FREGE:

SENSE AND REFERENCE ONE HUNDRED YEARS LATER

 PHILOSOPHICAL STUDIES SERIES
Founded by Wilfrid S. Sellars and Keith Lehrer

Editor
KEITH LEHRER, University ofArizona, Tucson

Associate Editor
STEWART COHEN, Arizona State University, Tempe

Board of Consulting Editors

LYNNE RUDDER BAKER, University of Massachusetts at Amherst
ALLAN GIBBARD, University of Michigan
DENISE MEYERSON, University of Cape Town
RONALD D. MILO, University ofArizona, Tucson
FRAN<;OIS RECANATI, Ecole Polytechnique, Paris
STUART SILVERS, Clemson University
NICHOLAS D. SMITH, Michigan State University

VOLUME 65
 FREGE:
SENSE AND REFERENCE
ONE HUNDRED YEARS LATER

Editedby

JOHNBIRO
University of Florida,
Gainesville, U.SA.
and
PETR KOTATKO
Institute of Philosophy,
Academy of Sciences of the Czech Republic,
Prague, Czech Republic

SPRINGER SCIENCE+BUSINESS MEDIA, B.V.

A C.I.P. Catalogue record for this book is available from the Library of Congress.

ISBN 978-94-010-4184-3 ISBN 978-94-011-0411-1 (eBook)

DOI 10.1007/978-94-011-0411-1

Printed on acid-free paper

AII Rights Reserved

© 1995 Springer Science+Business Media Dordrecht
Originally published by Kluwer Academic Publishers in 1995
Softcover reprint of the hardcover 15t edition 1995
No part of the material protected by this copyright notice may be reproduced or
utilized in any form or by any means, electronic or mechanical,
including photocopying, recording or by any information storage and
retrieval system, without written permission from the copyright owner.
 TABLE OF CONTENTS

Preface ........................................................................................................................................vii

THOMAS BALDWIN
Three Puzzles in Frege's Theory of Truth ..................................................................................... .1

GABRIEL SEGAL
Truth and Sense ............................................................................................................................. 15

BARRY SMITH
Frege and Chomsky: Sense and Psychologism ........................................................................... 25

PETRKOTATKO
Meaning and the Third Realm ...................................................................................................... .47

DAVID WIGGINS
Putnam's Doctrine of Natural Kind Words and Frege's Doctrines of
Sense, Reference and Extension: Can They Cohere? ................................................................. .59

ANTHONY C. GRAYLING
Concept-Reference and Kinds .......................................................................................................75

FRANCOIS RECANATI
The Communication of First Person Thoughts ........................•.................................................... 95

TOM STONEHAM
Transparency, Sense and Self-Knowledge .................................................................................. 103

CHRISTINE TAPPOLET
The Sense and Reference of Evaluative Terms .......................................................................... .113

PETER SIMONS
The Next Best Thing to Sense in Begriffsschrift ......................................................................... 129

DAVID OWENS
Understanding Names ..................................................................................... ,........................... 141

EROS CORAZZA & JEROME DOKIC

Why Is Frege's Puzzle Still Puzzling? ...................................................................................... .151

MARTIN HAHN
The Frege Puzzle One More Time .............................................................................................. 169

JOHN BIRO
The Neo-Fregean Argument ....................................................................................................... 185

Index .......................................................................................................................................... 207

 PREFACE

Considering the importance of Frege's "Uber Sinn und Bedeutung" in the

origins and subsequent course of analytic philosophy, the centenary of its
publication (in 1892) went relatively uncelebrated. One of the few special
events organized to commemorate the event was a conference sponsored by
the Institute of Philosophy of the Czech Academy of Science, held in
Karlovy Vary in September of 1992. The present volume contains a
selection from the papers presented at that conference.
The papers fall into several groups. The first explores questions about
Frege's somewhat scattered and sometimes puzzling remarks on truth. Tom
Baldwin discusses a number of the puzzles, coming to the conclusion that
they can be resolved so as to allow us to attribute to Frege a coherent theory
of truth. That theory, according to Baldwin, sees us as able to grasp the
truth-conditions of thoughts -- and thus of sentences -- independently of a
grasp of their senses, thus opening the way to a truth-based theory of
meaning of the familiar Davidsonian sort. Both Baldwin and Gabriel Segal
see a so-called interpretive truth-theory as able to serve as a theory of
meaning for at least the context-insensitive parts of language. Segal makes
the additional suggestion that we might see a theory of meaning of this sort
as a theory of linguistic competence in the Chomskian sense. Doing so
solves what he calls "the information problem," namyly, the problem of how
a theory that does not wear its interpretiveness on its sleeve can be taken as
representing speakers' knowledge of the language it in fact interprets. The
proposal is that there is a simple brute-force solution to the problem: the
knowledge that the theory is interpretive need not be represented in the theory
or, indeed, anywhere else, as long as the speakers whose language it
characterizes are hard-wired so as to treat its theorems as interpretive.
As Segal notes, a theory of this sort leaves unexplained how we deploy
the competence represented in our actual use of language. That explanation
is to be provided, presumably, by theories appealing to psychological and
conventional principles. Barry Smith and Petr Kotatko, respectively, explore
the prospects for theories of these kinds. Like Segal, Smith draws on
Chomsky'S ideas in setting out and defending his proposals for how we
should see the relation between abstract characterizations of language and the
psychology of linguistic competence. He argues that while linking the two
in the close way he advocates does lead to a psychologism that is in some
ways un-Fregean, the putative virtues of the anti-psychologist position --

vii
viii
most importantly, the ability to account for normativity and the non-
subjectivism it presupposes -- are not lost. Kotatko focuses on the
conventional elements both in language and in language use, arguing that
even the latter is best understood in a non-psychologistic way. He relies on
ideas from speech-act theory to defend a strong version of the autonomy of
language and of its independence from psychology. He illustrates his claims
by presenting a sketch of a theory of assertion in which the contingent
psychological states of individual asserters play no part. A crucial aspect of
this approach is a shift of focus to the hearer's perspective, specifying
semantic values in terms of conventionally-based reasonable interpretation.
David Wiggins asks whether Frege's and Putnam's respective pictures of
meaning and reference can be fitted together, and gives an answer in the
affirmative. In the course of his discussion, he develops an account of the
sense of a referring expression as its user's conception of the object to which
it refers. What such a conception, in turn, comes to is a certain
"identificatory capacity," a capacity that varies even among competent
speakers depending on the amount of information they possess about the
object. Extended to predicate expressions, the account yields a picture on
which the grasp of concept-words in general and the possession of the
concepts they denote are spelled out in terms of such capacities.
In a companion piece, Anthony Grayling, while generally approving of
Wiggins' Fregeanization of Putnam, rejects what he sees as a Putnamization
of Frege resulting from the emphasis on actual extensions as exemplars in a
deictic account of the meaning of kind-terms. For Grayling, the fact that the
information subserving individual identificatory capacities is public, shared,
to one degree or another, by competent speakers, takes the place of actual
extensions in grounding those capacities; his claim is that this adjustment
yields a more general and less ontologically committed -- and, in both these
ways, more Fregean -- theory.
Francois Recanati and Tom Stoneham are both concerned with questions
about the transparency of one's thoughts to one. Recanati's interest is in
what he cans "the paradox of the first person," that is, the puzzle of how it is
possible to communicate thoughts that others cannot share (such as thoughts
involving 'I' and, perhaps, other indexicals). Recanati reviews a number of
possible answers to this question: Frege's own ip terms of differing modes
of representation, a Dummettian one that relies on a distinction between
having a thought and merely recognizing what thought someone else has,
and a "Russellian" one that traces communicability to common elements in
the respective thoughts of speaker and hearer. He opts for a version of the
 ix

last, locating the common elements in partially shared "dossiers of

information" and argues that such a solution is at once Fregean and superior
to Frege's own.
Stoneham asks whether one's knowledge of the meanings of the words
one understands is epistemologically privileged. Are those meanings
transparent to one in the way a Fregean appeal to cognitive states to
individuate senses may suggest? Stoneham's answer is that transparency has
limits: there is an asymmetry between knowing that two words differ in
meaning and knowing that they do not. Difference in cognitive state is
sufficient for the former, but sameness of cognitive state is not sufficient for
the latter. Recognizing this asymmetry is important, since doing so enables
us to see that Frege's grounding of differences in sense in differences in
cognitive value cannot be invoked to support claims, such a Dummett's, of
full transparency.
Christine Tappolet strikes out in a different direction: she is interested in
the application of Fregean ideas to evaluative discourse. She defends a
suggestion, following in part one put forward recently by David Wiggins, on
which an evaluative term, no less than a non-evaluative one, has a sense, a
reference, and an extension. Her argument for the third, most problematic,
part of this claim appeals to our emotions as revealing the "evaluative facts"
that serve as the extensions of evaluative terms in the way that horses do as
the extension of 'horse'. (If this is right, realism follows. There is a
connection here with Grayling's worry: what price in ontological
commitment should we be willing to pay for a unified semantic theory?)
In one way or another, all the remaining papers address the bearing of the
cognitive states of language users on the semantics of the language they use.
Peter Simons discusses the way in which Frege's own views on this question
developed from his early discussions of informative identities in the
Begriffsschrift of 1879 to the mature theory of "On Sense and Reference".
While recognizing the greater generality and power of the later theory,
Simons finds more continuity between it and the earlier one than is
sometimes allowed. One important way in which the later theory is usually
thought to be superior is in offering another option for the treatment of
identity statements besides the alternatives of thinking of them as being
about the referent (which leaves their informativeness puzzling) or taking
them to be about the names (which leaves us with the problem of
arbitrariness). Simons argues that Frege in fact came closer to seeing that
third option earlier than is generally recognized.
x

David Owens takes up the question of what two speakers have to know
about each other's use of names in order to be able to communicate by using
them. Owens argues that a requirement that the two must think of the
referent of the name in the same way (under the same mode of representation)
would be too strong. He suggests a weaker requirement as sufficient: a
speaker must merely ensure that his hearer knows which name he is using.
As long as this is done, each party may, as is indeed often the case, identify
the name's bearer in a different way.
The last three papers offer detailed treatments of the so-called Frege
puzzle, discussing between them most of the approaches to it available in the
literature. Eros Corazza and Jerome Dokic opt for an interpretation of the
puzzle on which what matters is what inferences the use of a sign licenses
for its user. They suggest that this function of signs has not been
sufficiently investigated and that until it is, the puzzle stands unsolved.
Martin Hahn urges that the real challenge of the puzzle is not to direct-
reference theories, as is often thought, but to any semantic theory that does
not include "a theory of the contents of thoughts and how these determine the
thoughts' referents". On the other hand, meeting that challenge by
attempting to make one theory do double duty is a mistake, one argua~ly
committed by Frege himself. John Biro also worries about forging too close
a link between theories addressing cognitive and semantic questions,
respectively. He suggests that a careful separation allows a dis-solution of
the puzzle along (in a certain sense) meta-linguistic lines.
The inevitable overlaps notwithstanding, the issues taken up by these
essays cover a large part of the territory still being explored by philosophers
of language and of mind. Their breadth and variety, and the liveliness of the
current debates as represented here, bear witness to the fecundity and
continuing vitality of Frege's ideas.

The editors wish to express their thanks to all those who have assisted them
with this volume. First and foremost, of course, they are grateful to the
authors for their papers. In addition, they would like to acknowledge the
generous help given to them in the preparation of the manuscript by Barbara
Jones and Marin Smillov, and Virginia Dampier.
 THOMAS BALDWIN

THREE PUZZLES IN FREGE'S THEORY OF TRUTH

I want to discuss three related puzzles that arise from Frege' s numerous
discussions of truth. The first concerns his thesis that truth is indefinable.
Frege includes the thesis that "What true is, I hold to be indefinable" (Was
wahr sei, halte ich jur nicht erkliirbar) among his Kernsiitze, his "17 Key
Sentences on Logic". The Kernsiize are dated "1906 or earlier" in the standard
edition of his Posthumous Writings (Frege 1979, p. 174), but Dummett has
argued (Dummett 1981, p. 523) that they should be dated much earlier,
around 1880. Dummett's reasons for the early date include the hypothesis
that the Kernsiitze express Frege's reaction to Lotze's Logic (Lotze 1884).
Now it was, of course, Sluga who first raised the question of the relationship
between Frege and Lotze, and, in effect anticipating Dummett's re-dating of
the Kernsiitze he maintained that Frege's thesis of the indefinability of truth
is simply an extension of Lotze's criticisms of the correspondence theory of
truth (Sluga 1980, p. 114). Dummett has argued that this seriously
misrepresents Frege (Dummett 1981, p. 526); but as long as it is recognised
that Frege's line of thought goes well beyond Lotze's conclusion, it does no
harm to approach Frege's position via that of Lotze. Lotze's argument
(Lotze, sections 305-6) is essentially epistemological: he argues that,
confronted by sceptical doubts about the truth of our beliefs, a
correspondence theorist's advice that we should find out whether our beliefs
correspond to the facts is useless -- for in inquiring what the facts are, all we
acquire are further beliefs, which we have as much, or as little, reason to
doubt as we had to doubt our original beliefs. Hence, Lotze concludes, for the
purposes of epistemology the criterion of truth has to be internal coherence.
As we shall see, Frege' s argument for the indefinability of truth does indeed
start from the uselessness of the correspondence theory when one seeks to
determine the truth; but, unlike Lotze, Frege proceeds to argue that no
definition of truth can be of any use in determining what is true, and hence
that truth itself is absolutely indefinable.
In the Kernsiitze themselves Frege offers no reason for his thesis that
truth is indefinable. But in his 1918 paper 'The Thought' he offers the
following argument:

1
J. Biro andP. Kotatko (eds.), Frege: Sense and Reference One Hundred Years Later, 1-14.
© 1995 Kluwer Academic Publishers.
2 THOMAS BALDWIN

Can it not be laid down that truth exists when there is correspondence in a certain
respect? But in which? For what would we then have to do to decide whether
something were true? We should have to inquire whether it were true that an idea
and a reality, perhaps, corresponded in the laid-down respect. And then we should
be confronted by a question of the same kind and the game would begin again. So
the attempt to explain truth as correspondence collapses. And every other attempt to
define truth collapses too. For in a definition certain characteristics would have to be
stated. And in application to any particular case the question would always arise
whether it were true that the characteristics were present. So one goes round in a
circle. Consequently, it is probable that the contentof the word 'true' is unique and
indefinable. (Frege 1967a, p. 19)

Although this is, I think, the first published use of the argument, it occurs
much earlier in Frege's writings, most notably in the 'Logic' manuscript of
1897, where the argument is called 'The Treadmill' (Frege 1979, pp. 128-9,
134).
It is this argument that provides my first puzzle. For at first the argument
appears to be mere sophistry. Frege seems to be arguing that there is a
vicious circle in any attempted definition of truth, since in attempting to
apply a definition of truth in the course of asking whether a judgment is true
we cannot avoid reintroducing the notion we seek to define, for we are bound
to ask whether it is true that the judgment satisfies the conditions specified
in the definition. As Frege puts it in the 1897 'Logic' manuscript: "And so
we should be in the position of a man on a treadmill who makes a step
forwards and upwards, but the step he treads on keeps giving way and he falls
back to where he was before" (Frege 1979, p. 134). But to all this, someone
who seeks to define truth can object, first, that it is not obvious that one has
to reintroduce a reference to truth: further argument is required to show that
in considering whether a judgment, say, corresponds to the facts, one has to
consider whether it is true that the judgment corresponds to the facts;
secondly, even if one does introduce this further question, it needs to be
shown why one cannot just apply the definition to this question.
In the light of this response Dummett has proposed that we should
reinterpret Frege's argument as an argument to the effect that definitions of
truth give rise to a vicious regress (Dummett 1973, pp. 442ff.). His idea is
that once it is conceded that, in applying any definition of truth, one can
always raise the further question as to whether it is true that the conditions
specified in the definition are satisfied, we find ourselves launched on an
indefinite series of questions. But is the regress here vicious? The prospect of
an indefinite series of questions only gives rise to a vicious regress
concerning the possibility of answering even the first question where the
answer to this question depends upon answers to questions later in the series,
 THREE PUZZLES IN FREGE'S THEORY OF TRUTH 3

and so on indefinitely. Ironically, the young G. E. Moore's conception of

truth as an indefinable property of propositions, whose possession of this
property is constitutive of facts themselves, gives rise to a regress of this
kind (Moore 1899, p. 192). For on this view to find out whether it is a fact
that snow is white one has to determine whether the proposition that snow is
white is true; but (unless the truth of propositions is treated as a special case)
to determine this one will have to determine whether it is a fact that the
proposition that snow is white is true, i.e., whether the proposition that the
proposition that snow is white is true is itself true; and so on ad nauseam.
As we shall see, although Frege thinks that truth is indefinable, he
certainly does not think it is a property of this kind; so he is not vulnerable
to it. But are correspondence theorists themselves vulnerable to this
objection? Dummett claims that for a correspondence theorist questions
about truth form a vicious series, because, on the one hand, such a theorist
cannot plausibly deny that the question whether p gives rise to the question
whether it is true that p; but, on the other hand, on their account of truth,
this latter question does not just reduce to the former, since it makes
essential reference to new questions about the existence of facts and the
correspondence of ideas to them. (Dummett 1973, p. 444) The
correspondence theorist's response to this will surely be to dispute the
unrestricted validity of the "upwards" move in this argument, where the issue
of truth is explicitly introduced. Because such a theorist attaches some
substance to questions about the existence of genuine facts, he will separate a
question whether p from that of whether it is a fact that p. For example, he
may be happy to answer the question whether he should take a swim without
being able to say whether it is a fact that he should take a swim. As a result,
he will hold that there are conditions, concerning the existence of genuine
facts, on the move from 'p' to 'it is true that p'; but this is no objection to
his position, for it is precisely the point of the theory that, by linking truth
to the existence of facts and thereby giving some substance to the concept,
one should be able to place limits on its application. It is indeed the question
of whether these conditions are satisfied which, as Dummett rightly
observed, implies that for the correspondence theorist, the question whether it
is true that p does not reduce to the question whether p; what he should have
equally noted is that the same questions arise as conditions of the validity of
the upward move from a question whether p to a question whether it is true
that p, and thus inhibit the generation of any vicious regress of questions.
Dummett will respond that this just shows that a substantive definition
of truth, such as the correspondence theorist offers, is incompatible with the
4 THOMAS BALDWIN

unconditional validity of the "upward" move from 'p' to 'it is true that p',
which is essential if we are to make sense of the connection between truth
and attitudes such as belief and inquiry (Dummett 1973, p. 444). This now
shifts the burden of argument onto this last thesis, that in all cases a
question whether p is a question whether it is true that p. It seems to me that
the correspondence theorist can fairly object that it remains to be shown why
the connection between truth and belief or inquiry cannot be sustained by his
own account of the conditional validity of the upward move. He will allow
that where a question is admitted not to concern a matter of fact, the question
is not an inquiry into truth; but he can invoke familiar strategies of 'quasi-
realism' to explain how attitudes that are in the first instance tied to genuine
matters of fact come to be extended to other issues where truth is not
applicable. Furthermore, Dummett's reconstruction of Frege's argument
appears to be vulnerable to a quite different line of attack -- namely, that even
if one supposes that the argument does raise a doubt concerning substantive
definitions of truth, such as correspondence, it does not thereby establish
Frege's thesis that truth is indefinable. For there remains the alternative of
modest definitions of truth, such as the redundancy theory, which certainly do
not set off any vicious regress since they precisely equate a question whether
p with a question whether it is true that p. But perhaps Frege (and Dummett)
have other reasons for rejecting this alternative; and this question leads me to
my second puzzle concerning Frege's theory of truth.

II

Frege frequently makes statements in which he appears to be propounding a

redundancy theory of truth; for example, in 'On Sense and Reference' he
writes that:

One can, indeed, say: 'The thought, that 5 is a prime number, is true'. But closer
examination shows that nothing more has been said than in the simple sentence '5 is
a prime number'. (Frege 1960, p. 64)

Again, in the 1897 'Logic' manuscript he writes:

Again in the two sentences 'Frederick the Great won the battle of Rossbach' and 'It
is true that Frederick the Great won the battle of Rossbach', we have, as we said
earlier, the same thoughtin a different verbal form. In affirming the thought in the
first sentence we thereby affirm the thought in the second, and conversely. There
are not two different acts of jUdgment, but only one. (Frege 1979, p. 141)
 THREE PUZZLES IN FREGE'S THEORY OF TRUTH 5

But he never suggests these identities constitute even a partial definition of

truth; instead he seems to hold that they exhibit the fact that ascriptions of
truth are predicative expressions of assertoric force. Thus he writes in the
same 'Logic' manuscript that:

it is really by using the form of an assertoric sentence that we assert truth, and to do
this we do not need to use the word 'true'. Indeed we can say that even where we
use the form of expression 'it is true that' the essential thing is really the assertoric
form of the sentence. (Frege 1979, p. 129)

But the difficulty with this suggestion, as Frege here recognises, is that there
is no unequivocal link between use of the expression 'it is true that' and
assertoric force. In 'On Sense and Reference' he makes the point by alluding
to the use of language by actors on stage: the passage quoted above continues

The truth claim arises in each case from the form of the declarative sentence, and
when the latter lacks its usual force, e.g. in the mouth of an actor upon the stage,
even the sentence 'The thought that 5 is a prime number is true' contains only a
thought, and indeed the same thought as the simple '5 is a prime number'. (Frege
1960, p. 64)

Yet Frege never altogether abandoned the thesis that in some way truth is
essentially linked to assertion. In a late, and perhaps not altogether serious,
note ('My basic logical insights', 1915) he seems to characterise the notion
of truth as the inevitably unsuccessful attempt to express assertoric force
predicatively:

All one can say is: the word 'true' has a sense that contributes nothing to the sense
of the whole sentence in which it occurs as a predicate. But it is precisely for this
reason that the word seems fitted to indicate the essence of logic .... So the word
'true' seems to make the impossible possible: it allows what corresponds to the
assertoric force to assume the form of a contribution to the thought. And although
this attempt miscarries, or mther through the fact that it miscarries, it indicates what
is characteristic of logic .... 'True' makes only an abortive attempt to indicate the
essence of logic, since what logic is really concerned with is not contained in the
word 'true' at all but in the assertoric force with which a sentence is uttered. (Frege
1979, pp. 251-2)

This is clearly a mess, an indication that Frege was not able to bring the
intended issue sharply into focus. So the puzzle here is what essential
connection there is supposed to be between truth and assertoric force, given
that ascriptions of truth can occur unasserted. Might one not turn what Geach
called 'Frege's point' (cf. Geach 1972, p. 255) against Frege's own
conception of truth?
6 THOMAS BALDWIN

There is of course what I called before the "upward" move, that whatever
is asserted is asserted to be true. Even setting aside doubts about this move
grounded in substantive theories of truth, however, it is hard to see that this
provides grounds for the thesis that truth is specially connected with
assertion; for, as Frege's remarks about acting show, one can equally well
say that whatever is imagined is imagined to be true. Yet there is something
to Frege's thesis that truth is distinctively linked to assertion, and thus to
belief and judgment. One can readily see this from the way in which Moore's
youthful thesis that truth and falsehood are simple, indefinable, properties of
propositions left Russell, who was initially content to follow Moore on this
matter, quite unable to account for our preference for truth over falsehood:

It may be said -- and this is, I believe, the correct view -- that there is no problem at
all in truth and falsehood; that some propositions are true and some false, just as
some roses are red and some white; ..... But this theory seems to leave our
preference for truth a mere unaccountable prejudice, and in no way to answer to
the feeling of truth and falsehood. (Russell 1973, p. 75)

I think it was Frege's recognition of the need to avoid this kind of absurdity
that led him to regard "our preference for truth", i.e., the distinctive role of
truth in judgment and assertion, as the essential feature of the notion of
truth, a feature which, however ineffable, gives to it its non-redundant
content.
Frege's unsatisfactory attempts to clarify his position seem to me to
show, however, that it is a mistake to regard the internal relations between
truth, judgment, and assertion as indicating that the concept of truth has
some special assertoric content. The notion of assertoric force is instead to be
understood by reference to the role of assertion as the expression of
judgment, or conscious belief, and truth has no special role in the first
instance in identifying attitudes such as belief: for the mark of such attitudes
is their practical role in the causation of action. Yet there is a connection
between the serious attitudes and truth, in that we want our beliefs, but not
our imaginings, to be true, and a proper understanding of the matter should
not leave us regarding our preference for truth over falsehood, where beliefs
are concerned, as 'a mere unaccountable prejudice'. But the explanation for
this can start from the practical role of the serious attitudes. Because we want
our actions to succeed, and recognise that they are much more likely to do so
when guided by true, rather than false, beliefs, we want our beliefs to be true.
Nothing more than a redundancy conception of truth is required here: if I am
looking for a book on a particular shelf because I believe the book is on that
 THREE PUZZLES IN FREGE'S THEORY OF TRUTH 7

shelf, I will find the book just in case the book is where I believe it to be --
i.e., just in case my belief satisfies the redundancy theorist's condition for its
truth. So by sharing the burden of explanation, between a pragmatist
conception of belief and a redundancy conception of truth, it is possible to
elucidate what Frege sought, unsuccessfully, to account for in terms of his
conception of truth as the expression of assertoric force.
This may seem an excessively reductive treatment of the connection
between truth and assertoric force. But its merits emerge when it is compared
with Dummett's recent attempt to revive a Fregean position by proposing
that "the root notion of truth is then that a sentence is true just in case, if
uttered assertorically, it would have served to make a correct assertion"
(Dummett 1991, pp. 165-6). As this passage indicates, it is not Dummett's
view that truth is simply the expression of assertoric force; instead his view
is that the concept of truth is to be understood by reference to the intuitive
concept of the correctness of an assertion. But it then turns out that an
account of the correctness of an assertion has to draw upon an account of the
sense of the asserted sentence, and this is given by a theory which provides
that specification of the sentence's truth-conditions which captures the
content of a speaker's understanding of the sentence. So truth recurs in the
account of the correctness of assertions, and the suggestion is not, as at first
appears, that truth is to be analysed in terms of the correctness of assertions.
Instead, the proposal is just that there is an essential link between truth and
assertoric correctness.
There is, I think, no reason to reject Dummett's suggestion. But, as it
stands, it is not very illuminating, for the concepts of correctness and truth
are obviously very closely linked. By contrast, the perspective proposed
before is potentially more helpful: for where Dummett talks of the
correctness of assertions, the pragmatist talks of the success of actions, and
this latter conception, unlike the former, does not appear to rest upon the
concept of truth. Hence, once one adds that assertion is the expression of
belief, the pragmatic perspective offers a genuine clarification of the link
between assertion and truth, of the fact that, as Dummett put it in his 1959
paper 'Truth', 'it is part of the concept of truth that we aim at making true
statements' (Dummett 1978, p. 2). But where Dummett just advances this
essentialist thesis without explaining it, the pragmatist shows that it is an
implication of the fact that 'it is part of the concept of action that we aim at
success'.
I conclude therefore, that the obvious explanation for Frege's failure to
endorse a redundancy account of truth is a muddle. Despite the fact that he
8 THOMAS BALDWIN

recognizes that truth cannot be the predicative expressions of serious

assertoric force, Frege persists in thinking that the notion somehow
incorporates the distinctive feature of the serious attitudes. In fact this
distinctive feature is their practical role, and once this is acknowledged, it is
easy to see how the genuine link between the serious attitudes and truth --
our preference for true belief, or the fact that in assertion we aim at the truth
-- arises from our preference for success in action and the fact, entirely
compatible with the redundancy theory, that true beliefs greatly increase the
chances of success.

III
My third puzzle concerns Frege's thesis that the Bedeutung, or "reference", of
a sentence is its truth-value. Some critics have observed that Frege does not
explain how the concept of truth can combine the roles of being the
expression of assertoric force with that of specifying the reference of
sentences (Dummett 1978, p. 2; Furth 1967, p. iii). Having criticised
Frege's account of the first of these roles, however, I shall not pursue that
issue; instead, what concerns me is the relationship between that aspect of
hispositi01l1 in which he comes close to presenting a redundancy theory of
truth (though without explicitly doing so) and his thesis that the reference of
a sentence is its truth-value. For there seem to be two conceptions of truth in
play here -- that of the truth of thoughts and that of the truth-values of
sentences -- and it is not clear how they are to be co-ordinated.
In 'On Sense and Reference' itself Frege's exposition explicitly employs
and connects both conceptions. He argues that the fact that the sentences '5
is a prime number' and 'The thought that 5 is a prime number is true'
express the same thought shows that in ascribing truth to a thought we do
not ascribe a "predicate, or property, to the thought, since the ascription of
truth adds nothing to the thought. Instead, he urges, we should see that in
ascribing truth to a thought, we pass from the level of sense to that of
reference, to the truth-value of the sentence which expresses the thought
(Frege 1960 p. 64). We can, I think, capture the first aspect of Frege's
position here by attributing to him an analogue of the disquotational
conception of truth: in attributing truth to thoughts we employ a "descent"
operator, whereby we descend from the level of sense to that of reference. But
it appears that this role of truth does not altogether exhaust the notion, since
it recurs at the level of reference in the specification of the reference of a
 THREE PUZZLES IN FREGE'S THEORY OF TRUTH 9

sentence as its truth-value. Yet how are these two conceptions of truth to be
comprehended in a single concept?
Since, as we have seen, in Frege's own account of the role of truth as
descent operator we are led on to the role of truth as the referent of true
sentences, it looks as though the latter might well be taken to be the
fundamental conception of truth. This is certainly a feature of Dummett's
account of Frege's position, in which he places great emphasis on the central
role of the truth-conditions of sentences in Frege' s theory of sense in the
light of Frege's remark in Grundgesetze (I §32) that the thought expressed
by a sentence is the thought that its truth-conditions are fulfilled. (I discuss
this remark in more detail below.) In Dummett's account, therefore, it is the
ascription of truth and falsity to sentences which is the primary employment
of the concept of truth, and the resulting position is well exemplified by
Dummett's remark, quoted above, that "the root notion of truth is then that a
sentence is true just in case, if uttered assertorically, it would have served to
make a correct assertion". I was somewhat sceptical earlier about the
prospects for an account of the correctness of assertions sufficiently detached
from the concept of truth to make this a significant thesis; but it is primarily
the emphasis on the truth of sentences within Dummett's position that
concerns me here, and one can detach this emphasis from the idea of an
elucidation of truth in terms of assertion. The philosopher whose work most
clearly exhibits this latter possibility is, of course, Davidson; and from a
Fregean perspective it is striking that Davidson associates the fundamental
explanatory role of truth in his theory of meaning with the thesis that truth
is indefinable (Davidson 1990, p. 314).
Despite the intrinsic importance of this general line of thought, however,
it conflicts with Frege's insistence that truth is fundamentally ascribed to
thoughts, and only derivatively to the sentences which express them. Thus in
the 1914 manuscript 'Logic in Mathematics' he writes:

For brevity I have called a sentence true or faIse though it would certainly be more
correct to say that the thought expressed in the sentence is true or false. (Frege
1979, p. 233)

Passages of this kind can be found in Frege's writings throughout his life,
from the early Kernsiitze to his late paper 'The Thought'. Furthermore, this
approach is not compatible with the thesis that the role of truth as descent
operator is a fundamental mark of the concept. For where truth is conceived
of as basically predicable of sentences, thoughts will be true only insofar as
they are expressible by true sentences, and it could then only be via the
10 THOMAS BALDWIN

derivation of Tarski-style T-sentences that the ascription of truth to a thought

could be represented as approximating the descent character that Frege
emphasized, without really attaining it. Finally, even though Davidson's
emphasis on the indefinability of truth fits well with Frege's similar
emphasis on its indefinability, their conceptions of truth differ in one crucial
respect: for Davidson, truth is an indefinable property, whereas for Frege, as
we have seen, truth is not a property at all. So, although Dummett's
immensely influential account of Frege provides a powerful and attractive
way of constructing a Fregean philosophy of language, it does not do justice
to aspects of Frege's position to which Frege himself attached great
importance.
The alternative is obviously to start out from the ascription of truth to
thoughts, where this is understood in accordance with Frege's conception of
truth as the descent operator. It is then easy to characterise a sentence as true
just where the thought it expresses is true. But this does not by itself capture
the content of Frege' s thesis that the reference of a sentence is its truth-value.
This thesis is notoriously problematic, having. been forcefully criticised by
Geach and Dummett (Geach 1975, pp. 150-1; Dummett 1973, ch. 12; for a
defence of Frege, see Burge 1986); but although I am broadly sympathetic to
Frege's critics I do not want to pursue the details of their arguments here.
Instead I want to look at a passage from the 1906 manuscript 'Introduction to
Logic' in which Frege argues for his thesis:

What of a whole sentence, does this have a reference too? If we are concerned
with truth, if we are aiming at knowledge, then we demand of each proper name
occurring in a sentence that it should have a reference. On the other hand, we know
that as far as the sense of a sentence, the thought, is concerned, it doesn't matter
whether the parts of the sentence have referents or not. It follows that there must be
something associated with a sentence which is different from the thought, something
to which it is essential that the parts of the sentence should have referents. This is to
be called the reference of a sentence. But the only thing to which this is essential is
what I can the truth-value - - whether the thought is true or false. Thoughts in myth
and fiction do not need to have truth-values. A sentence containing a proper name
without a reference is neither true nor false; if it expresses a thought at all, then that .
thought belongs to fiction. In that case the sentence has no meaning. We have two
truth-values, the True and the False .... The True and the False are to be regarded as
objects, for both the sentence and its sense, the thought, are complete in character,
not saturated. (Frege 1979 p. 194) 1

Frege's line of argument here has two stages: first he argues that we cannot
characterise our serious cognitive interest, where it exists, in a sentence
simply as an interest in the sentence's sense: For the sentences of myth and
fiction have a sense even though we know that they concern things which do
 THREE PUZZLES IN FREGE'S THEORY OF TRUTH 11

not exist and thus that they cannot be of serious cognitive interest to us.
Instead, the object of serious cognitive interest is just "the truth-value --
whether the thought is true or false" -- and this, he says, is "to be called the
reference of a sentence". The second claim is, then, that these truth-values are
"objects", because, like proper names and unlike predicates, sentences are
"complete in character, not saturated". It seems clear that these points can be
separated, and that in presenting the first of them Frege explicitly introduces
the notion of a sentence's truth-value as dependent on the truth or not of the
thought it expresses. So there is no threat here to the priority of the truth of
thoughts over that of sentences. It is the second point that has drawn the fire
from Frege's critics; as I indicated above, although I am sympathetic to their
arguments, it is not to my present purpose to pursue the issue. For even if
they are wrong, Frege's presentation here of his position makes it clear that
the characterisation of the object the True as such depends on its being the
object to which a sentence refers when it expresses a true thought. So, again,
there is no threat to the priority thesis that is my main concern.
A different point which may be felt to raise such a threat is Frege' s
remark, to which 1 alluded above, in the Grundgesetze (I §32) that the
thought expressed by a sentence is the thought that its truth-conditions are
fulfilled. For this is very naturally construed as the thesis that one can give
the meaning of a sentence by means of a canonical specification of its truth-
conditions, and since this thesis takes the truth of sentences to be the
primary phenomenon in terms of which their meaning is defined, it appears
to be incompatible with any redundancy-type position according to which
truth is primarily predicable of the thoughts which constitute the meanings
of sentences. Dummett himself pointed to this apparent conflict in his
famous paper 'Truth':

It now appears that if we accept the redundancy theory of 'true' and 'false' ... we
must abandon the idea which we naturally have that the notions of truth and falsity
play an essential role in any account either of the meaning of statements in general
or of the meaning of a particular statement. (Dummett 1978, p. 7)

Although it will, I think, emerge in the end that Dummett is right about
this, we need to proceed cautiously here. An initial question is this: is
Frege's remark in the Grundgesetze consistent with his descent conception of
the truth of thoughts? It is clear that it is: for suppose that, within an
appropriate semantic theory, it is implied that the sentence S is true if and
only if snow is white. If we now add that the truth of S is the truth of the
thought expressed by S, and that this thought just is the thought that the
12 THOMAS BALDWIN

specified truth-conditions of S are fulfilled, it then follows that this thought

is the thought that snow is white; in which case the initial characterisation
of the truth-conditions of the thought expressed by S can be re-expressed in
the terms of the descent conception of truth, as the claim that the thought
that snow is white is true if and only if snow is white.
Consistency, however, is not enough; the fact that the descent conception
of truth is derivable in this way might well be taken to indicate that it is not
the fundamental mark of truth that Frege claims that it is. Thus it can appear
that there is still a conflict here, concerning the order of priorities, between
the discent conception of truth and the truth-conditions conception of
meaning. But even this apparent conflict is, I think, avoidable. For suppose
it can be ascertained that an appropriate theory implies that the thought
expressed by S is true if and only if snow is white: if we now introduce the
descent conception of truth (without deriving it from anything), it now
follows that the theory implies that the thought expressed by S is true if and
only if the thought that snow is white is true. And from this it is very
natural to infer, even if it does not follow logically, that the thought
expressed by S just is the thought that snow is white. So one can argue for
the truth-conditions conception of meaning from the presumption of the
descent character of truth, and thereby preserve what appears to be Frege's
intended order of priorities.
In setting out this last argument I have supposed that one can ascertain
the truth-conditions for a sentence, as specified within an appropriate theory,
without a prior grasp of the meaning of the sentence. Clearly, it would be
idle to advance a truth-conditions conception of meaning if one thought that
an account of the truth-conditions of a sentence required a prior grasp of its
meaning, i.e., of the thought expressed by it. Now even though, on the
descent conception, truth is fundamentally predicated of thoughts, it does not,
I think, follow that someone like Frege who advances such a conception has
to hold that we cannot ascertain the truth-conditions of the thought expressed
by a sentence without a prior grasp of the thought itself. His procedure in the
Grundgesetze is precisely the opposite of this: he here develops a systematic
theory of truth for the language he defines and then goes on to remark that
"by our stipulations it is determined under what conditions the name denotes
the True. The sense of this name -- the thought -- is the thought that these
conditions are fulfilled" (Frege 1967b, p. 90). It may still be wondered how
such a theory can be properly constructed, how one is to determine what
"stipulations" are appropriate. The answer to this will, to some extent,
depend upon the language under study -- e.g., whether it concerns
 THREE PUZZLES IN FREGE'S THEORY OF TRUTH 13

mathematics, as in Frege's case. But the general framework for finding an

answer has been investigated by Davidson, Lewis, and others under the
question of "radical interpretation" (Davidson 1984, Lewis 1983) and it is
not, I think, necessary to pursue the matter here, except to observe that the
interpretative strategies recommended by these philosophers can be detached
from their particular conceptions of truth -- e.g., in Davidson's case from
the conception of truth as a simple indefinable property of sentences. So, for
example, there is no reason why someone who upholds Frege's descent
conception of truth should not endorse Davidson's account of the way in
which one might determine the conditions under which alien speakers "hold-
true" sentences of their language, and then advance a systematic theory· of
truth for the language by relying on the principle of charity to convert the
initial account of the conditions under which sentences are "held-true" into an
account of the conditions under which they are true.
It seems, then, that the initial impression of conflict between a truth-
conditions conception of meaning and the descent conception of truth can be
dissolved. But there is a further point to be made here which brings me back
to a question I have several times raised before, namely why, despite his
acknowledgement of the descent conception of truth, Frege does not convert
this into a definition of truth, in the manner of a redundancy theorist. Our
discussion of the truth-conditions conception of meaning provides a reason
for resisting this final step. As we have seen, a truth-conditions approach to
meaning requires some presumption about the way in which one might
ascertain truth-conditions in advance of a grasp of meaning -- Davidson's
principle of charity, perhaps. Furthermore, this presumption, whatever it is,
is not going to be derivable from the descent conception of truth alone; so if
it is proper to regard it as a fundamental feature of the concept of truth, it
will follow that that conception does not exhaust the concept of truth, and
cannot therefore yield a definition of it in the way in which the redundancy
theorist supposes. Equally, therefore, Dummett's claim, quoted above, that
one cannot combine a truth-conditions conception of meaning with a
redundancy theory of truth turns out in the end to be correct (though not for
the reason he actually gives). Whether one should regard the need for a
presumption such as the principle of charity as a fundamental feature of the
concept of truth is, of course, disputable; perhaps it is not so much a feature
of the concept of truth as of our understanding of each other. But I shall not
pursue that issue; it is enough for now to have found a way of reconciling
most of the initially puzzling features of Frege' s theory of truth.
14 THOMAS BALDWIN

NOTES

1 I have here departed slightly from the published translation by translating Bedeutung as
'reference' (or 'referent', according to context) instead of as 'meaning'. This change is only
intended to bring this passage into line with the conventional translation of Bedeutung as
'reference' which I am following; I myself think that 'significance' is actually a preferable
translation, but it is too late now to change the established convention.

