittgenstein to his own

proposal of explication of a proposition that rests on the analogy to a
picture. Formulating his own solution to the problem of falsehood, Witt-
genstein also critically used the consideration of Frege. For Frege, names
and propositions had both sense and meaning (see Frege 1892/2000:
151-180). Wittgenstein, as indicated in his 1913 Notes on Logic written
before the Tractatus, seemed to accept the idea that a proposition has
reference. Unlike Frege, instead of describing the truth-value of a propo-
sition as its reference, he proposed to treat the fact that corresponds to a
proposition as its reference (Wittgenstein 1961: 94). In particular, the
reference of a proposition ‘p’, if ‘p’ is true, is the fact that p; and respec-
tively, if ‘p’ is false it is a negative fact that ~p.
This explanation of the reference of a proposition implies essential
contrast between (i) the correlation of names and their referents, and (ii)
the correlation of propositions and their referents. In order to com-
prehend a name, it is necessary to know its referent, although “we under-
stand propositions without knowing whether they are true or false”
(Wittgenstein 1961: 91), that is, without the condition of knowledge of
their referents.
Additionally, Wittgenstein in his Notebooks, while discriminating
propositions from names, explicitly argued that false propositions do not
have any reference at all (Wittgenstein 1961: 24), even though he used
to speak about “the reality corresponding to” false propositions. This
contention is comprehensible in some sense, because as he put it in the
Tractatus, “[t]he propositions ‘p’ and ‘~p” have an opposite sense, but
there corresponds to them one and the same reality” (Wittgenstein
1922/1974: 44-45). What we comprehend in the case of the propositions
is their sense, not their reference (Wittgenstein 1961: 94).
A name can have only one relationship to reality: it either names something
or it is not a significant symbol at all. But a proposition has a two-way rela-
tion: it does not cease to have a meaning when it ceases to be true. (Kenny
1973/1975: 61)
As Wittgenstein wrote, using spatial metaphor, “names are points, prop-

ositions arrows” (Wittgenstein 1961: 97; 1922/1974: 22-23).
However, the problem of falsehood arises also on the ground of the
aforementioned difference between names and propositions. If there is
no object that names refer to, names are just meaningless; they are not
names at all, because they do not name (Wittgenstein 1922/1974: 22-23).
But propositions, in the case when there are no facts to which they refer,
are not meaningless but simply false.
Thus what a name signifies must exist, but what a proposition signifies need
not. (Keyt 1966: 379)
In other words, the problem is how propositions can be false without los-
ing sense.
4. The combinatorial arrangement of possibilities
Keeping in mind the above consideration about pictorial description of

propositions by Wittgenstein and the problem of false propositions, we
can ask at this stage, how a proposition depicts something even if what it
depicts does not obtain. What is Wittgenstein’s solution to the problem

of falsehood on the ground of his theory of propositions?2
According to Wittgenstein, what a proposition depicts is the possibi-
lity of a certain arrangement of its constituents. As he puts it:
One name is the representative of one thing, another of another thing and
they themselves are connected; in this way the whole images the situation–
like a tableau vivant […] ‘The connection must be possible’ means: The
proposition and the components of the situation must stand in a particular
relation […] in order for a proposition to present a state of affairs it is only
necessary for its components to represent these of the situation and for the
former to stand in a connection which is possible for the latter. (Wittgen-
stein 1961: 26)
Rejecting the Russellian solution, Wittgenstein insists that we do not

need to introduce the additional logical form or the relation between
propositional constituents. It is enough to emphasise that components of
a proposition might be combined in a certain manner. The crucial point
to his proposal is that the possibility of this combination is explained by
virtue of the intrinsic combinatorial possibility of arrangement of the
components that represent the things in reality for which they stand. For
a proposition (or rather propositional sign) to picture, there is no need of
a state of affairs to correspond to it as a whole. Nevertheless, something
must correspond to its components, i.e. must be the one-to-one correla-
tion between components of a propositional sign and components of the
situation it depicts. Moreover, it must be determined by the order of cor-
relation between propositional sign’s elements that depict and depicted
reality.
2
However, it should be noted here that the order of this exposition is reverse to an
order of development of Wittgensteinian ideas. Firstly, he faced some problems
with falsehood, and later proposed the theory of proposition as picture, amongst
others, to solve the Russelian problem of false judgements. In my opinion, how-
ever, in order to comprehend Wittgenstein’s ideas it is better to follow the pro-
posed ordo dicendi (or exponendi), to explicate reverse Wittgensteinian ordo co-
gnoscendi (or maybe more precisely, ordo creandi). In other words, I am rather
interested in the context of justification of Wittgenstein’s solution than the heu-
ristic context of discovery.
[T]he fact that the elements of the picture are related to each other in a de-
terminate way represents that the corresponding things are related to each
other in the same way, whether or not they actually are. (Glock 1996: 309)
For a proposition to picture falsely is to picture a non-existing combi-

nation of existing elements (Wittgenstein 1922/1974: 42-43; 1961: 31).
In other words, what a false proposition signifies is a non-existent (pos-
sible) arrangement of the (actual) existent objects (Wittgenstein 1958:
3).
However, it seems to be still justified to ask the question of how even
this potential arrangement is possible if there is no state of affairs that
refers to the content of the false proposition. According to Wittgenstein,
assuming the method of projection, the names of the propositional sign
are arranged themselves as the object would be arranged if the propo-
sition in question were true (Wittgenstein 1922/1974: 42-43; 1961: 26).
In fact, the arrangement itself is possible, because it stands for the ar-
rangement of names, but not for what the names stand. Therefore, unlike
Russell, Wittgenstein focuses on a theory of symbolism of propositions
(or syntax of the propositions) rather than on the theory of judgement
that expresses a certain propositional attitude. He considers “the linguis-
tic representations we use to express thoughts” (Ricketts 1996: 69).
Propositions then possess sense independently of whether the situation
depicted is actual or not (Wittgenstein 1961: 15). However, they must
share with the situation the possibility, which is actualized when the
proposition is true, and it is not actualized when the proposition is false.
This possibility shared by a proposition and a situation (in bipolar spec-
trum) could be regarded as a form.3 A proposition must imply this possi-
3
Wittgenstein offered an account of the sense of a picture as the possibilities for
the existence or non-existence of states of affaires. The term ‘possibility’ gives
rise to some interpretative problems. Eric Stenius (1960/1996: 29-36), for in-
stance, claims that the whole (logical) space of possible states of affairs should
be regarded as positive possibilia. On the contrary, Max Black (1964: 38-46) sus-
tains that the only state of affairs are actual ones, and the meaningfulness of false
propositions is granted by the doctrine of bipolarity of propositions, not by posi-
tive possibilia. That is, the sense of propositions is the same state of affairs that
makes their negation true and makes them false. The interpretation of this paper
bility literally, not only metaphorically (Wittgenstein 1922/1974: 16-17,

18-19; Black 1964: 78-79). The form is then (in analogy to a picture) the
possibility (as a disposition) of a certain configuration of things that
could be depicted. It makes also comprehensible the logical isomor-
phism between the combination of elements in the propositional sign and
the possible combination of objects in the relevant, depicted situation.
As Glock notices,
[r]epresentation is possible through a logical isomorphism, an agreement in

form between what represents – whether it be diorama, painting, musical
score, proposition or thought – and what is represented (this recalls Aristo-
tle’s idea that in thought the mind and its object, although made of different
matter, take on the same form. (Glock 1996: 300)
5. Conclusion
To conclude, I would like to emphasise the following essential points of

Wittgenstein’s position on the sense of a false proposition. Wittgen-
steinian theory of propositions rests on the idea of picturing in general.
According to him, the sense of a proposition is ‘what it represents’, i.e. a
possible state of affairs – an arrangement of objects or elements that may
or may not obtain according to whether the proposition in question is
true or false. Proposition, then, is essentially bipolar: it is capable of be-
ing true or false. Using the modern truth-conditional semantics we can
express this in the following way: the sense of a truth-function of a
proposition ‘p’ is a function of the sense of ‘p’, whereas negation just
reverses the sense of a proposition in question. Propositions ‘p’ and ‘~p’
have opposite sense, although one and the same reality corresponds to
them.
The proposition demonstrates (or, in the Wittgensteinian technical
term, “shows”) its sense, that is, “how things stand if it is true. And it
says that they do so stand” (Wittgenstein 1922/1974: 40-41). The sense
is a possibility – a potential combination of objects or names in a propo-
provides arguments in favour of Black’s exegetical position (see also Anscombe

1959/1971: 30, 51-63; Thomson 1966: 217-231).
sitional sign, which need not be obtained or realized. In other words, a

proposition can be false, and yet, have sense (without mirroring the ac-
tual, non-existing state of affairs) only because, even if no state of affairs
corresponds to it as a whole, it still consists of elements that are correlat-
ed with elements of reality (at least by means of possible syntactical ar-
rangement of names). Propositions ‘p’ and ‘~p’ refer to the same reality,
that is to the same elements of the reality in virtue of the meaning of
names as their components and, more technically, arguments. Truth and
falsehood of the propositions are functions of these arguments, that is,
possible arrangements of these components. In that sense, we can then
say that ‘~p’ assumes ‘p’, and the reality that corresponds to ‘p’. Nega-
tion indicates to us whether the state of affairs obtains or not, but if it
does not obtain, the proposition in question is still meaningful because it
shows the possible arrangement of the reverse state of affairs than that
asserted.
References
Anscombe, G.E.M. 1959/1971. An Introduction to Wittgenstein’s Tractatus. South

Bend, Indiana: St. Augustine’s Press.
Beaney, Michel (ed.) 2000. The Frege Reader. Oxford: Blackwell.
Black, Max 1964. A Companion to Wittgeinstein’s Tractatus. Cambridge: Cam-
Chisholm, Roderick (ed.) 1960. Realism and the Background of Phenomenology.
Glencoe: The Free Press.
Copi Irving M. and Robert W. Beard (eds.) 1966. Essays on Wittgenstein’s Tracta-
tus. London: Routledge and Kegan Paul.
Frege, Gottlob 1892/2000. On Sinn and Bedeutung. Transl. from German by M.
Black. In: M. Beaney (ed.), 151-171.
Glock, Hans-Johann 1996. A Wittgenstein Dictionary. Oxford: Blackwell.
Hacker, Peter M. S. 1972/1986. Insight and Illusion: Themes in the Philosophy of
Wittgenstein. Oxford: Clarendon Press.
Kenny, Anthony 1973/1975. Wittgenstein. Harmondsworth, Middleesex: The Pen-
guin Press.
Keyt, David 1966. Wittgenstein’s Picture Theory of Language. In: I. M. Copi and
R. W. Beard (eds.), 377-392.
Meinong, Alexius 1904/1960. The Theory of Objects. In: R. Chisholm (ed.), 76-
117.
Pears, David 1987. The False Prison: A Study of the Development of Wittgenstein’s
Philosophy. Vol. 1. Oxford: Clarendon Press.
Ricketts, Thomas 1996. Pictures, Logic, and the Limits of Sense in Wittgenstein’s
Tractatus. In: H. Sluga and D. G. Stern (eds.), 59-99.
Russell, Bertrand 1907. Review of A. Meinong’s Über die Stellung der Gegen-
standstheorie und Psychologie. Mind 16, 436-439.
Russell, Bertrand 1912/1991. The Problems of Philosophy. Oxford and New York:
Sluga Hans and David G. Stern (eds.) 1996. The Cambridge Companion to Wittgen-
stein. Cambridge: Cambridge University Press.
Stenius, Erik 1960/1996. Wittgenstein’s Tractatus. Bristol: Thoemmes Press.
Thomson, Judith J. 1966. Professor Stenius on the ‘Tractatus’. In: I. M. Copi and R.
W. Beard (eds.). Essays on Wittgenstein’s Tractatus. London: Routledge and
Kegan Paul, 217-229.
Wittgenstein, Ludwig 1922/1974. Tractatus Logico-Philosophicus. Trans. from
German by D. F. Pears and B. F. McGuinness. London: Routledge and Kegan
Paul.
Wittgenstein, Ludwig 1958. The Blue and Brown Books. Oxford: Blackwell.
Wittgenstein, Ludwig 1961. Notebooks 1914-1916 (with Notes on Logic). Ed. by G.
H. von Wright and G. E. M. Anscombe, trans. G. E. M. Anscombe. Oxford:
Blackwell.
Mieszko Tałasiewicz
University of Warsaw
[email protected]
Categorial Grammar and the

Foundations of the Philosophy of
Language
Abstract: The aim of this paper is to present a new approach to explaining produc-
tivity of language, a feature that is crucial for constructing a credible logic of natural
language and elucidating many key issues in the philosophy of language. The start-
ing point of the proposed approach is a combination of Fregean idea of functoriality
and the idea of bi-modal intentionality (taking some hints and observations from
early Husserl, but not restricted to his views). These two ideas are dealt with in a
way inspired by Strawson, notably by his idea that categories are roles rather than
kinds of expressions, and that the logical syntax of language is to be founded in
some transcendental features of our thinking about the world. As a result, our ap-
proach reveals philosophical foundations of Categorial Grammar (far deeper than
Ajdukiewicz ever explicitly acknowledged) and shows, among other things, that
Categorial Grammar should not be considered as more or less accurate description
of acceptability judgments that constitute the empirical base of linguistics, but ra-
ther as a calculus of intentional structure of human cognition.
Keywords: Strawson, Frege, Husserl, Ajdukiewicz, Categorial Grammar, subject,

predicate, intentionality, functoriality, unsaturatedness
0. Introduction
The core problem of syntax is to explain the unified meaning of com-

pound expressions: what it is and how it can be obtained (productivity of
language). There are three distinct aspects of the question, three tasks:
270 Mieszko Tałasiewicz
• How to distinguish well-formed expressions from ill-formed ones

• How to obtain compound meanings from atomic meanings (com-
positionality)
• How to distinguish a unified proposition from a mere list of terms
Note, that discriminating well-formed and ill-formed expressions is one

thing, and understanding compound expressions is another. Accordingly,
calculating some compound meaning is one thing, and having this mean-
ing unified and distinct from a mere concatenation, is another.
“A problem of syntax” is deliberately ambiguous here: syntax is either
a part of language, or a theory of this part of language. Accordingly
there are two ways of understanding what the problem of syntax is: we
may speak of “a problem” for language: to develop – in the process of
evolution – appropriate mechanisms enabling the users to deal with the
abovementioned tasks; and we may speak of a problem for theoretician:
to explain how these tasks are accomplished. Both kinds of unders-
tanding are relevant here. Actually, identifying a mechanism working in
a language is a way of explaining linguistic phenomena. Admittedly,
mere identification of mechanisms is not a very deep explanation: a
deeper explanation of phenomena would require some elaboration – not
mere identification – of the mechanisms themselves: their evolutionary
origin, their cognitive background etc. Yet it is an explanation – as long
as it classifies and unifies phenomena and allows one to answer many
questions about details – just as phenomenal laws of physics (like Boy-
le’s Law) were genuine explanations of facts long before they them-
selves became explained by general theories (like Quantum Mechanics).
The aim of the paper is to propose a new approach to the Core Prob-
lem of Syntax or – as it would be more appropriate to say – to adapt a
bunch of old ideas and present them as a unified and promising frame-
work. Before we start, I must acknowledge the inspiration I draw for my
approach from the syntactic work of Peter F. Strawson, which is now far
less popular than it deserves. In the numerous books and papers pub-
lished in the period of a quarter of a century, Strawson attempted to
solve the problem in all three aspects (under the heading of subject-
predicate distinction). I believe that ultimately he has failed (I must
postpone the detailed justification of this claim for another occasion) and
Categorial Grammar and the Foundations of the Philosophy of Language 271
I will not follow his path. But he cleared the field in an unprecedented
way and made two very helpful observations, which I will heavily rely
upon in my attempt.
The first observation is that subject and predicate are not kinds of ex-
pressions but rather syntactic roles (Strawson 1950/1971: 15; 1974: 3).
At first Strawson distinguished in the function of language two main
tasks: that of forestalling the question ‘What are you talking about?’ and
that of forestalling the question ‘What are you saying about it?’ The first
task is to refer, identify or mention. The second is the task of attribution,
description, classification, ascription (1950: 13). Accordingly, we have
two main roles or, as Strawson would call them, uses: the uniquely re-
ferring use and the ascriptive use (1950: 14). In later writings Strawson
would eventually come to maintaining that there are three primary roles.
The third one is the role of establishing a propositional unity within a
sentence (1974: 17).
The second observation is that explanation for the distinction of sub-
ject and predicate (and the distinctions between syntactic roles in gen-
eral) must be in a sense transcendental:
We assume that the subject-predicate duality […] reflects some fundamen-

tal features of our thought about the world. Strawson (1974: 11)
We are dealing here with something that conditions our whole way of talk-
ing and thinking, and it is for this reason that we feel it to be non-contin-
gent. Strawson (1959: 29)
I use the word ‘transcendental’ in the sense defined by the cited phrases;
it is loosely connected with Kantian meaning (Strawson was the author
of The Bounds of Sense); but I am not committing myself to all nuances
historians of philosophy may wish to associate with the term.
What I am going to do is to establish the three main syntactic roles on
a transcendental foundations and derive the rest of the syntactic system
of the language from these grounds. The methodology is similar to
Strawson’s: we are looking for some fundamental features of our think-
ing about the world. Different are the starting intuitions: in the place of
the ontological distinctions from Individuals I shall take and raise to the
rank of our transcendental foundations the two ideas presented in the fol-
lowing section.
1. Two foundational ideas
1.1. Functoriality
The first idea is the well-known Fregean Principle of Functoriality. The