REFERENCES

Burge, Tyler: 1986, 'Frege on Truth' in L. Haaparanta & J. Hintikka (eds.), Frege
Synthesized, Reidel, Dordrecht.
Davidson, Donald: 1984, 'Radical Interpretation', in Inquiries into Truth and Interpretation,
Clarendon, Oxford, pp. 125-141.
Davidson, Donald: 1990, 'The Structure and Content of Truth', Journal of Philosophy 87, pp.
279-328.
Dummett, Michael A. E.: 1973, Frege, Philosophy of Language, Duckworth, London.
Dummett, Michael A. E.: 1978, Truth and Other Enigmas, Duckworth, London.
Dummett, Michael A. E.: 1981, The Interpretation of Frege's Philosophy, Duckworth, London.
Dummett, Michael A. E.: 1991, The Logical Basis of Metaphysics, Duckworth, London.
Frege, Gottlob: 1960, Translations from the Philosophical Writings of Gottlob Frege, ed. &
trans. P. Geach & M. Black, 2nd. ed., Blackwell, Oxford.
Frege, Gottlob: 1967a, 'The Thought', trans. A.M. & M. Quinton, in P. Strawson (ed.),
Philosophical Logic, Oxford University Press, London.
Frege, Gottlob: 1967b, The Basic Laws of Arithmetic, ed. & trans. M. Furth, University of
California Press, Berkeley.
Frege, Gottlob: 1979, Posthumous Writings, H. Hermes, F. Kambartel & F. Kaulbach (eds.),
trans. P. Long & R. White, Blackwell, Oxford.
Furth, Montgomery: 1967, 'Editor's introduction' to Frege (1967b).
Geech, Peter: 1972, Logic Matters, Blackwell, Oxford.
Geech, Peter: 1975, 'Names and Identity', in S. Guttenplan (ed.), Mind and Language,
Clarendon, Oxford.
Lewis, David: 1983, 'Radical Interpretation', in Philosophical Papers 1, Oxford University
Press, London, pp. 108-121.
Lotze, R. H.: 1984, Logic, trans. B. Bosanquet, Clarendon, Oxford.
Moore, G. E. : 1989, 'The Nature of Judgement', Mind, n.s. viii, pp. 176-193.
Russell, Bertrand A. W.: 1973, 'Meinong's Theory of Complexes and Assumptions III', in D.
Lackey (ed.), Essays in Analysis, Allen & Unwin, London.
Sluga, Hans: 1980, Frege, Routledge, London.

Clare College
Cambridge CB2 lTL, UK
 GABRIEL SEGAL

TRUTH AND SENSE

In Frege's account of reference we find that the referent of a singular term is

the object that it applies to, the referent of a sentence is its truth value and
the referent of a predicate is a function that maps referents of singular terms
onto truth values. It is attractive and natural to think of a Tarskian truth-
theory ("T-theory") for a particular language as providing a theory of
reference for that language (Tarski 1956). A T-theory for a language, L,
consists in a finite number of rules assigning semantic values -- referents --
to the simple expressions of L, a finite number of rules for deriving the
values of complex expressions from the values of their simpler components
and their syntactic configuration, and a finite number of deduction rules for
carrying out such derivations.
The T-theory tells us as much as we would want to know about the ref-
erence of all the various expressions. For example, it deals with singular
terms by telling us which objects they apply to. It deals with predicates by
telling us which objects satisfy them: and, for a Fregean, this is much the
same as telling us which function from objects to truth values they refer to.
And, though it doesn't tell us directly what the truth-value of every sentence
is, it does give us a T-theorem for every sentence, which at least gives us an
extensionally correct application condition for the predicate "is true" to that
sentence. It won't tell us, for example, that the French sentence "Les
chameux ont des bosses" is true. But it might well tell us:

(1) "Les chameux ont des bosses" is true iff camels have humps

And, of course, the whole theory provides an extensionally correct definition

of the predicate "is true" for every sentence in the language.
A T-theory, then, can provide a satisfying theory of reference for a lan-
guage. But what would provide a theory of sense for a language? In this
paper I want to examine a very interesting answer to that question. The
answer is that a T-theory can transcend its duty as a mere theory of reference
and do duty as a theory of sense.
I begin with a point made by Gareth Evans, elaborating on a remark of
Michael Dummett's. Dummett (1973, p. 227) points out that:

[W]hen Frege is purporting to give the sense of a word or symbol what he actually
states is what the referent is .... The sense of an expression is the mode of

15

J. Biro and P. Kotatko (eds.), Frege: Sense and Reference One Hundred Years Later, 15-24.
© 1995 Kluwer Academic Publishers.
16 GABRIEL SEGAL

presentation of the referent: in saying what the reference is, we have to choose a
particular way of saying this .... In a case in which we are concerned to convey ...
the sense of an expression we shall choose that means of stating what the referent is
which displays the sense.

Evans (1982, p. 26) goes onto explain:

The closest we shall get to a systematic theory of sense is a systematic theory of

semantic value [Le. reference] which, however, identifies the semantic values of
expressions, including whole sentences, in the way in which competent speakers of
the language identify them.

The idea of specifying the semantic value of a sentence by using a sentence

that has the same sense is of course prominent also in Tarski's work. Tarski
tells us that the extensional correctness of a T-theory can be assured by the
requirement that in each T-theorem the metalinguistic sentence appearing on
the right of the biconditional translates -- or, we might say, has the same
sense as -- the object-language sentence mentioned on the left. (1) meets this
requirement, since "Camels have humps" translates "Les chameaux ont des
bosses". Let us call any T-theorem that meets the translation requirement an
"interpretive" T-theorem and any T-theory that entails only interpretive T-
theorems an "interpretive" T-theory. So, an interpretive T-theory entails an
interpretive T-theorem for every object-language sentence, and entails no
uninterpretive T-theorems. We can now reformulate Evans' point as the
suggestion that an interpretive T-theory can do duty as a theory of sense.
Frege's notion of sense and the intuitive notion of meaning are obviously
kindred. Abstracting from the details of Frege's theory, the sense of a
sentence is its content, it is what the sentence says, or what is understood by
someone who understands the sentence. The sense of "Les chameaux ont des
bosses" -- what it says -- is that camels have humps. And, of course, that is
also what it means. Fregean sense and meaning are close enough, I think, to
warrant reformulating the question about a theory of sense as one about a
theory of meaning: What would provide a theory of meaning for a language?
The answer then becomes reformulated as: "An interpretive T-theory for a
language can do duty as a theory of meaning for that language".
What we have arrived at, of course, is a claim that has been defended at
length in the work of Donald Davidson. 1 What I want to do in the main part
of this paper is to defend a version of Davidson's thesis. But it is a version
that is very different in style and spirit from Davidson's own. Indeed it is a
version that I suspect Davidson himself would not want to associate with. At
 TRUTH AND SENSE 17

the end I shall briefly return to discuss the relationship between a theory of
meaning of the sort I shall be describing and a theory of Fregean sense.
What, then, is a theory of meaning for a particular language? For
Davidson it is a theory knowledge of which would suffice for understanding
the language. So if you knew the theory, you could use it to deduce the
meaning of every sentence of the language. The theories of meaning that I
shall be interested in are a subset of theories of meaning in Davidson's sense.
For I think it is reasonable to suppose that the actual semantic abilities of
real speakers are owed to their knowledge of theories of meaning for their
language. What explains, for example, my understanding of English
sentences is my knowledge of a body of rules and principles and of a
deductive apparatus that between them generate interpretations of all the
sentences I understand. This knowledge is certainly not conscious. For I am
simply not aware of any body of knowledge remotely sophisticated enough
to account for my understanding of English. But unconscious knowledge is
none the worse for being unconscious. So, now, if my semantic competence
is constituted by my (unconscious) knowledge of a theory of meaning, the
question arises as to what form this theory takes. And, following Davidson, I
suggest that it takes the form of an interpretive T-theory.
Now Davidson's suggestion that knowledge of an interpretive T-theory
for a language would suffice for understanding the language runs immediately
into a large difficulty. Suppose that you were in possession of an interpretive
T-theory for French, written out in English. And suppose further that this T-
theory provided all the information you had about French semantics. Would
you know enough to understand French? The answer is: "certainly not".
Suppose you came across the French sentence "Les chameaux ont des
bosses". Can you use your theory to deduce what the sentence means? No.
You can derive the T-theorem (1). But that tells you virtually nothing about
the meaning of the sentence. It just tells you something about its truth
value. For the T-theorem is purely extensional: the "iff' is the material
biconditional and requires only that the sentence on its right have the same
truth value as the sentence on its left. And the "is true" means no more than:
is true .. If you know that camels have humps, you can deduce that the French
sentence is true. But that's about all. And that doesn't distinguish "Les
chameaux ont des bosses" from "Les oiseaux ont des plumes" (birds have
feathers) or "Les fleurs sont perissable" (flowers wilt).
The problem is simple and sharp. Although your truth-theory is interpre-
tive, nothing in the theory itself says or implies that it is. 2 If there is one
interpretive T-theory for French in English, there are also many
18 GABRIEL SEGAL

uninterpretive ones. And you have no way of knowing whether or not yours
in one of those. Given this, you are in no position to deduce anything about
what any French sentence means. The problem is that the T-theory contains
far less information than you need to understand French. So let us call the
problem "The Information Problem".
The gap between what a T-theory enables you to deduce and what you
would need to know in order to understand French is not smalL For consider
the further knowledge you would need before you could use your T-theory to
deduce interpretations of French sentences. You would need to know that
your T-theory is interpretive -- that each of its T-theorems is not only true,
but that its right-hand side provides an interpretation of the object-language
sentence it mentions on the left. But knowing this is just knowing such
things as that "Les chameaux ont des bosses" means that camels have humps
and does not mean, for example, that flowers are perishable. And now it
looks as though what you would need, in addition to your truth-theory, is a
theory that would tell you what French sentences mean. And that would be a
theory of meaning for French. So it looks as though the information you
would need to add to your T-theory to enable you to use it as a theory of
meaning would be a theory of meaning. But then your T-theory is out of a
job: you could just use the theory of meaning directly to interpret French
sentence& and have done with the T-theory altogether.
So it appears that Davidson's suggestion that an interpretive T-theory can
do duty as a theory of meaning and my suggestion that interpretive T-
theories actually do duty for theories for meaning are completely scotched.
Davidson's own response to the problem is perhaps not entirely clear.
Certainly, different commentators have offered very different interpretations
of what Davidson has to say on the issue. Unfortunately I do not have the
space here to discuss this worthwhile topic. 3 What I want to do instead is to
propose a response of my own,one that is quite different from Davidson's.
The hypothesis that our semantic competence is constituted by our
unconscious knowledge of a body of semantic theory fits naturally within the
larger framework of the empirical discipline of cognitive linguistics as de-
veloped by Noam Chomsky.4 Chomsky'S guiding idea is that our linguistiC
competence as a whole is constituted by unconscious knowledge of a lin-
guistic theory. The linguistic theory will contain sub-theories pertaining to
different areas of language. For example there will be a body of specifically
syntactic rules that assign syntactic structures to the surface strings of the
language. These rules will enter into the explanation of our judgements of
 TRUTH AND SENSE 19

grammaticality. They will help explain, for example, why we take (2) to be
grammatical and meaningful, and why (3) appears ill-formed:

(2) What did John say?

(3) * What do you believe the claim that John said?

And there will be a body of phonological rules that allow us to identify se-
quences of sounds as strings of words and help explain the sound patterns we
produce. For example, they will explain why the plural ending in "cats" is
pronounced differently from the plural ending in "dogs". It is tempting to
believe that syntax and phonology are not the only systematic and rule-
governed areas of language. It is tempting to think that semantics should be
included as well -- although Chomsky himself appears not to think so.
My solution to the problem of how a mere T-theory can do duty as a
theory of meaning hinges on the point that there is a considerable gap be-
tween linguistic knowledge, including semantic knowledge, and the
linguistic acts of speech and understanding the knowledge allows us to
perform. Chomsky has emphasised the existence of this gap between
competence and performance. Before linguistic knowledge issues in
performance, it has to be applied. And the application of knowledge to a
particular task may itself be a complex process. The two main areas of
application of linguistic knowledge are speech production and the
understanding of perceived speech. The question of how linguistic knowledge
is applied in the production of speech is deeply perplexing. Indeed, virtually
nothing is known about it. When we speak, we utter words in order to
express our communicative intentions. Someone wants to know if Madonna
and Shaun Penn are still married. She says "Is Madonna still married to
Shaun Penn?" Surely her knowledge that those words express the question
she wishes to ask enters into the explanation of why she uttered them. But
how? What processes were involved in selecting those particular words?
Why, for example, didn't she ask "Are Madonna and Shaun Penn still
married?" As far as I am aware, virtually nothing is known about the
processes of speech production.
The processes of understanding perceived sentences are slightly less mys-
terious. We can distinguish at least three different factors here. First, there is
parsing: when you hear an utterance of a sentence, you need to identify the
speech sounds as words, you need to identify the syntactic configuration in
which the words are combined, and you need to identify the meanings of the
20 GABRIEL SEGAL

words and deduce the meaning of the sentence uttered. What is not so clear is
how all this is done. For example, we might imagine that the process is se-
rial: first you identify the phonetics, then the phonology, then you assign a
syntactic structure, then you compute the semantics. But equally the process
might be staggered or parallel. A theory of parsing would explain these mat-
ters.
The second factor involved in understanding involves the application of
purely linguistic knowledge -- knowledge of the language -- to the particular
context at hand. Purely linguistic knowledge is just knowledge of the more
or less permanent and context-independent features of the language you
know. BUlt this knowledge by itself is often insufficient to tell you even the
strict and literal meaning of an utterance in context. For example if someone
utters (4), your linguistic knowledge will tell you, very roughly, that some
contextually identified female gave some contextually demonstrated object to
some contextually identified male at some time prior to the time of
utterance:

(4) She gave this to him

But knowledge of the language alone will not tell you who the giver is, nor
the recipient, nor the gift. Thus to arrive at a full interpretation of the
utterance -- something that has a truth value -- you need to combine your
knowledge of language with your knowledge of the context.
The third factor in understanding is pragmatics. When you hear an utter-
ance, you are often able to discern not only the literal meaning of the words
uttered in the context. You are able also to discern the communicative inten-
tion behind the utterance: what the speaker is saying. To borrow an example
of Bill Lycan's: Peter Wimsey addresses his man, Bunter, with the words:

(5) It is cold in here

Bunter understands this as a command to close the window and immediately

complies. Here again there is an obvious gap between Bunter's knowledge of
language and his understanding Wimsey's command.
How does all this help with The Information Problem? Well, the reflec-
tions on the application of linguistic knowledge suggest the following pic-
ture. Linguistic knowledge is represented somehow, somewhere, in the brain.
This knowledge is deployed by various processors that draw upon it and use
it to figure out the solutions to various problems. Some of these processors
 TRUTH AND SENSE 21

are directly involved in drawing out the immediate consequences of the

linguistic knowledge: these processors just perform whatever deductions are
directly licensed by the stored knowledge. Call these, "purely linguistic
processors". Let's suppose that there are, inter alia, a phonological processor,
a syntactic processor and a semantic processor. These processors -- serially or
in parallel or however it is -- are able to draw out all, or at least many, of the
direct consequences of the stored linguistic knowledge. But, as we have seen,
there remains a large gap between the outputs of the purely linguistic
processors and our linguistic actions: for example, judgements of the literal
meanings of the words uttered in a context, judgements about what a speaker
is saying when she utters a particular sentence in a context, and our own
choices of words to express our thoughts in speech. Further processors are
required to draw on the outputs of the purely linguistic processors. These
further processors receive the outputs of the linguistic processors as inputs,
and use these, along with other information, to produce further outputs of
their own. It is these latter outputs that are close to the surface of our
behaviour: judgements of what is said, choices of words and so on. This
picture now allows for a solution to The Information Problem.
An interpretive T-theory can do duty as a theory of meaning if the job of
the semantic processor is merely to deploy an interpretive T-theory to
produce interpretive T-theorems. If these outputs -- interpretive T-theorems--
could serve as inputs to the further processors involved in speech and
understanding, then that is all that is required. My suggestion is that, indeed,
such outputs are all that is required. They will do just fine. They will do just
fine provided that the further processors that draw upon them assume that the
T-theorems are, indeed, interpretive. That is to say, the further processors
must be hardwired to treat the right hand sides of the T-theorems as giving
the meaning of the object-language sentences on the left. Given aT-theorem
as input, the further processor will deploy it on the assumption that it is
interpretive: it will use the right hand side as an interpretation of the sentence
on. the left.
The proposal, then, is this. Semantic knowledge is knowledge of an in-
terpretive T-theory. Learning the semantics of a language is building up an
internal representation of a T-theory. Using semantic knowledge involves
deploying the T-theorems of your internalized T-theory on the assumption
that the theorems are interpretive. This assumption is hardwired into the
processors that use the outputs of the semantic module to help you speak and
understand what people are saying. That is how a mere interpretive T-theory
22 GABRIEL SEGAL

can do duty for a theory of meaning. In fact, on this view, a theory of

meaning just is an interpretive T-theory.
On this proposal The Information Problem is solved by a sort ofJiat: the
information that your internalized T-theory is interpretive is just hard-wired
into your functional architecture. This can make it look as though it is all
just too easy. Above we saw that the gap between possession of an
interpretive T-theory and knowledge that it is interpretive is a large one. It
appeared to require a whole theory of meaning to plug it. So perhaps it looks
as though I am appealing to something like magic: where we seem to need a
theory of meaning to do a job, I am saying we have a simple assumption
that we get one for free. But a little reflection should allay this worry. For
the hard work in ensuring that the T-theory you have is interpretive all goes
into the process of learning the language. Learning a language involves
acquiring axioms for the words and constructions of the language that will
yield interpretive T-theorems. And doing this is evidently quite a feat. To
take a very simple case, when you learn the word "rabbit", you will need to
acquire some rule along the lines of (6):

(6) (x)(x satisfies "rabbit" iff x is a rabbit)

It's no good acquiring an axiom that, combined with the rest of your T-the-
ory, will yield uninterpretive T-theorems -- even if they are true. So, for ex-
ample, you must not end up with (7) instead of (6):

(7) (x)(x satisfies "rabbit" iff x is a rabbit and 2+2=4)

If you did, then your T-theory would likely have (8) as a consequence:

(8) "Cottontail is a rabbit" is true iff Cottontail is a rabbit and 2+2=4

In order to avoid ending up with (7) encoded in your semantic module you
must either be innately constructed so as not to make this extravagant hy-
pothesis in the first place, or you must be able to test the T-theorems that
result from it for interpretivity. You must be able to discover that these T-
theorems are not interpretive and revise the axiom accordingly.
I conclude by returning briefly to the original issue: the theory of sense.
Is a theory of meaning of the sort I have described a theory of Fregean sense?
The answer is: not exactly .. The sense of a sentence -- the thought it
expresses -- must be complete, it must be truth-evaluable (see, e.g., Frege
 TRUTH AND SENSE 23

1956). But sentence-types containing indexical elements (such as (4) above)

are not truth-evaluable independently of context. Only particular utterances of
such sentences in context are truth evaluable and express thoughts. But a
theory of meaning is only a theory of the context-independent and fixed fea-
tures of expression types. It doesn't by itself allow you to deduce the truth-
conditions of utterances in context. It therefore doesn't by itself provide a full
theory of the senses of utterances of sentences with indexicals. 5
A theory of meaning is really a theory of something like what David
Kaplan (1989) calls "character": a function from contexts of utterance to the
contents that are expressed by utterances in the contexts. The contents of ut-
terances are what correspond to Fregean senses.
There is no space here to examine the relations between character and
sense, nor to explore how one might attempt to derive a theory of sense from
a theory of character. But it is clear that in many cases there is no important
difference between the two. Many complex expressions and sentences are free
of indexical elements. These expressions express the same thing on each
occasion of their use. And for these expressions we can simply identify
character and sense. 6 So, when it comes to context-independent expressions,
we can straightforwardly see interpretive T-theories as theories of sense.?

NOTES

1 Davidson (in particular, 1967,1973, 1976). Evans draws the connection between Davidson's
proposal and Frege (Evans 1982, p. 34). John McDowell discusses it in McDowell (1977). For
Davidson's views on this matter, see the opening sections of Davidson (1967).
2 This problem was perhaps first dearly articulated by John Foster (1976).
3 Davidson responds directly in Davidson (1976), in a footnote (footnote 11) added to the
version of Davidson (1967) reprinted in Davidson (1984), and in the introduction to Davidson
(1984). In fact I think his response is just this: any compositional, true, truth theory that is both
lawlike and maximally simple is interpretive. Thus what you would need to understand a
language, L, is a theory that had all these properties, combined with the knowledge that it has
them. .
4 See, e.g., Chomsky (1965, 1975, 1980). The idea of integrating Davidson's T-theoretic
format for a theory of meaning into a larger cognitivist linguistic theory has been suggested by
Gilbert Harman (1972) and William Lycan (1974). The project has been pursued in depth and
detail by James Higginbotham in a series of papers (see, in particular, (1985), (1986».
5 The typical T-theoretic treatment of indexicals requires a modified format. The theory no
longer produces T-theorems that specify outright truth conditions for sentences. Rather it
produces theorems that function like recipes for deducing truth-conditions of sentences from
further contextual information. See Weinstein (1974); Burge (1974). Very roughly: we move
to relativised T-theorems along the lines of (i):
(i) If u is an utterance of "that is peculiar" and x is the object demonstrated by the utterer of u,
then u is true iff x is peculiar.
24 GABRIEL SEGAL

What we would need to arrive at something better resembling aT-theorem would be some
principled way of deriving sentences of the form of (ii) and thence (iii):
(ii) the object demonstrated by the utterer of u=o
(iii) u is true iff 0 is peculiar.
For discussion of the kind of theory that might achieve this, see Francois Recanati's
contribution to this volume.
6 Compare Kaplan (1989) p. 507.
7 All of the leading ideas relating to the recasting of Davidson's suggestions about truth and
meaning in terms of processing modules are drawn from Larson and Segal, forthcoming,
chapters one and two. I am much indebted to Richard Larson for all his contributions to that
portion of our joint work.

REFERENCES

Burge, Tyler: 1974, "Demonstrative Constructions, Reference and Truth", The Journal of
Philosophy.
Chomsky, Noam: 1965, Aspects of a Theory of Syntax, MIT Press, Cambridge, Mass.
Chomsky, Noam: 1975, Reflections on Language, Pantheon, New York.
Chomsky, Noam: 1980, Rules and Representations, Columbia University Press, New York.
Davidson, D.: 1967, "Truth and Meaning", reprinted in Davidson (1984).
Davidson, D.: 1973, "Radical Interpretation", reprinted in Davidson (1984).
Davidson, D.: 1976, "Reply to Foster", reprinted in Davidson (1984).
Davidson, D.: 1984, Inquiries into Truth and Interpretation, Oxford University Press, New
York.
Dummett, Michael: 1973, Frege: Philosophy of Language, Duckworth, London.
Evans, Gareth: 1982, The Varieties of Reference, Oxford University Press, Oxford.
Foster, John: 1976, "Meaning and Truth Theory", in G. Evans and J. McDowell, Truth and
Meaning, Oxford University Press, Oxford.
Frege, Gottlob: 1956, "The Thought: A Logical Inquiry", trans. A.M. and Marcelle Quinton,
Mind. (Originally published in 1918).
Harman, Gilbert: 1972, "Deep Structure as Logical Form", in D. Davidson and G. Harman
(eds.), Semantics of Natural Language, Reidel, Dordrecht.
Higginbotham, James: 1985, "On Semantics", Linguistic Inquiry.
Higginbotham, James: 1986, "Linguistic Theory and Davidson's Program in Semantics", in E.
LePore (ed.), Truth and Interpretation: Perspectives on the Philosophy of Donald
Davidson, Basil Blackwell, Oxford.
Kaplan, David: 1989, "Demonstratives", in J. Almog, J. Perry, and H. Wettstein, Themesfrom
Kaplan, Oxford University Press, Oxford.
Larson, R. and G.Sega1.: forthcoming, Knowledge of Meaning: Logical Form and Semantic
Value, MIT Press, Cambridge, Mass.
Lycan, William: 1984, Logical Form in Natural Language, MIT Press, Cambridge, Mass.
McDowell, John: 1977, "On the Sense and Reference of a Proper Name", Mind.
Tarski, Alfred: 1956, "The Concept of Truth in Formalized languages", in Tarski, Logic,
Semantics, Mathematics: Papers from 1923-38, Oxford University Press, Oxford.
Weinstein, Scott: 1974, "Truth and Demonstratives", Nous.

Kings College, University of London

Strand, London, WC2R 2LS, UK
 BARRY SMITH

FREGE AND CHOMSKY: SENSE AND PSYCHOLOGISM

In the introduction to the Grundlagen, Frege tells us that it is one of the fun-
damental principles of his theory of knowledge "always to separate sharply
the psychological from the logical, the subjective from the objective" (1884,
p. x). This methodological maxim was intended to produce a division of
labour between the philosopher and the psychologist, and to clear the ground
for the distinctively analytic approach to philosophy. For Frege believed that
we-could study thoughts independently of thinkers, and that we could map
out the logical relations among thought-contents in a wholly objective way
that would purge philosophy of any empirical taint. By this method the
ideals of analytic philosophy were established: "rigour of proof, precise
delimitation of extent of validity, and as a means to this. sharp definition of
concepts". (1884, p. I)
Theories of thought were to be distinguished from theories of thinking,
which had nothing to do with the logician's task but belonged instead in the
domain of empirical psychology. But this separation of philosophy from
psychology left Frege with a puzzle whose solution required some acknowl-
edgement of the limitations on the analytic programme. In what follows I
shall describe some strands of Frege's anti-psychologism, accepting some
and rejecting others. and in this way I shall argue that we have to temper the
original anti-psychologistic claims if we are to see Frege's doctrines as stiU
shaping the domain and methods of the analytic philosopher.
The puzzle for Frege arose with his conception of mind-independent
thoughts. Thoughts, and their constituents, senses. were not the contents of
consciousness. But grasping a sense was a mental act. So how could the
mind of a thinker come to be related to those mind-independent entities? The
answer to this question also provided the key to the question of how we were
supposed to study these pure and objective thoughts: viz., we have mastery
of a language in which such thoughts can be expressed. Let us look at this in
more detail.
Senses are not mental contents but grasp of sense is a mental act. So,
then, just how can an act of mind put us into contact with that which lies
wholly outside it? This may seem less of a problem to those who have re-
membered that this IS precisely what our perceptual systems do for us all the
time. Perception is a process by which the mind relates us to things in the
world which lie outside us and do not depend on us for their existence. But
this solution is not available to Frege. Firstly, senses, for Frege, are abstract

25
1. Biro and P. Kotm/w (eds.), Frege: Sense and Reference One Hundred Years Later, 25-46.
C 1995 Kluwer Academic Publishers.
26 BARRY SMITH

entities, and it is far from clear what a perceptual encounter with one of these
is meant to be like. Secondly, Frege, unlike GOdel, had no truck with the
idea of a special faculty of the mind equipped to give us quasi-perception of
the abstract. Moreover, in the ordinary cases of perceiving spatio-temporal
particulars, Frege took it that perceiving an object was one thing, but
thinking of it was another. When we think of an object, we always think of
it in a particular way and think something about it. So if perception of the
external world is to be a source of thoughts about objects in that world, our
perceptions must be accompanied by a conceptual act in which we grasp the
sense or mode of presentation in which the object is given to us. So even in
the case of perceptually-based thoughts, grasp of a sense is in addition to
perception of an object, and not reducible to it.
But, some will be inclined to ask, is it any concern of Frege's, or the
philosopher's, what it is to grasp a thought? After all, in the division of
labour doesn't this belong to the theory of thinking and thus fall within the
domain of the empirical psychologist? Frege wavers at this point, for he
writes:

... grasping is a mental process! Yes, indeed, but it is a process that takes place on
the very confines of the mental and which for that reason cannot be completely
understood from a purely psychological standpoint. For in grasping ... something
comes into view whose nature is no longer mental in the proper sense, namely the
thought; and this process is perhaps the most mysterious of all. (1979, p. 145)

So there is work here for the philosopher to do.

Frege's solution to his puzzle about grasp owes a lot to his mathematical
epistemology in the Grundlagen. Numbers were not spatio-temporal objects
of the sort with which we could have a perceptual encounter; they could only
be known to us though our comprehending use of numerical terms used to
refer to them. In similar fashion, our grasp of pure thoughts or senses was
mediated by our mastery of a language in which such thoughts and senses
were expressed. This meant that our access to, and indeed our only means of,
analysing thoughts was via the analysis of a language in which they could
be expressed. Thus the methods of the analytic philosopher involved taking a
linguistic tum.
Frege's interest in language and his fashioning of a conceptual notation
for rendering proofs with absolute precision was a response to his repudiation
of psychologism and his need to delimit the methods of objective inquiry.
This linguistic tum did not involve any claim about the priority of language
over thought: the relative priority of language was in the order of explanation
 FREGE AND CHOMSKY 27

only. So although Frege's real interest was in thought-content, he conceded

that his only route to studying it was via an understanding of the language in
which those thoughts were clothed. Thoughts are not essentially linguistic,
since as Frege tells us, "there is no contradiction in supposing there to exist
beings that can grasp the same thought as we do without needing to clothe it
in a form that can be perceived by the senses." (1979, p. 269) But it is not
like this for us: "That a thought of which we are conscious is connected in
our mind with some sentence or other is for us men necessary" (ibid.) So
while thoughts are not themselves essentially linguistic, they are essentially
linguistic for us. 1
That Frege was somewhat dissatisfied with these limitations on the
analytic project is well-known. He wanted to penetrate to thought-content
proper without the unwanted distractions of a logically imperfect language.
He was keen to remind us that

The use of language requires caution. We should not overlook the deep gulf that yet
separates the level of language from that of thought, and which imposes certain
limits on the mutual correspondence between the two levels. (op. cit., p. 259)

But however imperfect it may be to have to work with an intermediary, some

linguistic vehicle is required to go proxy for the Platonistically conceived
realm of thought; and the study of the structure of thought had to proceed,
with due caution, via the study of the semantic structure of language, itself
something which was not fully present to the naked or untutored eye.
It is worth noting at this point that the Fregean principle which asserts
the explanatory priority of language over thought is not restricted to what we
might call 'language-first theorists', those philosophers, that is, who wish to
explain our very capacity to think in terms of our capacity for using and
understanding a public language. For even Jerry Fodor, in Psychosemantics,
subscribes to the Fregean principle, by advocating a language of thought.
The argument for the LOT hypothesis, remember, depends on an analogy
between the productivity and systematicity of thought and the productivity
and systematicity of natural language. As Fodor has it, there is a certain
property that linguistic capacities have which is due to the fact that natural
languages have a combinatorial semantics, and since thought has this
property too, thought must also have a combinatorial semantics (Fodor
1987, p. 148). So certain 'thought-first theorists' will still come within the
scope of the arguments I give below. I say this to draw attention to what
would be required if one were to give up what Michael Dummett has called
Frege's fundamental principle of analytic philosophy -- that any, account of
28 BARRY SMITH

thought must go via an account of the language in which that thought is

expressed. Simply moving to the level of thought is not enough.
Having characterized the way in which, for Frege, both the project of the
analytic philosopher and the epistemology of the ordinary thinker bear an
essential connection to language, we are now in a position to consider the
dangers to philosophy, as he sees them, of psychologism. For Frege, sense
determines reference; and so if sense owes anything to the psychological fac-
tors involved in our grasp of sense, then the determination of reference, and
therefore the settling of truth, would have to depend on psychological con-
siderations also.
As an example, consider Bertrand Russell's account of reference in his
logical atomism. This can be termed psychologistic, since the semantic
property of an expression, whether it refers or not, as well as the semantic
category it belongs to, depends on the relation of the speaker to the object of
reference -- on whether he is acquainted with it or not. Thus reference rela-
tions depend on cognitive relations, and the determination of reference goes
psychological considerations as to whether the object in question is in-
corrigibly known or not. Russell's logical atomism thus still seems to owe
something to psychological atomism.
Why is this thought to be pernicious? Well, in answering this, we should
do well to note that there are many theses Frege had as the targets of his anti-
psychologism. Chief amongst them was the threat to the objective subject
matter of mathematics that Frege foresaw if the contents of our judgements
(or later, the senses of our terms) were to be identified with the psychological
states of the thinkers who grasped those contents. There could be no
communication and no genuine disagreement if what our terms referred to
were·the subjective items of consciousness.
But why should we think this likely just because we identify the thoughts
our words express with certain psychological states of speakers and hearers?
After all, there is no restriction on the nature of psychological states which
obliges them to be about nothing other than the subjective experiences of the
thinker who is in those states. In Section 27 of the Grundlagen Frege comes
close to stipulating an understanding of the psychological as restricted to
subjective ideas, sensations, mental images and the like. In this vein he
argued that while it is true that I cannot share your sensations or have your
experiences, you and I could think the very same thought. A thought, since
it could be shared by more than one thinker was not the property of a
thinker, but was to be seen as independent of the minds of all thinkers,
"standing over against all who grasp it". (1979, p. 133). Thus Platonism
 FREGE AND CHOMSKY 29

about the contents of judgements was the response to the privacy of mental
items when compared with the communicability of shared thoughts.
But why should we be restricted to subjective items of mental life? There
are more objectively specifiable psychological states, such as beliefs and de-
sires, which are not always, or not obviously, just occurrent conscious
states. Rather, they can take on a more dispositional character, extending
constraints to future actions and setting up logical relations to other mental
states. Frege himself tells us that what is objective is what is subject to
laws, and arguably in empirical psychology there are psychological laws.
Therefore nothing said so far impugns the wide range of law-governed psy-
chological states with semantically evaluable contents which are about
things in the world other than themselves and the conscious mental
experiences of those who have them.
So we can agree with Frege that any variety of psychologism that equates
the psychological with the purely subjective is too restrictive and would have
to be rejected. The real trouble is. that his response to psychologism appears
to be wide of the target. For the claim that thoughts do not depend for their
existence on any particular thinker does not licence the inference that
thoughts are therefore independent of all thinkers. This mind-independence
and its platonistic shadow can be rejected in favour of a notion of a more
moderate intersubjectivity, as Dummett points out. 2 Thoughts are not tied
essentially to those who think them, but thought can still be mind-
dependent.
A second strand of psychologism also threatens the objective subject
matter of mathematics, as Frege sees it. This is the identification of the laws
of logic with the laws of thought, or more precisely, with the laws of
thinking. This would prove disastrous for Frege's project of providing firm
foundations for mathematics. For if the laws of logic were treated as psycho-
logical laws describing the workings of the mind, then the attempt to reduce
arithmetic to logic would collapse into a set of empirical hypotheses that
would impugn the necessity of arithmetical truth. So the heart of this objec-
tion to psychologism is that it reduces the status of mathematical truth to
mere contingency: if logic depends on the way the mind works, then if this
were to change, so would the truths of logic and therefore of mathematics.
One might respond to the immediate threat as follows. In any naturalistic
account of the mental, the principles'which govern the workings of the mind
must be derived ultimately from the laws of nature. These laws are not so
susceptible to change. Indeed some have wished to argue that natural laws are
necessarily true. I shall not follow up this liue of response, but I point it out
30 BARRY SMITH

just to forestall any quick move from the identification of the laws of logic
with psychological laws to a conclusion that the laws of logic and the truths
of mathematics could easily change.
More lPertinent to my concerns, though, is a version of Frege's objection
that focuses on the distinction between the descriptive claims of psychology
and the prescriptive force of logic. For, according to Frege, the laws of logic
do not aim to tell us how people actually reason, but, rather, how they
should reason when drawing valid inferences. The identification of the laws
of logic with psychological laws of reasoning might seem to threaten this
prescriptive or normative aspect of logical practice. This is an important
issue and I shall return to it below.
Third, Frege voices his most general concern with psychologism when he
insists that process is irrelevant to product. That is, a theory of thought
products can be given without making any reference to, and without knowing
anything about, the processes by which we arrive at those thoughts. Once
again, this amounts to a plea to preserve the subject matter of an objective
logical enquiry. For Frege warns us not to confuse questions about whether
an assertion can be justified with questions about the way we happened to
arrive at the judgement in the first place. (1884, p. 3). The latter concern the
process of thinking by which we arrived at the thought as product; they are
the concern of the psychologist and have nothing to contribute to a proper
study of thought-contents and their logical relations.
So, to recap, we have three related strands in Frege's anti-psychologism:

(1) The need to avoid subjectivism and the erosion of truth.

(2) The need to preserve the standard of logic against the contingencies of
human reason.
(3) The needto distinguish a theory of thought-processes from a theory of
thought-products.