principle states an intuitive condition under which any compound ex-
pression can be a meaningful whole rather than just a heterogeneous sum
of its parts (‘list of terms’). To achieve this, the compound must consist
of an ‘unsaturated’ part and of another part which can serve as an argu-
ment of the first one and in so doing ‘saturate’ it:1
For not all the parts of a thought can be complete; at least one must be ‘un-
saturated’, or predicative; otherwise they would not hold together (Frege
1892a: 54).
1
The idea of functoriality works at the semantic level as well: the ‘unsaturated’
part denotes a function – an unsaturated entity itself – while the part which ‘satu-
rates’ it denotes some object which the function takes as its argument. A uniform
denotation of the whole compound expression is the value of the function.
Thanks to this we are able to understand why the denotation of a compound ex-
pression, say ‘Socrates’ father’, is not a sum (or product) of denotations of the
constituent expressions (‘father’ and ‘Socrates’) but a completely different object
(here Sophroniscus). This semantic aspect of functoriality is sometimes referred
to as ‘compositionality’: the denotation of a compound is a function, denoted by
the functor, of the denotations of the arguments of the functor. For reasons fully
explained in Tałasiewicz (2010), ‘functoriality’ should be preferred to ‘composi-
tionality’ in the present context. Compositionality nowadays has some air of an
epistemological thesis, saying that when we analyze meanings we must first have
access to atomic constituents and only then we can obtain the compound expres-
sion as their function. Functoriality, on the other hand, is a purely formal notion:
any compound expression is a function of its atomic constituents, and by the
same token any atomic constituent is a (reversed) function of the compound and
other constituents. Which is first given for analysis, is not decided formally, it is
a matter of fact. Sometimes it may be the case that it is the meaning of some
compound expression what is given first, and the meaning of some of its consti-
tuents is only derived. Thus, functoriality is fully consistent with both composi-
tionality and contextuality (and it is not accidental that Frege endorsed simulta-
neously the two principles). For analogous use of ‘functoriality’ see e.g. Simons
(1981: 88).
Statements in general […] can be imagined to be split up into two parts; one
complete in itself, and the other in need of supplementation, or ‘unsatura-
ted’ (Frege 1891: 31).
Strawson recognized this idea but found it, rightly, insufficient. ‘Unsatu-
ratedness’ by itself is hardly intelligible: as Strawson stressed (after
Ramsey) both subject and predicate are incomplete or unsaturated as
parts of a sentence, on the level of syntax. The semantic level of the no-
tion of ‘unsaturatedness’ works no better, either. According to Frege,
unsaturated expressions denote unsaturated objects, that is functions. But
since everything can be predicated of, functions notwithstanding, ex-
pressions referring to unsaturated objects may be subjects as well as
predicates. Our criterion cannot be built up with these bricks.
It turns out however that the subject-predicate problem can be solved
by supplementing Frege’s idea with another one, namely that of lan-
guage being founded in two kinds of intentional acts. I take this idea as
stemming from Edmund Husserl’s Logical Investigations, particularly
Investigation V (Husserl 2001), but the whole conception I set out here
is not fully Husserlian.
1.2. Bi-modal intentionality
I assume after Husserl that expressions have meanings because they are
founded in intentional acts, whereby they appear as being directed at
something:
The concrete phenomenon of the sense-informed expression breaks up, on

the one hand, into the physical phenomenon forming the physical side of the
expression, and, on the other hand, into the acts which give it meaning […]
In virtue of such acts, the expression is more than merely sounded word.
It means something, and in so far as it means something, it relates to what is
objective. (Husserl 2001, Vol. 1: 191-192)
There are two kinds of intentional acts: nominal acts and propositional
acts. In language, they correspond to names and sentences. It is not
enough to say that our thinking or its verbal expression is intentional or
directed at something, for our thinking and its verbal expression can be
directed at it in two different ways:
Nominal acts and complete judgements never can have the same intentional
essence, and […] every switch from one function to the other, though pre-
serving communities, necessarily works changes in this essence. (Husserl
2001, Vol. 2: 152)
Naming and asserting do not merely differ grammatically, but ‘in essence’,
which means that the acts which confer or fulfil meaning for each, differ in
intentional essence, and therefore in act-species. (Husserl 2001, Vol. 2:
158)
According to Husserl this is a primary, independent differentiation of

intentionality, not based on any other differentiation; we can say that our
intentional attitude is simply the way it is: when we reflect on our think-
ing and speaking, it presents itself to us as being directed at something in
either one way or the other, as mentioning something or stating some-
thing about it. The dichotomy of names and sentences as two fundamen-
tal, different and non-interchangeable basic roles of expressions, that of
expressing a nominal act and that of expressing a propositional act, is
thus fully justified as grounded in this primary foundational distinction.
The status of the distinction between nominal and propositional acts is
comparable to the intended status of Strawson’s distinction between par-
ticular and concept and thus it may be called transcendental in the
abovementioned loose sense: we might regard it as a further inexplicable
– perhaps metaphysically contingent but epistemologically necessary–
fact about the way in which people think about the world.
The question of what is the profound nature of intentional acts, and
how, exactly, meaning is constituted in such acts, needs not concern us
here as long as we keep observing some basic principles of intentional
account of meaning.2 What deserves our attention in the first place is the
fact that names-sentences distinction, being based upon distinct kinds of
intentional acts, has nothing to do with content (meaning) or denotation
(semantic correlates) of these expressions. We can speak about the very
2
Although surely it is a very interesting question. A rough sketch how to deal with
this matter – in a quite naturalistic way, far from Husserlian origins – is given in
Tałasiewicz (2012).
same objects using names and using sentences (concrete examples de-
pend on many further assumptions; let us consider for the sake of rela-
tively uncontroversial illustration a following one: ‘The cat is on the
mat’, which is a sentence, is correlated with exactly the same part of re-
ality as ‘The fact that the cat is on the mat’, which is a name). It is the
mode of intentionality, not the object intended, that counts. Being a
name depends on the role in which the expression is used (the role of
expressing a nominal intentional act) rather than on the kind of expres-
sion or the kind of its designate. Names include proper names, nominal
phrases with or without articles, definite/indefinite descriptions, pro-
nouns, demonstratives, nominalized sentences (in fact, nominalized eve-
rything), and so on.3
In particular, names need not have designates at all. An intentional act,
therefore an expression, can be directed at empty space. Expressions ap-
pear as directed at something, but not necessarily are so directed. Hus-
serl lays great emphasis on this:
Relation to an actually given objective correlate, which fulfils the meaning-

intention, is not essential to an expression. (Husserl 2001, Vol. 1: 199)
Intentional experiences have the peculiarity of directing themselves in vary-
ing fashion to presented objects, but they do so in an intentional sense. An
object is ‘referred to’ or ‘aimed at’ in them […]. This means no more than
that certain experiences are present, intentional in character […]. There are
(to ignore certain exceptions) not two things present in experience, we do
not experience the object and beside it the intentional experience directed
upon it. […] only one thing is present, the intentional experience, whose es-
sential descriptive character is the intention in question […] If this experi-
ence is present, then, eo ipso and through its own essence […], the inten-
tional ‘relation’ to an object is achieved, and an object is ‘intentionally pre-
sent’ […]. And of course such an experience may be present in conscious-
ness together with its intention, although its object does not exist at all, and
is perhaps incapable of existence. The object is ‘meant’, i.e. to ‘mean’ it is
an experience, but it is then merely entertained in thought, and is nothing in
reality.
3
This view is consistent with Geach’s approach whereby, for example, general
terms can be used as predicates and names, and that these are different uses
(Geach 1980a, section 34).
If I have an idea of the god Jupiter, […] this means that I have a certain pre-
sentative experience, the presentation-of-the-god-Jupiter is realized in my
consciousness. This intentional experience may be dismembered as one
chooses in descriptive analysis, but the god Jupiter naturally will not be
found in it. The ‘immanent’, ‘mental object’ is not therefore part of the de-
scriptive or real make-up of experience, it is in truth not really immanent or
mental. But it also does not exist extramentally, it does not exist at all. This
does not prevent our-idea-of-the-god-Jupiter from being actual, a particular
sort of experience or particular mode of mindedness (Zumutesein), such that
he who experiences it may rightly say that the mythical king of the gods is
present to him, concerning whom there are such and such stories. If, how-
ever, the intended object exists, nothing becomes phenomenologically dif-
ferent. It makes no essential difference to an abject presented and given to
consciousness whether it exists, or is fictitious, or is perhaps completely ab-
surd. I think of Jupiter as I think of Bismarck, of the tower of Babel as I
think of Cologne Cathedral, of a regular thousand-sided polygon as of regu-
lar thousand-faced solid. (Husserl 2001, Vol. 2: 98-99)
1.3. Frege and Husserl: cooperation not competition
The idea of bi-modal intentionality has a high explanatory potential by

itself. It is only after we have distinguished two kinds of intentional acts
that we can explain the concept of a sentence. The problem which arises
here attracts surprisingly little attention: Strawson posed the question
about what the subject and the predicate are in a sentence, treating the
sentence itself as something quite obvious. He shares such a light-
hearted attitude towards the notion of sentence with many of his critics,
notably Bob Hale, who in (1979: 283) says: “The notion of sentence
may reasonably be taken for granted…”. Yet in fact there is nothing ob-
vious about it, especially if we reject Frege’s original conception that a
sentence is an expression denoting one of the two truth-values (while
Strawson, for example, and every exponent of the truth-value-gap theory
rejects the necessary link between a sentence and its truth value). The
notion of assertion, which perhaps may be thought of as sufficiently
strong to replace truth-values in the role of an indicator of sentence-
hood, is no good, either. Sentences may be correlated with both positing
and non-positing acts: assertions as well as mere suppositions. It is clear
that we can utter a grammatical sentence without a hint of an assertion

(and make it so that it does not have any truth-value).4 How can we dis-
tinguish a sentence from a non-sentence then? The popular claim that a
sentence is an expression that may have a truth-value hardly explains
anything. Instantly it begs the question: what makes some expressions
enabled to have a truth-value and others not? Bi-modal intentionality
provides a good and simple answer to this question: a sentence is an ex-
pression that expresses a propositional intentional act.
Furthermore bi-modal intentionality explains the otherwise rather ob-
scure notion of ‘unsaturatedness’. Interestingly, Husserl adopted some
kind of saturatedness-unsaturatedness distinction himself. In particular,
he distinguished independent from non-independent expressions, as well
as complete from incomplete ones. Complete expression is a kind of ex-
pression which is syntactically coherent, with a unitary meaning, for ex-
ample, ‘a cat’, ‘The cat sits on the mat’, or ‘quite good’. Incomplete ex-
pression is an expression that lacks this internal syntactic coherence: ‘the
on cat quite’. This, so far, has nothing to do with Frege’s saturated/ un-
saturated expressions (despite the fact that Frege himself sometimes
called saturatedness-unsaturatedness ‘completeness-incompleteness’).
However, Fregean saturatedness-unsaturatedness in turn has quite a lot
in common with Husserl’s differentiation between independent and non-
independent expressions. Independent expressions are e.g. ‘a horse’,
‘I’ve seen a ghost’, ‘a green cow’, while non-independent expressions
such as ‘is green’ or ‘quite good’, are functors which have unitary but
unsaturated meanings and therefore require objects:
Several non-independent meanings […] can be […] associated in relatively

closed units, which yet manifest, as wholes, a character of non-indepen-
dence. This fact of complex non-independent meanings is grammatically
registered in the relatively closed unity of complex syncategorematic ex-
pressions. Each of these is a single expression, because expressive of a sin-
gle meaning, and it is a complex expression, because expressive part by part
of a complex meaning. It is in relation to this meaning that it is a complete
expression. If nonetheless we call it incomplete, this depends on the fact
that its meaning, despite its unity, is in need of completion. Since it can on-
4
For a forceful arguments against assertion as a sentence-hood indicator, see
Geach (1972).
ly exist in a wider semantic context, its linguistic expression likewise points

to a wider linguistic context, to a completion in speech that shall be inde-
pendent and closed. (Husserl 2001, Vol.2: 57)
Thus the explanation of unsaturatedness, available to Husserl but not to

Frege, is the following:
An expression is saturated when it expresses a complete intentional

act: propositional or nominal (thus we have two kinds of saturated
expressions: names and sentences). Other expressions are unsatu-
rated.5
It is a very simple explanation, which very easily side-steps Ramsey’s

charge: indeed both subject and predicate fall short of being complete
sentences, but a name in the subject position expresses complete inten-
tional act (nominal), whereas predicate does not express any complete
intentional act, neither nominal nor propositional.6
This explanation accounts in turn for Frege’s ‘horse concept paradox’.
Frege, who took pains to distinguish the object (as a correlate of a satu-
rated name) from the concept (as a correlate of an unsaturated predicate),
ran into considerable difficulties trying to express his conception. Dum-
mett (1993: 55) calls these difficulties a paradox, which sounds like a
rather serious description, but even Frege himself felt the situation was
somewhat awkward:
It must indeed be recognized that here we are confronted by an awkward-

ness of language, which I admit cannot be avoided, if we say that the con-
cept horse is not a concept whereas e.g. the city of Berlin is a city and the
volcano Vesuvius is a volcano […]. In logical discussions one quite often
5
Unsaturated expressions do not express incomplete intentional acts but are corre-
lated, in some intermediate way, to certain structures of complete acts from
which their meanings have been abstracted, cf. Tałasiewicz (2010).
6
Compare: “A name has a complete sense, and can stand by itself in a simple act
of naming […]. [A] predicate never has a complete sense, since it does not show
what the predication is about; it is what is left of a proposition when the subject
is removed, and thus essentially contains an empty place […]” (Geach 1980a:
57).
needs to assert something about a concept […]. Consequently, one would

expect that the reference of the grammatical subject would be the concept;
but the concept as such cannot play this part, in view of its predicative na-
ture; it must first be converted into an object, or, speaking more precisely,
represented by an object (Frege 1892a: 46).
I admit that there is a quite peculiar obstacle in the way of an understanding
with my reader. By a kind of necessity of language, my expressions, taken
literally, sometimes miss my thoughts; I mention an object, when what I in-
tend is a concept (Frege 1892a: 54).
In terms of Husserl’s distinction, matters are clear. The saturation or un-

saturation of an expression does not depend on whether the denotation
of the expression is saturated or not (whether it is, ontologically, a par-
ticular or a concept or a function) but on whether it expresses a full in-
tentional act or not. A name is saturated, even if it is only a name of an
ontologically unsaturated object, such as a concept or a function. When
we speak about a concept then, we do not convert it in any way. We
simply name it, instead of predicating it of something. This is the differ-
ence of the logical role, not of the denotation.
Nevertheless, there is something about unsaturatedness that Frege has
got and Husserl has not. Husserl has got unsaturatedness explained,
whereas Frege has got it applied. Unsaturatedness in Frege’s conception
is employed in the important task of explaining compoundability of ex-
pressions, which is crucial for our purpose of establishing the general
principles of logical grammar. Saturatedness-unsaturatedness distinction
is namely a part of the more general notion of functoriality, whereby
(1) any process of compounding must involve the process of saturat-

ing some non-independent (unsaturated) parts (functors) in order to
obtain independent (saturated) wholes;
(2) the compounding process has the form of supplying arguments to

a function indicated by the unsaturated expression.
In Husserl’s account the notion of unsaturatedness appears surprisingly

idle. Husserl’s method of compounding meanings is much weaker than
Frege’s and does not rely on the saturatedness-unsaturatedness distinc-
tion. Where Frege provides one cardinal rule for compoundability, Hus-
serl does not have any single method for making compounds. He has
saturatedness-unsaturatedness at his disposal but he doesn’t employ this
distinction to characterize compoundability. In his words:
In a purely logical form-theory of meanings […] we must fix the primitive

forms of independent meanings, of complete propositions with their internal
articulations, and the structures contained in such articulations. We must
fix, too, the primitive forms of compounding and modification permitted by
the essence of different categories of possible elements. […] After this, we
must systematically survey a boundless multitude of further forms […].
[For instance] Any two propositions yield, when combined in the form M
and N, another proposition, any two adjectives another adjective […].To
any two propositions, M, N, there belong, likewise, the primitive connective
forms If M then N, M or N, so that the result again is a proposition. To any
nominal meaning S, and any adjectival meaning p, there belongs the primi-
tive form Sp (e.g red house), the result being a new meaning fixed by law in
the category of nominal meaning. We could in this manner give many other
examples of primitive connective forms. (Husserl 2001, Vol. 2: 69)
This is a piecemeal approach, according to which we must turn to the

grammars of particular languages for information on what forms a com-
pound expression may take. This cannot play the role of foundational
principle of logical grammar (for it would be question-begging) and it is
obviously inferior to Fregean functoriality, which is a single fundamen-
tal principle for compounding meanings and may be taken as a transcen-
dental, a priori principle in the Strawsonian sense.7
Thus, the full picture emerges only when we bring together Fregean
functoriality, which explains compoundability in terms of saturatedness-
unsaturatedness distinction, and Husserlian bi-modal intentionality,
which explains saturatedness-unsaturatedness distinction. Husserl and
Frege do not compete over which of them has managed to express the
idea of meaning better, contrarily to what has been suggested in Dum-
7
To a large extent, Husserl’s approach resembles Richard Montague’s grammar
which builds on a large set of grammatical principles given ad hoc, whereas
combined Fregean-Husserlian approach yields a classical Categorial Grammar,
or so we shall argue further.
mett (1993: 56).8 They complement one another, each drawing out a dif-
ferent aspect, and only together do their insights generate a powerful
framework in which we can understand logical grammar of the langu-
age, a framework identifying saturated names and sentences and unsatu-
rated functors as three fundamental syntactic roles (compare Strawson!).
2. A unified approach. Subject and predicate revisited
Now we have the resources to explain how the logical structure of a sen-
tence is to be built. The basic syntactical rule for all kinds of sentences,
or even all and any compound saturated expressions, follows:
Basic Syntactic Rule (BSR)

Every saturated compound expression (i.e. a name or a sentence)
parses into exactly one n-place unsaturated functor and n of its satu-
rated arguments (if any argument happens to be a compound expres-
sion itself, the rule applies accordingly).
In atomic sentences the functor is an n-place predicate, while its argu-

ments are n names (n ≥ 1).
(BSR) is a more fundamental insight into the logical structure of lin-
guistic expressions than mere subject-predicate distinction. The latter we
can get from the former:
A predicate is a(ny) sentence-forming functor taking nominal argu-

ments.
A subject is relativized to a sentence. In a sentence in which the func-
tor is one-place predicate, the subject is the name that is the sole ar-
gument of this functor.
With respect to sentences in which the predicate takes more than one
name, the notion of the subject needs some more elaboration.
8
For similar observation that Fregean conception of language achieves its full
ripeness only in combination with the insights of Husserl, see Smith (1994).
Let us start the labour with a short summary of what we know about
sentences, names, and functors. What we need to repeat and highlight is
that these are structural roles that different expressions may play in re-
flecting our intentional acts. We may call sentences, names and functors
‘categories’ for short, but only if we keep in mind that they are not kinds
of expressions – they are roles. Perhaps expressions of one kind are bet-
ter suited to play certain roles than expressions of another kind, but that
is a different story. The very same expression, say ‘yellow’, can play the
role of a name, as in ‘yellow is a bright colour’, or the role of a functor,
as in ‘a yellow car hit the wall’.
The most important consequence of this is that the structure of a sen-
tence reflects the structure of an intentional act. This should be con-
trasted with a common view that the structure of a sentence reflects the
structure of a corresponding part of the world (a situation). I am not say-
ing that the structure of a sentence has completely nothing in common
with the structure of a situation; but I do say that it might be so only in
an intermediate way. Immediately the structure of a sentence depends
solely on intentional structure of a thought. It is a reasonable assumption
though that the structure of a thought might be induced somehow by the
structure of a situation we are thinking about – but, again, that is a dif-
ferent story (the way in which situations induce structures in thoughts is
not exactly a straightforward one, I’m afraid).
The structure of intentional acts rests on two pillars: the choice of
names and the arrangement of names. The choice of names reflects how
we frame our presentations; it shows how we parse reality in order to
refer to it. The arrangement of names reflects the cognitive structure in
which we place the presentations. Compare:
(1) Mary | loves | John