All three strands contribute to Frege's doctrine of the mind-independence of

thoughts, and although I shall have something to say about the second
strand, it is the third of these claims I wish to challenge: the need to distin-
guish process from product. The first can be conceded readily enough without
embracing Platonist imagery or the wholesale mind-independence of thought.
This will emerge more clearly in discussing the third thesis, which I shall
now treat in more detail.
By ignoring the process of thinking in his theory of thought, Frege is
denying the relevance of all aspects of our psychological states and processes
 FREGE AND CHOMSKY 31

to the project of the analytic philosopher. He is relegating the details of our

cognitive processing to the domain of empirical psychology. But the
division of labour between philosopher and psychologist is not quite so clear
here as Frege seems to think. For while psychologists may answer 'how'
questions about the order of execution of certain mental operations,
philosophical questions still abound in this area: questions such as what
thinking is, what it is for us to be in certain kinds of mental states and what
it is for those states to have the particular contents they do. It is these 'what'
questions and not just 'why' questions which contemporary philosophers of
mind have set about answering. (The 'why' questions are best left to depth
psychology. )
Thus it is that Jerry Fodor tells us that mental processes are causal chains
of content-bearing states, i.e., computational processes that "presuppose a
medium for representing the structures over which the computational
operations are defined". This commits Fodor to syntactically structured men-
tal representations, i.e., a language of thought. Other philosophers (Dennett,
Stich, Clark, among others) have contested Fodor's conclusion, arguing for a
connectionist architecture of distributed representations and parallel
processing, and they do so on a mixture of philosophical and empirical
grounds. This makes it hard to see just what type of considerations should
hold sway or take priority in the theory of cognition. Still, when philoso-
phers claim that "mental processes are causal sequences of mental states",
they are making a contribution to the theory of grasp. However, these re-
marks still don't threaten Frege's priorities, even if they do disrupt the disci-
pline boundaries he is presupposing.
Let us return to the issue of how much, if anything, the nature of sense
owes to the psychological process of grasping sense. The first thing to ask
here is whether Frege's linguistic turn solves his puzzle. For if grasping
sense is a matter of knowing the meaning of an expression that expresses a
given sense, and if knowing the meaning of an expression is a mental state
while meaning is non-mental, then once again we are left with problem of
how the mind of the speaker relates him to something abstract and non-men-
tal. Understanding becomes the process by which one comes to grasp the
objective and mind-independent meanings of words. How is this done?
We know Wittgenstein's solution to Frege's problem. Understanding, ac-
cording to Wittgenstein, is not a mental process: it amounts to mastery of a
technique. This dissolving of the mentalist component of the mind's relation
to meaning went hand in hand for Wittgenstein with a full-scale revision of
the notion of meaning itself. The meaning of an expression was seen as a
32 BARRY SMITH

matter of use, and with this view went the widespread rejection of the theory
of reference as providing the semantic foundations for a theory of linguistic
significance.
I do not have room here to examine the relative merits and limitations of
this alternative to Fregean semantics; so I shall have to content myself by
observing (without argument) that Wittgenstein's account oflanguage in the
end provides no satisfactory account of a speaker's understanding and that the
nemesis of his conception lies in the unresolved and murky depths of what
have come to be called the rule-following considerations.
Returning to Frege and his followers, we see an admission that the nature
of linguistic understanding consists in a genuine state of knowledge of the
individual language user. Any account of the meanings of words and
sentences of our language must pick out the meanings we actually grasp. So
understanding, or knowledge of meaning, has a role to play in identifying the
range of contents to be described in any satisfactory account of a speaker's
language. Put in Dummett's terms, a theory of meaning should be a theory
of understanding: i.e., a theory of the meanings that anyone who understands
the given language must know. More generally, a theory of a given lan-
guage, if it is to serve its task of specifying the semantic, syntactic and
phonological properties of the language, must always be seen as addressing
the knowledge of language its speakers actually possess. Otherwise the the-
ory would fail to characterise their language.
Now to argue in this way is to insist that any theory of linguistic sense
is a~theory of the speaker's grasp of sense. But although grasp of sense may
be a psychological act, and although the speaker's grasp may play a role in
individuating the very contents we are keen to characterise in an account of
sense, this is still not yet to concede anything to psychologism. For a theory
of grasp can amount to a theory of what is grasped, rather than a theory of
our psychological means of grasping sense. And even when our epistemic
states are appealed to in order to pick out the contents in question, this still
doesn't mean that what is grasped is, or owes anything to, the psychological.
We may, like Dummett, want to know what it is to grasp the senses of
certain expressions, but Dummett will insist on a sharp distinction between
"the process by which we came to acquire a grasp of sense and what
constitutes such a grasp" (1973, p. 240). It may well be that what is grasped
must be cited in a proper characterisation of the psychological states of
language users, but this does not mean that what is grasped (i.e., a sense)
itself belongs to the domain of psychology.
 FREGE AND CHOMSKY 33

So what would it take to establish grounds for psychologism in the the-

ory of meaning or sense? At the very least it would require showing that
there are substantive psychological constraints on what is grasped, what is
meant or what is understood by a thinker. And for these factors to enter into
settling the content of what is grasped and not just be to conditions for
identifying such contents, we must show how cognitive considerations can
playa role in the determination of truth or reference. This takes us back to
the determination claim: for if sense determines reference, and sense is a
cognitive notion which owes some aspect of its character to the
psychological conditions for grasping sense, then psychological factors will
play an ineliminable role in fixing the reference relation.
But we are some way from establishing that conclusion. So it is ironic
that Putnam (in 'The Meaning of "Meaning"') criticises Frege's conception
of sense and grasp of sense as leading to a form of psychologism that is in-
compatible with the claim that sense determines reference. Putnam points
out that although Frege does not regard meanings as mental entities he does
suppose that "knowing the meaning of a term is just a matter of being in a
psychological state" (1975, p. 219); and in the light of this Putnam claims
that " ... whether one takes the 'Platonic' entity or the psychological state as
the 'meaning' would appear to be somewhat a matter of convention" (1975,
p. 222), since the psychological state must uniquely determine what is
meant. And yet "taking the psychological state as the meaning" is
incompatible with the claim that meaning determines reference, since the
psychological states of a speaker (narrowly construed) do not determine the
reference of his words, as the famous 'twin-earth' thought experiments set
out to show. Putnam takes this line of reasoning to be a reductio ad
absurdum of Frege's combined view of the Platonistic nature of sense, the
psychological nature of grasp of sense, and the thesis that sense determines
reference; and he concludes that we must prize meaning away from the
psychological state of understanding to ensure its role as a notion which
determines reference to things external to the speaker.
However, Putnam's line of reasoning is too quick here. There are at least
two ways the Fregean could resist it. First, it involves a fallacious inference
from the claim that the knowledge of the meaning of a term is a
psychological state to the claim that the meaning might as well be cast as
part of the thinker's mental make-up. It does not follow from the fact that
the meanings of words are grasped by a psychological act either that meaning
is itself psychological, or that it is limited to what lies within the confines
of a speaker's narrowly described psychological states.
34 BARRY SMITH

Platonists can retain the conception of meanings as abstract entities and

continue to think of psychological states as in the head while supposing that
mention of abstract entities is necessary when characterising the psy-
chological states of competent speakers. We might think of the platonistic
meanings as indexing states of linguistic understanding. In this way, the
meanings would continue to enjoy their ontological status independently of
their role in giving the content of the thinker's states. Grasp of meaning
would still be a psychological matter, but the meanings that indexed those
states would determine reference independently of us.
Another way to block the argument depends on accepting, pace the
Platonist, that meanings are in the mind of the speaker, while resisting the
claim that those psychological states cannot accommodate facts about refer-
ence. It all depends on how one chooses to individuate the relevant psycho-
logical states. Putnam intends the argument to indicate the consequences of
individuating them narrowly in accordance with the doctrine of methodologi-
cal soplisism: the assumption that our psychological states entail nothing
about the world around us and presuppose nothing but the existence of a
thinking subject. But why should a Fregean accept such a restriction upon
what someone who grasps the meaning of an expression knows? The psy-
chological state in which our understanding of a word consists can also be
characterised in externalist terms. For if factors outside the individual's body
-- his relations to a physical and social environment -- contribute to deter-
mining the meanings of his words, they can equally well contribute to the
individuation of his state of understanding. So on this view, pace Putnam,
psychological states can determine reference, by dint of the dependence of
one's psychological states on the meaning and reference one's words actually
have. One could not be said to enjoy the same states of linguistic knowledge
if the reference, and hence the meaning, of one's words were any different.
We should have different meanings in mind.
So by opting for either Platonism or a more worldly externalism, the
Fregean can have a notion of meaning which ensures that words refer to
things external to the speaker while still supposing that a theory of meaning
must be a theory of what speakers understand: the meanings they actually
grasp.3 No such reductio ad absurdum argument tells against Frege's
doctrines, and Putnam is mistaken to think that it is "somewhat a matter of
convention" whether meanings for Frege should be regarded platonistically or
psychologistically. We shall need to work harder that this if we are to
motivate a plausible psychologism within the Fregean camp.
 FREGE AND CHOMSKY 35

Perhaps the best place to press is the Determination Thesis: the claim
that sense determines reference. For without this, however much we
contaminate the notion of sense with the psychological, we shall not ensure
the relevance of psychological matters to the settling of truth and the fixing
of reference.
The difficulty we have here is that it is far from clear or uncontroversial
what is meant by the claim that sense determines reference. Some see the
reference as depending on the sense, while others see senses as object-depen-
dent. So for instance, on the former view, one might think that sense deter-
mined reference by providing a descriptive condition which had to be
uniquely satisfied by some entity for it to count as the reference of a term.
This would provide strong grounds for treating the beliefs or intentions of
the speaker who grasped the sense of an expression as providing the
satisfaction conditions for denoting a particular object. But no such
descriptive model of sense can be accepted as part of the Fregean picture. And
indeed, some commentators (such as Evans and McDowell)4 have insisted
that the particular states of mind a thinker is in, and just which thoughts she
is thinking, depend for their identity on the particular object to which the
thinker is related and can be said to be thinking of. For a certain range of
singular thoughts, the identity of sense, conceived here as a way of thinking
of the referent, depends on which object serves as the informational source of
the thinker's current perceptions or memories. The sense does not lead us to
the referent, the sense depends on the existence of the object and the thinker's
relation to it for the presence of that thought in the thinker's mind. So to
avoid these exegetical disputes I shall plump for the weakest interpretation of
the determination thesis, but one which I hope sheds light on the role of
psychological considerations in the fixing of reference. The weak in-
terpretation of sense's determining reference is a supervenience claim: there
can be no change in the reference of a term without a corresponding change at
the level of sense. So the claim is that the facts of reference supervene on
facts about sense. Without facts at the level of sense there would be no facts
about reference. There are just facts about the senses of our expressions, and
thereby reference is fixed. Without our presence in the world to grasp sense;
create perspectives and thus give rise to the one-sided illuminations of refer-
ents, there would be no relations of reference, no links from our words or
signs to the entities in the world. So the pressing question is this: what
constitutes sense, or more properly, what constitutes the facts about sense
which determine all the facts there are about reference? The facts about the
senses of our expressions must be facts about our grasp of sense. But as yet
36 BARRY SMITH

no psychologistic conclusions are warranted. All we can say is that our

access to facts about reference,. indeed the very existence of such facts, de-
pends on our capacity to grasp the meanings that make the possibility of
their being references for our terms. For creatures like us, reference is
brought about and sustained by linguistic cognition.
Hence, psychologism, and, in particular, a conclusion that process is rel-
evant to product, will only follow from a substantive thesis about our un-
derstanding of language and the role of our cognitive psychological states in
determining the form and character of what is said or expressed in our lan-
guage. And the materials for this are not far to seek. For insofar as con-
straints on the assignment of reference to terms partly determine what is said
and so contribute to settling the truth-values of sentences, they are part of the
sense or content of sentences. Factors that contribute to deciding how truth
and reference are settled for a sentence are clearly part of its content. So to
understand what is said by an utterance and to be able to evaluate it correctly
it.is essential that one grasp these conditions and understand how they
contribute to determining the sentence's truth or falsehood. Such conditions
are a key part of the identity of a sentence, of fashioning just what is ex-
pressed by the sentence. They are part of the way the facts of reference are
given to UlS, and so they belong to the senses of our linguistic expressions.
The constraints I have in mind, here, are the structurally specified con-
straints on co-reference known as the Binding Principles which dictate
permissible patterns of co-reference and disjoint reference between items in a
sentence. The principles derive from Noam Chomsky'S recent work on
Universal Grammar. 5 However, to call these conditions merely syntactic
would be to ignore their semantic relevance in settling the logical forms of
sentences and in helping to determine an expression's contribution to deter-
mining the truth-values of sentences in which it occurs. These are clearly
semantic considerations: part of what we could call an expression's semantic
role, where what we mean by the semantic role of an expression is the con-
tribution that an expression with its particular semantic value makes to the
semantic values of larger expressions in which it appears. It makes its se-
mantic contribution in concert with the semantic properties of its syntactic
neighbours, given the way they are related to one another. This aspect of a
term's semantic functioning is, properly speaking, part of reference: the part
that fixes the role of such a reference-bearing term within a sentence. (The
other part of an expression's reference is, of course, the 'vertical connection',
as it were, between the word and the world, the part concerned with the
referent of the expression.) Now semantic role is· a key part of reference for
 FREGE AND CHOMSKY 37

Frege, as his advocacy of the context principle shows. What that principle
says, in effect, is that there is no more to the reference of an expression than
the contribution it makes to determining the truth or falsity of the sentences
in which it occurs. This depends both on the referent of the expression and
its semantic role within larger constructions.
The precise nature of an expression's semantic role, and hence its contri-
bution to determining the semantic interpretation of whole sentences,
depends on its position within a sentence and its relation to other lexical
items. These grammatical relations are not just a matter of the linear order of
words but, rather, are fixed by the hierarchical relations of dependence and
dominance we can trace between elements at a level of syntactic
representation underlying the sentence's surface form. These structural
dependencies among sentence-constituents are usually represented in
generative grammar by a phrase-structure tree conforming to the universal
principles of X-bar theory.6 It is in terms of these structural configurations
(represented tree-geometrically) that we can formulate the binding principles
that govern the patterns of co-reference.
I shall illustrate cases of binding shortly, but, first, it is important to
remember their relevance to the issue of psychologism. The binding princi-
ples of universal grammar place constraints on the distribution of reference to
items in a sentence partly on the basis of grammatical relations which apply
at a level of syntactic representation other than the surface form of the
sentence'? I want to argue that it is our grasp, albeit tacit, of these relation-
ships that establishes the strong psychological conditioning of the contents
of our utterances: the cognitive psychological contribution to the meanings
of what we say. The adversion to psychology is crucial, since although these
principles of well-formedness apply at levels of syntactic structure not dis-
played directly on the surface of a sentence, yet speakers conform to the
requirements the binding principles impose in arriving at the semantic inter-
pretations they give to sentences; and equally importantly, they do not con-
sider semantic interpretations ruled out by them. But just how do speakers
conform to generalisations not predictable from surface structures? What
triggers their principled interpretations? The most plausible and best empiri-
cally motivated explanation is that these underlying syntactic forms are the
mentally represented structures speakers assign to the strings of sounds and
signs that confront them. It is these configurations which enable speakers to
understand those strings as the semantically interpretable sentences they are.
Without assuming that a speaker has tacit knowledge of these underlying
38 BARRY SMITH

linguistic structures there is no explanation at all of how human speakers

conform to the intricate generalisations of Universal Grammar.
A brief illustration of the binding principles and some indication of their
level of application should help us to focus the issue more sharply. Whether
certain noun-phrases take the same or a differentreference as some other noun-
phrases in a sentence can follow from the facts about a particular syntactic
configuration among elements in the sentence. This configuration is known
as c-command, and it accounts in part for the facts about whether an element
can be bound by an antecedent in the way a variable is bound by a quantifier
in the predicate calculus. An antecedent must c-command any anaphor or
pronoun it binds; the bound element must not c-command its antecedent. An
anaphor is a reflexive pronoun or reciprocal (e.g., 'himself, 'each other')
which must be bound; a pronoun is a referentially dependent noun-phrase
('she', 'her', 'he', 'him') which may be bound but need not be. Consider the
following sentences (where the indices indicate intended assignments of
reference and the asterisk represents an unacceptable interpretation):

(1) John i shaved himself i

(2) * John i shaved him i

(3) John i said that Billj shaved him i

(4) * Maryi said that Beth loved herselfi

(5) Maryi said that Bethj loved herselfj

(6) * Herselfi loved Maryi

(7) * Hei said that Johni loved Mary
(8) After her left, Johni saw Mary

Notice that 'John' is required to be the antecedent of 'himself in (1). In (2),

'John' cannot be the antecedent of 'him'. Whereas in (3), 'John' can be the
antecedent of 'him' but 'Bill' cannot be. By contrast, in (4), 'Mary' cannot
be the antecedent of 'herself, but 'Beth' must be as we see in (5). It seems as
 FREGE AND CHOMSKY 39

if the pronoun 'him' is too near 'John' in (2) for that item to serve as its an-
tecedent, but sufficiently far away in (3) to take 'John' as antecedent.
Conversely, 'Mary' is too far from the anaphor 'herself in (4) to serve as its
antecedent, but 'Beth' is close enough to play this role as we see in (5). Just
as with examples (1) and (2), if we substituted the pronoun 'her' for the
anaphor 'herself we should get the opposite pattern of acceptable and un-
acceptable readings. How are we to account for the data? Notice that the
nearness of constituents will not serve as the unacceptability of (6) and (7)
indicate. Nor are these ruled out as unacceptable just because the anaphor or
pronoun proceed the antecedent, as the acceptability of (8) demonstrates.
To explain the data we have to look at the syntactic configuration of c-
command. Where a, band g range over all constituents, a c-commands b iff a
does not dominate b and every g that dominates a dominates b. B C-command
holds between an element and all the material dominated by the first
branching category (or maximal projection, i.e., XP) dominating that
element, just so long as the element does not dominate any of that material
itself, as illustrated in Fig. 1:

Fig. 1

XP
I \
X' YP
I I \
X Y' ZP

Here the element labelled by X c-commands the elements labelled by YP, Y'
and ZP. Note that by the definition, Y' only c-commands ZP and vice versa;
neither c-command X or X' . 9
We can now say what is meant by 'binding'. Where a and b range over
all constituents, a binds b iff

(i) a c-commands b, but not vice versa.

(ii) a and b are co-indexed.

Co-indexing requires that elements share number, person and gender. For our
purposes, it also indicates the intended assignments of reference, which may
be permitted or excluded by the principles of binding.
40 BARRY SMITH

In the example below in Fig.2, the antecedent 'John' binds the anaphor,
'himself as we can check against the definition of c-command and binding: 10

Fig. 2

C'
/ \
C IP
/ / \
[that] / I'
Spec \
/ \
NP VP
I I \
N' V' \
/ / \ \
N V NP AP
John i shaved himself i quietly

In examples (6) and (7) above, the antecedent would occur within the VP and
so could not c-command (hence could not bind) the item in subject position.
However, in (8), where the pronoun belongs within an adverbial position AP
moved to the front of the sentence, the antecedent in VP could c-command
(and bind) it in its original position. Clearly we need to pay attention both to
c-command and to how near or far the binding antecedent is from the bound
element if we are to explain the legitimate patterns of assignment. The
binding principles sum up these requirements as follows:

Principle A
An anaphor must be bound in its local domain (governing category).

Principle B
A pronoun must be free in its local domain (governing category).

Principle C
An R-expression (name, definite description, etc.) must be free.

Consider the following examples, where the square brackets mark the local
domain of the antecedent or pronoun and, once again, the indices show the
intended assignments of reference:
 FREGE AND CHOMSKY 41

(9) * [NP(John's i friends) upset himself i)

(10) [NP(John's i friends) upset him i)
(11) * [John i despised him i ]
(12) [NP(John's father)i upset himself i)

The binding principles tell us that an anaphor (such as the reflexive

'himself) must be bound in its own domain or governing category, and for
this to happen, the antecedent must c-command the anaphor, and the anaphor
must not c-command its antecedent. But a pronominal like 'he', 'she', 'her'
or 'him', by contrast, must be free in its domain. So although 'John' c-
commands 'him' in (11), it is ruled out by Principle B of the binding theory.
Whereas in the case of (10), the very fact that the possessive subject 'John's'
does not c-command and so does not bind the pronoun 'him' leaves open the
possibility of co-reference, ruled out in (9) because the intended antecedent
does not c-command (and so does not bind) 'himself. Notice that we do have
binding of the anaphor in (12), because the whole NP c-commands (and so
binds) 'himself.
We are now able to appreciate how, according to Chomsky's
Government-Binding theory of syntax, the distribution of pronouns and
anaphors, and the possible and non-possible cases of co-reference, are dictated
to a large extent by facts about syntactic configuration, with properties like
c-command playing a key role. But the concept of c-command belongs to the
theory of syntax, and it is far from obvious that ordinary (or even
extraordinary) language users have explicit knowledge of the concept.
Nevertheless, the facts which the theory explains are known to us quite
readily. As mature users of the language, we know without reflection a great
deal about the distribution of certain word types: we know which words to
put in the right order to express our thoughts, and we know which thoughts
can and cannot be expressed by certain sentences. All the cases explained here
are already known to us, as we recognise when our attention is drawn to
them.
Now it is reasonable to suppose that these facts are syntactic facts -- facts
to be explained by constraints articulated within an adequate theory of syntax.
But it is also reasonable to suppose that these constraints bear importantly
on the notion of word and sentence meaning: they are constraints on co-
reference, a semantic matter par excellence. Such syntactic constraints playa
crucial role in determining what interpretations we give to one another's
utterances; and what is just as important, which fnterpretations we don't give
-- which interpretations we rule out and would never consider. This is
42 BARRY SMITH

important, for it suggests that there is less fumbling for interpretation than
we expect, less trial and error, less inductive learning and so on. Our re-
sponses to, and uses of, language are much more highly constrained than we
perhaps recognise unreflectively.
What lesson can we take from Chomsky's account of these phenomena?
He begins by asking "how the language learner knows these facts, which are
not [after all] necessary properties of any imaginable language and are surely
neither taught nor derivable by general principles of induction or analogy
from direct experience". (1986, p. 78) And he concludes that they are known
to the speaker because his or her knowledge of language is organised in ac-
cordance with certain specific principles of grammar. As we have just seen,
the binding principles are not constraints upon surface structures, but upon
an underlying level of syntactic representation (S-structure or LF). To be
apprised of these syntactic constraints on interpretations, the mental represen-
tation or knowledge of something other than surface structure is needed if
speakers are to appreciate the relevant semantic properties of their language.
Now Frege was prominent in insisting that the logical form of a sentence
was distinct from the surface grammar of the sentence, but the level of
structure to which these constraints apply is not posited as part of a philo-
sophical understanding of language: it is an empirically motivated level of
structure which enjoys psychological reality in the minds of speakers. I I
These constraints are universal and so open to empirical test. For
example, several facts that the speaker knows follow from the Principle P
(which we can take to be Principle B of the binding theory), which says that
a pronoun may not be dependent on an antecedent in its domain, together
with the Principle Q, which says that sentences are structured NP-VP and not
NP-V-NP, Without the latter principle we could not assume the same
relations of c-command to obtain, and we could not define the domain of a
constituent as we do. Now as Chomsky tells us:

If we take P and Q to be properties of the initial state So of the language faculty,

then we can explain how [someone] comes to know these facts, given the boundary
conditions set by experience; note that this is a genuine explanation. Given
reasonable assumptions about conformity across species it must hold for Japanese. If
not, we have counterevidence to P and Q; if it does, we have confirmation. (1987, p.
185)

These conditions are seemingly arbitrary in their way, and yet they do hold
for human language. What is more, to change just one of the principles
would threaten to unravel the tight-knit set of inter-dependent generalisations
 FREGE AND CHOMSKY 43

which predicts these universal features of human languages. Now since for
Frege, the identity of a thought depends on its semantic structure, and these
conditions give form to the semantic structure of sentences, they are part of
the proper characterisation of the thought-contents expressed in language.
There is simply no way to dispense with these conditions without also dis-
pensing with the conditions whereby a sentence says or expresses what it
does and is determined as true or false in accordance with its composition and
the semantic properties of its parts. The rub of course is that these conditions
apply solely to the mentally represented levels of structure known as S-
structure and LF which speakers assign to sounds and signs. These are
cognitive structures which have the character they do because of the internal
organisation of the minds of human language users. Thus such cognitive
psychological states have an essential part to play in characterising the
products that we as human language users produce.
Notice, however, that in arguing for this degree of psychologism I do not
intend to dispute Frege's claim that laws of logic and laws of content set
standards and do not merely describe what we actually say or do. For
although there are processes at work to produce the speech we utter and hear,
what we utter or hear is not always sanctioned as correct. We often find
ourselves making false starts, verbal slips, or grammatical infelicities, which
we correct on-line, so as to better realise what we are trying to say. And this
too points to a gap between the processes by which we actually produce and
comprehend speech and the system by which we monitor and check our own
performance. Here we say our adjustments are tailored to what our compe-
tence demands: competence being just one part of the contribution to per-
formance. And this distinction between competence and performance, how-
ever it is to be explained, ultimately allows us to respect Frege' s second
strand of anti-psychologism without giving up the important contribution
psychology makes to meaning. The only strand of Frege's anti-
psychologism we must reject is the claim that process is irrelevant to
product.
I finish with two caveats. Firstly, note that in arguing for this cognitive
psychological contribution to semantics I am not arguing for methodological
solipsism or a reduction of meaning to the purely internal processes taking
place within a thinker's skull. The intra-cranial factors which the linguist or
psychologist cites are just one of the many sets of factors which contribute
to the determination of meaning and reference. Causality, social norms of
language use, prior states of knowledge and memory, the cognitive dynamics
needed to track the objects of our environment through space and time; all
44 BARRY SMITH

these, together with our internal cognitive organisation, can make their
contribution felt. But these factors apply not directly at the level of reference,
but rather at the level of sense, and in particular in determining the contents
of the states of mind we are in which constitute our grasp of sense. It is the
states of mind we are in when we are said to grasp meanings that give us
licence to say that the sounds and signs we produce and respond to stand for
something other than themselves.
Secondly, when we talk of a speaker's tacit knowledge of the syntactic
constraints on reference, we are talking about a body of knowledge and not
about computational processes and mechanisms. It is a further task of cogni-
tive psychology to find out what form the mental representation of that
knowledge takes in the mind of the individual speaker; only then will we
begin to understand how it engages the mechanisms of speech. This is why
we should be wary not to confuse our representation of linguistic structure
with the structure of these representations. It is the content of our represen-
tations, what they are about, that affects our linguistic judgements, not the
formal or syntactic form of the vehicles of representation. Fregeans may try
to take comfort from the fact that we are talking about states of knowledge
rather than psychological processes, but the above arguments for the psy-
chological reality of grammar clearly show that the objects of linguistic
knowledge owe their form and character to the internal structure of the mind,
and insofar as these factors ~ontribute to the determination of truth and ref-
erence, we have grounds for a plausible psychologism. 12

NOTES

1 Though he acknowledges that there need be no necessary connection of a thought with one
particular sentence. (See 1979, p. 269)
2 See Dummett 1991.
3 Putnam's assumption that the Fregean is committed to methodological solipsism seems to
force the unwanted conclusion that sense cannot determine reference. But whether one
construes psychological states as states of the speaker's head or as relational states of some
kind, they seem to presuppose something other than the thinking subject, and it is this further
thing in terms of which we give their content; whether that thing be a realm of platonic
meanings or a physical or social world. Either way we have something outside the speaker's
head that determines reference. So do both options count as forms of externalism? It is a moot
point, since we need to decide whether platonism about meaning does or does not entail the
ejection of methodological solipsism and narrow content. I leave it open whether there can be
a view according to which one could have 'in-head' psychological states with narrow contents,
where the contents were calibrated by abstract propositional meanings which determine their
less discriminating referents. It depends on whether the narrow states require there to be, or
are merely characterisable in terms of, platonic con;elates.
 FREGE AND CHOMSKY 45

4 See Evans, Ch.l and McDowell.

5 See Chomsky 1986 and 1988 for informal accounts of his recent Principles and Parameter
approach to syntax. For a more technical earlier treatment of these concepts in the
Government-Binding framework see Chomsky 1981.
6 See the references in fn. 5 for an account of the role of X-bar Theory within the Principles
and Parameters approach.
7 For this reason referential dependencies between items within a sentence, like con-
siderations of scope, cannot be judged solely with respect to surface word order.
8 Subtleties arise here over whether we choose to see g as the first branching category, or the
the first branching category which is a maximal projection of a head (i.e., an NP, PP, VP, etc.)
The latter relation is referred to as m-command. We can ignore these details for our purposes.
9 For more discussion, see Chomsky 1986, p. 162.
10 Note that in recent syntactic theory, IP, which dominates the non-lexical category INFL
(which carries inflectional information about the verb) has replaced the category S.
11 In the end, it may tum out that the theory-internal. level of structures postulated by
syntacticians bear a close and systematic relation to what logicians mean by logical form. It is
a plausible assumption that the empirically motivated level of structure in contemporary
syntactic theory known as LF (logical form) may be the level of logical form postulated by
philosophers and linguists wQrking on the (truth-theoretic) semantics of natural language. In
addition to Chomsky, the most notable work here is by Higginbotham, (1983, 1987), May
(1985) and Neale (1993).
12 Thanks to Peter Simons, Tom Baldwin and Pavel Tichy for thought-provoking and error-
saving comments at the Conference, and to Peter Milne, Tim Crane, Jennifer Hornsby and
Sarah Patterson for comments afterwards.

REFERENCES

Frege, G.: 1884, Die Grundlagen der Arithmetik, translated as The Foundations of Arithmetic,
J. L. Austin, 1959, Blackwell, London.
Frege, G.: 1979, Posthumous Writings, Blackwell, London.
Chomsky, Noam: 1981, Lectures on Government and Binding, Foris, Amsterdam.
Chomsky, Noam: 1986, Knowledge of Language, Praeger, New York.
Chomsky, Noam: 1987, Reply to Review of 'Knowledge of Language' in Mind and Language.
Chomsky, Noam: 1988, Language and Problems of Knowledge, MIT Press, Mass.
Dummett, M.: 1973, Frege: Philosophy of Language, Duckworth, London.
Dummett, M.: 1991, 'Objectivity and Reality in Frege and Lotze' in Frege and Other
Philosophers, Oxford University Press, Oxford.
Evans, G.: 1982, The Varieties of Reference, Oxford University Press, Oxford.
Fodor, J.: 1987, Psychosemantics, MIT Press, Mass.
Higginbotham, J.: 1983, 'Logical Form, Binding and Nominals' in Linguistic Inquiry, 11.
Higginbotham, J.: 1987,'On Semantics' in E. LePore (ed.), New Directions in Semantics,
Kluwer, Amsterdam.
McDowell, J.: 1985, 'De Re Senses' in C. Wright (ed.), Frege: Tradition and Influence,
Blackwell, London.
Neale, S.: 1993, 'Logical Form and LF' in C. Otero (ed.), Noam Chomsky: Critical
Assessments, Routledge, London.
46 BARRY C. SMITH

May, R.: 1985, Logical Form: its Structure and Derivation, MIT Press, Mass.
Putnam, H.: 1975, 'The Meaning of "Meaning'" in Mind, Language and Reality, Philosophical
Papers Vol. 2, Cambridge University Press, Cambridge.

Department of Philosophy
Birkbeck College, University of London
Malet Street
London WCIE THX ,UK
 PETR KOT A TKO

MEANING AND THE THIRD REALM

If I were to identify the most fundamental phenomenon to be explained by

the theory of meaning, I do not know what else it could be other than the
following: it is a matter of fact that· we quite often make utterances
(typically, but not necessarily, consisting in the production of sounds or in-
scriptions) which have a special kind of effect, or as I shall prefer to say, a
special kind of value. We are used to acknowledging this kind of value by
saying that something has been asserted, promised, ordered, etc., in the utter-
ance. In short, a communicative act has been performed. The communicative
act performed in the utterance can quite naturally be called its meaning. A
theory which wants to be taken seriously as a candidate for a theory of
meaning must offer replies to questions like: what is it in general for an ut-
terance to have a meaning and what is it for an utterance to have a particular
meaning, e.g., to have the communicative value of an assertion that it is
raining? It seems natural to regard these questions as more fundamental than
the questions concerning sentence meanings and word-meanings: from this
perspective, the meaning of a sentence is its communicative potential which
amounts to the ability of the sentence to serve in some linguistic commu-
nity as a vehicle of meaningful utterances, and word meaning amounts to a
words's contribution to the communicative potential of sentences in which
the word appears as a component.
Frege is not a theorist led by this kind of motivation: but it is surely
legitimate to ask what light his work can throw on the questions mentioned.
My particular question in this paper will be whether his doctrine of Sinn can
help us to capture one of the most important features of utterance meaning,
namely its objectivity. Whatever utterance meaning is, it cannot, by the very
nature of communication, be anything in the private possession of the
speaker. Rather, it is a value which the utterance intersubjectively has in a
given linguistic community and which is accessible in principle in the same
way to any competent member of that community. Were this not so, we
could never have more than hypotheses about the meaning of the utterance
which could always be refuted by the speaker's verdict: but any speaker's
utterance would count as a particular verdict (e.g., as the statement that the
speaker had such and such meaning-constitutive intentions) in the same
provisional way, i.e., it would be always open to the speaker to refute our
interpretation of that utterance, and so on ad infinitum. But it surely belongs

47
J. Biro and P. Kotatko (eds.), Frege: Sense and Reference One Hundred Years Later, 47-57.
© 1995 Kluwer Academic Publishers.
48 PETR KOT A TKO

to the character of linguistic communication that, despite all ambiguities,

there are many cases in which an utterance counts in the community of
language users as a particular assertion, promise, order, etc., and no speaker's
additional manoeuvre can deprive the utterance of this value. Imagine what
would happen with the force of promising, with the practical importance of
distinguishing between true or false assertions, with the relation between the
orderer and the addressee of the order, etc., were this not so. This does not
imply that the analysis of utterance meaning should not include the notion of
speaker's intentions: it "only" implies that any theory which makes the
utterance meaning at least partially dependent on the actual speaker's
intentions, taken as something to which the speaker has privileged access,
cannot be right.
It may seem that the most direct and natural way to account for the ob-
jectivity of utterance meanings (and meanings in general) is to postulate
meanings as objective entities in some strong sense. The Fregean senses
("Sinne") -- in the case of whole utterances, Fregean thoughts ("Gedanken")
-- are surely candidates for this role. As far as I can see, they would play it as
follows: the speaker grasps a certain thought (Frege calls this performance
"das Fassen des Gedankens" and regards it as the nature of thinking), evalu-
ates it, let us say, as true (this is, from Frege's point of view, another act,
characterized as "die Anerkennung der Warheit des Gedankens" or "das
Urteilen")l and expresses the thought in an utterance with the assertive force
("Behauptungskraft") which reflects his attitude to the thought; finally, the
audience grasps the same thought and recognizes the force with which it was
brought into communication. The involvement of the thought in this proce-
dure is a matter of subjective attitudes or acts on the part of both speaker and
audience. What is objective is just the thought itself: but its objectivity is
really solid. Thought in itself has no essential relation to the mental acts of
the speaker or the addressee, nor to the actual communicative situation, nor
to the language in which it has been expressed, nor to any actual or possible
language at all, nor to any actual or possible human culture: its timeless be-
ing excludes any such relativization. Its special status deserves a special
realm of being in addition to those of physical things and mental phenom-
ena. Frege was neither the first nor the last to discover this wonderland: but
nobody can deny that he gave to it the most awful name: "das dritte Reich".2
When confronted with the traditional problem of how something timeless
and unchangeable can exercise any influence on anything else and so prove
its reality ("Wirklichkeit"), Frege finds the solution in the grasp ability of the
thought by our mind, which is, on our part, made possible by our pos-
 MEANING AND THE THIRD REALM 49

sessing a special intellectual capacity. This graspability of thought has been

found by some authors (for example Putnam, Kaplan and Perry) incompati-
ble with another feature ascribed by Frege to a thought, namely, its being the
extension-determiner and hence the ultimate truth-value bearer.
But our question is different: how can anything like Frege's senses help
us explain the objectivity of utterance meaning? To be sure, we have, by
fiat, introduced with Frege an objective entity to be regarded as the utterance
meaning: but this in itself does not solve the problem. What distinguishes a
meaningful utterance from other kinds of acts is that it is a move by which
something communicatively happens between the speaker and the audience.
And it is the objectivity of what communicatively happens through the
utterance that we are trying to account for. So we should now explain how
the postulated objectivity of the Fregean sense can be transferred to (or be
used to explain the objectivity ot) what happens through the utterance.
The picture we have drawn so far presupposes that the speaker and the
audience connect the same thought with the sentence (as uttered in a given
context), i.e., that their mental acts have the same correlate in the third
realm. But, as Frege himself shows in the case of proper names, in actual
communication this is often not so. We can connect different senses (modes
of presentation) with the same name and hence also different thoughts with
the same sentence containing that name. 3 This is bad enough, but, according
to Frege, bad for natural language, rather than for the theory. But even if we
get rid of what counts in Frege's eyes as disturbing features or imperfections
of natural language, even if we imagine that everybody is disciplined enough
to think, of, e.g., Brigitte Bardot, when referred to by that name, under the
same, officially prescribed, mode of presentation, I still fail to see how the
objectivity of the Fregean thought can be used to explain the objectivity of
the communicative value of the utterance. In our model, the speaker grasps
(privately) some thought, the audience grasps (privately) the same thought:
what more do we gain for our purpose than the fact that their mental states
(which are, according to Frege, private in the most radical sense4) have
something in common (and, perhaps, that this is mutually supposed by
them)? The perfect objectivity of the thought, which we have so admired,
does not have any special part to play here. When summarizing what
happens in the utterance, we must say: nothing in the world of thoughts
(which is by Frege's ruling protected from any change) and so nothing
objectively, since thoughts are the only objective element in the picture.
Something happens in the mental life of both participants, which may have
some effect on their behaviour. Of course, once we decide to specify the
50 PETR KOT A TKO

utterance~ meaning in terms of these psychological and behaviourial changes,

we do not need any help from the third world at all.
Somebody may say that the thought is, in the Fregean picture of the ut-
terance, not only grasped by the participants, but also, through being ex-
pressed, in some way objectively actualized in the communicative situation.
But if we ask what difference this makes to the communicative situation, we
are at a loss. Once we start to characterize the difference in terms of some ef-
fects of the utterance in this "sublunary world", we make the presence of
Fregean thought rather decorative; the effects are perfectly well tractable, for
example, as institutional consequences of the utterance in the given com-
munity, without any reference to the third world. This is, in fact, the way I
wish to follow.
The starting point is quite simple. According to the strategy we have fol-
lowed so far, the objectivity of utterance meaning should have been secured
by something so perfectly objective that it could not have been from this
world and could intervene in communication only through being subjectively
grasped. As a result, the radical objectivism collapsed into radical sub-
jectivism. So, why not look for the principle securing the objectivity of the
utterance meaning in the same world in which the utterances take place - in
the world of social communicative practice? This may be found extravagant
enough, but the idea is, alas, not original. The idea of defining the commu-
nicative value of utterances in terms of their consequences may be perhaps
attributed to the late Wittgenstein, if we risk attributing to him any project
in the theory of meaning at all. M. Dummett finds this idea "completely
programmatic" and complains that it is totally unclear what such a theory
would look like. Nevertheless, he knows in advance that it could not cover
assertion, because this, unlike "more formalized" communicative acts, as he
calls them (e.g. commands and promises), does not have conventionally de-
fined consequences. 5
In my opinion, the consequences in terms of which the communicative
value of the utterance, i.e., the communicative act performed (and, in the
second step, sentence meaning), can be defined are commitments incurred by
the speaker in the utterance. And in this respect I do not see any essential
difference between assertion and other types of communicative act. Here is
how I would generally characterize commitments incurred in an assertion:

(IAI is the proposed definition of assertion, IBI specifies some features of

assertion which I regard as typical but not necessary. fII specifies features
 MEANING AND THE THIRD REALM 51

which differentiate assertion from other speech act types, IIIf features which
assertion shares with other speech act types.)

IAI S asserts to A that p by his utterance of x if and only if his utterance

commits him to the following intentions and beliefs:

1If1 S believes that p is the case (with a certain specific degree of certainty, if
this is indicated in the utterance).
1If2 S intends to make A believe that p, or to take p' s being the case fully
into consideration (or to achieve another conventional point of assertion
which is relevant in a given context - cf e.g., replies to examination ques-
tions).
1If3 (implication of 1If2) It is not obvious to S that A already knows that p,
or that A takes p' s being the case fully into consideration (or that any other
conventional point of assertion which may be relevant in given context is
out of the game.)
IIIfl (implication of IIfl) S believes that the presuppositions implied by the
way in which the propositional content of the utterance (that p) is specified
are true.
IIIf2 S believes that the conditions of communication are such that they at
least do not exclude that A can receive and understand his utterance.

fBI Additional commitments incurred in asserting in standard contexts:

1If4 S regards the belief that p as justified (in the degree corresponding to the
degree of certainty indicated, if any, and in the way conforming to the stan-
dards ofjustification, if any, relevant to the type of discourse in question).
1If5 S believes that A regards him (or should regard him) as competent to
express his view on whether p is or is not the case (if S does not appeal to
some other authority).
/II/3 S rregards his utterance as ifitting into communication, i.e., as (in
some way) relevant with r.espect to the communicative context.
IIIf4 S believes that A does not disbelieve the presuppositions implied by the
way in which the propositional content of the utterance (that p) is specified.

Let us use the term "manifestation" for such a presentation of the speaker's
intentions or beliefs which commits the speaker to having them (but does
not Imply that the speaker actually has them). Then the speech act types can
be correlatively defined either in terms of the types of commitments
52 PETR KOT A TKO

established in them or in terms of the types of manifestations performed in

them. In the second case, the left side of the definition will be modified as
follows:

S asserts to A that p by his utterance of x if and only if his utterance counts

as a manifestation that... (and the rest will remain the same).

The characteristics of assertion I am suggesting can be approached as a set of

claims about what happens through the assertion (namely, about the
normative consequences of the assertion), each of which can be tested in the
following ways:

Ia! A performance of an assertion is liable to criticism from the point of

view of any of the commitments or manifestations listed, i.e., is open to the
objections to the effect that the speaker does not actually have some of the
intentions or beliefs manifested or that some of the manifested beliefs are
false or some of the manifested intentions improper.

fbI The combination of asserting with speaker's statement (or implication) to

the effect that he does not actually have some of the manifested intentions or
beliefs, or with speaker's behaviour which counts as an evidence of this,
results in some kind of communicative or behaviourial inconsistency. A
famous example related to IIIl is Moore's paradox.

Icl Any of the manifestations listed may become a proper speaker's motive
for the utterance.

Idl On the other hand, if the speaker for some reason does not want to make
some of these manifestations, this may be sufficient to prevent him from
making an assertion (which would otherwise be most appropriate to his in-
tentions).

I think that all claims included in my characteristics of assertion can meet

these tests. Nevertheless, there is a difference between the part IAI and fBI in
what the negative result of such tests would mean. If you give me enough
representative counter-examples to the claims in fBI, I shall have to admit
that commitments specified there require special circumstances to be estab-
lished by an assertion, and hence do not belong to the general characteristics
of asserting. Till then I believe, to the contrary, that these commitments are
 MEANING AND THE THIRD REALM 53

established by the assertion unless there are special circumstances which

would block them. I think that they should be mentioned in the characteris-
tics of asserting because they help to explain certain communicativephe-
nomena (those mentioned in laI-IdI above) .. But if you give me an example of
an act which counts in our communicative practice as an assertion but does
not establish some of the commitments listed in IA/, I will have to admit
that I do not understand correctly what assertion, or speech acts in general,
consist in. For I regard the commitments mentioned under I All as necessary
and sufficient to distinguish assertion from all other speech act types and
those under lAIlI as common to all speech-act types. Or, to put it another
way, if the development of our communicative praxis led to a state in which
there were no utterance types such that their use would quite regularly have
all those types of normative consequences together, I would have to say that
assertion, as I understand it, has disappeared from our life as one of the basic
speech act types and survives at most as an ad hoc phenomenon.
A characteristic feature of this account is that neither types of speech acts
nor utterance meaning (which is nothing other than the speech act performed
in the utterance) are defined in terms of the actual speaker's intentions or be-
liefs. In this it essentially differs both from the Gricean-type definitions of
utterance-meaning and speech-act types and from Searle's definitions of
speech-act types in terms of "conditions of success", some of them being
purely intentional 6. Our account is given in purely institutional terms, and
the question of the speaker's having or not having the manifested intentions
and beliefs is regarded as irrelevant with respect to the content and force of
his utterance.
Now it should not be difficult to see how our approach enables us to ac-
count for the objectivity of utterance meaning. From the point of view based
on the notion of commitment, utterance meaning does not appear as
something which is at the beginning possessed or grasped by the speaker and
than transmitted to the addressee. Utterance meaning is an institutional fact
established by the utterance and valid in the same way for both parties,
putting them into a specific institutional relation. What can be, for the
speaker, characterized as commitments vis-a-vis the audience, can be, for the
audience, characterized as an institutional power over the speaker. 'This power
can be exerted through criticism or other kinds of sanction over the speaker,
provided that something in the speaker's behaviour or in the context of the
utterance (in the broadest sense) counts as evidence that the commitments
incurred in the communicative act are not fulfilled.
54 PETR KOT A TKO

If utterance meaning amounts to the speech act performed in the utter-

ance, then the specification of the meaning of an utterance with assertive
force will differ from our definition of assertion just in the place of p: it
must be filled in by whatever is needed in that place to complete the specifi-
cation of the commitments incurred by the speaker. One can say that the lo-
cation of p in the definition marks the place which specification of the truth-
conditions has in the specification of the meaning of an assertive utterance:
the state of affairs whose being the case would make the assertion true is just
that state of affairs whose being the case the speaker is committed to
believing (cf. /Ill), and whose being the case the speaker is committed to in-
tending to be taken into consideration by the audience (cf.1II2), etc. Then the
location of p in the definition also shows how the truth-conditions of a
declarative sentence (i.e., the scheme of the determination of the truth-condi-
tions of utterances of that sentence in particular contexts) enter into its
communicative potential, i.e., into its capacity to establish certain com-
mitments when uttered in certain contexts, or, quite briefly, into its mean-
ing. And on this basis we can also understand why the specification of a cer-
tain word's contribution to the determination of the truth-conditions of sen-
tences in which the word occurs amounts to the specification of that word's
contribution to the meaning of those sentences.
When saying this, we, of course, evoke two important, but not yet men-
tioned, aspects of the Fregean notion of sense: its essential connection with
the notion of truth-conditions and its compositiona1ity. This also indicates
that something like a Davidsonian truth-theory, which accounts for these
features of language, can be integrated into the theory of meaning in our
sense to provide an account of one (crucial) aspect of the mechanism deter-
mining the communicative potential of sentences. This may sound a per-
verted version of the well-known idea of combining a Davidsonian truth-
theory with some theory of force in order to get a full-blooded theory of
meaning. But there is something more fundamental I would like to mention.
According to several authors,7 there is a very strong constraint which a
Davidsonian truth-theory for a language L must meet, if it is to be regarded
as a theory of meaning for L at all (i.e., if its theorems, which are material
equivalences of the form _ is true if and only if p, are to count as
something as strong as meaning specifications): the theory must, in tandem
with some other theories, contribute to our making sense of both the
linguistic and the non-linguistic behaviour of the speakers of L. So far this
is in perfect agreement with the thrust of the theory of meaning I have been
arguing for. I can only applaud claims such as "Meaning is whatever enables
 MEANING AND THE THIRD REALM 55

us to make sense of people's behaviour" and "There is no gap between

understanding the person and understanding the language". But that is all.
The specification of the postulated contribution of the theory of meaning to
our making sense of speakers' behaviour requires some account of the
functions of language and of the way in which it is anchored in the linguistic
community. This account is sometimes given via the definition of the so-
called actual-language relation (language L is an actual language of a
population G iff... ). When approaching the question what it is for a language
to be the actual language of some population, all the authors mentioned
above (in note 7) presuppose, as the main factor, an essential connection
between the meanings of utterances and the actual propositional attitudes of
speakers. Some of them explicitly suggested combining the Davidsonian
theory, in order to get the supplementation mentioned, with something like
the Gricean or Schifferian notion of speaker's meaning (or something
weaker) and the Lewisian notion of convention (or something weaker). What
I have said so far commits me to a radically different view. According to it, a
language is not anchored in the population through some systematic links
between its sentences and actual propositional attitudes of speakers who utter
these sentences, but through the institutional power of its sentences in that
population, in particular through their ability to establish (in a systematic
way) certain commitments, when uttered. Then a theory of some language L
will be a correct theory of meaning for a language spoken in a population G
just in case the meaning-specifications generated by the theory for sentences
of L help to interpret (better, or at least not worse, than any alternative
theory) what happens in particular cases of linguistic communication in G in
terms of the institutional facts mentioned. Then the theory may be expected
to help us make intelligible (or make sense of) the behaviour of the members
of a given population in such a way that it will enable us to re-describe (and
so to interpret) their behaviour in terms of certain institutional concepts,
rather than in terms of propositional attitudes behind their behaviour. This
reflects a conflict between two notions of understanding or making sense of
human behaviour: on the one side, psychological explanation, on the other
side, cultural interpretation. The account of the objectivity of meaning I have
argued for directly binds the notion of meaning to understanding in the
second sense. But I believe that it also creates a good basis for the
investigation of communicative intentionality. Once we have characterized
utterance-meaning in terms of the institutional consequences of the utterance,
it makes good sense to ask what it can mean deliberately to make an
utterance with such consequences and, from the other side, how such an
56 PETR KOT A TKO

utterance can serve (together with some background knowledge) as a ground

for attributing certain intentions and beliefs to the utterer.
Perhaps I should finish by stressing that the account of the objectivity of
utterance meaning I am advocating here does not imply any radical linguistic
conventionalism. Assertion in my account requires the institution of com-
mitment and some principles which enable us to say in particular cases
whether or not certain commitments have been incurred: but these principles
need not necessarily be linguistic conventions. They must be effective (give
clear results in most practically relevant cases) and have normative power
over the participants in the communication (so that any of them is obliged to
acknowledge those results on pain of making doubtful her ability to par-
ticipate im communication). Clearly, one cannot incur commitments in our
sense, i.e., in the sense of institutional consequences of the utterance, if the
utterance does not establish those commitments in virtue of some such
principles. So, one cannot participate in communication, either as a speaker
(incurring certain sets of commitments in his utterances) or as an addressee
(to whom the speaker is related by those commitments), unless one is
already committed to certain social principles. (There is no regress here, since
one is committed to respect these principles simply in virtue of having the
status of a competent member of the community: no extra act is required.)
But it is not at all easy to see how any such principles can be fixed in the
linguistic community: indeed, the well-known considerations about rule-
following show that this is one of the most difficult things to explain. To
get some understanding here we must go very deep into the complexities of
the social life and functioning of its institutions. This may be far from what
many philosophers speaking about meaning are ready to do; but I am
inclined to think that this is the only ground on which a serious theory of
meaning can be based. 8

NOTES

I "fIber Sinn und Bedeutung"; "Der Gedanke. Logische Untersuchung."

2 "Der Gedanke." Well-known philosophical analogues include H. Rickert's realm of values,
N. Hartmann's third layer of being ("die dritte Seinschichte") and, of course, K. Popper's third
world. But perhaps the most important thing to mention here is that Frege's formulations
concerning the relation of thinking (as mental activity) and thought (as an ideal unit) almost
plagiarize B. Bolzano's ideas about "sentences in themselves" ("Satze an sich"). (See Bolzano's
Wissenschaftslehre, par. 34.)
3 "Uber Sinn und Bedeutung"; "Der Gedanke."
4 "Der Gedanke"
5 M. Dummett, (1973, p. 361 f.)
 MEANING AND THE THIRD REALM 57

6 J. R. Searle, (1969, p. 57 f.)

7 Evans and J. McDowell (1976), "Introduction"; C. Peacocke, "Truth Definitions and Actual
Languages", (loc. cit.); D. Lewis, (1983); M. Davies, (1981, Ch. I); D. Wiggins (manuscript).
8 This is a modified version of the paper "Fregean Sinn and Utterance Meaning," to appear in
a volume edited by I. Max and W. Stelzner (1995).

REFERENCES

Bolzano, B.: 1837, Wissenschaftslehre, trans. & ed. Rolf George, 1972, University of
California Press, Berkeley.
Davies, M.: 1981, Meaning, Quantification, Necessity, Routledge, London.
Dummett, M.: 1973, Frege: Philosophy of Language, Duckworth, London.
Evans, G. and J. McDowell (eds.): 1976, Truth and Meaning: Essays in Semantics, Oxford.
Frege, G.: 1966, "Uber Sinn und Bedeutung," in Funktion, Begriff, Bedeutung, Vandenhoeck
and Ruprecht, Goettingen.
Frege, G.: 1966, "Der Gedanke. Logische Untersuchung," in Logische Untersuchungen,
Vandenhoeck and Ruprecht, Goettingen.
Lewis, D.: 1983, "Languages and Language", in Philosophical Papers, vol. 1, Oxford.
Peacocke, C.: 1976, "Truth Definitions and Actual Languages", in Evans and McDowell
(eds.), (1976).
Searle, J.: 1969, Speech Acts, Cambridge University Press, Cambridge.
Stelzner, W. (ed.): 1995, Logik und Mathematik. Frege-Kolloquium Jena 1993, De Gruyter.
Wiggins, D.: "Meaning, Truth-Conditions, Propositions: Frege's Doctrine of Sense Retrieved,"
(manuscript),

Institute of Philosophy, Academy of Sciences of the Czech Republic

Department of the Philosophy of Language and the Philosophy of Science
Jilslai.l, 11000 Praha 1, Czech Republic
 DAVID WIGGINS

PUTNAM'S DOCTRINE OF NATURAL KIND WORDS AND

FREGE'S DOCTRINES OF SENSE, REFERENCE AND
EXTENSION: CAN THEY COHERE?

1. Hilary Putnam has been apt to emphasize all the differences between the
deictic doctrine that he advocates for the understanding of our understanding
of natural kind substantives and the accounts of the meanings of these
expressions that would have had to be offered by his predecessors in the
philosophy of meaning. Delighting in iconoclasm, he has sought at various
times to include within the ambit of his entertaining criticisms of his
predecessors such figures as Aristotle, the Scholastics, Locke, Mill, Frege,
linguistic philosophers, analytical philosophers, philosophers of linguistics,
indeed practically everyone. 1 Frege would not have enjoyed the idea that he
might be thought to belong in such a list.
In this paper, I set out Putnam's proposal and show how, twenty years
ago, it broke the mould for one kind of philosophical analysis. But I also
try to show that we may deploy Putnam's proposal most convincingly if,
flying in the face of Putnam's own wish, we try to place it within the
framework of Fregean sense and reference. Then I try to show that, having
done that, we can improve our understanding further if we seek to integrate
the deictic proposal, significantly but desirably adjusted at one key point,
with an extant, neglected but even more time-honoured tradition of semantic
speculation, a tradition, not empiricist, in which there is already a clear place
for Putnam's insight into the functioning of natural kind words.

2. I begin by reminding you of the contents of "Is Semantics Possible?".

This is a paper that Putnam read to a conference in Brockport, New York in
1967.2 It was in this paper that he introduced the idea that to impart the
meaning of a natural kind term is to impart certain core facts: (1) the
stereotype -- consisting of the facts an ordinary speaker needs to know in
order to use a natural kind term -- and (2) the extension, the identification of
the latter being the province of experts. Or, as I would rather say (see,
Sameness & Substance, Preamble and see below), it was in this paper that
Putnam introduced the idea that the sense of thing-kind words standing for
natural kinds is reality-invoking or extension-involving.
Putnam led up to his conclusion by criticizing incisively and amusingly
the easy (or fall-back) supposition that the right way to give the meaning of

59

J. Biro and P. Kotatko (eds.), Frege: Sense and Reference One Hundred Years Later, 59-74.
© 1995 Kluwer Academic Publishers.
60 DA VID WIGGINS

'lemon', 'tiger', 'water' or whatever, would have to be by analysis into

simpler terms or by giving necessary and sufficient conditions. It was
disturbing perhaps that Putnam filed no report of an analytical philosopher
(contrast one philosopher of linguistics) with his trousers down actually
attempting such a thing. But maybe this did not matter. The thing that
mattered was that the problem of natural kind words was not to be solved by
the tact or good taste in which analytical philosophers then excelled, or by
the refusal to recognize that these expressions constituted a special question
for the philosophy of meaning.
On this matter, as on others, ideas were in short supply at the time of the
Putnam article. But it must be recorded that, for the significantly similar
case of proper names, there had existed from before 1960 an important
minority opinion (held by Geach, Anscombe, Michael Dummett and, under
Dummett's influence, by myself). This was that a proper name has its
meaning, and thus affects the truth-conditions of the sentences in which it
occurs, by standing for its bearer: and that there is no other way to give the
sense of a proper name than to say which object it is that the name stands
for. 3 It was not out of the question then, even in that distant epoch, to say
that the senses of proper names were reality-invoking or object-involving or
that a proper name had its sense by being assigned to something, not by
virtue of the laying down of a specification such that a bearer of the name
bears the name by meeting that specification. Nor, among those who held
.the minority opinion I have just described, was it an unfamiliar question
what it would involve to find room for this insight in Fregean semantics.
(The notion of Art des Gebebenseins seemed to be ready-made for such an
attempt. See §8 below.) Rather, that which was still missing in linguistic
philosophy and in the philosophy of science was any strong perception of
the need to say something similar for potentially predicative expressions like
'lemon', 'tiger', 'water' or to try to generalize from the direction of
semantical fit that is so strikingly exemplified by the assignment of an
ordinary name to something. 4
Back now to "Is Semantics Possible?". Not only, Putnam insisted, were
philosophical analyses of substantives such as 'lemon', 'tiger', 'water'
laughably inadequate. The only explanation of anyone's even supposing that
it might be possible for there to be such an analysis was sheer negligence of
the mise-en-scene, the whole social-cum-technological context, on which we
all depend in order to come to understand one another. This context is the
thing we depend upon even to understand what we find in a dictionary.
Philosophers who were ready to suppose that there could be a philosophical
 PUTNAM'S DOCTRINE OF NATURAL KIND WORDS 61

analysis of the meaning of 'lemon' or 'tiger' were ignoring the division of

mental labour and the role of the authorities or experts who sustain our
shared understanding of natural kind words. Philosophers should cease to
complain, Putnam said, about the fact the dictionaries are "cluttered up with
colour samples and stray pieces of empirical information (e.g., the atomic
weight of aluminum) not sharply distinguished from purely linguistic
information". They should take that fact seriously as a clue to the real
situation, the situation that Putnam himself wanted (however schematically)
to describe.

3. Confining oneself to the ideas of "Is Semantics Possible?", but departing

a little from his mode of exposition there, one can put the positive proposal
Putnam wanted to advance as follows: where the instructor's grasp of
extension is authoritative (or is downstream from an authoritative
identification of that extension), an instructor could initiate a learner into the
meaning of the word 'lemon' as follows:
This is a lemon. [Here the instructor displays a specimen.] A thing is a lemon if it
resembles this [the specimen] or this [another displayed specimen] or this [a third
specimen] in the relevant way. I say the relevant way. But to understand better
what that way is you must inquire, just to the extend necessary for your purposes,
into the nature of these three things that I am showing you.

The philosophical claim is that (however artificially) such a demonstration or

ostension reconstructs the ordinary teaching and learning of thing-kind words.
It reconstructs that which is essential to the transactions that take place
between those who know and those who do not know what a given
substantive means.

4. It is hard for those who have been introduced to philosophy by anyone

who understood the point of such a proposal to appreciate the novelty it once
enjoyed, at the time when Putnam introduced it. But before we come to the
question of the tractability of Putnam's doctrine to the theory of sense and
reference or its affinity to certain rationalist ideas, three points need to be
registered.
First, the doctrine, which is sometimes loosely called "the indexical
theory", does not, on a true understanding, imply any close similarity
between natural kind substantives and indexicals or demonstratives. If
'lemon' or 'tiger' or 'water' had any real resemblance to the indexicals 'this'
or 'that' or 'now' or 'today', these substantives might in other contexts, and
without change of lexical meaning, pick out other kinds of things than the
62 DAVID WIGGINS

kinds that we denominate lemons, tigers or water. But the point of the
theory is to attach the meaning of these words to the real natures, more or
less well known, of the actual lemons, tigers and water that we have
encountered. Therefore we must not compare 'water' to a demonstrative.
The theory is a deictic theory in just one sense: it is the theory of the deixis
by which we can, under special and favourable conditions, attach a word to a
kind of thing.
Secondly, the doctrine does not extend to all thing-kind words.
Occasionally, in later papers, Putnam was tempted to apply it to
substantives like 'pencil'. But that was a pity, indeed it threatened the
shipwreck of a good idea. 'Pencil' denominates a functional or instrumental
kind. 5 Such a kind might well be defined (even in the strict old fashioned
acceptation of 'define'); and this is just as well, because almost the only
nomological generalization one will discover by investigation of the class of
pencils is the undependable generalization that one can write or draw with a
pencil. There is little or no resemblance here to the case where the ostension
of a natural kind invites us to extrapolate freely across the observed
properties of its exemplars in the search for interesting generalizations about
its nature.
This point leads to a third, which might itself have been the occasion for
a whole paper, but is not the occasion for the whole of mine. In this area
there are problems of the underdetermination of meaning by deixis and
problems of the proper representativeness of specimens (problems analogous
to those that Goodman in The Structure of Appearance called "constant
companionship" and the "imperfect community"). Room must be made
within the theory of this demonstrative practice for an instructor to make his
pupil understand whether it is a species that is being indicated or a
subspecies/variety (is it rose or Rosa Rugosa?); and room must be made for
the instructor to make the pupil understand whether something specific or
something generic is to be identified (tiger, say, or Felis).6 Ostension had
better not be the magical solution to the problem of natural kind predicates.
We shall return to this point in § 13.

5. So much in outline for the new theory of natural kind words. Now one
might ask why it was so difficult for so long for our sort of philosophy to
arrive at this deictic or extension-involving conception of their meaning? In
the face of the question, I shall begin by quoting the review in Mind and
Language (Autumn issue, 1990) of the book Representations and Reality
which is Putnam's recent critique of many positions in present day
 PUTNAM'S DOCTRINE OF NATURAL KIND WORDS 63

philosophy, not least of functionalism. After expounding Putnam's deictic

theory in his own way, the reviewer writes:

But why should water beliefs, so understood, pose a problem for the functionalist?
He will not be able to capture the content of such beliefs by alluding to their
characteristic causal connections with experience alone. Rather, the important
connections will be those between the current water thought, a past demonstrative
thought about certain samples then in front of the subject, and the theoretical belief
that such substances fall under some fundamental classification which may be
revealed by investigating the causal properties of those substances. It is these
connections which will ensure that the thought is indeed about water and they are
part of that thought's causal role. (David Owens, loc.cit.)

When I first read that, the words "water beliefs" and "current water thoughts"
and the ambiguity of the latter seemed to awaken something in memory. I
read the passage again and I got the thing I was searching for. It was a
footnote of lL. Austin's: "There will not be books in the running brooks
until the dawn of hydrosemantics".1 But now that (for 'water' at least) we
have got ourselves some hydro semantics -- now that the running brooks are
in the books 8 (even if not yet vice versa) -- it must be time to consider the
question why in the 1950's and well into the 1960's it seemed so irresistibly
comical to contemplate the very idea of hydrosemantics. What was so
unthinkable in the idea of a hydro semantics even for 'hydor' ('water')? The
answer to the question is that this was unthinkable to most philosophers of
Austin's area because, even though everyone knew that Quine had exposed a
grave circularity in all exant attempts to explain from a standing start the
analytic/synthetic distinction, that did not seem to matter very much. It did
not matter because analytic philosophers thought that, in practice, they could
still reach principled consensus on what was analytic. So what good reason
was there to question the separation of language from the world, or let water
itself into the semantics of 'water'? What reason was there to modify the
idea that philosophy consisted in analysis? Analysis and the search for non-
circular necessary and sufficient conditions could continue. 9
By attacking the very idea of analyticity (a supposedly salutary
exaggeration), the Quinean onslaught had overextended itself and miscarried.
The point about Putnam's "Is Semantics Possible?" is that it was part of a
more modest, and far more effective, second onslaught on the idea of
analyticity. This second attack is not targeted on the existence of definable
single criterion concepts (vixen =female fox), (oculist = eye-doctor). Nor
is it an attack on the analyticity of 'a vixen is a female fox', 'an oculist is an
64 DAVID WIGGINS

eye-doctor', etc. That is scarcely worth attacking. What it is targeted upon

is the attempt to generalize from those, however undeniable, small successes:
"A theory which correctly describes the behavior of perhaps three hundred
words has been asserted to correctly describe the behaviour of tens of
thousands of general names."
In other words, what obstructed the discovery (or rediscovery, as I shall
claim) of the necessity for a deictic view of natural kind names was the
failure of our philosophical commnity at large to think through exactly and
to the end the problems and limitations of the analytic/synthetic distinction.
It is no accident that Putnam, who had spent as long as anyone in re-
examining the distinction, should have been the rediscoverer (or co-
rediscoverer with Kripke) of the deictic view of natural kind words. What is
strange, if anything is, is that these matters have still not been thought right
through and have still not impinged as they should on the practice of
philosophy.lO It is stranger still perhaps, that in the philosophy of science
those who now concern themselves with the meaning of theoretical terms
still show so little inclination to emancipate themselves from a purely
model-theoretic approach or to explore the possibility of sense-giving
relations between predicates and kinds that might mimic the direction of fit
we now take for granted in our understanding of the setting up of the relation
of designation. (See, again, paragraph 3 section 2 above.)

6. Rather than seek further explanation of what was happening at the time
of Putnam's introduction of his insight, it is now time for me to redeem my
promise to show how the extension-involvingness of natural kind words will
cohere with the theory of sense and reference.

Sentence Singular term Concept-word

~ ~ ~
Sense of the Sense of the Sense of the
sentence singular term concept-word
(Thought)

~ ~ ~
Reference of Reference of Reference of ~ Object(s)
the sentence, the singular term, the concept-word, that fall under
a truth value an object a concept the concept

On 24th May, 1891 Frege wrote Husser! a letter about the sense and
reference of predicates, and he included in it the above instructive and
 PUTNAM'S DOCTRINE OF NATURAL KIND WORDS 65

remarkable diagram. Let me first gloss this diagram in familiar ways. Then,
after proposing a small repair to Fregean doctrine under the third column, let
me try to show how neatly the deictic proposal can be assimilated within the
scheme.

7. About column 1, let us simply remark that grasping the sense of a

sentence is a matter of grasping under what conditions it attains to the True.
In preference to saying that, in the strict sense of 'designate' or 'refer', the
True is the reference or designation of a sentence, let us say that the True is
the semantic value -- Dummetts' term -- of the sentence. (Let us think of
reference itself as a special case of semantic value.)

8. Now column 2. Here let us bear it in mind that, strictly and literally
interpreted, the claim that "sense determines reference" is compatible with the
denial of the priority of either -- if, in addition, reference.
Because it is rarely quoted, let us start with Frege's remark, in
Ausfahrungen,. aber Sinn und Bedeutung, about the fictional names
'Nausicaa' and 'moly'. In so far as 'Nausicaa' is allowed any sense, Frege
says its literary signification depends upon its behaving as if (als ob) it
designated some girl. (Similarly, the kind of name 'moly' gets what sense it
has by behaving as if it designated a particular species of herb, namely the
herb that in Odyssey, Book X, Hermes gives to Odysseus in preparation for
his encounter with Circe. The name is still there, ready and waiting for any
pharmaceutical product with similar or comparable apotropaic powers.) Now
this is the case of fiction. But implicit in the claim about fiction is a
simpler claim about the case of fact. The simpler claim is that in the non-
fictional case a name has a sense by behaving as if it has -- or simply by
having -- a reference. A name has its sense then by somehow presenting its
object. To grasp the sense of a name is to know (in the manner correlative
with the mode of presentation that corresponds to this particular somehow)
which object the name is assigned to. To impart the sense is to show (in
that manner) which object that name stands for.
To speak as Frege does of sense as "mode of presentation" suggests
nothing less than this: there is an object that the name presents and there is
a way in which the name presents it, or a conception, as I shall say, of that
object. A conception of an object is an account of how things are, a body of
information (not necessarily correct in every particular), in which the object
itself plays some distinct and distinguishable role. I I Note, however, that
such a conception or body of information will normally be open-ended,
66 DA VID WIGGINS

imperfect and corrigible. Rarely, if ever, could it be condensed into a

completed description of the object. But why should one ever have expected
that there would be any description of x that is synonymous with a given
name of x? What it is reasonable to expect is at most this: that the
particular body of information about x to which the sense of some particular
name of x is keyed may generate various different descriptions of x that
might serve in a suitable context to identify x.
Properly possessed then, the conception of an object x that sustains the
sense of a name that is keyed to that conception is a way of thinking about x
that fixes (with the help of the world, the world being what helps create the
conception) which object the object in question is. Which object the object
is is precisely what is mastered by him who comes to understand the name
with the sense corresponding to that conception. You can say if you like
that what Frege means by the sense of a name is a certain sort of conception
of what the name stands for.

9. Now perhaps we can effect the transition to Frege's third column. Even
as grasping the sense of a name or singular term and its contribution to truth
conditions is a matter of grasping the particular conception of the object
correspondling to the mode in which the name presents what it stands for; and
even as we give the term's sense by saying in a manner congruent with that
conception what the term stands for; so grasping the sense of the predicate
and its contribution to truth conditions consists in grasping the predicate's
mode of presentation of the concept it stands for, and we give the predicate's
sense by saying what concept it stands for. And even as in the case of a
singular term we show or exhibit this sense in preference to that sense by
exploiting one mode of presentation rather than another to say which object
this is -- drawing upon one body of information in preference to another in
filling out our identification of the object -- so similarly, in the case of a
predicate, we show this sense in preference to that, e.g., the sense of 'horse'
rather than the sense of say 'Equus caballus', by exploiting one mode of
presentation rather than another to say which concept this is. We prefer (say)
the body of information one might expound by saying "A horse is a certain
animal with a flowing mane and tail; its voice is a neigh; and in the
domestic state it is used as a beast of burden and draught, and for riding
upon" over the body of information that identifies such creatures by
classifying them as perissodactyl quadrupeds, locates their species among the
genus Equus and the family Equidae and then dwells on other zoological
features.
 PUTNAM'S DOCTRINE OF NATURAL KIND WORDS 67

10. With this suggestion -- entailing that there can be different accounts of
one and the same kind of thing, namely horse or Equus Caballus -- I can
look forward to a possible conclusion abut stereotypes, namely this: that
Putnam's stereotypes approximate to those special conceptions that
correspond to the various senses of the various expressions that stand for
natural kinds; that stereotypes are particular special ways of thinking in
identificatory fashion about such kinds and their specimens. That will indeed
be my proposal. But there are confusions to be avoided here, and there is a
departure from Putnam. We must hasten more slowly.

11. To avoid confusion, we must take care to present Frege's theory in

column three as perfectly general with respect to all predicates and then to
say what is so special about natural kind words among predicates. It will
also be necessary to indicate at some point what sort of thing Fregean
concepts are. Finally, conceptions of a horse being in this picture
conceptions of the concept horse (i.e., conceptions of what it is for
something to be a horse), we must be as clear as possible about the
difference between a conception, something that goes with sense, and a
concept, which is something on the level of reference -- lest we fall into the
old error of confusing sense and reference.
So let us return to the analogy between names in general and predicates in
general, that is, the analogy to be made between columns two and three. If
we take the analogy in the manner proposed, then it makes good sense of
Frege's insistence, in the letter to Husser! and elsewhere, that the reference of
a predicate cannot be any object or objects that it is true of. Just as singular
terms without reference are unfitted to figure in the expression of a judgment
possessed of a significance that enables us to move forward, as we must, to a
truth value, just as a name capable of figuring in the expression of a
judgment that can constitute knowledge must have a reference, so must any
predicate that aspires to this status. But many predicates essential to the
expression of good information do not have anything they are true of.
Therefore their Fregean reference is not any object they are true of. Thus, as
Frege says in the letter to Husser! already cited,
With a concept word, it takes one more step to reach the object than it does with a
proper name, and this last step may be missing -- i.e., the concept may be empty --
without the concept words ceasing to be scientifically useful. I have drawn the last
step, from concept to object, horizontally in order to indicate that it takes place on
the same level, that objects and concepts have the same objectivity. In literary use it
68 DA VID WIGGINS

is sufficient if everything has sense; in scientific use there must also be

Bedeutungen.

So on this account of things, the claim that a predicate has a concept as its
reference does not bring out what is distinctive of natural kind terms or of
Putnam's proposal concerning them. 'Round square' has a reference, 'blue'
has a reference, 'pencil' has a reference. You come to know this reference,
which is nothing other than the concept, by corning to know what it would
take (whether that be possible or impossible) for a thing to be a round
square, blue, or a pencil. And now, as one special case of this, there are
terms, such as 'lemon' or 'tiger', where to grasp what it would take for
something to be a lemon or a tiger ore whatever it is, you need exposure to
the extension of the term. In this case -- if the argument of "Is Semantics
Possible?" was correct -- there is no grasping the reference or concept
otherwise than through the extension. Here (at least) we cannot have reached
the understanding that we do have by coming to grasp a strict lexical
definition. For there is no strict lexical definition. So whereas the reality-
involvingness of proper names amounts to their senses' being reference-
involving, the reality-involvingness of natural kind terms amounts to their
senses' being extension-involving. The schema itself of sense and reference
neither dlemands that idea about extension nor excludes it, but it
accommodates it.

12. On this account, we could say that he who understands 'horse' and knows
what Victor and Arkle both are, viz., horses, grasps a general rule for
accepting or rejecting the sentence 'x is a horse' according as it collects or
fails to collect for arbitrary item x the verdict True. And, equally we could
say that he who knows what Victor or Arkle are has got a grasp of the
concept horse. These two accounts of the concept come to the same thing. 12
At this point, I suppose I must digress for a moment to say one word
about one well known and entirely general difficulty in Frege's scheme, a
difficulty which has nothing to do with extension-involvingness, namely
Frege's supposed need to deny that the concept horse is a concept. Ifthe
view I take of these things is right, then the reference of 'horse' is indeed
something predicative but only in the following restricted sense. The term
'horse' can be combined with the copula and article to give the predicative
phrase 'is a horse'. The phrase 'is a horse' is indeed, in Frege's terminology,
an unsaturated expression. You can, if you wish, assign the phrase a
semantic value. But to say that is not to say that that the predicative phrase
has a reference. The predicative phrase isn't what has the reference. 'Horse'
 PUTNAM'S DOCTRINE OF NATURAL KIND WORDS 69

is what has the reference. In other words, the way back to the truth that the
concept horse is indeed a concept is to take the copula seriously and
distinguish 'horse' from 'is a horse'. If I say that there is something that
Victor and Arkle both are, then the thing they both are is horse, not is a
horse. What we quantify over are simply concepts. But a concept is the sort
of thing which we can connect to a subject of discourse by linking a subject
term and a predicate term by the copula. The concept has a predicative role
because it plays such a role in predication, but it is complete in itself, not
unsaturated. 13
There is much more to say on this subject, as there is on the analogy that
Frege wanted to see between predicates and functors, but the time has come
to sum up and move on to the next stage, which is to show how easily the
deictic theory can be placed at column three within the framework of Fregean
semantics.

13. In the case where we can only explain what is the reference of a predicate
by something like the method that Putnam describes, because no strict
analytical definition exists or could exist, the thing we who do know the
sense have to recapitulate for someone else who does not know the sense of a
given substantive is surely that which we have ourselves learned by
commerce in the world at large with the objects that satisfy or fall under or
exemplify the concept. In practice, and so in the theory of this practice, any
specification of such a concept will have to depend directly or indirectly on
exemplars. This is Putnam's point. But surely that point does not commit
Putnam to think that the exemplars themselves can be given by bare
unfocused presentation, unsupported by collateral explanation. (Cp. §4
above.) No presentation, one might say, without focus, and no focus
without elucidation. When we fix the sense of an expression with a
predicative role by giving its reference and give its reference by
demonstrating or alluding to exemplars, what we need to impart to one who
would learn the sense is both factual information and a practical capacity to
recognize things of a certain kind, the information sustaining and regulating
the recognition, and the recognition making possible the correction and
amplification of the very information that first sustained the recognition.
What we need to impart, we might say, is an identificatory or recognitional
conception. It is this that corresponds to the sense of a natural kind term.
What one must say then is that the sense of a natural kind term is correlative
to a recognitional conception that is unspecifiable except as the conception
70 DAVID WIGGINS·

of things like this, that and the other specimens exemplifying the concept
that this conception is a conception of.

14. At this point it may be said that what I have just claimed isn't the same
as what Putnam claimed. I shall attend to that in a moment. And there may
be a more basic disquiet -- disquiet at the fact that, if you follow my
suggestion abut accommodating the doctrine of extension-involvingness,
then you will find yourself saying: "that which supports the sense of the
substantive 'horse' is a certain identificatory conception of the concept horse.
Is this not an intolerable convergence in terminology?" I shall attend to the
first of these two worries in the course of allaying the second.
The convergence of 'conception' and 'concept' is ugly perhaps but it
doesn't signify any confusion. It is clear what the terms 'conception' and
'concept' mean here and it is clear what they have to do, namely, quite
different work. But not only that. It is perfectly possible to replace the term
'conception'. There are at least three ways of doing this, each of them
illuminating in its own way.