(2) John | loves | Mary
(3) Mary | loves John
Our cognitive structure expressed in (1) is such that we present to our-

selves two objects and declare them to be in some binary relation. The
cognitive structure in (2) is such that we present to ourselves the same
objects as in (1) and declare them to be in the same binary relation, but
in reversed order. The cognitive structure reflected in (3) is such that we

present to ourselves just one object – ‘John’ is not a name at this level! –
and ascribe to it some property (unary relation). In further refinements of
the structure we may notice that this property is nothing else than the re-
lation mentioned in (1) with the second place saturated with the second
object mentioned in (1). In such case the second name will reappear, but
at another level of analysis:
(3’) Mary | loves John

loves | John
It is worth noting that eventually the sentence ‘Mary loves John’ is struc-
turally ambiguous as it may reflect two different cognitive structures (1)
and (3). Therefore in the logical grammar we not only need categories of
sentences, names and functors, but also syntactic positions of parts of the
sentences: the position of the first argument of the functor, the position
of the second argument of the functor, the position of the third argument
of the functor, and so on. ‘Mary’ has in (1) different position than in (2)
although it belongs to the same category – in both cases it is a name.
Moreover, if any part is still a compound expression, we need another
level of analysis. Because all levels of analysis are governed by the same
(BSR), each level contains exactly one functor. There might be in fact
many functors in the whole expression, at different levels and in differ-
ent positions. In (3’) the functor ‘loves’ has different position than in (1)
or (2) although it belongs to the same category – in all cases it is a two-
place predicate. To keep trace of this, it is convenient to introduce a spe-
cific term of ‘operator position’ or ‘operator’ for short: the operator posi-
tion in an expression is the syntactic position that is occupied by the
constituent of the expression that is actually the functor precisely in this
expression (not in any part of it). Thus ‘loves’ is in the operator position
(or, for short, just is the operator) in (1) and (2) but not in (3).9
9
Precisely to this end of discriminating between categories and syntactic positions
Geach has introduced the distinction of predicate and predicable (Geach 1980a:
50). His predicable is our predicate in general; his predicate is our ‘predicate in
the expression’, or the operator. However, we will retain our terminology (bor-
Now we may turn back to the question of the subject. We’ve said that
in a sentence with one-place predicate the subject is the name (there is
just one name in such a sentence). In sentences with many names we
may attach the notion of the subject to some highlighted syntactic posi-
tion that may be taken by a name, precisely as in Strawson’s ‘The whale
struck the ship’, where both ‘the whale’ and ‘the ship’ are names but on-
ly ‘the whale’ is the subject. The subject may be highlighted logically –
usually it is the first argument of the predicate – but occasionally it may
be highlighted in some different way, e.g., pragmatically.10
Perhaps for a better understanding of the relation between types of ex-
pressions, logical roles and syntactical positions some visualization
would be helpful. Consider the task of cleaning the lawn of dead leaves.
It consists in executing the following roles: raking the leaves, gathering
raked leaves in a container, transporting the container with the leaves to
the compost-hole, emptying the container, transporting the container
back to the lawn, further raking the leaves, gathering leaves etc. until the
whole lawn is neat. These roles can be played by objects of various
kinds, some of which are better suited to play a given role than others.
For instance, raking the leaves is best performed with a rake. But house
broom would do as well, if necessary. Or one’s own hands. Thus ‘a rake’
rowed from Ajdukiewicz 1978), as the concept of operator is more general: it

works not only for predicates, but for all functors.
10
This complication with the notion of subject justifies to some extent one of
Geach’s charges against Strawson. Namely, Geach deplores Strawson’s use of
the term ‘name’ on a par with ‘subject’, by saying: ‘name’ ought to mean a logi-
cal category, ‘logical subject’ ought to mean a term’s role in a given proposition’
(Geach 1980b: 175). We see, that in a sense, Geach is right here: ‘subject’ does
not stand for a category of name, but stands for a specific syntactic position
available to names. However, we can see also that this is not a fundamental
charge: names (as well as all logical categories) are roles, and in sentences with
one-place predicate the subject is the (only) name. It’s just that roles have some
arrangement in a sentence that needs to be accounted for, and in sentences with
many names the notion of subject is better suited to indicate rather the arrange-
ment than just the nominal role. Subsequently, it is no longer so fundamental; we
may choose relatively freely which name we want to honor with this term (subj-
ect’s being the first argument of the predicate is a common and useful conven-
tion, but there is nothing logically compelling about it).
may stand for a kind of object designed to rake, but also it might stand
for a role in the process (which can be performed by virtually anything
that is physically capable of moving a leaf). And now, even if we are
talking about roles, not objects, we must properly arrange the execution
of these roles in order to accomplish the task, no matter what objects ac-
tually play these roles. If we first rake, then gather leaves in a container,
then empty the container, then transport the container to the compost-
hole, then transport the container back to the lawn and so on, the lawn
will never be neat. The distinction between roles and their arrangement
in performing a complex task should not blur the distinction between
roles and objects that can play them.
3. Towards Categorial Grammar
We have arrived at a point where we are in possession of a very simple

yet powerful framework capable of accounting for many facets of logical
grammar of language and describing the logical structure of highly com-
plicated expressions. We have seen that Frege’s Functoriality Principle,
just as it is, by mere cooperation with Husserl’s bi-modal intentionality,
generates a wide range of available structures. With bi-modal intentio-
nality at hand we can distinguish name-forming and sentence-forming
functors and cross this distinction with name-taking functors and sen-
tence-taking functors. We get sentential connectives (sentence-forming-
sentence-taking), predicates (sentence-forming-name-taking), reificators
or nominalizers (name-forming-sentence-taking) and logical adjectives
(name-forming-name-taking).
We can still enrich this framework if we take seriously Husserl’s in-
sight that unsaturated expressions may be compound themselves (so far
we were considering cases in which saturated parts of some compound
expression could be compound themselves). Guided by this insight we
may generalize the notion of functoriality:
Any compound unsaturated expression parses either according to

(BSR) or into one ‘more’ unsaturated expression and some ‘less’ un-
saturated expressions in such a way that the less unsaturated expres-
sions are the arguments of the more unsaturated expression and the
whole compound (unsaturated) expression is the value of such func-
tion.
In this way we get many other kinds of functors, such as functor-forming

functors, among which logical adverbs occupy a prominent place, and
eventually in case of some more complex sentences we need to handle a
fairly complicated arrangement of categories and positions at once.
At this point we may feel a need for some formal calculus. Let us have
the types (n) and (s) for names and sentences respectively and let us
have a fractional type (x/y1…yk) for an x-type-forming functor taking k
y-type-arguments (for the sake of simplicity we ignore functors that have
heterogeneous arguments; the complication would be only notational,
though). For instance, a two-place predicate may be noted as (s/nn), a
logical adverb over two-place predicate as ((s/nn)/(s/nn)), an adjective as
(n/n), a two-place connective as (s/ss) and so on. Syntactic positions may
be encoded to some extent in the graphical form of analysis.
(4) Polluted water affects heavily public health

| (n) | (s/nn)* | (n) |
| (n/n)* | (n) ||(s/nn)|((s/nn)/(s/nn))* || (n/n)* | (n) |
Formula (4), where (*) indicates the operator, reads that the sentence
‘Polluted water affects heavily public health’ parses into unsaturated
two-place predicate ‘affects heavily’ and two names: ‘polluted water’ as
the first argument of the predicate and ‘public health’ as the second ar-
gument; furthermore, the name ‘polluted water’ parses into unsaturated
one-place adjective ‘polluted’ and a name ‘water’ as its first and only
argument; the predicate ‘affects heavily’ parses into <more> unsaturated
functor ‘heavily’ and <less> unsaturated predicate ‘affects’ as its argu-
ment; the name ‘public health’ parses into unsaturated one-place adjec-
tive ‘public’ and a name ‘health’ as its first and only argument. There are
of course different notational conventions, better suited for complicated
cases but less intuitive for not-accustomed readers, so we will stick to
the above. Perhaps we may introduce only a more formal way of encod-
ing syntactic positions (as the graphical form sometimes may seem am-
biguous). Let us namely write the syntactic position of an expression in

form of a sequence of numerals such that to the whole analyzandum we
assign the sequence (1), to the operator on the first level of analysis
(1,0), to the first argument of the operator (1,1), to the second one (1,2),
to the third one (1,3) and so on. If anything of it is still a compound ex-
pression and has assigned the sequence (n,m), we assign (n,m,0) to the
operator in it, (n,m,1) to the first argument of the operator, (n,m,2) to the
second argument and so forth. In the above example it will yield (1,0)
for ‘affects heavily’, (1,1) for ‘polluted water’, (1,2) for ‘public health’,
(1,0,0) for ‘heavily’, (1,0,1) for ‘affects’, (1,1,0) for ‘polluted’, (1,1,1)
for ‘water’, (1,2,0) for ‘public’ and (1,2,1) for ‘health’.11
The functoriality principle may be now rephrased in the form of the so
called ‘multiplying out’ rule of functional application:
(MP) (x/y) y → x
This rule, if applied backwards, gives us some constraints for parsing

sentences and other compound expressions, although it usually does not
determine any particular parsing, since – as we have seen – sentences
usually are structurally ambiguous (because the same surface sentence
may express different structures of intentional acts). When applied for-
wards, the rule allows for unambiguous restoring of original expression
from its parsed elements. ‘Heavily’ ((s/nn)/(s/nn)) takes ‘affects’ (s/nn)
and yields ‘affects heavily’ (s/nn). ‘Polluted’ (n/n) takes ‘water’ (n) and
yields ‘polluted water’ (n). ‘Public’ (n/n) takes ‘health’ (n) and yields
‘public health’ (n). ‘Affects heavily’ (s/nn) takes ‘polluted water’ (n) and
11
For details, e.g. formal recursive definition, see Ajdukiewicz (1978). Still another
way of encoding syntactic positions is the so called Polish notation: the functor
goes first, followed by its arguments, from the first of them to the last one. With-
in compound constituents the same rule applies accordingly (cf. Ajdukiewicz
1935/1967). An example in Polish notation would look like this: ‘heavily affects
polluted water public health’. Apart from its being quite hard to visualize, Polish
notation is ambiguous if (Composition), see below, is adopted. A string (s/s) (s/n)
n can be parsed either as ((s/s) (s/n)) n or as (s/s) ((s/n) n). There are still more
options. Steedman (2000: 34) prefers for instance graphical encoding with left/
right arrows to encode the direction of functional application.
‘public health’ (n) and yields ‘Polluted water affects heavily public
health’.
We may perhaps inquire whether there are some further plausible and
insightful generalizations of functoriality principle. Let us recall that the
first generalization allowed for functions to be arguments of other func-
tions. At the level of expressions this amounted to the possibility of hav-
ing compound functors. Now we may entertain the idea of the so called
functional composition: instead of applying two functions to some ar-
gument one after another, we might first merge these functions into one
and only then apply the resultant function to the argument.
(Composition) (x/y) (y/z) → (x/z)
It allows for, e.g. a composition of sentential negation with a predicate:
(s/s) (s/n) → (s/n)
which would be syntactically incoherent under (MP) alone (note that