15. In the first place, we can replace the term 'conception' by an expression
of Evans' and we can say instead that there are "two ways of thinking about"
the concept of horse or about horses. Either one can think about what Victor
is as a familiar domestic animal (as a beast of burden, or mount) or one can
think of what Victor is as a perissodactyl quadruped belonging to the genus
Equus and the family Equidae. These are two different ways of thinking of
one and the same kind of thing (concept).1 4

16. The second possibility is surely -- is it not? -- to replace the word

'conception' by Putnam's word 'stereotype'. Or should this be doubted?
At the beginning of his article "Is Semantics Possible?", Putnam
introduced the deictic theory by saying that in order to impart the meaning of
a thing-kind word one has to impart certain central facts, the stereotype and
the extension. But what did Putnam mean by 'stereotype'? And what did he
think was the relation of stereotype and extension? When Putnam spoke of
the stereotypes that support our normal understanding of the meaning of
thing-kind words, he may sometimes have been thinking of something like
the little engravings one finds in dictionaries like Larousse. (Indeed such
engravings are full of theoretical interest in these connections. They are
reminders of what deixis has to do in bringing language up to the world and
the world up to language.) Normally, however, what Putnam means by a
 PUTNAM'S DOCTRINE OF NATURAL KIND WORDS 71

stereotype is a fund of ordinary information or a collection of idealized beliefs

that one needs to grasp in order to get hold of the meaning of a thing-kind
word. So a stereotype is a bit like what you find -- better, a bit like one
proper part of what you find -- in the text of a good dictionary or
encyclopedia (which is not of course to say that it resembles an analytical
definition).
What then must the relation be between the stereotype and the extension,
the core facts Putnam spoke of? (See Section 2 above.) Putnam himself
was insistent that the core facts that determine the meaning of thing-kind
word were two perfectly separate kinds of fact, as if the stereotype could be
explained or specified in a manner entirely independent of the question of its
extension. The extension was the special province of experts and what
belonged to it could be dark to possessors of the stereotype. But I find
something strange in this separation. Of course the stereotype is different
from the extension, but this is not to say that the first can be explained
without any allusion to the second. To defend his own way of separating
stereotype and extension, Putnam himself claimed that he possessed the
stereotype of elm and the stereotype of beech but didn't know the difference
between an elm and a beech.IS Reports have it that he has studiously
defended this ignorance over twenty years in order to defend his opinion that
one can understand a word by grasping the stereotype without having any
grasp of what is the extension. Nevertheless, if Putnam is as vague as he
says he is about the difference between an elm and a beech, then I think one
has a good right to doubt the degree of his comprehension of these words. If
he entirely lacks the capacity to tell elms and beeches and to distinguish
them, then there is something that he is missing. To insist that this thing
he is missing is nothing semantic might become an upholder of the
separation of linguistic meaning from the world or of the old understanding
of the analytic-synthetic distinction. But these are things that Putnam's own
work has done so much to discredit. Issuing from him of all people, such
insistence as this would be unbecoming. Would it not be better to say that
the stereotype is the stereotype of this or that concept and that grasping the
stereotype represents the beginning of an identificatory capacity, a capacity
that the expert manifests more completely than the non-expert? In the case
of the non-expert, the capacity can be rudimentary, but surely it is
identificatory. It is a capacity which could advance to the point where it
became the capacity of an expert.
72 DAVID WIGGINS

17. There is a third way to avoid saying that the conception which supports
the sense of the word 'horse' is a conception of the concept horse. This is to
redraft one's description ofthe whole situation in tenns inspired by Leibniz's
theory of clear and distinct ideas. The thing we most badly need here is what
Leibniz called a clear but confused (Le., non-distinct) idea.
In Leibniz's account of ordinary human knowledge, a clear idea of a horse
is not an image or a likeness of a horse. It is that by the possession of
which I recognize a horse when I encounter one. (What clarity in an idea
contrasts with is not confusedness or non-distinctness, but obscurity.) A
clear idea of a horse is confused (or non-distinct) if, even thought I can
recognize a horse when I encounter one, I cannot enumerate one by one the
marks which are sufficient to distinguish that kind of thing from another
kind of thing. My understanding is simply practical and deictic. What I
possess here I possess simply by having been brought into the presence of
the thing. ('Being brought into the presence of a thing' translates Leibniz's
own words. I6 ) Our idea of a horse will begin to become distinct as we learn
to enumerate the marks that flow from the nature of a horse and that
distinguish a horse from other creatures.
What Leibniz shows us how to describe here is nothing less than the
process by which clear but indistinct knowledge of one and the same concept
begins life anchored by a stereotype to examples that are grouped together by
virtue of resemblances that are nomologically grounded. But then, as it
grows, this knowledge gradually unfolds the concept in a succession of
different and improving ideas (or conceptions -- as we might have said).
Rather than tell again the old Lockean story of nominal and real essence and
face again the question what, if anything, makes them essences of some one
and the same thing, we can now describe the process by which a clear
indistinct idea becomes a clear distinct idea, then a clear adequate idea. 17
Which brings me to my last remark. When we reconstruct the first
stages of the process, from the moment where there is ground to credit
language-users with possession of a stereotype for horse, and ground to credit
them with possession of a stereotype of the very same thing we have a better
and more infonnative stereotype for, and when we reflect that at that stage
there could scarcely have been any experts, well we may think that strictly
speaking, what Putnam should have stressed was not the necessity of experts
but the necessity of the possibility of experts. I8
 PUTNAM'S DOCTRINE OF NATURAL KIND WORDS 73

NOTES

See, for instance, the recent exposition that he gives in Representations and Reality,
Cambridge, Mass. (M.I.T.), 1988.
2 Eventually, this paper was published in Language, Belief and Metaphysics, Munitz and
Kiefer (eds.), SUNY Press, Albany 1970, and in Metaphilosophy, 3, 1970, pp. 187-201.
3 See G.E. M. Anscombe, Introduction to Wittgenstein's Tractatus, Hutchinson 1959, pp. 41,
42,44; Michael Dummett "Truth" Proc. Ar. Soc. 1958-9, pp. 141-3; David Wiggins, "Identity
Statements", in R.R. Butler (ed.) Analytical Philosophy II, Oxford, Blackwell, 1965.
4 For "direction of fit", see lL. Austin, "How to talk: some simple ways" (reprinted in
Collected Papers, Oxford,) and MJ. Woods, as cited in my Sameness and Substance, (Oxford,
Blackwell, 1990).
5 In discussion, Hilary Putnam pointed out to me that at this point in his development of the
doctrine he had been reporting a discussion with Rogers Albritton in which they had been
envisaging circumstances under which it was found that all extant pencils had a certain
microstructure. Here I ask: But all pencils, however manufactured and for whatever
specialized purpose? This would be magic. Or (more likely) there would have to have been
a practical joke somewhere in this tall story.
6 These problems are more tractable than the problems they superficially resemble and that
were thrown up by the resemblance (or no-universals) theory of universals defended by
Russell and H.H. Price. They are more tractable because here deixis can be supported by
context and by verbal explanations that are unconstrained by special requirements of
ontological parsimony. See the explanations envisaged below, especially § 13.
7 "Truth", Proc. Ar. Soc. XXIV, Suppl. vol., 121.
8 For a closer approximation to the hydrosemantics or the natural meanings that the exiled
Duke must really have had in mind at As You Like It, ll.i.12; cpo Paul Valery's celebration
(however ill calculated to evoke Austin's approval) of the Source Perrier.
9 Indeed it still continues, to judge by the aims and ambitions that are prescribed by most of the
participants in the group effort to solve for x -- if necessary by brute force -- in the equation:
(knowledge) = (belief + x). In easel painting, pointillisme was a short-lived experiment. In
philosophy, it bids fair to continue for ever.
10 Not anticipating that Kripke would be charged with violating some supposed distinction
between metaphysics and the philosophy of language (another case of the distinction Quine
and Putnam subverted) and falsely supposing that, after Kripke's excellent observations on the
differences between the statuses of necessity, analyticity and a priority, all these things would
inevitably be thought through to the end, I rashly elected in Sameness and Substance
(Blackwell 1980) to call certain truths and necessities that were neither analytically nor
formally nor combinatorially guaranteed, but which rested on what things (objects or kind)
both individuatively are and cannot help but be (that is, on the individuation of objects under
certain concepts of the sort Putnam had described), conceptual truths. I little thought that my
"conceptual" would be read as an evasive synonym of "analytic" or that I would be seen as
seeking to save some of the most implausible theses of linguistic philosophy. Why should a
concept such as man, horse, tree, be something that arises on the language side of a barrier
that keeps the world from flowing into the word? It is expressly denied in Sameness and
Substance that language can be protected by such an exclusion zone.
11 See John McDowell, "On the sense and reference of a proper name", Mind, 86 (1977) pp.
59-85; David Wiggins, "Frege's Problem of the Morning Star and the Evening Star" in M.
Schirn (ed). Studies on Frege, II, Logic and Philosophy of Language, Stuttgart Bad Canstatt,
Fromann-Holzboog 1976.
12 What, then, does it tum on whether concept Cl is or is not the same as concept C2? The
question is difficult, but the difficulty is not one that we bring down on ourselves by exercising
an escapable option to speak of concepts. Concepts are not philosophical artifacts. They are
74 DAVID WIGGINS

general things we are already committed to thinking about when we quantify (as we
frequently do) over what predicates stand for. Once we see this, we shall not rush to offer
any perfectly general answer, along the lines of Axiom V of Frege's Grundgesetze, or some
modalization of this. Who asks nowadays for a unitary criterion of identity for substances or
continuants? For some important contributions to the proper (that is, the piecemeal) treatment
of the problem, see Hilary Putnam, "On properties', Philosophical Papers, vol. 1, Cambridge,
1975.
13 For the semantics of the copula and other aspects of the difficulties that Frege encountered
with the concept horse, see David Wiggins, "The Sense and Reference of Predicates: a Plea
for the Copula", Philosophical Quarterly, 1984.
14 See Gareth Evans, The Varieties of Reference, ed. John McDowell, Oxford, 1982, Chapter
One.
15 See "The Meaning of "Meaning'''' in Philosophical Papers, vol. 2, Cambridge, 1975.
16 See "Meditationes de Cognitione, Veritate et Ideis" (Gerhardt IV) p. 422; Discourse on
Metaphysics (§24-5); New Essays, pp. 254-6.
17 But only in a sense of 'adequate' that must (I hold) be purged of certain Leibnizian
preoccupations, e.g., the idea that, at the limit, as human knowledge approximates to God's
knowledge, a posteriori knowledge will be able to be replaced by a priori demonstration.
More generally, and against the idea that extension-involvingness itself only reflects a state in
the development of scientific understanding, see Sameness and Substance, pp. 210-213.
18 An earlier version of this paper was given in French at the Institut d'Historie et Philosophie
des Sciences et des Techniques, in Paris 1989. It is published in Revue de theologie et des
philosophie 1992/3. An English version was first published in Adrian Moore (ed.), Meaning
and Reference (Oxford University Press, 1993); also in Reading Putnam, Clark and Hale
(eds.), Blackwell, 1993.

New College, Oxford University,

Oxford, OX1 3BN, UK
 A. C. GRA YLING

CONCEPT-REFERENCE AND KINDS

In "Putnam's Doctrine of Natural Kind Tenns and Frege's Doctrine of Sense,

Reference and Extension: Can They Cohere?" David Wiggins proposes
placing Putnam's suggestions about natural kind terms into a Fregean
framework of sense and reference, adjusting both Putnam and Frege in
interesting ways in the process. 1 A consideration of these adjustments
suggests to me, in turn, a way of defending an aspect of the kind of view
about natural kind tenns that Putnam and Wiggins are at one in rejecting.
Let me make it clear that I am not defending the kind of view they reject:
only one aspect of it, but a central one, having to do with tlie place of
epistemic constraints on mastery of expressions in a language. If the relevant
aspect of that view can be defended, it makes an important difference to the
question of what conception of sense we can award ourselves if we agree, as
we surely should, with those like Wiggins who argue that natural kind tenns
have sense.

To establish a point of approach I wish to begin with some considerations

about a closely neighbouring matter. Recall briefly what "Twin Earth"
examples are alleged to illustrate. Familiarly, my twin and I drink, bathe in
and make remarks about phenomenologically indistinguishable stuff we both
call 'water', but on his earth the stuff so named has the chemical makeup
XYZ whereas here it is H20. Putnam took this to upset a brace of
traditionally platitudinous theses about meaning, one being that meaning
detennines extension, the other that meaning is determined by the content of
certain mental states of speakers (let us abbreviate the discussion required
here by saying: speakers' knowledge and beliefs).2 The Twin Earth case
suggests that these traditional platitudes contradict each other, for the first
implies that 'water' has a meaning in my twin's mouth different from the
meaning it has in mine, while the second -- granting the atom-for-atom
identity of our twinhood and the detennination of psychological states by
states of central nervous systems -- implies the opposite.
The usual responses to the Twin Earth dilemma consist in denials of or
modifications to one or both the traditional platitudes. Enrich the first
platitude by relativisationto context, and contradiction disappears. Or

75

J. Biro and P. Kotatko (eds.), Frege: Sense and Reference One Hundred Years Later, 75-93.
© 1995 Kluwer Academic Publishers.
76 A. C. GRA YLING

distinguish "wide" from "narrow" meaning: the first, extension-determining;

the second, restricted to what governs a speaker's use of an expression by
being psychologically manifest for him in his grasp of it; and again,
contradiction disappears. Or one could even adopt the currently unfashionable
strategy of denying the second traditional platitude, by holding that mental
states are not related to central nervous systems in the premised way. Here
contradiction does not even threaten.
But questions about Twin Earth cases might arise at an earlier point. One
gets drawn into discussing them only because one has already accepted certain
theses about natural kind tenns which load the dice against, in particular, any
version or cognate of the second traditional platitude. Chief among them is
the thesis that natural kind tenns designate rigidly and in virtue of facts about
the real natures of kinds of things -- their essences -- which need not and well
might not be known to users of expressions designating them.
I have had occasion to argue elsewhere that this is a not entirely
convincing doctrine. I rehearse just one of the considerations briefly. Recall
Kripke's proposals about the nature of essence as internal structure. 3
"Tigers," he says, "cannot be defined simply in tenns of their appearance; it
is possible that there should have been a different species with all the
external appearances of tigers but which had a different internal structure and
therefore was not the species of tigers."4 Note first that internal structures
are nested: tigers have internal organs, these are composed of cells, which in
turn are composed of molecules in their turn composed of atoms, and so on
down through elementary particles to so far merely hypothesised structures --
"superstrings" were a recently fashionable candidate -- and perhaps beyond.
Genes must count as a fairly high level of structure in this hierarchy. Now,
suppose we ask which level of structure constitutes the essence. Suppose
two tigers have the same structures all the way down to a certain level --
say, the atomic -- but differ thereafter. Which is the tiger? It could be replied,
as a geneticist might reply, that all levels of structure are uniformly
correspondent, so that similarity at anyone of them entails similarity at all;
this indeed is how geneticists justify identificatory claims about an
individual's species based on investigation of small samples of its tissue.
But why then should this correspondence not run up to the highest level of
structure, namely external structure, so that it -- that is, appearance -- counts
as sufficient for the detennination of kinds? Kripke and Putnam reject this,
but unless some level of internal structure is privileged as uniquely kind-
detennining, independently of whatever competing higher-level structures
might supervene upon it, it is no longer clear why. In practice we rely on
 CONCEPT-REFERENCE AND KINDS 77

our theories about the genetic level in the case of living kinds; but that is
not a resource in the case of elements like gold and stuffs like water; and
anyway our doing so for living kinds does not settle the matter, it only starts
the required debate about what counts as telling the difference between, on the
one hand, correctly identifying the real natures of things, and on the other
hand, classifying them according to our own theoretical convenience and
interests.
Further, if difference of structure entails difference of kind, a
counterintuitive result follows, which is that each individual is its own kind.
For every individual is unique -- this tiger has longer claws than that one --
and such differences result from differences at more fundamental levels of
structure, including those underlying the chemistry of DNA. This
disagreeable result can be blocked by saying that it is a logical sum of the
internal properties which count, the sum of inclusive disjunction. Kripke and
Putnam reject this for external structure, that is, appearance; Putnam allows
only that this can enter into fixing a "stereotype" of the kind in question; but
here what gets disjoined are properties which rarely if ever figure in
stereotypes because they are much less accessible and far more theory-
dependent than those which typically do.
Similar arguments can readily be adduced to unsettle us over other
candidates for essence, for example "origin".5 What they jointly suggest is
that the notion of a natural kind, and hence of a natural kind term, is not well
defined. It was Quine who taught us to be suspicious of claiming that our
way of carving the world follows its joints; which must be a reason why he
finds so bizarre the notion of essence, without which the notion of a natural
kind is -- perhaps insuperably -- difficult to make clear. 6
These thoughts in turn prompt a further: that ill-definition of the concept
of a natural kind makes a difficulty for the view that correct use of natural
kind terms does not depend upon what spea~ers know or do, but on the way
the world is, independently of speaker~. Note what this turns on: a
commitment to the view that rigid designators, like air guns and golf balls,
have correct uses but -- because use and sense fall apart -- no sense: the first
platitude had it that sense determines extension, but on this view extension,
quite literally, determines itself: and on occasions of reference we proceed
only with the help of the rule-of-thumb provided by something like
Putnam's "stereotypes" until the experts arrive. And even then -- supposing
we are ingenuously realist about their sciences, and that we leave aside
questions about the inherent defeasibility of their theories and whether such
78 A. C. GRA YLING

theories are not irreducibly instrumental anyway -- even then it cannot be

guaranteed that the experts will know enough.
This view is, I suggest, to some extent de stabilised by scepticism over
kinds, essence and our ability to determine whether we are carving the world
its way or ours. For it is hard to know what to do with a theory of a given
sort of term which is crucially dependent on a conceptually unsteady category
of things, claimed to be often unknown and potentially unknowable, which
the theory itself postulates and which putatively constitute the determiners of
those ternlS' correct use.
I wish however to carry forward only two aspects of this thought to a
discussion of Wiggins's proposals, given that, in contrast to Putnam's
austerity of view, he embraces a notion of sense for natural kind terms. The
ill-defined character of natural kinds is one aspect. The other is that it is at
least deeply implausible -- one is inclined to agree with a tradition stemming
from Locke that it is impossible -- that our employment of a term can be
determined by something of which we are ignorant. In the framework of the
traditional platitudes, this might be put as a requirement that the second of
them -- that what we know or believe enters constitutively into what our
expressions mean (that what we mean by them enters into what they mean: a
Griceful thought) -- controls their relation to what they range over, at least as
much as they are controlled by it in tum. This point is central in what
follows later. At this juncture it need only be remarked that a large part of
the pressure towards thinking this way comes from the reflection that just as
it is intolerable to suppose that speakers are uninvolved in settling what the
expressions they utter apply to, so it would be intolerable to countenance a
third-realmist realism about senses which regards their objectivity as such
that an expression may have a sense which no current user of the language
attaches to it: that indeed an expression may have a sense which has, say,
been forgotten and will never be recovered, and that speakers who attach a
different sense to it are accordingly just mistaken'?
There is one more point I must mention before turning to Wiggins's
account. At the outset of thinking about these matters, there is a natural
enough hope that one might be able to give a univocal account of
expressions having potentially predicative use, and it is a discovery of some
moment, if discovery it is, that this is not so. In the case of natural and non-
natural kind terms it may turn out that a univocal account of them is
impossible precisely for Putnam's reason, that whereas one might be able to
define the latter in terms of conditions which are both necessary and
sufficient for their application, no such definitions are possible for the
 CONCEPT-REFERENCE AND KINDS 79

former. We are reminded here of a thought about grounds for attributing

mastery to speakers in the case of colour terms, as an example of the cases of
at least many general terms -- not only characterising terms but sortals too.
Plausibly, speakers may be credited with grasp of a colour term on the joint
evidence of these three marks: that they recognise what counts as a focal case
of the colour in question, that they consistently reject application of the term
in focal cases of what is not that colour, and that they agree by and large
with fellow-speakers in hesitating over application of the term on the
indeterminate boundaries between that colour and its neighbours. It is open
to us to suggest that mastery of natural kind terms is best understood in
rather this way; that the chief difference between mastery of this sort and
knowledge of necessary and sufficient conditions is degree of open-endedness
of the concept in play; and that in both cases application of terms is
epistemically driven in quite definite and straightforward ways. Once again,
this point re-emerges in due course, and is by no means at odds with the
modified Putnamian account Wiggins wishes to give. s

II

I turn now to Wiggins's view, which I describe and discuss in detail because
I propose to define my own view by differing from it in what I shall argue
are crucial ways at crucial points.
For Wiggins the attraction of Putnam's theory of natural kind terms is
that it attaches their meaning to the real natures of kinds by a deixis, a means
of assignment by exemplification in which a term is attached to a kind under
"special and favourable" circumstances for conveying what it is about
samples that can lead on to a richer grasp of the term's extension. Deixis is
not to be understood so that natural kind terms are compared to indexicals or
demonstratives; 'water' must not be compared to 'this'. Nor must it be
understood as making mere ostension the key. The comparison Wiggins
invites us to draw is, rather, to proper names, understood in terms of a view
in which they have their meaning, and make their contribution to truth-
conditions, by standing for their bearers; and in which the only way to give
their senses is to say which objects they stand for. (Wiggins, p. 60)
Moreover, the senses of proper names are "reality-invoking or object-
involving", and they have their senses "by being assigned to something, not
by the laying down of a specification such that a bearer of the name bears it
by virtue of meeting that specification." (ibid.) The prospect held out by
80 A. C. GRA YLING

Putnam's proposal, Wiggins suggests, is that something very similar can be

said in the case of natural kind terms.
In Wiggins's view the reason why philosophers found it difficult to arrive
at a deictic or extension-involving conception of meaning for natural kind
terms was their conviction that, despite Quinean strictures, there could still
be well-founded agreement on what counts as analytic. And accordingly they
still found it possible to attempt analyses and to look for non-circular
necessary and sufficient conditions. Putnam's proposal constitutes a more
moderate and effective second onslaught on analyticity. It does not deny that
definition by necessary and sufficient conditions can be given of single-
criterion concepts, but it refuses to generalize from them. Once this point is
taken, the possibility offers of exploring the idea that there are "sense-giving
relations between predicates and kinds that could mimic the direction of fit
we now take for granted in our understanding of the setting up of the relation
of designation." (Wiggins, p. 64)
Wiggins's aim is to show how the extension-involvingness of natural
kind terms coheres with the theory of sense and reference. He begins by
glossing the diagram Frege drew for Husserl in a letter of 24 May 1891.
What he says about the middle column is important because his account of
concept-reference -- the third column -- centres upon it.

Sentence Singular term Concept-word

~ i i
Sense of the Sense of the Sense of the
sentence singular term concept-word
(Thought)

i i i
Reference of Reference of Reference of ~ Object(s)
the sentence, the singular term, the concept-word, that fall under
a truth value an object a concept the concept

The notion of the sense of a singular term is to be understood in the light of

Frege's remarks about fictional names. Frege says that the name 'Nausicaa'
has sense, in so far as it does, by behaving as if it designated some girl.
Wiggins takes this to license a simpler claim for the non-fictional case: "a
name has sense by behaving as if it has -- or simply by having -- a reference.
A name has its sense then by somehow presenting its object. To grasp the
sense of a name is to know (in the manner correlative with the mode of
presentation that corresponds to this somehow) which object the name is
 CONCEPT-REFERENCE AND KINDS 81

assigned to." (Wiggins, p. 65) Taking the sense of a name as its mode of
presentation of an object means that we have two things here: an object that
the name presents, and a way in which it is presented. This latter Wiggins
calls the "conception" of the object. This is "a body of information --
typically open-ended and imperfect, and hence rarely if ever condensable into
a complete description of the object -- in which the object itself plays a
role." We must be careful to insist on the open-endedness of conceptions in
order to distance the account from one in which some description or set of
descriptions of a thing is taken to be synonymous with its name. The most
that we can reasonably expect is that the conception of something X to
which the sense of X' s name is "keyed" might generate various descriptions
of X which in suitable circumstances would help us to identify it. So:

Properly possessed then, the conception of an object X that sustains the sense of a
name that is keyed to that conception is a way of thinking about X that fixes (with
the help of the world, the world being what helps create the conception) which
object the object in question is. Which object the object is is precisely what is
mastered by him who comes to understand the name with the sense corresponding to
that conception. (Wiggins, p. 66)

With these points made, Frege's third column is glossed by paralleling

concept-words and singular terms. The parallel is claimed to be direct. Just as
grasping the sense of a name and its contribution to truth-conditions is to
grasp the conception -- "corresponding to the name's mode of presentation of
what the name stands for" -- of the object, and just as giving the name's
sense is to say, "in a manner congruent with that conception", what object it
stands for, so grasping the sense of a predicate and its contribution to truth-
conditions is to grasp its mode of presentation of the conceptit stands for.
Giving its sense is done by saying what concept it stands for. We
demonstrate preferences for one sense over another by preferring one mode of
presentation or body of information over another; examples of the two ways
of expounding bodies of information about, say, horses, might be found
respectively in a dictionary and a zoological textbook. (Wiggins, pp. 66-7)
Thus armed, Wiggins can anticipate concluding that Putnam's stereotypes
approximate to the conceptions which sustain the senses of natural kind
terms; stereotypes are "particular special ways of thinking in identificatory
fashion about such kinds and their specimens". (Wiggins, p. 67) But this is
so far only an anticipation: there are adjuncts to the view which make a
significant difference. These become manifest upon closer examination of the
analogy between columns 2 and 3.
82 A. C. GRAYLING

As the diagram makes clear, predicates do not refer to objects but to

concepts. In his letter to Husser!, Frege points out that it takes one more
step to reach an object from a concept-word than from a singular term, but
that it may also happen that the concept word is scientifically useful even
though the concept it refers to is empty. Thus his reason for putting objects
off to the side in his diagram, but at a level with concepts, is to show that
they share the same objectivity. Thus the thesis that predicates refer to
concepts is wholly general, and neither brings out what is distinctive of
natural kind terms nor what a theory like Putnam's makes of them. On
Putnam's theory the point is that one has to be exposed to the extension of
the term to acquire mastery of it; there is no grasping the reference or concept
otherwise. Understanding cannot be acquired by grasping a "strict lexical
definition", for no such thing is available. "So whereas the reality-
involvingness of proper names amounts to their being reference-involving,
the reality-involvingness of natural kind terms amounts to their being
extension-involving." (Wiggins, p. 68) And this permits us to say that what
it is to have a grasp of say, the concept horse is to know that Red Rum and
Arkle are horses, that is, to know the truth-conditions of 'Arkle is a horse' .
What is being welcomed in Putnam's account is the thought that natural
kind terms can only be mastered deictically; specifying the concepts they
refer to depends on giving examples. But Wiggins insists that this can only
happen if what the speaker thereby acquires is a recognitional capacity,
sustained by a "conception" or body of information. In teaching a tyro a
natural kind term, what is imparted is just such an identificatory or
recognitional conception.

It is this that corresponds to the sense of a natural kind term. What one must say then
is that the sense of a natural kind term is correlative to a recognitional conception
that is llnspecifiable except as the conception of things like this, that and the other
specimens exemplifying the concept that this conception is a conception of.
(Wiggins, p. 69)

In sum (taking the term 'horse' as example) the doctrine is that "that which
supports the sense of the substantive 'horse' is a certain identificatory
conception of the concept horse". (ibid.)
A conception and a concept are of course different. A conception is a way
of thinking about a concept (or about the objects falling under it). More to
the point (and here is where Fregean and Putnamian considerations are
merged), a conception is -- with one important adjustment -- a Putnamian
stereotype. The adjustment is this. Recall that by "stereotype" Putnam
 CONCEPT-REFERENCE AND KINDS 83

means, in Wiggins's formulation of the view, "a fund of ordinary

information or a collection of idealized beliefs that one needs to grasp in
order to get hold of the meaning of a thing-kind word". Now Putnam insists
that the stereotypes and the extension of natural kind terms are quite
independent. A speaker may have a stereotype for 'horse' but no knowledge
of the extension; such knowledge is the preserve of experts. He supports this
by laying celebrated claim to arboreal ignorance: I know the stereotype of
elm and the stereotype of beech, he says, but I cannot tell them apart. One
might naturally be sceptical about the segregation; there must be some link
of the kind sketched earlier when conceptions were introduced. What we
should hold is that a stereotype is the stereotype of a given concept, and that
grasping the stereotype "represents the beginning of an identificatory capacity
... which could advance to the point where it became the capacity of the
expert." (Wiggins, p. 71)

III

Changed emphases, and a few crucially located adjustments, yield a picture

which differs from this otherwise attractive account in significant ways -- not
least, as suggested, in offering a theory of the sense of concept-words which
respects epistemic constraints.
First I note that Wiggins accepts or at any rate does not question the view
of natural kinds central to Putnam's account; in particular, he does not
question how well-defined the notion is, and whether, as I gave reasons for
thinking, it is in fact not so, and this because kinds themselves are not; and
whether therefore the open..:endedness of concepts of kinds, and the correlative
impossibility of giving "strict lexical definitions" of them, is a function of
this fact. For Wiggins, it would seem, the imperfection of the bodies of
information in which conceptions consist is a function of our ignorance
about something that is, independently of us, quite objective and determinate,
namely, the extension of the term whose sense that conception sustains.
This view is opposed not just to anti-realist theses which might be variously
nominalist or conventionalist about kinds (opposing the objectivity claim),
but by versions of realism about kinds in which they are just nqt very neat
(opposing the determinateness claim). The world, on such a v~ew, might
simply be genetically -- or, more basically, chemically -- ambiguous in ways
that made kinds more like colours than like artefacts, most of which are
completely determinate, a characteristic of non-natural kinds as the point
about single-criterion applicability of their designaters shows. E~ther on the
84 A. C. GRAYLING

anti-realist or realist possibilities here mooted, a difference is made to our

understanding of kind terms.
Having registered these thoughts, let us leave them in the background for
the present. The first foreground anxiety concerns conceptions and their
relation to senses. There is something uneasy about this. A conception,
recall, is a body of information. At first Wiggins says, in speaking of
Frege's claim that a sense is a "mode of presentation", that this involves the
existence of two things: an object -- the reference of the name -- and a way
the object is presented. But it immediately appears that the way the object is
presented is not, as one would expect, the sense, but a conception, that is, a
more or less rough body of information in which the object figures. 9 The
move that troubles is that alongside 'sense of the singular term' in Frege's
diagram, and off to one side, we must make an additional entry; we must
write 'conception'. The conception is more or less, in effect, what a cluster
of descriptions is in the theory which equates such clusters with senses; but
Wiggins keeps the two apart, and speaks of senses being "keyed" to
conceptions, or "sustained" by them, or as "corresponding" or being
"congruent" to them; and that we give a term's sense "by saying in a manner
congruent with that conception what the term stands for" (Wiggins, p. 66).
Both the relation and the difference between senses and conceptions is
hard to specify. In the passage under consideration what troubles is the
immediate move to saying that terms have both reference and sense but that
what presents the reference to thought in a certain identificatory way is a
third thing, the conception: "To speak as Frege does of sense as 'mode of
presentation' suggests then nothing less than this: there is an object that the
name presents and there is a way in which the object is presented, or a
conception, as I shall say, of that object" (Wiggins, p. 65) Here 'modes of
presentation' has shifted from being what senses are to being what
conceptions are. Conceptions thereafter do all the work; they fix the reference
("with the help of the world"), present the object, and serve as what must be
taught a tyro in his acquisition of mastery of the term.
It is clear that Wiggins is trying to preserve the advantages of something
similar to a description theory of sense, for this makes a good marriage with
Putnam's notion of stereotypes, while trying to avoid the disadvantages of
equating the sense with a given body of information. But if the informational
aspect of grasp of a term is hived off into a detachable package, namely the
conception, what is there left for the sense to be? As it stands, Wiggins has
dieted Frege's notion of sense to something blade-thin. A name's sense is
what it stands for, "congruent with the conception" of the object which
 CONCEPT-REFERENCE AND KINDS 85

"sustains" that sense. It is clear also that Wiggins seeks to avoid espousing
a simple denotative theory in which the meaning of a name is the object it
stands for, by having conceptions "sustain" the relation of "standing for" --
to clothe the nakedness, so to say, of the denoting relation. But at the same
time he does not wish this package of information to constitute the sense,
because that would be to espouse the equally questionable description theory.
So he seeks to manouevre between the two by having something of both,
held apart yet linked by the presence of the metaphorical relation of
"sustaining", "corresponding", "being congruent with".
It is not a criticism of Wiggins to say that he is seeking some way of
having the best of both sorts of view in a satisfying account of sense, for if
he succeeds he provides a legitimate addition to the options. The problem
lies with the vagueness of the relation between conceptions and senses. I
asked: if conceptions are detachable packages of information, what is there
left for senses to be? Whatever the answer to this question is, it will have in
important part to turn on the nature of the relation between conceptions and
senses. As long as that relation remains vague, so does the notion of sense.
One is inclined to suspect that more express characterisations of the relation
would let information bleed from conceptions into senses: which would call
into question Wiggins's grounds for thinking that the relation between them
is much short of identity.
A complication here is that it is anyway not clear whether the relation we
are seeking to understand obtains between a one-to-one pair, a conception and
a sense, or whether it is many-one, several conceptions correlating with a
single sense. 10 One possibility is that sense is something minimal. The
sense of a given designating term N consists in its having a given object X
as its reference; the sense of N is 'that X is its reference'. Then if some other
term M has X as its reference also, Nand M have the same sense, viz, that X
is their reference; and what makes an identity statement 'N = M' informative,
if it is so, is that they differ in the conceptions associated with them. As
Frege's view is usually understood, what Frege meant by Sinn is in this
respect at least similar to what on this construal is meant by 'conception'.
However, Wiggins is intent upon seeking to exploit Putnam's notion of a
stereotype, a notion particularised to the case of natural kind terms; and this
Putnamian notion is rather different from a notion of conceptions which can
stand many-one to a sense, for it would seem that there can only sensibly be
one stereotype per natural kind term, since it is hard to see how a given
stereotype for a given kind could properly serve as such if a different body of
information could also serve as the stereotype for that kind. Given this, and
86 A. C. GRA YLING

because it is natural kind terms we are thinking about, at least in this case it
would seem that Wiggins must or should think that there is one conception
per sense; consistently with this interpretation he speaks throughout of the
conception keyed to or sustaining the sense of a given natural kind term.
Either way, however, we are left with the difficulty that the relation
between conception and sense is unclear, and that so also therefore is the
notion of sense itself. These difficulties can be entirely avoided if we
recognise that there is already a way of identifying senses and conceptions
without inviting the difficulties inherent in theories like the description
theory. Recall that a conception is a body of information, imperfect and
open-ended, which can be extended to the point of expertise by refinement
and accretion of information. But despite its imperfection and open-
endedness, the conception -- considered as Putnamian stereotype adjusted so
as not to be independent of the extension -- has to be such that it gives
ordinary users of the term a mastery of it sufficient for the purposes, surely
exigent enough, of ordinary communication. Possession of the conception-
stereotype confers a capacity on the possessor to identify the reference of a
term. Now, what is the objection to cutting through the multiplication of
notions here by saying that this is what a term's sense is: an open-ended,
extensible body of information, possession of which enables speakers to
identify the term's reference? This keeps better faith with Frege's intentions,
while not laying claim to the existence of a specific piece of information --
some one description or cluster of them -- which is the term's synonym.
What might be gained by thinking this way? Consider the most salient of
facts about language. Because it is a public instrument of communication,
for the functioning of which it is a necessary condition that senses be public
and stable, it follows that the bodies of information in which, on this view,
senses consist must be shared by users. On Putnam's view a stereotype is a
core of information, which implies a minimum basis for mastery. Wiggins
adopts this for conceptions, noting that beyond it there is a large area for
additions up to expert level. There seems to me no more difficulty in holding
that the shared public sense of a term can be given as the open-ended core of
information required for identifying its reference, than there is in requiring as
much of Putnam's stereotypes or Wiggins's conceptions. This is exactly
what dictionaries and entries in reference works do.
The point can be explicated as follows. In saying what is involved when
a tyro is taught a natural kind term, and arguably any designating expression,
it would be implausible to hold that the act of fixing the reference of the
term for him stops short there, for this is to hold that there can be such a
 CONCEPT-REFERENCE AND KINDS 87

thing as bare baptismal knowledge of the reference of a term, possession of

which is all that mastery of the term need consist in. For in coming to
identify something X as the reference of N the tyro acquires information and
beliefs about X beyond the mere fact that N designates it; information and
beliefs of just the sort Wiggins has in mind as conferring a recognitional
capacity. It would be bizarre to hold that when Fred tells Joe 'this is a tiger,
and so is that,' and so on in the manner of giving Joe a conception for
'tiger', Joe thinks, believes or conceptualizes nothing further about the
individuals so named and the collection to which they are thus asserted to
belong. Indeed the fullest way to reconstruct Joe's musings as Fred teaches
him to talk about tigers is surely to describe them as amounting to
something like this: 'so, these large striped meat-eating four-footed growling
animals are called "tigers".' (I say 'to reconstruct fully' because Joe might,
say, be too young to put matters just this way. But this is more or less what
he has to know when he knows what he is talking about when he is talking
about tigers.) In so musing, Joe has the correctable and extensible makings
of something, knowledge of which relates that word in his language to those
things. That something I am calling the 'sense'. Further, if it really is a
grasp of the term he can thus be credited with, what he has is therefore what
all competent users of the language have, for only then are he and they in
communication.
Espousers of theories in which there is no more to be said about
designating expressions, in this connection, than that such-and-such is
sufficient for fixing their references, fail to take into account certain
complexities which attach to them in different ways. In the case of a natural
kind term, use of it has to be understood in a twofold way: concatenated with
an article or demonstrative, it is used to refer to members, qua individuals, of
a kind; but it is also understood that a certain relation obtains between
individuals so referred to, which significantly groups them together; the
relation being some kind of relevant similarity. One can see a case for saying
that such notions as relevance and similarity are irreducibly epistemic, and
that therefore the senses of such terms have fairly rich graspable descriptive
content from the outset. But even if it were held that the relevant similarities
in virtue of which individuals collect into kinds are recognition-transcendent,
it remains the case that speakers have to be able to identify individuals, and
identify them as members of their kind, in virtue of the (perhaps unknown)
basis for this grouping, from which definite entailments follow, to govern
what speakers can say and do in related connections. So here too we have
88 A. C. GRA YLING

cause for thinking that built into the senses of such expressions is what
"stereotypes" or "conceptions" are claimed to do for users of them.
My first suggestion, then, is that the sense of a term is what Wiggins
describes as a conception, and that there is therefore no reason to multiply
entries in Frege's diagram by having 'conception' pencilled alongside 'sense'.
But this is not to suggest that there is any straightforward parallel between
the second and third columns in Frege's diagram; as some questions
prompted by Wiggins's account suggest, matters may be otherwise.
The most obvious first thought is that when a parallel between singular
terms and natural kind terms is mooted, this will have natural kind terms
designating kinds, suitably hypostatized to rank alongside the middle
column's objects. But no suggestion is made that in addition to individual
tigers the world contains this extra object, the kind tiger. Rather, what a
natural kind term refers to is the concept of a kind. Now Frege is able to
keep his diagram neat, and to have concepts as references in the third column,
because he treats them as having the same objectivity as the objects which
fall under them when any do. But two difficulties which immediately press
are these: are we clear about what we are committed to if we go along with
Frege's ontology of concepts, relations and functions? and, if we make an
independent assessment of the comparative statuses of, say, a tiger (the beast
out there in the jungle) and our ordinary concept of a tiger, are we happy to
regard whatever parallels subsist between the expressions respectively
designating them as direct?
We have at least been alerted by Dummett to the difficulty of pressing
analogies between singular terms and concept-words too far.ll The sense in
which in Frege's view the latter can be said to have reference, granting
already that the existence of concepts (relations and functions) commits us to
admitting second-order quantification, is one in which, says Dummett, "the
notion of reference does not play the same role in regard to them as it does in
regard to proper names". (Dummett, p. 245) What is crucial in understanding
predicates, relational terms and functional expressions is, familiarly and
respectively, being true of an object, holding between objects, and yielding
given values for given arguments. If one thinks of reference in terms not of
its semantic role -- that is, in terms of contribution to truth-conditions -- but
of an identification of the reference of an expression with its bearer, as in the
prototypical case of naming, then the claim that predicates refer simply looks
implausible. (Dummett, pp. 210-11) The adapted Putnamian account of
natural kind terms under consideration is based on an analogy with this latter
 CONCEPT-REFERENCE AND KINDS 89

aspect of reference, the relation of a name to its bearer. Perhaps this is the
source of at least one of the worries it prompts.
A discussion of the complexities thus uncovered can be shirked if we
leave Fregean exegesis aside and simply ask what is the least that we can
allow as constituting grasp of a predicate term. And here we are helped by the
thought that to understand a predicate term is at least to know how to apply
it: grasp of its sense is grasp of a criterion of application for it, the core
capacity thus provided being an ability to judge the truth-value of basic
predications effected by its means. More may well be required, depending on
the kind of predicate in question; but we should find it hard to attribute to a
speaker grasp of the sense of any such term which did not endow him with at
least this. 12 In tum, this suggests that what is to be grasped crucially about
their senses is what it is for them to apply to or be true of items of suitable
types. Taking Wiggins's notion of a body of information again, we might
say that the sense of a concept-word is the core information we must have to
enable us to apply it in exemplary cases -- more generally put, perhaps: to
recognise the truth-values of sentences in which such expressions occur on
exemplary occasions. The 'exemplary' here shows that we are keeping faith
with the idea of the openness of senses to addition and development to expert
levels, while at the same time respecting the constraint represented by the
demand that speakers share a core grasp of senses if language is to work as a
public instrument of communication.
To this we can add a further point. Theories of sense typically
underestimate or ignore the creative ability of speakers to grasp rules for the
use of terms in new or variant applications. This is not a merely empirical
claim; it exploits another requirement for the possibility of language,
namely, that since speakers have to learn language for severely practical
purposes, and with finite capacities in a limited time, and moreover since
vagueness and open-endedness are essential properties of many expressions
and their senses respectively, it is clear that it is a necessary condition for the
existence of language that speakers should be able to extrapolate correct uses
from sample cases. It is therefore no surprise that senses may have the open-
ended character here attributed to them. So we can specify that a learn ability
condition has to apply as adjunct to the publicity condition just sketched, and
that for this, senses have to have the character here c1aimed.13
These thoughts do not amount to a theory of the senses of concept words
in general; perhaps they approach the identification of constraints to be
respected inthe construction of one. But even as such they entail a redrawing
of the third column in Frege's diagram, and at the very least the weakening
90 A. C. GRA YLING

of the analogy between the second and third columns which Wiggins finds
attractive for the case of natural kind terms. Since the task of offering the
detail of a positive alternative theory is not to the point here, just one remark
is pertinent: it is that on the minimum specification given for the grasp of
the sense of a concept word, any concept word which applies to nothing
retains its sense because what is known by one who understands it is what
would count as an exemplary instance of its application if ever one offered.
This at least squares with Frege's view that concept-words which refer to
empty concepts can still be useful, even scientifically so. And it also squares
with the thought that one should treat the reference of concept words
primarily in the semantical sense -- as explaining how such words contribute
to the truth-conditions of sentences in which they figure -- and not on that
aspect of the analogy with singular terms in which the reference of a term is
identified with its bearer.
Wiggins remarks that it is distinctive of natural kind terms that a
necessary condition for grasping their senses is exposure to their extensions.
Such grasp, Wiggins insists, cannot be acquired by means of a "strict lexical
definition". This second point has already been embraced; it seems not
merely right but crucial to an understanding of the notion of sense for such
terms that they be taken as open-ended and approximate. But what is rejected
here is not the idea of a lexical definition, but the idea of strictness. For it is
clear that talk of "exposure to the extension of natural kind terms" cannot be
taken to be a demand that only those who have been to the zoo or had an
adventure in the jungle grasp the sense of 'tiger'. As we would colloquially
say, I know the meaning of many kind words despite not having been
perceptually exposed to individual members of those kinds; a grateful
example is 'rattlesnake'. But if exposure to the extension of such terms can
mean the kind of lexical definition which conveys core information and thus
a recognitional capacity for exemplary cases, while leaving open further and
variant possibilities, then there can be no cavil over "exposure". Generalising
from natural kind terms, we might wish to say that concept words which, in
Frege's terminology, refer to empty concepts, can nevertheless be
understood, because we can be (so to say) lexically exposed to -- it is more
accurate to say: given an understanding of what it would be for something to
fall into -- the extensions they would, in better or fuller worlds, have.
A point for Wiggins's proposals here is this: on Frege's view we can
grasp the senses of concept words which lack extensions. Wiggins points out
that it is distinctive of natural kind terms, a subclass of concept words, that
exposure to the extension is a necessary condition for grasp of them. So a
 CONCEPT-REFERENCE AND KINDS 91

further divergence between Wiggins's diagram and Frege's appears: the

account of concept words is not univocal. This is not by itself a criticism,
merely an observation. But note that Wiggins's analysis makes natural kind
terms more like singular terms than other concept words are. So perhaps
natural kind terms should appear in a fourth column, intermediate between
the second and third on Frege's diagram. This emendation to the scheme is a
function of realism about natural kinds, expressing itself through the demand
that the extension be integral to the determination of sense (on Wiggins's
view,via conceptions). But now recall the anxiety over the well-definedness
of natural kinds and hence terms designating them. Suppose for a moment
that the way we carve up the world is a response not to its contours but to
our capacities and interests, and that therefore kinds are conceptual artefacts.
Then natural kind terms are concept words; and may refer to empty concepts;
and the account to be given of them and their senses is the same as for
concept words in general. Then no fourth column needs to be added to
Frege's diagram. We get a univocal and hence more economical account, we
avoid problems over the doubtful aspects of natural kind theory -- and even if
everything realists assume about natural kinds turns out to be true, our
theory would still explain how we grasp, as manifestly we do, the senses of
expressions used in talk of them. This view is less committal than
Wiggins's, and makes no ontological assumptions about natural kinds; so
even if the view of natural kinds that Wiggins assumes turns out to be true,
this theory is saved as part of the stronger theory that would result.
But explanation in terms of core conceptions or stereotypes -- just as for
concept words which refer to empty concepts -- of how natural kind terms
have publicly shared senses, is not going to be made somehow more adequate
by adding to 'this is how to recognise an X if there are any' the clause 'and
there really are some!' This extra fact does not make one understand more or
better what X is. Wiggins took it that the extension-involvingness constraint
ensured the realism of the reality-involvingness he took this to entail: these
thoughts show he is wrong.
Perhaps the point which will be least congenial to some in this response
to Wiggins is the invocation of a notion of sense which says that natural
kind terms have senses consisting in describable informational content
which, as grasped by speakers, is what presents the terms' references to them
in a way that is common to at least the very great majority of such speakers.
I iterate the point that Wiggins is surely right in arguing that whatever
information a speaker needs to acquire mastery of such a term, it must be.
correctable and extensible, and that ordinary employment of the term differs
92 A. C. GRA YLING

from expert employment only in degree -- in quantity of knowledge or theory

-- not in kind. But to this must be added the condition that whatever
information does the trick in giving a learner masery of the term, that
information must be the same as, or at the minimum must mesh with, the
information guiding any other ordinarily competent user of the term. If such
information constitutes sense, then the point is to be put by saying that
senses must be public (Frege talks more strongly of senses being
"objective"). The modality here is straightfoward: senses must be public,
shared and relatively stable because language is an instrument of
communication, and it is a necessary condition of its existence as such that
at very least the great majority of speakers understand the same thing by the
same expressions at the same time. 14 This means that something stabilizes
senses; that something counts as -- at anyone time -- authoritative, despite
long-term changes, as an account of an expression's meaning. This
authoritative or standard sense of an expression is the product of agreement in
use between speakers, and is what gets reported by lexicographers when they
write their dictionaries. It is for this reason that, as we have all noticed from
time to time, dictionaries are so useful. It is also for this reason that we are
prepared to defer, when it comes to improvement of our grasp of natural kind
terms, to those who are most informed in their uses of them. 15

NOTES

1 Wiggins, D. 'Putnam's Doctrine of Natural Kind Terms and Frege's Doctrine of Sense,
Reference and Extension: Can They Cohere?' in (ed) A.W. Moore, Meaning and Reference,
Oxford, 1993. At the 1992 Kar10vy Vary conference, my paper was a response to an earlier
version of Wiggins'S paper. Although at the conference Wiggins made some adjustments to his
argument in his response to my response, the main points at issue remain eminently pursuab1e.
I am grateful to Wiggins for discussion of them, then and since.
2 Putnam, H., 'The Meaning of "Meaning'" in Mind, Language and Reality, Cambridge, 1975
3 Kripke, S., Naming and Necessity, Harvard, 1980. For this response, see Grayling, A.C.,
'Internal Structure and Essence', Analysis, 1982.
4 Kripke, ibid., p. 156.
5 Since all living things apparently derive from primeval slime, some node on the evolutionary
tree will have to be selected as the place where, say, man's origin lies. Where is that? When
one is selected, presumably on genetic grounds, the question already encountered becomes
pressing: does this reflect our classificatory decisions, or the real nature of things? How do we
tell which? And so forth.
6 There are other ways of making the point about the ill-definedness of the concept of a
natural kind. One is to question whether designators of kinds are truly rigid (cf. W.A. Collins,
'Types, Rigidity and A Posteriori Necessity', Midwest Studies XII). Another is to say that we
may have variable conceptions of kinds not dependent on essential facts; so XYZ might just be
another sort of water, microstructural differences from H20 notwithstanding. (cf. Tim Crane,
'All the Difference in the World', Philosophical Quarterly 41 (1991»
 CONCEPT-REFERENCE AND KINDS 93

7 What prevents this from being the logically extreme consequence of holding sense and idea
apart as Frege does?
8 It is worth wondering in passing whether or not on the Putnam-Kripke view of natural kind
terms, it should be held that from the vantage point of an ideal epistemic agent (say, God),
necessary and sufficient conditions for the application of any natural kind term can be given.
Presumably so; which would make our account of natural and non-natural kind terms
univocal, the apparent difference between them being merely a function of the ignorance
suffered by finite epistemic agents. This view indeed conforms neatly with the principle in
traditional syllogistic that the difference between natural and non-natural kinds is that the
former fall under an indefinite number of predicables (perhaps too many for us to know) and
the latter under just one. Alex Orenstein suggested to me a different way of effecting the
assimilation of the two kinds of kind-terms: since God made all the natural kinds, they are non-
natural (artefact) kinds anyway, and this underwrites the univocality of the account to be
given of both.
9 This too raises a question: why not say directly that a conception is a body of information
about the object? Perhaps Wiggins chooses the less committal formulation to insure against
convergence with description theories of names.
10 Michael Woods suggested this point to me in discussion.
11 Dummett, Frege: Philosophy of Language, 2nd ed. (1981), Ch. 7, passim.
12 Dummett, p. 233. Dummett brings out for example the difference between common nouns
and adjectives as to ingredient criteria of application and identity.
13 This aspect of grasp of sense may contain materials for dealing with the problems which,
Wiggins acknowledges, infect deixis. (Wiggins, p. 62)
14 See my "Publicity, Stability, and 'Knowing the Meaning"', in A.C. Grayling and Petr
Kotatko (eds.), Meaning, Oxford, forthcoming.
15 I am grateful to those at Howard Robinson's summer vacation meeting group in Oxford,
especially Alex Orenstein, Michael Woods, Howard Robinson, Anita Avramides, Michael
Lockwood, John Foster, and Michael Martin, for their comments.

Birkbeck College, University of London

Malet Street
London, WCIE 7HX, UK
 FRANCOIS RECANATI

THE COMMUNICATION OF FIRST PERSON THOUGHTS

In his article 'The Thought' , Frege writes:

Every one is presented to himself in a particular and primitive way, in which he is

presented to no one else. So, when Dr. Lauben thinks that he has been wounded, he
will probably take as a basis this primitive way in which he is presented to himself.
And only Dr. Lauben himself can grasp thoughts determined in this way. But now he
may want to communicate with others. He cannot communicate a thought which he
alone can grasp. Therefore, if he now says 'I have been wounded', he must use the
'I' in a sense which can be grasped by others, perhaps in the sense of 'he who is
speaking to you at this moment' ... (Frege 1918-19, pp. 25-6)

From this passage, two important ideas can be extracted. The first one is
commonly accepted nowadays:

(a) First person thoughts concerning a person A can be grasped or en-

tertained only by A Another person, B, can entertain thoughts about A,
but not first person thoughts about A: only A can think of himself in the
first person. To be sure, B can also entertain first person thoughts; but
these thoughts will be about B, not about A (Even if B falsely believes
that he is A, that would not make his first person thoughts thoughts
about A)

Once we accept (a), a problem arises, which I call 'the paradox of the first
person'. First person thoughts are private, hence incommunicable; yet we do
communicate them, by uttering first person sentences. How do we manage to
do this? Frege's second idea is meant to solve the paradox:

(b) There are two sorts of senses or modes of presentation associated

with the first person. Let us call the 'special and primitive' mode of pre-
sentation which occurs in first person thoughts 'Ego' or rather 'Egox'
where 'x' stands for the name of the person thinking the thought (for ex-
ample 'EgoLauben' in the case of first person thoughts about Lauben).1
This mode of presentation must be distinguished from the mode of pre-
sentation associated with the word '1' in communication ('he who is

95

J. Biro and P. Kotatko (eds.), Frege: Sense and Reference One Hundred Years Later, 95-102.
© 1995 Kluwer Academic Publishers.
96 FRANCOIS RECANATI

speaking to you at this moment'). The latter can be grasped by others,

the former cannot.

The paradox is solved because, according to Frege, we do not communicate

the original, incommunicable first person thought involving the mode of
presentation Egox , but a different thought involving the other sort of mode
of presentation.
One might think (and some have said) that the distinction between the
two modes of presentation is ad hoc and designed only to solve a particular
problem. I, on the contrary, thinkit is an essential distinction -- one which
lies at the very foundation of the theory of language use. In earlier writings
(Recanati 1990, 1993) I have drawn a similar distinction on quite independent
grounds. The mode of presentation 'he who is speaking to you at this
moment' -- technically, 'the utterer of this token' -- closely corresponds to
the conventional meaning of 'I' (at least if the token-reflexive analysis of
indexicals is correct), yet it is clearly distinct from the mode of presentation
that occurs in our first person thoughts, i.e., the special and primitive way in
which everyone is presented to himself.2 This distinction between what I
called 'linguistic' and 'psychological' modes of presentation is general (it
affects every indexical, not merely 'I'), and it holds whether or not we like
Frege's solution to the paradox of the first person.
Even though it is far from ad hoc, Frege's solution is not altogether sat-
isfactory, for it is sketchy and incomplete. What is the relation between the
two sorts of mode of presentation mentioned in (b)? What makes it possible
for the intersubjective sense associated with 'I' in communication to stand
for the private sense Egox which cannot be directly communicated? These
are important questions which Frege does not address, let alone answer.
There is another solution to the paradox, which does not rely on the dis-
tinction between the two sorts of mode of presentation but on another dis-
tinction (Dummett 1981, pp. 122-3). The speaker's first person thought is
not, and cannot be, 'communicated' because the hearer does not, and could
not, come to entertain that very thought as a result of the communication
process. Still, it may be argued, the speaker's first person thought which is
expressed by the utterance 'I have been wounded' can be recognized as such
by the hearer. Even if the speaker's thought is unavailable to the hearer, the
utterance may inform the hearer that the speaker entertains a certain type of
thought, which he himself (the hearer) is unable to entertain. The speaker
who says 'I have been wounded' expresses a first person thought, to the
effect that he himself has been wounded. The hearer, upon understanding the
 THE COMMUNICATION OF FIRST PERSON THOUGHTS 97

utterance, can only form a different thought: 'He has been wounded'. Unlike
the speaker, the hearer does not think of the referent (i.e., the speaker) in a
first person way. So the speaker's thought has not been 'communicated' in
the strong sense of the term. Yet it has been communicated in a weaker
sense: Leo Peter knows which thought Lauben has expressed in saying 'I
have been wounded'. Along these lines, it may be found unnecessary to dis-
tinguish between two sorts of thought, that which Lauben privately enter-
tains and that which he communicates. To account for the communication of
first person thoughts despite the 'incommunicability' of the latter one needs
only to draw a distinction between two forms of communication: an
utterance can 'express' a thought (weak communication) even if that thought
is not thereby made available to the hearer (strong communication).
My primary aim in this paper is to fill the gap in Frege's account so as
to make it satisfactory. But I will start by considering the alternative account
and raising an objection to it. To meet the objection I will suggest an
improvement of the alternative account which makes it indistinguishable
from Frege' s. The problem of the relation between the two modes of presen-
tation will be solved along the way.

Even though it departs from Frege's own solution, the alternative account is
Fregean in spirit. From a Fregean point of view, it has the great merit of
allowing one to maintain the equation of thought and semantic content.
The semantic content of an utterance -- that which the utterance expresses
and which must be grasped for it to be correctly understood -- is by definition
an 'objective' property of that utterance which can be recognized by both
speaker and hearer and which remains stable in the process of com-
munication; but the required stability cannot be found at the level of
thoughts, or so it seems. The first person thought which the speaker ex-
presses by saying 'I have been wounded' differs from the hearer's thought
formed upon understanding the utterance -- they involve different modes of
presentation of Lauben. The sentence means the same thing for speaker and
hearer, and the statement that is made -- to the effect that Lauben has been
wounded -- is also the same for both, but the associated thoughts change as
communication proceeeds from speaker to hearer. This is what makes the
Russellian notion of a (singular) 'proposition' an arguably better candidate
for the status of semantic content than the Fregean notion of a thought. For
the proposition ('what is said', the 'statement' that is made) remains constant
98 FRANCOIS RECANA TI

from one person to the next, in contrast to the thought. As John Perry says,
'one reason we need singular propositions is to get at what we seek to pre-
serve when we communicate with those who are in different contexts' (Perry
1988, p. 4).
The alternative account disposes of this objection to Frege's equation of
thought and semantic content. Even though the speaker's thought is tied to
his own point of view and cannot be entertained by someone else (e.g., the
hearer), still, it is this thought which is expressed by the utterance and can be
recognized as such by the hearer. Its being publicly recognizable confers
sufficient objectivity on the speaker's thought, despite its essential subjec-
tivity, to make it a plausible candidate for the status of semantic content.
Let us analyze the theoretical move at work here. Two points of view are
involved in the communication process: that of the speaker and that of the
hearer. In Frege's example, the speaker's thought includes the mode of pre-
sentation lEgox' while the hearer's thought, formed upon understanding the
utterance, is a demonstrative, third person thought: 'He has been wounded'.
As long as the speaker's thought is seen as on the same footing as the hear-
er's, it is tied to a particular point of view and lacks the sort of objectivity
needed to equate it with the utterance's semantic content. The move consists
in privileging the speaker's thought and giving primacy to his point of view
over the hearer's. On the 'alternative account' I have sketched (following
Dummett), it is the speaker's first person thought rather than the hearer's
third person one which is objectively 'expressed' by the utterance and
recognizedl as such by all participants in the speech episode.
One possible objection to this move is that it is somehow arbitrary. How
do we choose the particular point of view to be privileged? On intuitive
grounds, it seems natural to select the point of view of the speaker, yet there
are also reasons to select the hearer's point of view. As Evans emphasized,
what matters, when we want to individuate semantic content, is what would
count as a proper understanding of an utterance (Evans 1982, p. 92, 143n,
171, etc.); but 'understanding' defines the task of the hearer. Thus it is the
hearer's point of view Evans privileges. As a result of that choice he is led to
deny that 'I' expresses the concept Egox: according to him, 'I' expresses a
demonstrative concept akin to that expressed by the demonstrative phrase
'that person'.
To overcome the difficulty, one may try a slightly different route. Instead
of privileging a particular point of view (that of the speaker or that of the
hearer), we may decide to focus on what is common to both points of view.
This more or less corresponds to the Russellian strategy. According to the
 THE COMMUNICATION OF FIRST PERSON THOUGHTS 99

Russellian, what is common to the speaker's thought that he himself has

been wounded and to the hearer's thought that that man, Lauben, has been
wounded, is the singular proposition: <Lauben, the property of having been
wounded>, that is, the state of affairs which both thoughts represent (their
common 'incremental truth-conditions', in Perry's terminology [Perry
1990]). Now this commonalist strategy can also be opted for in a Fregean
framework, for there is more that is common to both thoughts than merely
the state of affairs they represent at the level of 'incremental truth-condi-
tions'. In particular, there is more that is common to the modes of presenta-
tion under which Lauben and his addressee respectively think of Lauben than
merely the reference, i.e., what these modes of presentation are modes of
presentation of.

Most authors in the field, whether Russellian or Fregean, draw a distinction

between descriptive and nondescriptive modes of presentation. Indexicals and
proper names typically express nondescriptive modes of presentation, in
contrast to (attributively used) definite descriptions, which express descriptive
modes of presentation. 3
What are nondescriptive modes of presentation? Like a number of con-
temporary authors, I construe 'nondescriptive' modes of presentation as
dossiers of information. Thus EgoLauben is Lauben's dossier for whatever
information he gains about himself. What is 'special and primitive' about
this sort of dossier is that Lauben -- like all of us -- has a particular way of
acquiring information about himself, such that (i) only Lauben can acquire
information about Lauben in this way, and (ii) Lauben can acquire in-
formation in this way only about Lauben. An 'Ego' -dossier serves as repos-
itory for information gained in this particular way (the first person way).
Qua dossier, a nondescriptive mode of presentation contains information
about whatever the dossier concerns. This allows for the following possibil-
ity: two dossiers which differ by their global content and/or by the sort of
dossier they are may nevertheless have something in common, namely part
of their content -- some particular piece of information which they both
contain. This is what happens in Frege's example. Both the thought of the
speaker and that of the hearer include a mode of presentation which corre-
sponds to their dossier concerning Lauben. The modes of presentation in
question are quite different from each other: the speaker's is a first person
mode of presentation (i.e., it corresponds to a dossier based on the special
100 FRANCOIS RECANA TI

way of acquiring information mentioned above), while the hearer's is a third

person mode of presentation. Nor do they contain the same information:
there are things which Lauben knows about Lauben which his hearer does
not know, and the other way round. But there are also pieces of information
which both dossiers contain -- there are things which both Lauben and his
hearer know about Lauben. The latter provide identificatory facts which
Lauben and his hearer can appeal to in order to secure reference when com-
municating about Lauben. In particular, both Lauben's and his hearer's
dossier concerning Lauben include the information that Lauben is the utterer
of this token of 'I have been wounded'. That is part of Lauben's current no-
tion of himself as much as it is part of his hearer's current notion of Lauben:
Lauben is conscio.us of being the utterer, and the hearer also knows that
Lauben is the utterer, the man speaking to him at this moment. That
information is part of both dossiers, even though one is a first person dossier
and the other a third person dossier. Now that specific aspect common to
both the speaker's and the hearer's notion of the reference is, I suggest, what
is expressed by the linguistic expression 'J'. The reference of 'J' is presented
as being the utterer of this token (linguistic mode of presentation). That
linguistic mode of presentation is intersubjective, unlike the psychological
mode of presentation which is subjective (i.e., the notion of himself, on the
speaker's side, or the notion of that man, on the hearer's side); but the former
may be construed as an aspect or part of the latter, an aspect (or part) which
is common to the speaker's and the hearer's point of view.
Note that the identificatory fact which Lauben appeals to in order to
secure reference to himself in communication belongs to a special category
of identificatory facts: the category of cqmmunication-specific identificatory
facts. Those facts do not exist independently of communication but are cre-
ated in the very process of communication (Benveniste 1956). They are as-
pects of the speech situation, and as such they are automatically (and mutu-
ally) known to both speaker and hearer qua participants in that situation.
Thus both the speaker and the hearer (in a normal conversational setting)
know that the speaker -- say Lauben -- is the speaker, that the hearer -- say,
Leo Peter -- is the hearer, and so forth. This enables the speaker to use these
mutually manifest facts in referring to the speaker, the hearer and other
aspects of the speech situation. Indexicals are conventional means of doing
so: the linguistic modes of presentation conventionally expressed by indexi-
cals such as 'J' or 'you' ('the utterer', 'the addressee') correspond to facts
about their referents which are created by the speech situation itself and are
therefore mutually manifest to participants in the speech situation.
 THE COMMUNICATION OF FIRST PERSON THOUGHTS 101

Since the identificatory fact appealed to by virtue of the linguistic sense

of the indexical is mutually known to the speaker and his hearer, it belongs
to their respective dossiers concerning the reference -- the speaker's first
person dossier and the hearer's third person dossier. The linguistic sense of
the indexical can therefore stand for both dossiers through a cognitive process
of 'synecdoche' (Recanati 1993): the part stands for the whole. The linguistic
sense of '1' ('the utterer of this token') stands for the speaker's notion of
himself because it corresponds to an aspect of that notion, to some in-
formation which the speaker's Ego-dossier contains (the information that he
is the utterer of this token). Interpreting the utterance -- the hearer's task --
consists in going back from the part to the whole; but it i~ not the same
whole at both ends of the communicative process. We start with the
speaker's thought which involves his first person dossier. The latter does not
go into the semantic content of the utterance, which contains only the
linguistic sense of 'I', corresponding to an aspect or part of the original
dossier. At the other end we find the hearer's thought, formed by
'interpreting' the utterance's semantic content and somehow enriching it with
the hearer's own dossier concerning the speaker, which dossier also contains
as a component part the identificatory fact appealed to by the indexical.
On this view Frege was right: that which is communicated -- the utter-
ance's semantic content -- is not quite the speaker's original thought. It in-
volves a mode of presentation ('the utterer of this token') which is closely
related to the linguistic meaning of 'I' and differs from the special and primi-
tive mode of presentation Egox which occurs in the speaker's first person
thought that he himself has been wounded.

NOTES

1 This notation is Peacocke's. See Peacocke 1981, 1983.

2 This is demonstrated by, inter alia, the fact that the utterer of a token might not realize that
he (he himself*, as Castaneda would say) is the utterer of this token. To be sure, such a
situation would be quite extraordinary, but it is by no means impossible.
3 The modes of presentation in question, whether descriptive or nondescriptive, are all
'psychological' modes of presentation in the sense of section 1.
102 FRANCOIS RECANA TI

REFERENCES

Benveniste, E.: 1956, La nature des pronoms. Reprinted in his Problemes de linguistique
genera Ie, Gallimard, Paris (1966).
Dummett, M.: 1981, The Interpretation of Frege's Philosophy, Duckworth, London.
Evans, G.: 1982, The Varieties of Reference, l McDowell (ed.), Clarendon Press, Oxford.
Frege, G.: 1918-19, The Thought: a Logical Enquiry. English trans. by A. and M. Quinton in P.
Strawson (ed.), Philosophical Logic, Oxford University Press, Oxford (1967).
Peacocke, C.: 1981, Demonstrative Thoughts and Psychological Explanation, Synthese 49:
187-217.
Peacocke, c.: 1983, Sense and Content, Clarendon Press, Oxford.
Perry, l: 1988, Cognitive Significance and New Theories of Reference, Nous 22,1-18.
Perry, l: 1990, Individuals in Informational and Intentional Content, In E. Villanueva (ed.),
Information, Semantics and Epistemology, Basil Blackwell, Oxford.
Recanati, F.: 1990, Direct Reference, Meaning, and Thought, Nous 24,697-722.
Recanati, F.: 1993, Direct Reference: From Language to Thought, Basil Blackwell, Oxford.

CREA
1 rue Descartes
75005 Paris, France
 TOM STONEHAM

TRANSPARENCY, SENSE AND SELF-KNOWLEDGE

O. If you understand me, then you know what I am saying, that is, you
know what my words (and sentences) mean. Understanding is an epistemic
notion roughly equivalent to knowledge of meaning. The philosophy of
language is particularly interested in this kind of knowledge, concentrating
much energy on the question: what is it for someone to know what a sen-
tence means? In this paper I am going to concentrate upon one particular
aspect of that project, namely the question of whether knowledge of meaning
is epistemologically distinguished.
When we say 'Fred knows the meaning of "confabulate"', that sounds
very much like 'Fred knows his cat'. But it is consistent with the latter that
Fred should sometimes fail to recognize his cat, should sometimes take an-
other cat for his, or his cat for another's. With knowledge of meaning this is
not so obvious: if I attach the wrong meaning to a word, then surely I do not
know what it means, I do not understand it. The related claim that if I know
the meaning of two words, then I must know if they are synonymous or not,
is the claim that meaning is transparent. This is an epistemological claim
which makes knowledge of meaning privileged compared to some other
knowledge we have. Knowledge of (one's own) meaning is also a form of
self-knowledge, so the transparency claim squares well with the thought that
self-knowledge is privileged.
I want to argue that Frege's Law or Principle, often known as the
Intuitive Criterion of Difference of Sense, which relates meanings of sen-
tences to the cognitive state of one who understands those sentences, does
not entail that meaning is transparent. It does, however, provide an impor-
tant epistemological asymmetry between knowledge of language and knowl-
edge of, say, cats.

1. Frege appears to be employing his criterion in the following passage:

[T]he thought in the sentence 'The morning star is a body illuminated by the Sun'
differs from that in the sentence 'The evening star is a body illuminated by the Sun'.
Anybody who did not know that the evening star is the morning star might hold the
one thought to be true and the other to be false. (1892, p. 62)

What distinguishes the two sentences here is that someone can take conflict-
ing attitudes towards their truth. We might state this principle thus:

103
J. Biro and P. Kotatko (eds.), Frege: Sense and Reference One Hundred Years Later, 103-112.
© 1995 Kluwer Academic Publishers.
104 TOM STONEHAM

(IC) If (and perhaps only if) it is possible to fully understand two sentences and
coherently believe/desire what one expresses while not believing/desiring what the
other expresses, then those sentences have different meanings.

This is not a principle of individuation but an a priori constraint upon theo-

ries of sense. For Frege this claim is justified because he identifies the
meanings of sentences with the contents of the propositional attitudes. In
the past twenty years it has often been suggested that there is no one notion
of sense that can serve both these functions. I shall not enter into this de-
bate, assuming that Fregean senses are desirable if possible and therefore
worth investigating. IC is an essential property of a Fregean Sense. The
criterion is 'intuitive' not only in its plausibility, but also in its concern
with the ordinary linguistic intuitions of ordinary speakers. Having made
clear exactly what this link is, we can ask to what extent it gives the indi-
vidual speaker an epistemologically privileged position with respect to the
meanings of his sentences (and if we identify those with the contents of
(some of) his thoughts, with respect to the contents of his thoughts).

2. It is important now to distinguish the various interpretations of IC which

are sometimes confused. The obvious question to ask is: possible for
whom? In these days of descriptive lexicography it is natural to ask whether
everyone's opinions count, or whether a majority rules or perhaps whether
refined academicians can dictate? There must be a quantifier implicit in the
claim, and it matters which one. Compare the following:

A) If (and only if) it is possible for someone who fully understands two sen-
tences (simultaneously) to take conflicting attitudes to what they express,
then the sentences have different senses.

B) If (and only if) it is possible for anyone who fully understands two sen-
tences (simultaneously) to take conflicting attitudes to what they express,
then the sentences have different senses.

Here we see a problem generated by English which leads to a certain amount

of confusion in the area, for it is quite possible to read these two claims as
synonymous. We often use the phrase 'someone who' not as an existential
quantifier equivalent to 'at least one person who', but as expressing indiffer-
ence to which of the people who fulfill the ensuing condition ('who fully
understands .. .'), is being talked about. But this is to universally quantify
over a domain. To see this consider the normal, as opposed to the
 TRANSPARENCY, SENSE AND SELF-KNOWLEDGE 105

existential, understanding of 'Someone very tall will bang their head coming
through that door'. To avoid both the confusion and the clumsy expression
'at least one person', I shall use some symbols. Let 'xAy' be the relation,
between a person and a sentence, of taking some specific attitude CA' is
schematic) towards the truth of what is expressed by that sentence, 's l' and
's2' can be names of sentences and the squiggly brackets can be used to refer
to the sense of a sentence (so '{sl}' reads 'the sense of "s("). '0' is not
used strictly, but to express our interest in what speakers can do and the
requirement of simultaneity is ellided for simplicity. We then get:

[1] (:Ix 0 (xAsl & -xAs2) ~ -({sil = {s2}»

[2] (\Ix 0 (xAsl & - xAs2) ~ -({sil = {s2}))

These are distinct but related. The antecedent of [2] is a special instance of
the antecedent of [1], so the truth of [1] guarantees the truth of [2], but not
vice versa. We must also be careful to distinguish the biconditionals.
Consider the converse of [1] and [2]:

[3] -({sil = {s2}) ~ (:Ix 0 (xAsl & -xAs2»

[4] -({sil = {s2}) ~ (\Ix 0 (xAsl & -xAs2»

Here the entailment goes the other way, for [4] entails [3] but not vice
versa. For ease of reference, I shall call these four claims, respectively,
Someone Is Sufficient (SIS), Everyone Is Sufficient (EIS), Someone Is
Necessary (SIN) and Everyone Is Necessary (EIN). Now, the imprecision of
English and the fact that if SIS then EIS and if EIN then SIN, can easily
conspire to generate an unwitting confusion of 'Someone Is Necessary and
Sufficient' (A) with 'Everyone Is Necessary and Sufficient' (B) as interpreta-
tions of IC. These are clearly incompatible, the disagreement resting on the
possibility of

that is, on whether there could be a difference in sense even though some
speakers who understand both sentences cannot take conflicting attitudes.
The plausibility of this, and thus the preference for SIS over BIN, is
heavily dependent on how we choose to understand what a person can or
cannot do. That decision also affects the epistemological relevance of IC, for
on a too generous account of what one can think it would be possible to
106 TOM STONEHAM

remain unaware of one's capabilities. If so, they are irrelevant to epistemic

questions.

3. Many might think that the failure to distinguish SIN and EIN is not a
confusion but perfectly deliberate. The two are equivalent if we hold the fol-
lowing principle (which I shall dub egalitarianism):

(E) What one person can do, any person can do.

Of course, when we are considering the question of coherently taking con-

flicting attitudes to a pair of sentences, this is prima facie absurd, as Frege
made clear in the passage quoted earlier. Taking his example, if someone
believes that the evening star is the morning star, it is not rational or coher-
ent for them to take conflicting attitudes to the sentences 'The morning star
is a body illuminated by the Sun' and 'The evening star is a body illuminated
by the Sun'. For any two sentences which differ only in distinct but co-
referential component senses, there will be some epistemic situation in
which it is not rational to doubt the truth of one but not the other.
However, it may seem that the case of the astronomer who knows the
identity is not very interesting to us, because (a) knowing the identity is not
essential to understanding the two sentences (someone who disbelieved it
could still understand them), so it is possible rationally to take conflicting
attitudes, and (b) since the identity is a posteriori, the astronomer can cer-
tainly conceive of it being false and thus of it being rational to take the con-
flicting attitudes. This objection introduces two notions of what a person
can do. The first appeals to what is logically possible for a person to do
given only that they comprehend the sentences in question. As such it ab-
stracts from anything particular to an individual. Egalitarianism is thus true
of this notion of what an individual can do. The second appeals to the
thought that though it is not rational for someone who believes a=b to take
conflicting attitudes to 'a is F' and 'b is F', they might be able to conceive
of taking those attitudes. In other words, the possibility of taking conflict-
ing attitudes is psychologically real for that person, even if they actually do
not take them. If this notion of what it is possible to believe is distinct
from what it is logically possible to believe, that will be because there is an
attitude it is logically possible for a rational person to take which at least
one (other) person cannot conceive of taking. If such were the case,
egalitarianism would not be true for the attitudes that a person can imagine
taking to a pair of sentences.
 TRANSPARENCY, SENSE AND SELF-KNOWLEDGE 107

4. I shall assume that IC is true when we read it as referring to logical pos-

sibilities. We can now ask whether psychologically real possibilities are
distinct, whether egalitarianism is true for them and what interpretation of IC
is true for them?
There is clearly no reason to suppose that the conflicts of attitude any in-
dividual can unreflectively conceive of should delimit the logical possibili-
ties. I want to make a slightly stronger claim, namely, that the conflicts of
attitude countenanced by an individual's linguistic intuitions, i.e., his
intuitive judgements of the applicability of the words he understands, do not
delimit the logically possible rational conflicts of attitude.
In other words, there is at least one case where a coherent conflict of atti-
tudes is logically possible but at least one competent speaker, in full com-
prehension, cannot conceive of taking conflicting attitudes, even when
prompted. The failure to conceive of a situation in which it is rational to
believe that something is an F while not believing that it is a G might be
due to a lack of imagination, an inability to exploit what one knows when
one knows the meaning of '_is an F' and '_is a G' (if one knows the
meaning of '_is an F', then one knows what it is for something to be an
F). Alternatively, what one needs to know in order to be said to know the
meanings of the terms may not make it clear whether being an F is exactly
the same thing as being a G, though it may make it reasonable to believe
that it is.
I shall argue by example. The example is of a pair of predicates in very
common use, which nearly all English speakers would normally be said to
understand, namely '_is a brook' and '_is a stream'. The question is
whether they are synonymous. On a straw poll, I have found that most
people's working knowledge of English leads them to believe so, for they
can think of no case wh~re it would be correct to describe something as one
but only dubiously correct to describe it as the other. Interestingly enough, a
lot of people are cautious about this judgement. Anyway, for present pur-
poses we can accept that an imaginary, but not unlikely, person, Fred, can-
not conceive of taking conflicting attitudes.
Now, a brief check in the dictionary and a little bit of research will reveal
that the terms 'brook' and 'stream' are not even co-extensive, let alone syn-
onymous. The predicate '_is a stream' picks out, albeit under a rather
prejudicial mode of presentation, a natural kind, the same natural kind as is
picked out by '_is a river'. So the Vltava in Prague is a stream (notice that
'stream' and 'river' are not synonymous, though they both name the same
108 TOM STONEHAM

property). The predicate '_is a brook' does not pick out a natural kind at
all, but only a (socially determined) subset of the members of the kind (the
ones that babble?).
My interpretation of the situation here is that Fred incorrectly thinks that
the two terms are synonymous because he cannot practically conceive of dis-
tinguishimg them by conflicting attitudes. They are intersubstitutable ev-
erywhere, as far as he knows. It is easy to tell the story such that it would
be unreasonable to deny that Fred understands both terms, but there remain
two common objections to my interpretation of Fred's case. The first is that
any competent speaker could tell apart 'stream' and 'brook' by consideration
of such cases as the Gulf Stream. This is questionable. For a start, 'the
Gulf Stream is a stream' is not a normal tautology, but can easily be read so
as to be false. The problem arises from nominalizations of the verb 'to
stream', which can be distinguished from the predicate we are concerned with
by such phrases as 'a stagnant stream'. For those who remain doubtful, we
can switch the example to the predicates '_is a bottle' and '_is a flask'.
My dictionary actually says that these are synonymous, but we should not
take that as proof of anything. The important point is that people who
understand both words perfectly well can rationally debate whether they are
synonymous or not.
The second complainant insists that what I have shown is that in Fred's
idiolect the terms 'brook' and 'stream' are synonymous and that his idiolect
might differ from the shared language (whatever that might be). This objec-
tion is, ultimately, question-begging. If Fred's case is meant to provide a
reason to introduce idiolects, then it does so only on the assumption that
there must be some notion of meaning which is transparent. The only ob-
vious motive for such an assumption is IC and I am in the business of
showing that IC does not provide such a motive. Alternatively, the objector
might have an independent reason for introducing idiolects. But where is the
reason for thinking that idiolects, any more than shared languages, must be
transparent? Similarly, we should not think that we can here distinguish
speaker meaning from literal meaning. Ex hypothesi, Fred understands the
literal meanings of the terms; there are no independent grounds for finding a
general divergence between what he means by them and what the words
mean.
If one refuses to accept my interpretation of Fred's case, I can only offer
an argument to a stronger conclusion. There is a type of case in which it is
impossible to imagine believing what one reasonably takes someone else to
believe. It may even be impossible to imagine believing something which
 TRANSPARENCY, SENSE AND SELF-KNOWLEDGE 109

one once believed oneself. These claims follow from the thought that we are
unable to imagine what we believe to be impossible, despite our ability
unwittingly to imagine what is de facto impossible. If I believe that a=b,
then I cannot imagine believing that -(a=b), even though someone else
might believe that. For imagining that -(a=b) is imagining it to be true and
that would either involve imagining a situation of non-self-identity, which is
absurd, or imagining two objects, at least one of which would not be the one
I call, variously, 'a' or 'b'.
This argument should make us take seriously Frege's caveat about the
person who knows that the morning star is the evening star. It is not
strictly true that he can conceive of rationally taking conflicting attitudes to
the two sentences. At best he can conceive that someone might get them-
selves into the position of believing the impossible. There is no reason to
believe that egalitarianism is true for the intuitive judgements that an indi-
vidual would make based upon what he can conceive as rational.