(s/n) is not an argument of (s/s)). Such a possibility is much welcome,
because in terms of it we can account for what was so forcefully ad-
vanced by Geach (1950): the asymmetry of subject and predicate with
respect to negation. Under (Composition) it is nearly self-evident: sen-
tential negation with a predicate gives another predicate, whereas with a
name – (s/s) n – it gives incoherent string. Names sometimes can be ne-
gated, but only through name negation (n/n), which is categorially dif-
ferent from sentence negation.
This is pretty much all we need for logical grammar of language, ex-
cept for a still better formal elaboration. Luckily enough, such a calculus
has already been elaborated in the finest detail and is called Categorial
Grammar (CG). It was initiated in 193512 by Kazimierz Ajdukiewicz (cf.
12
Actually 1935 was the year of the first publication of a full-blown grammatical
calculus. Ajdukiewicz was presenting the core ideas as early as in 1925, at the
meetings of the Polish Philosophical Society. Abstracts were published in the
proceedings as (Ajdukiewicz 1925a, 1925b). It is worth noting that in these lec-
tures Ajdukiewicz acknowledged very clearly the distinction between nominal
and propositional intentional acts and stressed the need to keep this distinction
Ajdukiewicz 1935/1967), who explicitly acknowledged Husserl’s con-

tribution to the idea (Frege’s contribution was perhaps too obvious to be
mentioned).13 Ajdukiewicz developed the calculus up to the (MP) prin-
ciple. In 1970 Geach introduced philosophically important (Compo-
sition) principle (cf. Geach 1970). Further modifications of the calculus
introduced over the years, eventually resulted in a bunch of sophisticated
formal theories which now constitute one of the leading frameworks in
the logical syntax of today.14
However, when we speak about Categorial Grammar in the context of
the present purpose – i.e. the purpose of establishing in a philosophically
compelling way the right criterion for the basic distinctions of logical
grammar – a qualification is needed. Ajdukiewicz conceived his calculus
as a grammar, plausible but rather tool-like device for parsing sentences.
He never insisted explicitly on any particular philosophical importance
of the ideas he took from Frege and Husserl, although it may be argued
that implicitly he treated his grammar as philosophically important. Con-
sequently, the developments of the calculus did not always heeded the
philosophical underpinnings of this calculus.
Now, the point of the present paper is that the joint insights of Frege
and Husserl do reveal the actual (and necessary, in a transcendental
sense) conditions of our thinking about the world, and thus CG is the
logical grammar of natural language – as long as it is faithful to the
foundations.15 CG, as opposed to other syntactic theories, in particular
generative grammar or Montague grammar, is not a generalization or
idealization of a traditional grammar of a particular language like Eng-
separate from the distinction of assertions and mere suppositions. This topic was
never so overtly discussed in his further articles.
13
Ajdukiewicz was concerned only with Logical Investigations. Later works of
Husserl, perhaps decisive for phenomenology, were beyond the scope of his in-
terests, as they are beyond the scope of ours here. For arguments that Husserl’s
early theory from Logical Investigations is a different theory, and a better one,
than his theory developed later, especially in Ideas, see Smith (1994: 177).
14
See e.g. Carpenter (1997); Steedman (2000).
15
From this point of view, the present status of CG as a formal theory does not
seem quite satisfactory. The reasons why it is so and some criticisms of unjusti-
fied developments in CG are presented in Tałasiewicz (2009, 2010).
lish, Polish or Latin. It is not a reservoir of distilled linguistic compe-

tence of native speakers of some language, either. It is a cognitive frame
that shapes the whole way in which we might approach the syntax of any
human language whatsoever – a frame that, admittedly, in many cases
may be hidden deep beneath actual surface realization. Actual grammar
of a given language is a product of interference of many factors, logical
as well as semantic and pragmatic. CG reveals only logical aspect of it;
empirically not-detachable from other aspects. We might say that Cate-
gorial Grammar is a calculus of intentionality, which in turn makes the
latter term less obscure than it is normally taken to be, even if it does not
explain it to the bone.
4. A comparison with Simons’ account
We have been arriving at the formulation of CG step by step from the

initial insights to the full-blown calculus in an annoyingly slow pace in
order to make this transcendental character of CG absolutely clear.
Keeping this character in mind is particularly important for assessing
Peter Simons’ approach to unsaturatedness (cf. Simons 1981).
Simons’ enterprise has arguably narrower scope than ours. He tries to
explain just unsaturatedness, without making use of it in the project of
elucidating the subject-predicate distinction.16 Yet unsaturatedness is a
big part of the project and his approach is superficially very similar to
ours. Simons, namely, explicitly addresses the notion of unsaturatedness
as a place in Fregean conception where one ‘shall look to Husserl for a
fuller theory, which can illuminate some of the obscurity surrounding
Frege’s account’ (Simons 1981:75). Moreover, he acknowledges the
Ramseyan type of charge against it (‘all subsentential expressions fall
short of being complete sentences’, cf. Simons 1981: 85) and, finally,
entertains a type of solution to the problem of unsaturatedness that we
have adopted, namely based upon Husserl’s distinction of complete and
incomplete intentional acts. Eventually, however, he rejects this solution
with a claim that
16
Although he makes use of it in some partial elucidating of the notion of the predi-
cate in the context of the dispute between Dummett (1973) and Geach (1975).
Husserl’s [distinction][…] may be question-begging, since it seems unlikely

that we can classify mental phenomena into the requisite broad classes
without bringing in, by the back door, the grammatical classification of the
expressions apt to put such experiences into words: is not judgment, for ex-
ample, simply analogous to the silent internal assertion of a declarative sen-
tence? (Simons 1981: 85)17
Simons provides solution that he thinks is a different one: he decides to

take ‘grammatical classification of expressions’ as the ultimate expla-
nans and introduces Categorial Grammar as ‘an intrinsically linguistic
account’ of it (1981: 85). With this in mind, he derives saturatedness-
unsaturatedness distinction from the distinction of basic and functor cat-
egories in the technical sense of CG (which is quite obvious at this
point), and claims, that wherever CG “might fail to be adequate to the
linguistic data […] there […] we should also have little idea whether to
describe the expression as complete or incomplete” (1981: 86).
Now, in the light of what we have said above, and in particular in the
light of Strawsonian analysis of language, this manoeuvre seems entirely
futile. Categorial Grammar, if treated as a summary of common gram-
matical intuitions, looses much of its explanatory force and becomes it-
self in need of explanation (just as other superficial grammatical features
considered by Strawson). Besides, CG is far too distant from the empiri-
cal corpus to be such a summary (arguably it is counterintuitive in some
17
Notably, Simons acknowledges also Husserl’s and Frege’s importance for CG:
“Both Frege and Husserl are in a sense intellectual grandfathers of categorial
grammar: Frege by example, Husserl by precept. Frege’s concept-script is effec-
tively built on categorial principles […]. Husserl, on the other hand, first sug-
gested and outlined the general idea of a categorial grammar […]. It is therefore
no accident that the Polish thinkers who first developed categorial grammar were
influenced by both Frege and Husserl” (1981: 86). A comment is due. Indeed
Frege and Husserl are both ancestors of CG, but in a different sense than Simons
depicts it: they are rather like grandfather and grandmother (leaving aside the
question who is who): their intellectual gametes had to copulate in the womb of
Ajdukiewicz’s mind before he in turn could conceive CG. Neither of Frege nor
Husserl alone was doing exactly Categorial Grammar, explicitly or implicitly. As
we have seen, Fregean functoriality by itself yields hardly any grammar at all;
Husserl devised a sort of Montague grammar rather than CG.
derivations).18 The explanatory power of CG is derived, or so we have

argued, from the fact that CG is founded in some fundamental, basic in-
sight about language: it is a formal and detailed elaboration of this in-
sight. And it is precisely this insight we have tried to reconstruct in the
present paper: the confluence of Husserlian bi-modal intentionality and
Fregean functoriality.
References
Ajdukiewicz, Kazimierz 1925a. Nazwy i zdania (autoreferat). Ruch Filozoficzny 9,