5. So we should distinguish two readings of Ie, the logically possible and

the psychologically realistic, or intuitive, and distinguish them on the ques-
tion of whether egalitarianism is true.
What is psychologically realistic is possible, so it is logically possible;
thus the reading of Ie as logical possibility entails that SIS and SIN apply
to intuitive judgements. If someone can conceive of the conflicting atti-
tudes, then there is a difference of sense, and if there is a difference of sense
then at least one person can conceive of the conflict. It is a further question
whether SIN requires that one of the actual speakers can so conceive or only
some possible speaker. That, however, makes no difference to the argument
of this paper. If SIS is true for a sense of possibility which is not egalitar-
ian, then EIN is false, that is:

I take Fred to have provided an example of this situation.

6. Ie, when taken as referring to logically possible conflicts of attitude,

does not entail transparency, since transparency is an epistemic claim and
epistemic claims should concern themselves only with what is available to
the putative knower (e.g., Bonjour 1985). Nor does the entailment go
through when Ie is taken to involve a thicker notion of which attitudes can
be taken, though it does give us a certain type of privileged self-knowledge.
110 TOM STONEHAM

To see this we had better look a little more closely at what the transparency
claim actually is. In Dummett we find: "It is an undeniable feature of the
notion of meaning -- obscure as that notion is -- that meaning is transparent
in the sense that, if someone attaches a meaning to each of two words, he
must know whether these meanings are the same." (1978, p. 131)
This is a claim to the effect that sameness (and thus difference) of sense is
self-intimating (upon understanding, of course). The essence of the trans-
parency claim is the sufficiency of understanding the words in question when
making intuitive judgements of synonymy. There is no possibility that a
judgement of sameness of meaning, based solely on understanding the two
words/sentences, could be shown to be wrong by appeal to something not
necessarily grasped in understanding those words/sentences. When anyone
understands a sentence, he has at his disposal everything there is to know
about its meaning.
Ie, conceived of as ensuring the logical possibility/impossibility of con-
flicting attitudes to non-synonymous/synonymous sentences, will not guar-
antee this: transparency of meaning. Transparency is a claim about the epis-
temological status of the individual's actual judgements of synonymy based
upon understanding alone, but I have argued that the logically possible con-
flicts required by Ie need not be achievable for all who understand. It seems
that conceivable conflicts of attitude consequent on understanding do not dis-
criminate all senses for all speakers, so Ie does not ensure transparency.
Specifically, it does not rule out the possibility that someone might under-
stand two non-synonymous words and fail to conceive of two sentences, dif-
fering only in respect of those words, to which she can take conflicting atti-
tudes.
There is an interesting and important asymmetry here, between what Ie
guarantees with respect to judgements of difference, as opposed to sameness,
of sense. Aware only of what attitudes I can take, I can sometimes conclu-
sively establish that a conflict of attitudes is possible, and thus by SIS that
there is a difference of sense. Since egalitarianism is not true for the atti-
tudes an individual can take, being aware only of my failure to take conflict-
ing attitudes, I establish nothing about the impossibility of someone doing
so. It foHows that, as an individual speaker of a communal language, as-
suming understanding and rationality, my judgements of non-synonymy,
based upon what conflicts of attitude are possible for me, are authoritative,
but in my judgements of synonymy I must be prepared to defer to others.
(Though I may actually be pretty good, I have no guarantee.) This should
 TRANSPARENCY, SENSE AND SELF-KNOWLEDGE 111

not be surprising, for it needs but one point of difference to undermine an

identity claim, while no amount of similarity can establish it conclusively.

7. Frege's Criterion constrains a theory of sense to respect the intuitive

judgements of speakers of the language. It does not, however, constrain the
theory by granting, merely on the grounds of comprehension, infallible
knowledge of all features which might distinguish the senses of words (and
sentences).
This paper has been primarily concerned with the effect of IC on a theory
of linguistic meaning, but it is a Fregean thesis that we can equate the senses
of sentences with the contents of mental states. If what I have said is
correct, then it follows that a theory of content constrained by IC is not
thereby required to find no difference in the content of two thoughts merely
because the thinker cannot conceive of taking conflicting attitudes to them.
But it is constrained to find a difference where the thinker can so conceive.
Similarly, the theory of content is not constrained to find each identity claim
that a thinker finds indubitable to be a logical truth. His concepts may be
more fine grained than he thinks they are, for someone else may be able to
distinguish them. Egalitarianism does not apply here.

8. The nearest IC brings us to transparency is to ensure that judgements of

difference of meaning are privileged (though not immediate). David Wiggins
has recently mentioned the project of 'refining in application Frege's
intuitive criterion for sameness and difference of sense' (1992, p.90). The
first refinement must be that, in so far as the criterion has applicability on
the level of the speaker's linguistic intuitions, it is only a criterion of
difference of sense.

ACKNOWLEDGEMENTS

In thinking about these issues, I have been greatly helped by conversations

with Martin Davies, Jonathan Knowles, Greg McCulloch and Barry Smith.

REFERENCES

BonJour, L.: 1985, The Structure of Empirical Knowledge, Harvard University Press,
Cambridge, Mass.
Dummett, M.: 1978, 'Frege's Distinction between Sense and Reference', in Truth and other
Enigmas, Duckworth, London.
112 TOM STONEHAM

Frege, G.: 1892, 'On Sense and Reference', translated and reprinted in Translations from the
Philosophical Writings of Gottlob Frege, P. Geach and M. Black (eds.), Blackwell, Oxford
(1952).
Wiggins, D.: 1992, 'Meaning, Truth-conditions, Propositions', Dialectica, vol. 46, pp. 61-90.

Sub-Faculty of Philosophy,
University of Oxford,
10, Merton Street,
Oxford OX1 4JJ, UK
 CHRISTINE TAPPOLET

THE SENSE AND REFERENCE OF EVALUATIVE TERMS

What account of evaluative expressions, such as 'is beautiful', 'is generous'

or 'is good', should a Fregean adopt? Given Frege's claim that predicates can
have both a sense and a reference in addition to their extension, an interesting
range of only partially explored theoretical possibilities opens to Frege and
his followers} My intention here is to briefly present these putative
possibilities and explore one of them, namely David Wiggins' claim that
evaluative predicates refer to non-natural concepts and have a sense which is
sentiment-involving. In order to -defend this claim against objections which
aim at showing that evaluative concepts do not really exist, I shall suggest
that our awareness of evaluative concepts involves affective (or emotive)
states. I shall start with a brief account of Frege' s view of predicates.

It is beyond doubt that Frege thought that the distinction between sense and
reference applies also to the case of predicates: an expression such as 'is a
horse' has both a sense and a reference. According to Frege, the reference of a
predicate is not constituted by the objects it is true of -- that is, by its ex-
tension -- but by what he calls a concept. 2 Thus, in addition to its sense, the
predicate 'is a horse' has a reference -- the concept horse -- and an extension --
the things which fall under the concept horse.
At least one point concerning the distinction between sense and reference
at the level of predicates would need clarification. The question is whether
Frege was right to claim that a concept under which an object falls is the
same as its property. 3 For reasons of space, I shall sidestep this question, as
well as the related question of the unsaturatedness of predicates. Thus, I shall
leave open both the question whether genuine predicates are of the form 'is
an F' or of the form 'F' and the related question whether concepts are proper-
ties or not. 4 It is worth noting that the main claims I shall make are inde-
pendent of these questions.
Given Frege's claim that the sense of a singular term corresponds to the
mode of presentation of its reference, the mode of presentation being con-
tained (,enthalten' in the original text) in the sense, iUs natural to take the
sense of a predicate as corresponding to the mode of presentation of a con-
cept. 5 A consequence of this claim is that just as singular terms can be co-

113
J. Biro andP. Kotatko (eds.), Frege: Sense and Reference One Hundred Years Later, 113-127.
© 1995 Kluwer Academic Publishers.
114 CHRISTINE TAPPOLET

referential, two or more predicates can refer to the same concept in different
ways, provided that they have different senses. As Wiggins has noted, the
predicate 'is a horse' refers to the same concept as 'is an Equus caballus', but
they present their reference in different ways, making appeal to different con-
ceptions of the concept: if we were to explain the difference between the two
predicates, we should have to refer to the different bodies of information upon
which the predicates draw, the former being the commonsensical conception
of horses, the latter corresponding to their biological conception. 6
It has to be noted that it is plausible that just as Frege claimed that
proper names have what he called a tone (or a colouring), he would have
attributed a tone to predicates as well. The tone of a singular term is a
subjective element such as a mental image associated with the term. Its
subjectivity .lies in the fact that it varies from individual to individual, and
even for the same individual it can be different at different times.? I see no
reason why one should deny that predicates have tones. Thus, one could
suggest that the evaluative element of evaluative predicates lies in the tone or
tones we associate with them. More precisely, one could claim that the tone
of evaluative predicates consists in some associated affective state. 8 But what
about the subjectivity of tone? Contrary to whata Fregean adopting this line
would have to say, it does not seem correct to claim that any term can be
associated with any affective state. There might be ways of dealing with this
worry. However, I shall not pursue this question here and shall prescind from
tone in the following, concentrating instead on sense, reference and
extension.

II

Where does this leave us as to the semantics of evaluative terms? Given that
within the Fregean framework it is not possible for a term to have an exten-
sion without having a reference or a reference without having a sense, it ap-
pears that if we adopt such a framework, we have the following possibilities:

1) Evaluative terms have a sense, a reference, and an extension.

2) Evaluative terms have a sense and a reference, but no extension.
3) Evaluative terms have a sense, but no reference and no extension.
4) Evaluative terms have no sense, no reference, and no extension.

The interesting fact about these four possibilities9 is that if they are genuine,
they would show that the common distinction between realism and irrealism
 THE SENSE AND REFERENCE OF EV ALUA TIVE TERMS 115

about values -- or evaluative concepts, in Frege's terminology -- is too

simple. Realism about values is usually thought to be the.claim that there
are genuine evaluative properties -- that values are real in the sense that they
are part of the world and cannot be reduced to non-evaluative properties. So,
if we suppose that Fregean concepts are properties, realism about evaluative
concepts would be the view that evaluative concepts are real. 10 As irrealism
about evaluative concepts is just the negation of realism, it follows that
irrealism about values amounts to the claim that evaluative concepts are not
real. On the first two possibilities, it would seem that provided evaluative
statements are construed non-reductively, there are real evaluative concepts.
In other terms, evaluative concepts could be both real and empty. If the
second possibility counts as a form of realism, it follows that it is wrong to
think, as many do, that realism about values entails that at least some of the
propositions expressed by evaluative sentences are true. I I For on this
possibility, evaluative terms could have a proper reference, even though no
evaluative sentences were true, given that nothing would fall under evaluative
concepts. I2 However, this is less than clear: a realist might want to require
that evaluative terms have an extension.13 If so, the second option would
not count as a form of realism; only the first would.
If one opts for one of the two last possibilities, it is natural to deny that
there are real evaluative concepts. For if one were to maintain that there were
any, one would be committed to the claim that we cannot talk about them,
given that the terms best suited to do the job would not refer to them. It has
to be noted that if the third possibility is a genuine one, it would be wrong
to claim that an irrealist is bound to deny that evaluative sentences have a
sense. I4 In other words, it appears that it is too simple to oppose (as would
someone who has an undifferentiated notion of meaning) realism plus the
view that evaluative sentences have a meaning on the one hand and irrealism
plus the view that evaluative sentences have no meaning on the other hand. I5
Instead of fleshing out all four putative possibilities and discussing their
vices and virtues, I shall focus on one particular suggestion.

ill

The suggestion I have in mind has been put forward by David Wiggins.
According to Wiggins, evaluative predicates have both a sense and a refer-
ence. If we take him, as seems likely, to be committed to the claim that
evaluative terms have an extension, Wiggins' account corresponds to the first
116 CHRISTINE TAPPOLET

possibility. Wiggins' conception of evaluative predicates can be summarized

by the following two claims: a) the sense of evaluative predicates is
'sentiment-involving'16 and b) the reference of evaluative predicates is a non-
natural concept, that is, a concept which is not one which pulls or will pull
its weight in the natural sciences or that can be reduced to concepts which
doP
Concerning the former claim, Wiggins writes: "Roughly speaking, value
terms have their sense by being annexed to that which in the object calls for
certain shareable responses of feeling and action, the responses that it makes
appropriate."18 In other words, evaluative predicates have the sense they
have because the concepts they refer to are concepts which include in their
extensions things that, in virtue of being the way they are, make certain
affective states appropriate. The sense of 'is beautiful', for instance, is
determined by the fact that the things which fall under the concept beautiful
are such as to make an affective state like admiration or attachment
appropriate.
A natural way to understand the notion of sentiment-involvement is to
see it manifested in the fact that to elucidate evaluative terms, we have to
refer to affective states. 19 On this view, a consequence of the sentiment-
involvement of the predicate 'is beautiful', for instance, would be that the
sentence 'x is beautiful' is best elucidated in terms of some affective state
made appropriate by x, such as admiration or attachment. 20 Indeed, on a
minimal account of the notion of sense, one could even say that the
sentiment-involvement in question just consists in the fact that evaluative
terms have to be elucidated along these lines. (I shall leave open the
exegetical question whether or not this is what David Wiggins had in mind.)
As many have noted, there are two importantly different ways to elucidate
evaluative predicates. One can elucidate them with respect to our actual ap-
propriate affective responses or in terms of any possible affective response.
In possible-worlds jargon, the first option can be formulated as follows:
something has a certain value at a world if and only if in the actual world it
makes appropriate a certain affective state. This option has been dubbed the
'rigid reading' of the biconditional because it keeps the affective responses
the same at all possible worlds. By contrast, the second option is called the
'non-rigid' reading. Here is how it would go: something has a certain value
at a world if and only if at that world it makes appropriate a certain affective
state. 21 On the first option, an inhabitant of a possible world who had dif-
ferent affective responses from ours could be wrong about evaluative judge-
ments (and this even if the external conditions in which he had his affective
 THE SENSE AND REFERENCE OF EV ALUA TIVE TERMS 117

response were suitable). Thereby the scope of errors about evaluative ques-
tion is extended. This is why many realists favour this option.
What is particularly appealing about the suggestion under consideration is
that it is in a position to fulfill two desiderata which are not easily recon-
ciled: it can make sense of the relation between evaluative statements and af-
fective and/or motivational states while it is perfectly able to fit the grammar
of evaluative terms. Emotivism, for instance -- the view that evaluative
statements are essentially expressions of affective states -- easily fits the first
desideratum but flounders on the second one. 22 And, as has been noted by
many, unguarded versions of realism have problems making room for the
first one. 23
As is generally recognised, evaluative statements have a privileged link to
affective and/or motivational states: evaluative statements more often than
not arise out of such states, and they more often than not arouse affective re-
sponses. Furthermore, evaluative statements are related to action in the sense
that the judgements or beliefs on which they are based are at least generally
motivating. 24 These are facts which a semantic account of evaluative terms
needs at least to be compatible with.
Now, if the sense of evaluative predicates is sentiment-involving, we can
see why evaluative statements often arise out of affective states and also
arouse such states. If I approve of some action, for instance, and also think
that my approval is appropriate, I will be inclined to say that the action is
good, given that something is good just in case it makes approval appropri-
ate and that I, knowing the sense of the terms I use, know that. And a likely
effect of my utterance consists in the fact that my audience reacts to what I
say. For having its attention drawn to the evaluated action, the audience is
likely to manifest its attitude with respect to the action, either displaying the
same attitude as mine towards it or expressing a different one.
It has to be noted that on the account of sentiment-involvement under
consideration, it does not follow that each time we utter an evaluative term,
we need to be in the corresponding affective state. If this were the case, the
account under consideration would be in trouble, for at least on some occa-
sions we use evaluative terms without being in any particular affective state.
But the fact that the sense of evaluative prydicates is sentiment-involving --
that on the minimal interpretation suggested above, evaluative predicates
have to be elucidated in terms of certain affective states -- does not have as a
consequence that each time we utter a sentence including an evaluative predi-
cate, we need to have any sentiment, even though we often do.
118 CHRISTINE T APPOLET

Similarly, the relation between the serious utterance of evaluative

sentences and motivation is also readily explained, given the affinity between
affective states and motivation. There are two main ways to see the relation
between affective states and motivation: either we suppose that affective
states can be motivating as such -- to fear a dog is, inter alia,to be motivated
to run away, for instance -- or affective states have necessary connections
with motivational states, such as desires. However that may be, there is little
question as to the fact that affective states are, directly or indirectly,
motivating., But then, someone who believes that something is good will
also believe that it makes approval appropriate, given his understanding of
evaluative language, that is, given his knowledge that what is good makes
approval appropriate. And to believe that something makes approval
appropriate is to believe that it makes appropriate a state which motivates to
act in certain ways. But this amounts to believing that it is appropriate to be
motivated to act in certain ways. And someone who believes such a thing is
himself likely to be motivated to act in certain ways. Indeed, he will act
accordingly, unless he is weak-willed. So much for the fIrst desideratum.
As I said above, the attractive fact about the suggestion under
consideration is that it meets the first desideratum while perfectly fItting in
with the grammar of evaluative terms, that is, meeting the second
desideratum. As many have stressed, evaluative terms linguistically behave
in much the same way as non-evaluative terms do, so that it would be odd to
say that the former merely express affective states, whereas the latter are
genuinely assertoric. Take 'odious' and compare it with an uncontroversially
non-evaluative term,such as 'square' (in its literal use). On the face of it,
'odious' appears to be a bonafide predicate. Just as 'square', 'odious' can be
used to construct well-formed sentences such as 'This is odious'. Such
sentences appear to express propositions which allow for appraisal in terms
of truth and falsehood. They can figure in that-clauses used to ascribe
propositional attitudes such as belief or knowledge. And they can appear in
the scope of the full range of logical operators, such as when they are negated
or when they fIgure in the antecedents of conditionals. 25
This is, of course, just what is to be expected on the account under
consideration: apart from the difference at the level of sense and from the fact
that the concepts to which they refer are non-natural, evaluative predicates are
taken to be much the same as non-evaluative predicates.

IV

Concerning more specifically the second part of the suggestion under con-
 THE SENSE AND REFERENCE OF EV ALVA TIVE TERMS 119

sideration, it has to be asked whether or not evaluative concepts are real and
whether or not they really are non-natural. As we have seen, two predicates
could have the same reference but differ as to their senses. So, it is possible
that an evaluative predicate refers to the concept referred to by a non-evalua-
tive predicate. However, it is more plausible that they should be non-natural.
The reason is primarily the difficulty of finding adequate identifications of
evaluative concepts with natural concepts. What would be the physical
feature or set of features common to all beautiful things that would make
them fall under some natural concept? And what would be common to all
good things apart from being good? We have, as yet, no answer to such
questions, and there seems to be little hope of finding any.
But do non-natural concepts really exist? Here is how an argument in
favour of an affirmative answer to this question might go. A distinctive
feature of the Fregean framework as formulated above is that truth and
reference go hand in hand: the singular term and the predicate of a truth-
assessable sentence necessarily have a reference. This is a consequence of the
two Fregean claims according to which sentences refer to truth-values and the
reference of a sentence is determined by the reference of its parts. 26 Thus, the
fact that one can construct truth-assessable sentences with an evaluative
predicate would be sufficient to show that the evaluative predicate has a
reference.
The same conclusion can be arrived at in a different way. Suppose the
sentence 'x is F' has a truth-value. Then, from 'x is F', it can be inferred
either that there is something which x is or that there is something which x
is not, depending on whether 'x is F' is true or false. So in both cases, we
quantify over F. Thus here, too, to prove that there are evaluative concepts,
one would have to show only that evaluative sentences have truth-values,
provided that such sentences cannot be construed reductively. But as we have
seen above, the propositions expressed by evaluative sentences appear to
admit of assessment in terms of truth. And the difficulties of finding satis-
factory indentifications between evaluative and non-evaluative concepts sug-
gest that the prospects of uncovering reductions are meager. One has to con-
clude, it appears, that there are real evaluative concepts. 27
This will seem too easy to many.28 That at least many evaluative
sentences have truth-values is uncontroversial. But that it follows from this
that the concept to which the predicates refer exists is far from clear, and this
is so even if the sentence cannot be construed non-reductively. It might well
be necessary for a term to have a real reference that it can figure in sentences
120 CHRISTINE TAPPOLET

which are truth-assessable, but this is surely not sufficient -- or so it will be

claimed. With Crispin Wright, we might want to distinguish between a thick
and a thin notion of truth, that is, roughly, between truth which is based on
independent truth-conferring states of affairs -- correspondence truth -- and
truth which is not, that is, merely disquotational truth. 29
Can one make room for such a distinction within Frege's framework? It
might be suggested that expanding Wright's distinction between thick and
thin truth, we distinguish between twO kinds of reference which we might
call thick and thin reference; predicates which can be used only in sentences
which have thin truth-values would have a thin reference, whereas a predicate
which can figure in sentences admitting thick truth would have a thick refer-
ence. In Philip Pettit's terminology, we might say that whatever is referred
to in discourse is posited. 3D However, not all the things which are posited are
necessarily part of reality. So, the question is whether evaluative statements
call for thin or thick truth and, correspondingly, whether the reference of
evaluative predicates is thin or thick.
What is needed for thick truth? Following Wright, we can make two re-
quirements. 31 The first is that for being thick, truth needs to command assent
by any individual who is fully endowed with cognitive powers. A con-
sequence of this claim is that disagreement about a statement that admits of
thick truth has to be explained either in terms of a difference of information
concerning the facts or in terms of a difference in the assessment of the facts.
The second requirement is that for being thickly true, a statement needs to be
about something to which we have access; in particular, for evaluative
statements to admit of thick truth, there has to be a satisfactory account of
how it is possible for us to be aware of evaluative concepts. We shall see
that the first requirement can be satisfied only if the second is, so that the
issue turns on whether or not there is an epistemology of values. Let me
start with the first requirement.
The crucial question is whether disagreement about an evaluative issue
can be rationally irresoluble in the sense that it persists even if the two
disagreeing parties have access to the same facts and assess them similarly.
Given the widespread character of disagreement about questions of values, it
might seem easy to conceive of a case of a dispute which is not due to a
misapprehension of facts. But as Wright notes, all depends on what one
counts as a fact. 32 It will always be open to a realist to claim that
disagreement is based on a misapprehension of evaluative facts, for to
exclude such facts from the outset is question-begging. So, the realist might
claim that if disagreement persists, it must be because at least one of the two
 THE SENSE AND REFERENCE OF EY ALUA TIYE TERMS 121

disputants misapprehend the evaluative fact, being, so to speak, evaluatively

blind.
Well, what is it to be blind as to evaluative matters? This leads us to the
second requirement: a realist who claims that persistent disagreement about
values is due to a misapprehension of evaluative facts owes us an account of
how we can come to be aware of evaluative facts and of how we could fail to
recognise such facts.

The natural answer for a realist who has gone so far is to claim that in anal-
ogy with colours and colour perception, we are aware of evaluative facts in
virtue of our affective states. The idea is that just as colour sensations nor-
mally allow us to be aware of colours, certain affective states -- that is, sen-
timents (or emotions) -- normally allow us to be aware of the fact that some-
thing has a value; they normally present us with evaluative facts or, to put it
differently, they represent evaluative facts. 33 The qualification 'normally' is
needed to take into account cases where affective states misfire, failing to
presents things as they are. The affective states which are plausible candidates
for such a presentational or representional function are intentional, occurrent
states which are essentially felt. Examples of such states are instances of
(conscious) fear, anger, love, hate, delight or amusement. Thus, the
suggestion would be that my states of fear normally present me with the fact
that what I am afraid of is dangerous.
An important point in favour of such a suggestion is that it appears to
correspond to the phenomenology of values. As McDowell notes with
respect to our awareness of aesthetic values, 'Aesthetic experience typically
presents itself, at least in part, as a confrontation with values: an awareness
of values as something residing in an object and available to be
encountered.' 34 Much the same, I think, is true of moral experience.
A second point which speaks for the claim that we are aware of evaluative
facts in virtue of our affective states is underlined by the fact that many of us
-- philosophers and laymen -- have been, and are still, inclined to say that
there is something like moral or aesthetic vision and blindness, that is, that
we might see or fail to see values. 35 If sentiments present us with evaluative
facts, it is legitimate to say that in a sense, we feel values. And from this, it
is only a short step to saying that we see values.
A third point in favour of the suggestion that our sentiments normally
present us with values is that we assess sentiments with respect to their
122 CHRISTINE TAPPOLET

appropriateness and inappropriateness. Thus, we think that it is appropriate

to fear danger, for instance, while it would be inappropriate to fear innocuous
things. In addition to being appropriate with respect to the value they pre-
sent, sentiments can also be appropriate with respect to the degree of the
value. Just as values admit of degrees, sentiments have degrees. Thus, an in-
tense fear is appropriate with respect to a serious danger, while it would be
inappropriate with respect to an insignificant danger. But what makes fear
appropriate or inappropriate? The simplest answer, I think, is that it is be-
cause fear normally presents us with danger that fear is appropriate with re-
spect to dangerous things; in other words, it is appropriate to fear danger be-
cause we have to be aware of how things are. And it is appropriate to have an
intense fear with respect to a serious danger because the function of such a
fear is to present a danger of that kind. More generally, one could say that if
x is V, then e (with intensity i) is appropriate with respect to x's being V (to
degree d) just if e is a token of t the instances of which normally present x' s
being V (to degree d) (where 'x' ranges over particulars, 'V' is an evaluative
predicate, 'e' an emotion token, 'i' its intensity, 'd' the degree of the value an
't' an emotion type).36 On this suggestion, emotions are assessed in terms of
their correspondence to what they present.
There might seem to be a difficulty here. How can we explain what is
meant by the requirement of normality without relying on the concept of
appropriateness itself?37 It is interesting to note that a similar difficulty
besets a more Humean suggestion, that is, the suggestion that an emotion is
appropriate if and only if it is felt by a good judge. 38 There appears to be no
way to explain what a good judge is without making appeal to the
explanandum: what makes someone a good judge if not getting things right,
i.e., having appropriate emotions?
Yet, should we expect a non-circular explanation of appropriateness? I do
not think so. We should keep in mind the fact that 'appropriate' appears to
be an evaluative predicate. This should make us suspicious of the demand of
non-circularity: we should not expect a reduction of appropriateness to non-
evaluative concepts. In other terms, it would be misguided to look for an
explanation of the requirement of normality that would dispose of its evalua-
tive character. But then, how can we discriminate between appropriate and
inappropriate emotions? My suggestion is that we should distinguish be-
tween two endeavours: it is one thing to give a philosophical explanation of
appropriateness and another to have the means to discriminate between ap-
propriate and inappropriate emotions. We have seen in the preceding para-
graph what could be said as to the former task. As far as I can see, the latter
 THE SENSE AND REFERENCE OF EVALUATIVE TERMS 123

task is not that of philosophy, but that of the practictioners of evaluative

language. It is their business to search for the conditions which allow a good
grasp of values and those that impair our awareness of values. The effect of
certain drugs or of mental disease such as depression might have to count as
distorting, for instance, but whether they do or not is not a matter to be
settled by philosophers.
A closely related point is made by Philip Pettit. He claims that the
normal conditions invoked in the elucidation of colour terms have to be
defined with respect to a discounting practice: normal conditions are those
which have not been discounted in the relevant practice of those who use the
terms. 39 The philosopher can do nothing more than register the relation
between a concept of that kind and a practice.
It has to be noted that if it is the function of sentiments to present us
with evaluative facts, we have a different way of arguing in favour of the
claim that evaluative terms have a sentiment-involving sense. As we have
seen, Frege emphasizes that the sense of a singular term corresponds to the
mode of presentation of its reference. Now if evaluative facts are presented by
affective states, and if a fact is the falling of an object under a concept, one
can say that evaluative concepts are presented by affective states. If so, the
mode of presentation of evaluative concepts involves affective states or sen-
timents. But then it just follows that terms which refer to evaluative con-
cepts have a sentiment-involving sense. So, the view that sentiments nor-
mally present us with evaluative facts entails that the sense of evaluative
predicates is sentiment-involving. The converse, however, does not hold: one
might well, I think, embrace the view that the sense of evaluative predicates
is sentiment-involving without believing that sentiments present us with
evaluative facts.
The claim that our awareness of evaluative facts is based on affective
states is far from unproblematic. Here are some of the questions which ought
to be answered. Is there a proper causal story relating values to our
sentiments? Does the fact that there are wide disparities amongst our affective
responses not indicate that it would be wrong to say that such responses
present us with anything? Does it make for a problem that apparently
contrary to the case of perception, we do not justify our evaluative
statements with respect to our sentiments? And would our account not have
as a consequence the implausible thesis that any sentiment could present any
value? The picture I have sketched so far would have to be completed and
defended against objections. Instead of doing so, I shall close with a
comment on extension.
124 CHRISTINE TAPPOLET

VI

Pending a more detailed picture of how affective states represent evaluative

facts, the most that can be affirmed at this stage is that there seems to be a
way to argue that the truth of evaluative statements is thick and corresponds
to thick references. In other words, if the suggestion that we are aware of
evaluative facts in virtue of our affective states proves correct, the two re-
quirements can be satisfied. But if there are evaluative concepts, what about
their extension? The question is whether or not evaluative concepts are
empty. If out of the four possibilities enumerated above only the first is
equivalent to realism, this is in fact the real question of realism. For if this
is so, it seems that one could accept our story about the sense and reference
of evaluative terms while rejecting realism.
As far as I can see, we have reasons to think that most evaluative
concepts have an extension. When discussing the beauty of some sonata, for
instance, we take ourselves to be making claims which are true or false.
Indeed, we are often convinced that our evaluative statements are true and
correspondingly believe that we really disagree with those who have a
different opinion. But if nothing fell under the concept of beauty, these
intuitions would have to be explained away as illusory by some error-theory
a La Mackie. 40 The problem with that kind of account is that it is difficult to
see why we should ever choose to fall into error, once we have discovered
it. 41 It might be that the error is unavoidable, that even a clear-minded
philosopher would be bound to commit it. Still, we should be suspicious of
such a theory and should only embrace it we have to. And, so far, nothing
forces us to do so.42
By way of conclusion, let us look back to the four putative possibilities
we started with. If what I have said is on the right lines, the first possibility
is the one which is realised: evaluative predicates have a sense, a reference
and an extension. So, not only does the Fregean framework open our eyes to
interesting possibilities, but it actually allows us to formulate what I take to
be a very attractive account of evaluative predicates.

ACKNOWLEDGEMENTS

Many thanks to John Biro, Hugh Mellor, Kevin Mulligan, Philip Pettit,
Tom Stoneham and David Wiggins for helpful comments, and special thanks
to Dennis Hall for correcting the English.
 THE SENSE AND REFERENCE OF EVALUA TIVE TERMS 125

NOTES

Among the few explorers are David Wiggins (1992), esp. p. 642, which is a longer version
of (1991) and from a non-Fregean perspective, Darwall, Gibbard and Railton (1992).
2 Cf. for instance (1969), pp. 205-6. See Dummett (1981) for the textual refutation of the
claim that Frege did not want to extend the distinction to the case of predicates.
3 For this claim, see Frege (1892).
4 For a discussion of these points, see Dummett (1981), pp. 211 and following, and Wiggins
(1984), (1993), and (1995 [this volume]).
5 Cf. Frege (1892). Thus, it would be wrong to think that Frege viewed the relation between
sense and mode of presentation as that of identity.
6 See Wiggins (1984), p. 313, as well as his contribution to this volume.
7 See Frege (1892), pp. 43-4.
8 For a closely related suggestion, see Blackburn (1992).
9 It has to be noted that, as has been suggested to me by Philip Pettit, one can at least partly
transpose these four claims into a non-Fregean terminology. Thus, one could try and equate
(Fregean) sense with meaning, (Fregean) reference with possible extension and (Fregean)
extension with actual extension. The obtained pattern would come quite close to the Fregean
one, were it not be for the third possibility. Indeed, it would be difficult to make sense out of
the third possibility: how could a predicate have a meaning without having at least a possible
extension? This result would not be unwelcomed by neo-Fregeans. See fn. 13.
10 Note that if concepts are different from properties, and if we stick to the thesis that realism
is a claim about properties, it is not clear which possibilities count as realist. Both senses and
extensions seem ill-suited to play the role of properties.
11 Pace Mackie (1977), Schiffer (1990), p. 602 and Smith (1986), p. 289.
12 This claim needs to be qualified: negative existential such as 'there is nothing beautiful'
would be true.
13 lowe this point to John Biro.
14 A neo-Fregean would deny this, given that he takes sense to depend on reference. See
Evans (1982). For predicates, see Wiggins (1984), p. 313, esp. note 9.
15 This is close to what Schiffer argues. Cf. Schiffer (1990).
16 Wiggins (1992), p. 647.
17 Ibid., p. 644 and p. 646. This definition goes back to G.E. Moore, (1966).
18 Wiggins (1992), p. 642.
19 Johnston's notion of response-dependence is defined along these lines. See Johnston
(1989); See also Pettit (1991).
20 See Wiggins (1987), p. 187; McDowell (1985), esp. p. 118.
21 Cf. Davies and Humberstone (1980) and Wiggins (1987), p. 206.
22 See Wiggins (1992), p. 638.
23 See Hume (1978), III, i, 1.
24 See Wiggins (1992), p. 638.
25 For this last point, see Geach (1960), (1965). For the other points, see Blackburn (1984)
and Wright (1988), esp. pp. 30-1.
26 See Frege (1892), p. 48.
27 It has to be noted that Frege himself would have balked at this: evaluative terms are perfect
examples of predicates which are not defined for every argument; 'is generous', for instance,
is not defined for inanimate objects, so that one can form sentences with it which have no
truth-value. However, given that one can also form sente~ces which have truth-values, it
seems that Frege is wrong to claim that such predicates have no reference. As Dummett notes,
the most that we can conclude from such cases is that the concepts referred to are not fully
determinate. See Dummett (1981), pp. 169-70.
28 See Wright (1988), (1989) and Pettit (1991) for such a line of thought.
126 CHRISTINE TAPPOLET

29 Wright (1988), pp. 35-6. For disquotational truth, see Horwich (1990).
30 Pettit (1991), esp. pp. 588-9.
31 For these two requirements, see Wright (1988), pp. 37-42. They have their origin in
Mackie's argument from disagreement and the epistemological part of the argument from
queerness (see [1977]). Wright himself adds a third requirement: thick truth need to be about
something which is elucidated in a detective biconditional, that is, a biconditional read from
right to left. See Wright (1988), p. 44-6. I think that consideration of the second requirement
sett les this issue.
32 See Wright (1988), p. 40.
33 This suggestion goes back to Scheler (1954) and to Meinong (1917). For more recent
versions, see McDowell (1985), p. 121; de Sousa (1987). See also Hume (1965).
34 McDowell (1983), p. 1.
35 See for instance McNaughton (1987), p. 205.
36 See Meinong (1917), pp. 130-1 for a closely related suggestion.
37 I am thankful to Hugh Mellor for drawing my attention to this problem.
38 See Hume (1965).
39 See Pettit (1991), p. 603.
40 See Mackie (1977), p. 35.
41 See Blackburn (1985), p. 2.
42 Schiffer's argument from the possibility of irreducible disagreement to the claim that
evaluative properties cannot be instantiated suffers from the fact that he does not consider that
the disagreement might be based on a misapprehension of evaluative facts. See Schiffer
(1990), pp. 608-11.

REFERENCES

Blackburn, S.: 1984, Spreading the Word, Clarendon Press, Oxford.

Blackburn, S.: 1985, 'Errors and the Phenomenology of Value', in T. Honderich (ed.),
Morality and Objectivity, Routledge and Kegan Paul, London, pp. 1-22.
Blackburn, S.: 1992, 'Morality and Thick Concepts', Proceedings of the Aristotelian Society,
supp!. vol. 66, pp. 285-99.
Darwall, S., A. Gibbard, and P. Railton: 1992, 'Towards Fin de siecle Ethics: Some Trends',
The Philosophical Review, vol. 101, no. 1, pp. 115-89.
Davies, M. and L. Humberstone: 1980,'Two Notions of Necessity', Philosophical Studies, vol.
38, pp. 1-30.
Dumrnett, M.: 1981, Frege, Duckworth, London.
Evans, G.: 1982, The Varieties of Reference, Oxford University Press, Oxford.
Frege, G.: 1892, 'Uber Begriff und Gegenstand', Vjschr. fiir wissenschaftliche Philosophie,
vol. 16, pp. 192-205. Reprinted in G. Frege, Funktion, Begriff, Bedeutung, Vandenhoeck
and Ruprecht, G5ttingen (1980) (References are to the re-edition.)
Frege, G.: 1892, 'Uber Sinn und Bedeutung', Zeitschrijt for Philosophie und philosophische
Kritik, vol. 100, pp. 25-50. Reprinted in G. Frege, Funktion, Begriff, Bedeutung,
Vandenhoeck and Ruprecht, G5ttingen (1980). (References are to the re-edition.)
Frege, G.: 1962, 'Ausfiihrungen tiber Sinn und Bedeutung' in G. Frege, Nachgelassen
Schriften, vol. I, Meiner, Hamburg, pp. 128-36.
Geach, P.: 1960, 'Ascriptivism', The Philosophical Review, vol. 69, pp. 221-5.
Geach, P.: 1965, 'Assertion', Philosophical Review, vol. 74, pp. 449-65.
Horwich, Paul: 1990, Truth, Blackwell, Oxford.
Hume, D.: 1965, 'Of the Standard of Taste', in lW. Lenz (ed.), Of the Standard of Taste and
other Essays, Indianapolis.
 THE SENSE AND REFERENCE OF EV ALUA TIVE TERMS 127

Hume, D.: 1978, A Treatise of Human Nature. Clarendon Press, Oxford.