nr 1-2, 23b-24b.
Ajdukiewicz, Kazimierz 1925b. Składniki zda (autoreferat), Ruch Filozoficzny 9,
nr 9-10, 164a-164b.
Ajdukiewicz, Kazimierz 1935/1967. Syntactic Connexion. In: Storrs McCall (ed.)
1967. Polish Logic 1920-1939. Oxford: Oxford University Press, 207-231 [first
published in 1935 in Studia Philosophica 1, in German as Die Syntaktische
Konnexität].
Ajdukiewicz, Kazimierz 1978. Syntactical Connections between Constituents of
Declarative Sentences. In: J. Giedymin (ed.), 269-281.
Black, Max and Peter T. Geach (eds.) 1960. Translations from Philosophical Writ-
ings of Gottlob Frege. Oxford: Basil Blackwell.
Carpenter, Bob 1997. Type-Logical Semantics. Cambridge, Mass.: The MIT Press.
Dummett, Michael 1973. Frege: Philosophy of Language. London: Duckworth.
Dummett, Michael 1993. Origins of Analytical Philosophy. Cambridge, Mass.:
Frege, Gottlob 1891. Function and Concept. In: M. Black and P.T. Geach (eds.), 21-
41.
Frege, Gottlob 1892a. On Concept and Object. In: M. Black and P.T. Geach (eds.),
42-55.
Frege, Gottlob 1892b. Sense and Reference. In: M. Black and P.T. Geach (eds.), 56-
78.
Geach, Peter Thomas 1950. Subject and Predicate. Mind 59, 461-482.
Geach, Peter Thomas 1970. The Program of Syntax. Synthese 22, 3-17.
Geach, Peter Thomas 1972. Czemu zdanie nie jest nazw [Why a sentence is not a
name] Studia Semiotyczne 3, 13-21 [published in Polish].
Geach, Peter Thomas 1975. Names and Identity. In: S.D. Guttenplan (ed.), 139-158.
18
For a discussion of CG as a linguistic theory, see Tałasiewicz ( 2010).
Geach, Peter Thomas 1980a. Reference and Generality. Ithaca, NY: Cornell Uni-
versity Press.
Geach, Peter Thomas 1980b. Strawson on Subject and Predicate. In: Z. van Straaten
(ed.), 174-188.
Giedymin, Jerzy (ed.). 1978. The Scientific World-Perspective and Other Essays,
1931-1963. Synthese Library Vol. 108. Dordrecht/Boston: D. Reidel.
Guttenplan, Samuel D. (ed.) 1975. Mind and Language. Oxford: Clarendon Press.
Haaparanta, Leila (ed.) 1994. Mind, Meaning and Mathematics. Dordrecht: Kluwer.
Hale, Bob 1979. Strawson, Geach and Dummett on Singular Terms and Predicates.
Synthese 42, 275-295.
Simons, Peter 1981. Unsaturatedness. Grazer Philosophische Studien 14, 73-95.
Smith, Barry 1994. Husserl’s Theory of Meaning and Reference. In: L. Haaparanta
(ed.), 163-183.
Stalmaszczyk, Piotr (ed.) 2012. Philosophical and Formal Approaches to Linguistic
Analysis. Frankfurt a. M.: Ontos Verlag.
Steedman, Mark 2000. Syntactic Process. Cambridge, Mass.: MIT Press.
Stirton, William R. 2000. Singular Term, Subject and Predicate. The Philosophical
Quarterly 50, 191-207.
van Straaten, Zak (ed.) 1980. Philosophical Subjects: Essays Presented to P. F.
Strawson. Oxford: Oxford University Press.
Strawson, Peter F. 1950. On Referring. Mind 59, 320-344 [cited from Strawson
(1971), 1-20].
Strawson, Peter F. 1959. Individuals. London: Routlege.
Strawson, Peter F. 1966. The Bounds of Sense: An Essay on Immanuel Kant’s ‘Cri-
tique of Pure Reason’. London: Methuen.
Strawson, Peter F. 1971. Logico-Linguistic Papers. Aldershot: Ashgate.
Strawson, Peter F. 1974. Subject and Predicate in Logic and Grammar. Aldershot:
Ashgate.
Tałasiewicz, Mieszko 2009. Geach’s Program and the Foundations of Categorial
Grammar. Acta Semiotica Fennica XXXIV, 1731-1742.
Tałasiewicz, Mieszko 2010. Philosophy of Syntax. Foundational Topics. Dordrecht:
Springer.
Tałasiewicz, Mieszko 2012. A New Perspective for the Theory of Intentionality. In:
P. Stalmaszczyk (ed.), 513-525.
Index
Ajdukiewicz, Kazimierz 247, demonstratives 99, 102, 103,
269, 284, 287, 288, 289, 291 104, 105, 107, 108, 111, 162,
ambiguous illusions 53, 54 172, 275
analytical philosophy 11, 71, diachronic historical linguistics
116, 119, 124 11, 14, 15
assertion 29, 33, 34, 35, 45, 79, Donnellan, Keith 5, 99, 100, 111,
85, 115, 116, 119, 120, 122, 161, 165, 168, 169, 170, 177,
123, 125, 133, 136, 229, 230, 178
239, 260, 276, 277, 291 duck/rabbit 4, 53, 54, 56, 65, 66,
Austin, John L. 11, 12, 13, 16, 67, 68, 69
19, 22, 26 Dummett, Michael 4, 71, 72, 73,
belief ascription 53 74, 76, 77, 78, 80, 140, 185,
bipolarity 29, 30, 31, 41, 49, 264 186, 187, 188, 189, 213, 226,
Bradley, Francis Herbert 6, 181, 227, 230, 233, 278, 281, 290
182, 183, 184, 185, 189, 214 falsehood 29, 31, 32, 33, 34, 36,
Carnap, Rudolf 2, 139, 140, 142, 172, 184, 255, 256, 259, 260,
194, 195, 196, 197, 209, 228 261, 262, 263, 266
Categorial Grammar (CG) 8, first and second order functions
269, 280, 285, 288, 289, 290, 181
291, 292 formal semantics 193
Chomsky, Noam 109, 139, 145, Frege, Gottlob 1, 2, 3, 4, 5, 7, 71,
149, 150, 152, 243, 244, 249 72, 73, 74, 75, 76, 77, 78, 79,
conceptual analysis 11, 12 80, 81, 82, 83, 84, 85, 86, 87,
conceptual clarification 3, 11, 12, 88, 89, 90, 91, 92, 93, 94, 115,
13, 14, 15, 17, 21, 22, 25, 26 116, 117, 119, 121, 130, 139,
copula 183, 225, 239, 240, 241, 146, 164, 171, 181, 185, 186,
242, 245 187, 188, 189, 190, 194, 196,
Davidson, Donald 139, 141, 147, 197, 204, 208, 209, 213, 214,
149, 151, 154, 181, 184, 185, 216, 217, 218, 219, 220, 221,
214, 234, 235, 236 225, 226, 227, 228, 229, 230,
definite descriptions 1, 5, 6, 99, 231, 232, 233, 236, 239, 240,
100, 101, 102, 104, 105, 106, 241, 243, 244, 245, 249, 255,
107, 108, 109, 111, 161, 164, 260, 261, 269, 272, 273, 276,
165, 169, 170, 174, 178 277, 278, 279, 280, 289, 291
296 Index
function 42, 46, 91, 116, 119, Kripke, Saul 5, 99, 100, 111,
125, 131, 146, 161, 162, 163, 161, 164, 165, 166, 167, 168,
170, 171, 173, 179, 186, 187, 169, 170, 171, 173, 175, 178
188, 189, 190, 209, 216, 217, linguistic disposition 139
218, 225, 230, 231, 232, 233, linguistic turn 3, 6, 12, 76, 193,
236, 239, 242, 244, 245, 247, 194, 195, 196, 197, 198, 200,
248, 258, 265, 271, 272, 274, 202, 203, 206, 210, 226, 227
279, 286, 288 logic 2, 6, 8, 79, 84, 88, 93, 119,
functor 236, 272, 281, 282, 283, 124, 156, 157, 189, 193, 194,
286, 287, 291 195, 196, 197, 198, 200, 203,
functoriality 8, 269, 272, 279, 204, 205, 206, 208, 209, 210,
280, 285, 287, 288, 291, 292 217, 226, 228, 229, 231, 233,
Geach, Peter Thomas 275, 277, 234, 255, 269
278, 283, 284, 288, 289, 290 logical form 4, 29, 38, 39, 43, 44,
generative grammar 3, 5, 7, 109, 46, 71, 75, 81, 91, 100, 257,
150, 225, 226, 236, 237, 243, 258, 261, 263, 280
244, 246, 289 Mill, John Stuart 161
Heidegger, Martin 5, 115, 124, ontological commitment 115,
125, 126, 133, 134 127, 260
historical commitment 115 ontologico-historical
Husserl, Edmund 121, 124, 202, understanding 5, 115, 135
228, 229, 231, 269, 273, 274, ontology 7, 68, 148, 184, 210,
275, 276, 277, 278, 279, 280, 213, 214, 237, 238
281, 289, 290, 291 optical illusions 53
impressions 53, 58 ordinary language philosophy 2,
inconsistency 5, 115, 139, 182, 3, 11, 12, 13, 14, 15, 23, 25
261 paradox of the concept horse 7,
indeterminacy 139, 144, 145, 213, 214, 221
146, 147, 148, 154 perception 3, 4, 53, 54, 55, 56,
intentionality 8, 269, 273, 274, 58, 60, 62, 64, 66, 68, 69, 198
275, 276, 277, 280, 285, 290, phases 99, 109
292 picture theory of language 255
judgement 29, 31, 32, 36, 37, 38, pragmatics 100, 170, 173
39, 40, 42, 43, 44, 47, 87, 88, predicate 69, 80, 89, 90, 91, 118,
168, 173, 175, 176, 181, 182, 129, 148, 155, 167, 183, 185,
183, 188, 190, 214, 216, 229, 187, 189, 216, 217, 218, 219,
230, 239, 242, 260, 261, 264 220, 225, 231, 232, 233, 234,
Kant, Immanuel 3, 117, 118, 235, 237, 238, 239, 240, 241,
120, 121, 132, 133, 188 242, 244, 245, 247, 248, 269,
Index 297
270, 271, 273, 276, 278, 281, 232, 237, 242, 255, 259, 260,
283, 284, 286, 288, 290 261, 262, 279
predication 1, 7, 125, 157, 213, relational properties 213, 215
214, 225, 226, 227, 231, 234, relations 2, 15, 19, 20, 32, 37, 38,
235, 236, 237, 238, 239, 240, 42, 45, 48, 73, 75, 85, 89, 118,
241, 242, 243, 244, 245, 246, 121, 124, 135, 148, 181, 182,
247, 248, 249, 278 183, 184, 186, 190, 205, 216,
proceduralism 5, 115, 116, 117, 228, 230, 233, 236, 237, 238,
118, 119, 120, 121, 122, 123, 241, 244, 245
126, 127, 128, 129, 130, 132, Russell, Bertrand 1, 2, 3, 4, 6, 7,
133, 134 25, 29, 30, 31, 32, 33, 34, 35,
proposition 5, 7, 30, 31, 32, 33, 36, 37, 38, 39, 40, 41, 42, 43,
34, 35, 36, 37, 39, 40, 41, 42, 47, 48, 49, 63, 99, 116, 119,
44, 45, 46, 47, 48, 49, 63, 74, 130, 139, 156, 161, 162, 163,
77, 80, 81, 89, 94, 99, 100, 164, 165, 166, 167, 168, 170,
101, 102, 111, 140, 142, 162, 171, 172, 173, 174, 175, 176,
167, 172, 181, 189, 190, 199, 177, 178, 179, 181, 182, 183,
213, 214, 220, 227, 235, 236, 184, 185, 189, 190, 195, 196,
237, 239, 240, 242, 244, 245, 197, 209, 215, 255, 260, 264
247, 248, 255, 256, 258, 259, Sainsbury, R. M. 161, 163, 165,
260, 261, 262, 263, 264, 265, 166, 167, 168, 170, 171, 172,
269, 278, 280, 284 176, 177
propositional unity 29, 191, 271 Saussure, Ferdinand de 15, 18,
quantification 99, 166 198
Quine, Willard Van Orman 2, 5, semantic-pragmatic distinction
139, 140, 141, 142, 143, 144, 161
145, 146, 147, 148, 149, 150, semantics 5, 7, 11, 15, 99, 100,
151, 152, 153, 154, 155, 156, 101, 103, 104, 105, 107, 108,
157, 194, 195, 208, 209, 234, 109, 110, 111, 139, 140, 141,
235, 242 142, 147, 148, 149, 150, 154,
reference 1, 5, 6, 33, 92, 93, 99, 157, 170, 195, 197, 199, 205,
100, 101, 106, 115, 116, 118, 206, 208, 209, 210, 217, 225,
119, 120, 122, 123, 125, 128, 226, 229, 231, 232, 234, 235,
130, 131, 136, 139, 140, 141, 237, 239, 240, 245, 246, 247,
146, 147, 148, 150, 151, 152, 248, 255, 265
154, 157, 161, 165, 166, 167, Strawson, Peter F. 8, 12, 66, 119,
168, 169, 170, 171, 173, 174, 157, 168, 235, 269, 270, 271,
175, 176, 177, 178, 183, 226, 273, 274, 276, 281, 284, 291
298 Index
structure 3, 4, 6, 7, 8, 40, 43, 45, 220, 221, 227, 228, 229, 230,
71, 72, 73, 74, 75, 76, 77, 79, 236, 238, 239, 265, 271, 272,
81, 84, 87, 90, 100, 105, 106, 275, 276, 282
109, 110, 112, 133, 135, 139, truth 1, 3, 4, 29, 31, 32, 33, 34,
143, 148, 150, 151, 154, 181, 35, 36, 42, 43, 44, 46, 47, 48,
185, 186, 187, 189, 194, 195, 49, 86, 92, 94, 134, 136, 140,
207, 208, 209, 213, 215, 216, 141, 142, 144, 146, 147, 148,
219, 220, 221, 226, 236, 241, 152, 157, 163, 167, 172, 183,
244, 249, 255,256, 257, 258, 185, 186, 187, 189, 191, 199,
269, 281, 282, 285 200, 203, 208, 209, 214, 229,
syntax 8, 150, 154, 217, 218, 230, 233, 235, 236, 239, 241,
228, 248, 258, 264, 269, 270, 256, 258, 259, 261, 265, 276
273, 289, 290 underdetermination of theory
theory of description 6, 161, 163, 139, 144, 145
164, 165, 168, 172, 178, 179 unity of the proposition 185, 214,
theory of judgement 3, 6, 29, 30, 236
31, 32, 37, 39, 40, 42, 48, 181, unsaturatedness 240, 269, 273,
185 277, 278, 279, 280, 290, 291
theory of names 161, 164, 171, Wittgenstein, Lugwig 1, 2, 3, 4,
173, 177 5, 11, 12, 15, 16, 17, 18, 19,
thought 4, 33, 41, 43, 44, 69, 71, 21, 29, 30, 31, 32, 37, 39, 40,
72, 73, 74, 75, 76, 77, 78, 79, 41, 42, 43, 44, 45, 46, 47, 48,
80, 81, 82, 83, 85, 86, 87, 88, 49, 53, 54, 63, 64, 65, 69, 78,
89, 90, 91, 92, 93, 116, 117, 115, 116, 117, 124, 125, 126,
118, 121, 122, 123, 124, 126, 130, 131, 133, 134, 139, 153,
128, 129, 130, 131, 133, 135, 155, 182, 189, 194, 196, 202,
142, 144, 156, 162, 165, 166, 203, 206, 208, 209, 210, 230,
168, 170, 171, 177, 181, 182, 255, 256, 257, 258, 259, 260,
183, 184, 185, 186, 187, 188, 261, 262, 263, 264, 265
189, 190, 196, 214, 216, 218,

Frege, Gottlob - Russell, Bertrand - Stalmaszczyk, Piotr - Wittgenstein, Ludwig - Philosophy of Language and Linguistics - The Legacy of Frege, Russell, and Wittgenstein (2014, de Gruyter) PDF

Uploaded by

Copyright:

Available Formats

Frege, Gottlob - Russell, Bertrand - Stalmaszczyk, Piotr - Wittgenstein, Ludwig - Philosophy of Language and Linguistics - The Legacy of Frege, Russell, and Wittgenstein (2014, de Gruyter) PDF

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Frege, Gottlob - Russell, Bertrand - Stalmaszczyk, Piotr - Wittgenstein, Ludwig - Philosophy of Language and Linguistics - The Legacy of Frege, Russell, and Wittgenstein (2014, de Gruyter) PDF

Uploaded by

Copyright:

Available Formats

What is the book about?

What is the book about?

What philosophers are discussed in the book?

What philosophers are discussed in the book?

Piotr Stalmaszczyk (Ed.

Herausgegeben von / Edited by

Library of Congress Cataloging-in-Publication Data

Bibliografische Information der Deutschen Nationalbibliothek

© 2014 Walter de Gruyter GmbH, Berlin/Boston

The Legacy of Frege and the Linguistic Theory of Predication

Philosophy of Language and Linguistics:

All philosophy is a ‘critique of language’.

Before we can understand language, we must strip it of its

The present volume investigates selected aspects of the legacy of Gott-

ticularly characteristic of categorial grammars, Montague grammar, and

1. Contents of the volume

The collection brings together contributions by philosophers, logicians

tered under its three versions in The Principles of Mathematics (1903),

Thomas J. Hughes defends the thesis that definite descriptions should

tries to separate them into the interlinguistic and intralinguistic, suggest-

uation in terms of distinguishing two perspectives: the expression-as-ob-

constructing a credible logic of natural language and elucidating many

I am grateful to Mr Ryszard Rasi ski for comprehensive editorial assistance, and to

Arendt, Hannah 1978. Thinking. New York: Harcourt Brace Jovanovich.

Mapping the Ancient City: Historical

Keywords: Wittgenstein, conceptual clarification, conceptual analysis, diachronic

Our language can be regarded as an ancient city: a maze of lit-

been attacked in various ways. Most of all, the objections focussed on

If agreement about usage cannot be reached within so restricted a sample as

Thus, it seems that conceptual clarification would eventually turn into

What diachronic linguistics studies is not relations between coexisting

Diachronic linguistics is concerned with the phenomenon of language

1. Conceptual clarification as a method

everyday speech, competent speakers hardly become confused by lan-

Language sets everyone the same traps; it is an immense network of easily

In the following, I do not wish to dwell on a defence for this method, as

2. Why the past might not matter

At first sight, to combine the idea of conceptual clarification with a his-

movements: the piece that is allowed to move only horizontally or verti-

3. First reason: clarifying the starting point

the expressions involved is mentioned quite regularly – which actually

Transitive consciousness lies at the confluence of the concepts of

4. Second reason: understanding language change

The second point shows another advantage of a diachronically enriched

5. Conclusion: philosophy as an appendix of empirical linguistics?

The fears are baseless. Etymological investigations can no more repla-

Russell and Wittgenstein on Proposition,

Keywords: Russell, Wittgenstein, judgement, truth, falsehood, sense, logical form,

Recent work on the objection that Wittgenstein raised to Russell’s theory

for a difficulty internal to the Russellian theory of judgement itself, ra-

limitation: he thinks that Wittgenstein did not provide a complete solu-

1. The evolution of Russell’s theory and its problems

Russell’s clearest difficulty in developing his theory was the necessity to

Line 1: To consider truth and falsehood as properties of propositions

Line 2: To consider truth and falsehood as properties of judgements,

It is crucial that in Line 1, propositions are primitive with respect to psy-

There are three important features of Russell’s conception at this early

(1) Obama is younger than Bush

where (2) is composed of the following constituents: the things [Obama]