Johnston, Mark: 1989, 'Dispositional Theories of Values', Proceedings of the Aristotelian
Society, suppl. vol. 63, pp. 139-74.
McDowell, J.: 1983, 'Aesthetic Value, Objectivity, and the Fabric of the World', in E. Shaper
(ed.), Pleasure, Preference and Value, Cambridge University Press, Cambridge, pp. 1-16.
McDowell, J.: 1985, 'Values and Secondary Qualities' in T. Honderich (ed.), Morality and
Objectivity, Routledge and Kegan Paul, London, pp. 110-29.
McNaughton, D.: 1987, Moral Vision, Blackwell, Oxford.
Mackie, J.L.: 1977, Ethics: Inventing Right and Wrong, Penguin, Harmondsworth.
Meinong, A.: 1917, 'Uber Emotionale Priisentation', Kaiserliche Akademie der Wissenschaft
in Wien, vol. 183, 2nd part.
Moore, G.E.: 1966, Principia Ethica, Cambridge University Press, Cambridge.
Pettit, Philip: 1991, 'Realism and Response-Dependence', Mind, vol. 100, pp. 585-626.
Scheler, Max: 1954, Der Formalismus in der Ethik und die Materielle Wertethick, Francke
Verlag, Bern.
Schiffer, S.: 1990, 'Meaning and Value', The Journal of Philosophy, vol. 87, no. 11, pp. 602-
18.
Smith, M.: 1986, 'Should we Believe in Emotivism', in G. McDonald and C. Wright, Fact,
Science and Morality, Blackwell, Oxford, pp. 289-310.
Sousa, Ronald de: 1987, The Rationality of Emotions, MIT Press, Cambridge, Mass ..
Wiggins, D.: 1984, 'The Sense and Reference of Predicates: A Running Repair to Frege's
Doctrine and a Plea for the Copula', The Philosophical Review, vol. 34, no. 136.
Wiggins, D.: 1987, 'A Sensible Subjectivism', in D. Wiggins, Needs, Values and Truth,
Blackwell, Oxford.
Wiggins, D.: 1991, 'Ayer's Ethical Theory: Emotivism or Subjectivism?', in A. Philipps
Griffiths (ed.), A.J. Ayer: Memorial Essays, Cambridge University Press, Cambridge, pp.
181-96.
Wiggins, D.: 1992, 'Ayer on Morality and Feeling: From Subjectivism to Emotivism and Back',
in L.E. Hahn (ed.), The Philosophy of A.f. Ayer, Open Court, La Salle, III., pp. 633-660.
Wiggins, D.: 'Putnam's Doctrine of Natural Kind Words and Frege's Doctrines of Sense,
Reference and Extension: Can They Cohere?', this volume.
Wiggins, D.: 1993, 'Sinn, Bedeutung, et les mots d'especes', Revue de Philosophie et de
Theologie, vol. 125, pp. 225-237.
Wright, Crispin: 1988, 'Realism, Antirealism, Irrealism, Quasi-Realism', Midwest Studies in
Philosophy, vol. XII, pp.25-49.
Wright, Crispin: 1989, 'Miscontrual Made Manifest: A Response to Simon Blackburn',
Midwest Studies in Philosophy, vol. XIV, pp. 48-67.

Seminaire de philosophie, Universite de Neuchatel

Espace Louis-Agassiz 1
CH - 2001 Neuchatell
 PETER SIMONS

THE NEXT BEST THING TO SENSE IN

BEGRIFFSSCHRIFT

1. THE IMPORTANCE OF BEING IDENTICAL

"On Sense and Reference" begins with a famous question about identity, or
"equality", Gleichheit, as Frege calls it. Is identity a relation? Is it a relation
between objects or signs for objects? If it is a relation between objects, how
can identity statements be infonnatively true or false rather than trivially and
evidently true or false? We know what comes next. Frege goes on to
distinguish sense from reference, and to apply the theory in outline to a range
of questions in the philosophy of language. We know that in "On Sense and
Reference" itself, he is concerned with the sense/reference distinction as
applied to names of objects only, clauses or names of truth-values being a
special case. In the posthumous Ausftihrungen tiber Sinn Und Bedeutung
Qand in his correspondence with Husserl about the matter, he expands in
greater detail on the sense and reference of concept words. Why, though, did
he begin with identity? Knowing that Frege rarely did something without a
good reason, we must suppose that identity is either very important
theoretically for him, or that it was didactically a good place to start in
introducing sense and reference. I think that in fact both considerations
apply. Didactically, identity is an excellent place to begin considering the
sense/reference distinction, because it is so obvious that some identities are
self-evident and others are not. Identity clauses, apart from what to us now
seems like their obvious importance; were important to Frege for many
reasons, not least because of their crucial role in his logicism. Much of the
positive discussion of the counting numbers presented in Grundlagen is
concerned with numerical identities. Frege is the first to demand criteria of
identity for objects if these are to betaken ontologically seriously. In §62 of
Grundlagen, Frege says:

How is a number supposed to be given to us if we can have no idea or intuition of it?

Only in the context of a sentence do words signify anything. So it will be a question
of explaining the sense of a sentence in which a number word occurs. That leaves at
first a good deal arbitrary. But we have already established that independent objects
are to be understood under number words. We are thus given a kind of sentence
which has to have a sense, the sentences which express recognition
(Wiedererkennung). If the sign a is to designate an object for us, we must have a
criterion (Kennzeichen) which decides in all cases whether b is the same as a, even
if it is not always within our power to apply this criterion. l

129

J. Biro and P. Kotatko (eds.), Frege: Sense and Reference One Hundred Years Later, 129-140.
© 1995 Kluwer Academic Publishers.
130 PETER SIMONS

He soon thereafter rejects contextual or abstractive definition because it

cannot answer all identity questions as thus required, e.g., whether 'the
number n = Julius Caesar' is true for some number n. This is now known,
following this example, as the Julius Caesar Problem. 2 Frege's way of
defining the numbers to get around this problem is to identify them with the
extensions of certain concepts. Many of the deepest problems concerning the
introduction of courses of values in Grundgesetze are connected with identity:
Basic Law V notoriously fails to determine the identity of courses of values,
and Frege's attempt to get around the problem (which is just the Julius
Caesar problem again) by conventional stipulations cannot work because of
the antinomy in his system. Identity is a primitive notion in both of Frege's
logical systems, and a brief look at the formulas of both Begriffsschrift and
Grundgesetze shows that both systems are swarming with identities.

2. IDENTITY IN BEGRIFFSSCHRIFT

The importance of identity in general and of informative identities in par-

ticular was already well known to Frege in Begriffsschrift. §8 of that work
is called "Equality of Content". By 'content' or, in full, 'conceptual content',
Frege means that which affects the consequences that can be validly drawn
from judgements. This is why he does not assign any logical importance to
the first or subject position in a sentence, since the same content can be
expressed with different expressions as subject. Among contents, we recall
that some, like that unlike magnetic poles attract, can be judged, whereas
others, like house cannot.
Frege said later, in commenting on the distinction between sense and
reference, that it was a distinction made within what he previously called
conceptual content. So in his pre-1891 writings, we find Sinn and Bedeutung
used more or less interchangeably. Sometimes one, sometimes the other
seems like the more plausible way of making what he says more precise, and
indeed he suggests to Hussed in correspondence which way to interpret
certain passages of Grundlagen. 3 Now I think the idea that content was split
up into sense and reference is an oversimplification and does disservice to the
complexity of the dialectic underlying Frege's move from his earlier views in
Begriffsschrift to the mature view of "On Sense and Reference". Sense does
not arise from content alone, but also from something which in
Begriffsschrift is external to content. The idea that sense and reference arise
from content is I think most plausible for the case of sentences: there it does
 THE NEXT BEST THING TO SENSE IN BEGRIFFSSCHRIFT 131

seem that Frege's earlier judgeable content is an amalgam of thoughts or

propositions on the one hand and truth-values on the other. In the
Grundgesetze Frege introduces the distinction between sense and reference as
arising from content first in connection with sentences,4 and he also
mentions to Husserl that judgeable content has been divided into thought and
truth-value. 5
From the fact that Frege considered sense and reference to have been ex-
tracted from conceptual content alone, we should intuitively expect there to
be little or no inkling of the distinction in the early writings. In
Begriffsschrift however, Frege comes very close indeed to making the dis-
tinction, in fact about as close as he could possibly come without actually
making it in the way he later did. The passage in question is precisely in the
section on identity.
Frege considers a geometrical example of an informative identity state-
ment. Let A be a point on the circumference of a circle, and let a ray (half-
line) from A intersect the circle in a point B. Which point B is in a given
case depends on the angle the line makes to the diameter through A. As the
angle varies, the point of intersection varies continuously with it. We can
see this by simple trigonometry. If the line makes an angle q with the di-
ameter, B is a distance d.cos(q) from A, where d is the circle's diameter, and
where q = ±90°, this distance is zero, so B is the point A itself. (See figure.)

-+----~A

In this case, Frege says, the name B has the same content as the name A, but
we cannot use the same name from the start, since the justification for doing
so is only given by the answer (to the question which point B is in this
particular case). The point is determined in two different ways, immediately
through intuition (i.e., the point A on the circumference), and as the point B
which is the point where the ray from A intersects the circle, in the case
where the ray is at right angles to the diameter through A. Each mode of
determination (Bestimmungsweise) corresponds to one of the names. Since
the same content (here: point, i.e., the referent) can be fully determined in
more than one way, or, as he also says, the same thing (Dasselbe) is given
132 PETER SIMONS

by two modes of determination, when we say this we express a judgement.

In order to express this judgement, we need two names, one going with each
mode of determination, and they need to be connected in the judgement,
hence we need a sign for identity to express this judgement. It follows, says
Frege, that having different names for the same content is not always merely
a matter of form, but, since they go with different modes of determination,
they concern the matter itself, and a judgement of identity of content is
synthetic in Kant's sense. 6 Frege also mentions that a sign for content
identity can also be used to introduce abbreviations. He summarizes his
discussion thus:

I- (A::B) then means: the sign A and the sign B have the same conceptual
content, so that B can be put everywhere in place ofA and vice versa.

This obviously makes an identity judgement metalinguistic, a judgement not

about the content of the signs A and B but about the signs A and B
themselves. Frege indeed begins §8 by saying that this is one place where
signs, instead of standing for contents, stand for themselves, i.e., we have
what the Scholastics called suppositio materialis instead of suppositio for-
malis, though Frege does not use the medieval terminology. He calls the
need for signs to have to stand now for their contents, now for themselves,
rather dramatically, "the bifurcation in the meaning of all signs".

3. MODE OF DETERMINATION VS. MANNER OF BEING GIVEN

Frege's very terminology, 'mode of determination', Bestimmungsweise,

strongly recalls what he says later in "On Sense and Reference" about the
sense of a sign:
It is natural to associate with a sign [... ], apart from that which it designates, which
can be called the reference of the sign, also that which I should like to call the sense
of the sign, which contains the manner of being given (worin die Art des
Gegebenseins entlUllten ist)'?

In "On Sense and Reference", as we know, Frege distinguishes three seman-

tic levels for any sign: the sign itself, its referent, and finally its sense,
which, if not precisely identical with the manner in which the referent is
given, at least contains this, as in the quotation. (Whatever else the sense
contains apart from this is not something on which Frege enlightens us.) As
in Begriffsschrift the different ways of determining a content allow identities
 THE NEXT BEST THING TO SENSE IN BEGRIFFSSCHRIFT 133

to be informative, so in "On Sense and Reference" the different ways in

which a referent is given allow identities to be informative. If we apply the
same way of counting levels to Begriffsschrift as to "On Sense and
Reference", we indeed find three, not two. We have the sign, its content, and
the way in which the content is determined by the sign. Let me call the mode
of determination of an expression its meaning. (So I am emphatically not
using 'meaning' for Frege's Bedeutung, which I do not do in any case.) We
have sign, meaning and content, instead of the later sign, sense and reference.
If we do not actually have sense in Begriffsschrijt, we seem to have the next
best thing.

4. THE CRITICISM OF THE METALINGUISTIC THEORY IN "ON

SENSE AND REFERENCE"

What then is wrong with this theory and why does Frege come to prefer the
one in "On Sense and Reference"? It would seem that the best way to find
out is to examine how Frege argues against a metalinguistic theory in "On
Sense and Reference". He says:

What one wants to say with a = b seems to be that the signs or names "a" and "b"
refer to the same thing, and then we should indeed be talking about these signs; a
relation between them would be asserted. But this relation would hold between the
names or signs only in so far as they name or designate something. It would be a
relation mediated by the connection of each of the two signs with the same
designatum. But this is arbitrary. No one can be forbidden to adopt any arbitrarily
producible event or object as a sign for something. A sentence a = b would thereby
no longer concern the matter itself, but merely our mode of designating; we would
not thereby eXI>ress a genuine item of knowledge. But that is just what we want to do
in many cases. 8

The idea is fairly clear. If all we have are the sign and the referent, and iden-
tities assert that two signs have the same referent, then the truth or falsity of
identities depends on the merely linguistic facts about which signs happen to
have been used to designate which objects, and so cannot convey the kind of
substantial information that identities are supposed to do.
This argument is convincing enough in itself, but one might be forgiven
for thinking that it has rather little to do with Frege's own previous theory
in Begriffsschrijt. It deals only with the relation of designation between signs
and their referents, and completely overlooks the meaning or mode of
determination which was crucial for Frege's argumentation in Begriffsschrift.
Frege then introduces the sense/reference distinction with another geometrical
example: if a, band c are the medians of a triangle, i.e., the lines joining the
134 PETER SIMONS

vertices to the midpoints of the opposite sides, the point of intersection of a

and b is the same as the point of intersection of band c (see figure):

It is of course not immediately evident, but has to be proved geometrically,

that the medians intersect in one point; the example is rather better than the
one in Begriffsschrift. Frege then goes on to call the different ways in which
the intersection point is given the senses of the expressions 'the point of
intersection of a and b' and 'the point of intersection of band c'. In other
words, now he lets something which in Begriffsschrift he would have called
the mode of determination be the sense, and goes on to work with that view.
Where then is the difference, apart from the terminological one that now he
calls the content the referent and the mode of determination the sense?

5. AN ADV ANTAGE OF THE OLD THEORY?

There even seems to be an advantage to the old way of speaking over the
new. If the sense of a proper name (singular term) is the way in which its
referent is given, then if there is no referent to be given, there is no way for
it to be given, and hence no sense for an expression without a referent. We
know of course that Frege did not think this: he thought that fictitious and
other expressions in ordinary language can have a sense and not refer, and one
can given examples from mathematics using improper decsriptions or
functions undefined at certain points, e.g., 'the largest prime number', 'the
square root of 4', '5/0', although Frege wanted to let all expressions in sci-
ence have a reference. Indeed Gareth Evans argued for the reason I have given
that Frege should have accepted that there is no sense without reference.9
The way of speaking in Begriffsschrift does not give rise to this problem,
or at least not so obviously. There can easily be a mode of determination
which fails to determine anything, or does not determine a unique thing, and
so fails of reference. And I do not imagine that Frege would have disputed
this. But I do not think that Evans's argument is very convincing; certainly
it hangs too much on Frege's rather informal elucidation of sense, and
 THE NEXT BEST THING TO SENSE IN BEGRIFFSSCHRIFT 135

ignores Frege's own acceptance of signs with sense but withour reference. So
the advantage of the old way of speaking is a superficial terminological one,
not a real one.

6. WHAT IDENTITY STATEMENTS ARE ABOUT

The real objection that Frege conceived to the earlier Begriffsschrift theory is
obviously not to the idea of modes of determination, which are, I suggest,
little more than senses by another name. Rather it is to the idea that identity
statements are about the signs and say that they have the same content, or, as
he would later say, the same referent. What then, are identity statements
about in "On Sense and Reference"? It looks as though Frege has already
rejected the idea that they are about the referents: that idea was the first to go,
and was what led Frege to the metalinguistic theory in the first place to
explain the informativeness of identity statements. But Frege does not say in
"On Sense and Reference" that identity statements are about the senses of
expressions: to talk about the sense of a name we do not simply use it modo
recto: either we use it modo obliquo or we use the expression "the sense of
the expression so-and-so" to refer directly to it.
The solution is that by introducing the senselreference distinction Frege
henceforth rejects the simple idea that there is one item which a name or
expression can be said to be about. Suppose I overhear someone saying that
the present Emperor of Austria is a German citizen, and know he happens to
be a Habsburg legitimist, and someone who has only heard a fragment of the
conversation asks me, "Whom was he talking about?", I can answer in one
of two equally legitimate ways. I might answer, "the present Emperor of
Austria", offering back an expression with the same sense as the one the
speaker used, or I might answer, "Dr. Otto Habsburg", offering back an ex-
pression denoting the same referent as the one referred to by the speaker.
There is, we might say, sense-aboutness and reference-aboutness. In
Begriffsschrift, Frege had only content-aboutness. Since content-aboutness
tends in Frege's discussion of names to gravitate towards reference-aboutness
(especially in the discussion of identity), content-aboutness would appear to
give the wrong answer to the question whether identities can be informative.
Having already isolated the mode of determination as that which mediates
between sign and content and allows identities to be informative, rather than
split the content into this and referent as he was to do later, Frege falls back
on the only other thing that is connected with the mode of determination and
which can plausibly be supposed to be what identity statements are about (if
136 PETER SIMONS

every sign can be about at most one thing), namely the signs themselves.
Since it is the signs that express modes of determination, and true
informative identities are informative by virtue of the names flanking the
identity sign having different modes of determination of the same thing
(content), there is only the sign left. It is precisely the rejection of this idea
that there is just one thing that a sign is about, one objective side to its
meaning, that constitutes what we and Frege take to be the advance of the
new theory on the old.

7. WHAT'S THE DIFFERENCE?

The old theory, apart from the misleading metalinguistic interpretation of

identity statements as about signs, has almost the same resources as the new.
And its misleading aspect does not interfere with Frege's treatment of the
logic of identity: the laws and the derivations go through perfectly well, and
indeed the earlier logic of identity is much more perspicuous than the later
theory in Grundgesetze, where Frege sacrifices the nicely modulated stages of
Begriffsschrift in favour of a more compact and ingenious yet in some ways
more baroque system. In any case, sense plays very much a background role
in Grundgesetze. So what's the difference?
One account of the difference between the old theory and the new might
be this. In Begriffsschrift, the relation between sign and mode of determina-
tion is too intimate: we need modes of determination to float free of signs
and our use of them as something objective. This does not seem right to me.
It is true that in Begriffsschrift Frege lays some stress on having signs
connected with modes of determination in order to be able to express identi-
ties, but this requirement holds for the later theory as well. As I read §8 of
Begriffsschrift, there is no indication that modes of determination are neces-
sarily tied to signs any more than the later senses are. On the contrary, it
seems as though because "we must supply two different names corresponding
to the two modes of determination", the relation between sign and mode of
determination is precisely the kind of external, arbitrary one which Frege
criticises in "On Sense and Reference".
It may be, on the other hand, that the use of geometrical examples, with
their clearly arbitrary use of letters for lines, angles, points etc., is mislead-
ing. Closer attention to the way in which the modes of determination are
actually expressed reveals that they are often given by complex phrases.
Complex phrases must derive their meaning in some way from their con-
stituents: given the meanings of the constituents and the way in which they
 THE NEXT BEST THING TO SENSE IN BEGRIFFSSCHRIFT 137

are combined, the meaning of the phrase is given. It may be -- and here I am
speculating beyond what the text says or entails -- that for all the talk of
modes of determination in Begriffsschrift, Frege was unwilling to there
hypostatize or reify them, and so was unwilling to conceive of designation as
a relation mediated by an intermediate entity. The recognition of
hypostatized, mediating senses in "On Sense and Reference" opens the way
to a fully-fledged systematic compositional theory of objective sense, which
the brief discussion of meaning and content in Begriffsschrift does not allow.
Frege's excursion into the philosophy of language in "On Sense and
Reference" is clearly intended to show the power of his distinction in dealing
with difficult cases, such as oratio obliqua. I am convinced that it was the
sense of a whole range of phenomena clicking into place which persuaded
Frege that the distinction between sense and reference as the basis of a
thoroughgoing theory was preferable to the ad hoc solution to the apparently
isolated problem of informative identities in Begriffsschrift.

8. COMPLICATIONS

Nevertheless, the recognition of sense and reference as distinct from content

or meaning and content does not eliminate the "bifurcation of meaning"
which afflicted signs in Begriffsschrift. Although sense and reference consti-
tute in their own way a bifurcation, it is one which Frege is happy to rec-
ognize as legitimate and useful. There is no longer any need to distinguish
suppositio materialis and suppositio formalis. If we want to say that two
signs have the same referent we can use quotation mark names, so distin-
guishing

The Morning Star = the Evening Star

which uses but does not refer to the two expressions referring to Venus, and
is true because they both refer to Venus, from

The referent of 'the Morning Star' = the referent of 'the Evening Star'

which refers directly to the expressions, but to Venus only indirectly, via the
expressions and the function 'the referent of ... ', and, for all we know, could
be true if both referred to some other thing than Venus. On a Russellian
theory of descriptions, indeed, the second example would not even refer to
138 PETER SIMONS

Venus indirectly. The two "say quite different things", as a translation

shows: the first translates as

Der Morgenstern= der Abendstern

whereas the second translates as

Die Bedeutung von 'the Morning Star' = die Bedeutung von 'the Evening
Star'.

The secol1ld, but not the first, mentions English phrases.

But Frege has not hereby abolished all bifurcations. For one thing, the
analysis of indirect contexts shows that expressions can have their normal or
direct sense as referent, and so must have an indirect or second sense in such
contexts. Whether this leads to an unending hierarchy of senses or can be
stopped there is an issue I shall not go into, but either way the theory is not
so simple as appears at first sight.
There is another and more tantalizing suggestion that things might not be
as simple as they appear. The first public mention of the sense/reference
distinction occurs not in the essay "On Sense and Reference" itself but in the
essay "Function and Concept", delivered as a talk in Jena early in 1891. In
this essay Frege not only presents his newly refashioned function/object
distinction, he also runs us through all the primitive functions of
Grundgesetze with the exception of the description function. In particular, the
Basic Law V for courses of values is presented. He says
in 'e (e 2 - 4e) = a(a.[a - 4])' we have the expression for [sa2'ing] that the first
course of values is the same as the second. [... The expression] 'x - 4x = x(x-4),
[understood as a generalization, i.e., universally quantified] expresses the same
Sinn, if we understand it as above, but in another way. It presents the Sinn as the
generalization of an equality, whereas the newly introduced expression is simply an
equality, whose right side as well as its left has a complete Bedeutung in itself. lO

This is very suggestive. Just as the sense of an expression is the way in

which its referent is given to us, so it would appear that senses can be ex-
pressed in different ways by different expressions. This would appear to lead
to complications of the same order as, and perhaps not unconnected with,
indirect senses in oratio obliqua. Now it might be conjectured that Frege was
simply being sloppy here and failing to distinguish sense and reference again.
Since the distinction is actually made in the essay, I rather doubt whether
Frege was being sloppy. An indirect indication that the two sides of Basic
 THE NEXT BEST THING TO SENSE IN BEGRIFFSSCHRIFT 139

Law V were thought by Frege to have the same sense is provided by his
discussion of the law in §10 of Grundgesetze. As will be recalled, Frege
points out that Basic Law V, by only stating when courses of values are
identical, and not explicitly laying down which objects they are, does not of
itself fix their reference. He shows this by considering an arbitrary
permutation X of the objects, and shows that this, too, preserves Basic Law
V. In introducing the argument, Frege says that, because X is a bijection of
all objects, 'X(e:<l>(e» = X(a'l'(a», is coreferential (gleichbedeutend) with
'-U-(<l>(x) = 'I'(x»', but adds in a footnote "That is not to say that the sense
is the same."ll. This strongly suggests that the sense of '( e: <l>( e) = a'l'(a»'
is the same as that of '-U-( <l>(x) = 'I'(x» , , whether or not it is expressed in
the same way. But since the syntactic structure of the two clauses is quite
different, how can the sense be expressed in the same way each time? So it
seems the same sense can be expressed by syntactically quite different
expressions, which is, after all, what happens in definitions.

9. SUMMARY

I have suggested that Frege got as close as he could have done in the context
of Begriffsschrift to the distinction between sense and reference, but only for
names of non-judgeable contents, and only locally, in connection with
thinking about identity. The drive to a sense/reference distinction as
something made within content seems to have come from thinking about
judgeable contents. Its advantages as a general theory were soon apparent to
Frege, and he reapplied the new distinction to cover much the same ground as
in the Begriffsschrift discussion of identity, as well as indicating how it
would apply in many other areas. In doing so, however, he not only avoided
some of the complications of the earlier theory: he also opened the door to
more.

NOTES

1 Frege 1953, p. 73. The translation here, as elsewhere, is my own.

2 So called in Dummett 1991, pp. 160 ff. after the example at Frege 1953, §56, p. 68. The
actual statement of the problem, at §66, pp. 77-8, uses the example "the direction of the
Earth's axis = England".
3 Frege 1980, p. 63.
4 Frege 1893, Preface, p. 10.
5 Frege 1980, p. 63.
6 The allusion to informative identities as Kantian synthetic judgements recurs at the
beginning of "On Sense and Reference", Frege 1984, p. 157.
140 PETER SIMONS

7 Frege 1984, p. 158.

8 Frege 1984, p. 157.
9 Evans 1982, pp. 22 ff.
10 Frege 1984, p. 143. I have deliberately left Sinn and Bedeutung untranslated here.
11 Frege 1893, p. 16n.

REFERENCES

Dummett, M.: 1991, Frege. Philosophy of Mathematics, Duckworth, London.

Evans, G.: 1982, The Varieties of Reference, 1. McDowell (ed.), Clarendon Press, Oxford.
Frege, G.: 1879, Begrijfsschrift, eine der arithmetischen nachgebildete Formelsprache des
reinen Denkens, Nebert, Halle. Reprinted in Begrijfsschrift und andere Aufsiitze, I.
Angelelli (ed.), 2. Auf!. Hildesheim: Olms, 1977, pp. VII-88. English translations: (1) S.
Bauer-Mengelberg (Trans.) in 1. van Heijenoort (ed.), From Frege to G6del. A Source
Book in Mathematical Logic, 1879-1931. Cambridge, Mass.: Harvard University Press,
1967, pp. 5-82. (2) Trans. T. W. Bynum, in G. Frege, Conceptual Notation and Related
Articles, T. W. Bynum (ed.), Oxford: Clarendon Press, 1972, pp. 101-203.
Frege, G.: 1893, Grundgesetze der Arithmetik, begrijfsschriftlich abgeleitet. Vol 1. Jena: Pohle,
1893. Reprinted Darmstadt: Wissenschaftliche Buchgesellschaft, 1962. Partial English
translation as Basic Laws of Arithmetic. Exposition of the System, ed. & trans. M. Furth,
Berkeley: University of California Press, 1964.
Frege, G.: 1953, The Foundations of Arithmetic, trans. J. L. Austin. 2nd ed. Oxford: Blackwell.
Frege, G.: 1980, Philosophical and Mathematical Correspondence. Blackwell, Oxford.
Frege, G.: 1984, Collected Papers on Mathematics, Logic, and Philosophy, Blackwell, Oxford.

University of Salzburg
Franziskanergasse 1,
A-5020 Salzburg, Austria
 DAVID OWENS

UNDERSTANDING NAMES

Accordingly, with a proper name, it is a matter of the way the object so designated is
presented. This may happen in different ways, and to every such way there
corresponds a special sense of a sentence containing the proper name. The
different thoughts thus obtained from the same sentences correspond in truth value,
of course; that is to say, if one is true, then all are true, and if one is false then all are
false. Nevertheless, the difference must be recognized. so we must really stipulate
that for every proper name there shall be just one associated manner of presentation
of the object so designated. It is often unimportant that this stipulation should be
fulfilled but not always)

Frege's stipulation leads to trouble. Take almost any proper name and you
will find users of the name who identify its referent in quite different ways.
Are we to conclude that such people speak different languages with the same
words? This is what Frege implied. Recently, several philosophers have
sought to avoid this awkwardness by being less pedantic. For example,
Evans: "the single main requirement for understanding a use of a proper
name is that one think of the referent ... but is there any particular way in
which one must think of the object? It would appear not."2 This relaxed
attitude is enough to avert the fragmentation of our language into a thousand
peculiar idiolects. But before relaxing, we must ask ourselves why Frege
held it was sometimes important that his stipulation be fulfilled.
I suspect Frege was worried that people who identified the bearer of a
name in quite unrelated ways could not be said to understand one another. To
understand your interlocutor, you must at least know what his words mean.
But if the meaning he ataches to them is quite unrelated to the meaning you
attach to them, there is no guarantee that merely hearing his sentence will
enable you to latch onto the thought behind it. Therefore, to guarantee
understanding, we must demand that people attach similar meanings to
similar words. This line of thought is very rough and ready -- people who
speak different languages can still understand one another and even within a
language we must distinguish what a speaker means from what his words
mean -- but I think we can tighten it up by noting a crucial connection
between the notion of understanding and the concept of knowledge. 3

141

J. Biro and P. Kotatko (eds.), Frege: Sense and Reference One Hundred Years Later, 141-149.
© 1995 Kluwer Academic Publishers.
142 DAVID OWENS

THE PUZZLE

If someone is a reliable authority abouta certain subject matter and I know

this and they tell me something about it, then, ceteris paribus, I can come to
know that what they said is true merely by understanding what they said.
This suggests that one way to test an account of understanding is by asking
the following question: given that the account's conditions for understanding
that p are fulfilled by a certain hearer, can the hearer come to know that p
when he is told that p by an epistemic authority? If not, then the proposed
conditions on understanding are insufficiently stringent. Frege' s point can
then be expressed as follows: any account of understanding sufficient to
guarantee the transmission of knowledge by testimony will have to insist
that interlocutors attach a similar sense to each of the words they use.
Two characters, Blind and Deaf have beliefs about a certain woman which
they express by means of the proper name 'Margaret Thatcher'. Each
character is surprisingly ignorant of contemporary British politics, neither
can produce a uniquely identifying description of this 'Margaret Thatcher' .
They know she is a female British politician of some importance and
longevity but neither could name the governmental offices she has held etc.
Some philosophers may deny that Blind and Deaf really understand the name
"Margaret Thatcher' on the grounds that neither can say to whom the name
refers, but I shall suppose Blind and Deaf do know the referent because they
are able to recognize her, Deaf by her facial appearance and Blind by her
voice. To avoid irrelevant complications, let us further assume Blind and
Deaf can tell her apart, by these separate means, from every other human
being.
Here, Frege must insist on distinguishing the language of the deaf from
the language of the blind. since Blind and Deaf think of the referent of
'Margaret Thatcher' in different ways, they mean different things by
'Margaret Thatcher' and when Deaf tells Blind that Margaret Thatcher is
becoming unpopular, there is no one thought which Deaf expresses by that
sentence and Blind understands. But is Frege not being unduly fastidious? It
is true that Blind has auditory information about Margaret Thatcher which
Deaf lacks and that Deaf has visual information about Margaret Thatcher
which Blind does not possess and, as a result, they identify her in different
ways. But when Blind hears a characteristic political speech of hers and Deaf
subsequently learns from Blind that Margaret Thatcher will become
unpopular as a result, why scruple to deny that Deaf knows what he is
agreeing to? Our epistemological principle provides a possible answer.
 UNDERSTANDING NAMES 143

Blind knows the contents of the speech and he tells Deaf that Margaret
Thatcher will become unpopular. Bind is sincere and reliable on the subject
of politicians and Deaf is aware of this. Further, the other epistemically
relevant features of the situation are as one would wish and Deaf is aware of
this also. Surely, all that is now required for Deaf to come to know about
Margaret Thatcher's unpopularity is that he understand what Blind is telling
him -- understanding is here sufficient for knowledge. So if Deaf can't learn
of Thatchers's unpopularity, this can only be because he doesn't understand
Blind.
Deaf certainly acquires a true belief, one which he would express with the
sentence 'Margaret Thatcher is unpopular', as a result of what Blind tells
him. But such a true belief cannot be knowledge if that belief is true only as
the result of an accident. Now the truth of the belief which Blind would
express by means of this sentence ensures the truth of the belief that Deaf
would express by means of this sentence only if this sentence refers to the
same person in the mouths of both Blind and Deaf. And its seems little
more than coincidence that Blind and Deaf refer to the same woman when
they use this sentence. Since the mechanism by which Blind identifies the
referent is quite independent of that by which Deaf identifies the referent, it
looks like an accident when these mechanisms pick out the same individual.
If so, Dears belief will not be knowledge.
Imagine that a third party, neither blind nor deaf, is presented with a
photograph of Margaret Thatcher. There would be no way of inferring
Thatcher's penetrating voice, somewhat modulated by extensive coaching,
from her peroxide hairdo and receding throat. Knowledge of Thatcher's visual
appearance alone would give this person no reason to suppose that the speech
he is now hearing on the radio was being delivered by Thatcher. Usually,
one must witness the conjunction of the face and the voice before one can
establish a connection between them -- exactly what Blind and Deaf cannot
do. Equally, it is impossible to infer Thatch's visual appearance from her
voice. And if one person cannot make this move, how can two do any
better? Blind's asserting the sentence 'Margaret Thatcher is becoming
unpopular', because he has heard a speech delivered in that penetrating voice,
is no reason for Deaf to assert it also.
Someone might complain that I am operating with an unacceptably
internalist conception of knowledge. Surely, what is required for
communication between Blind and Deaf is simply that 'Margaret Thatcher'
refer to the same person in both of their mouths. It is not further required
that Deaf be able to produce some reason for believing that Blind's Thatcher
144 DAVID OWENS

is his Thatcher. Now, in my view, this is a moot point, but I can perfectly
well concede it in the present context. For the externalist must agree to at
least this much: if Blind is to transmit knowledge to Deaf it cannot be a
mere accident that they end up thinking of the same person.
Now in any sense relevant to the acquisition of knowledge, it is a
coincidence that the woman with that particular appearance has a voice which
sounds just that way, and therefore it is a coincidence when two people, one
looking for that voice and the other for that face, alight on the same person.
Only with vast amounts of background knowledge of a genetic,
physiological and sociological nature could one guess the face from the voice
or vice versa. If the externalist claims that even someone without this
background information could infer from the appearance of a certain face to
the sound of a certain voice, then almost any inference from one property of
a person to another would yield knowledge. Surely this cannot be correct.
Perhaps we can invent a method of Thatcher-identification common to
Blind and Deaf by focusing on the name itself. Could Deaf and Blind each
think of Margaret Thatcher as 'the person called 'Margaret Thatcher" and
thereby restore communication between them? There is the difficulty that
many women are called 'Margaret Thatcher', but Blind can get around this
simply by making it clear that he is using the name of a British politician.
However, as Evans points out, there is a great difference between thinking
that Thatcher is unpopular, on the one hand, and thinking that a British
politician called 'Thatcher' is unpopular, on the other. 4 To think that
Thatcher is unpopular, one needs to know who Thatcher is, and it is this
thought which Blind is trying to convey to Deaf.
Apparently Frege has been vindicated by Blind and Deaf. Blind and Deaf
identify the reference of the name 'Margaret Thatcher' in quite independent
ways and can't convey information abut her to each other. So if sharing a
language is to be sufficient for communication, we must conclude that Blind
and Deaf do not share a language because they don't share a way of thinking
about Margaret Thatcher. However, I shall suggest that we can restore
communications between Blind and Deaf without insisting that they pick out
Thatcher in the same way.

A RESOLUTION

Evans insists that, when referring by means of a name, the speaker must
indicate which name he intends to be using. 5 I take it that a name is to be
individuated by reference to the practice of applying that name to a given
 UNDERSTANDING NAMES 145

person. Since there are many people called 'Margaret Thatcher' , people who
are rarely, if ever, confused, we must conclude that there are many different
name-using practices which employ names of the same phonetic and
orthographic type. The speaker must indicate which of these names he
means to be using when alluding to Thatcher, only then will he have made a
definite assertion.
Note that there may be two quite distinct name-using practices involving
the same person and the same type of name. 6 In Stevenson's novel, no one
knows that the nasty night-time Dr. Jekyll and the nice day-time Mr. Hyde
are the same person, and so the practice of applying the name 'Hyde' is quite
distinct from that of applying the name 'Jekyll'. But the same could be true
even if our character were called 'Mr. Hyde' both before and after dark: there
would still be two quite distinct practices of applying a 'Hyde' -name to the
same person and therefore two distinct names for that person. I shaH urge
that, to be understood, a speaker must make clear which name he is using,
for it is only by so doing that he can convey which person he is talking
about.
But how is this new requirement going to solve our problem; won't it
simply impose an extra burden on us? In actual fact, Blind will have done
enough to identify the word he is using simply by making it clear that the
British politician is in question. So Blind can indicate which name he is
using. without having to convey to Deaf the method by which he identifies
that name's referent, a requirement he can't possibly fulfill. Once it is clear
to Deaf that Blind is using that 'Thatcher' name, Deaf can bring to bear his
own way of identifying the referent of the name 'Margaret Thatcher' and thus
succeed in having the desired thought 'Margaret Thatcher is unpopular',
rather than some metalinguistic surrogate. So Blind and Deaf can
communicate despite thinking of Margaret Thatcher in quite different ways.
What Evans has done is to replace the issue of how Blind can indicate to
Deaf which person he is referring to with the issue of how Blind can indicate
to Deaf which name he is using. The thought is that you can identify a
particular name 'N' in a way which ensures that your interlocutor latches
onto that name even if you have no method of identifying N to your
interlocutor other than as the bearer of the name 'N'. Once your interlocutor
knows which name is in play, he can then connect the name with the bearer
in his own idiosyncratic fashion. Thus, two people can exchange
information about N while identifying N in quite unrelated ways, provided it
is no coincidence that they both latch onto the same name. Since Blind and
Deaf identify the name 'Thatcher' in the same way, namely as the name of a
146 DAVID OWENS

famous British politician, it is no coincidence that they end up employing

the same name.
If I am right, where two speakers identify a given name in quite
independent ways, communication will be undermined. Suppose that
Margaret Thatcher decides, upon her retirement, to become a radio game
show host. To ensure a convivial atmosphere on the show, her political past
is never mentioned. Smith knows nothing about politics and listens only to
radio game shows. Jones followed Thatcher's political career on television
but never listens to the radio. Smith informs Jones that Margaret Thatcher
has a good radio voice. Jones takes it that Smith is speaking of the
politician called Thatcher and concludes that she has a good radio voice. Has
Jones understood Smith? Can Jones learn from what Smith says that
Thatcher has a good radio voice?
I would say not. It is certainly true that there is only one name in play
here. The same women is both a politician and a game show host, and the
practice of calling her 'Thatcher' was carried over from the political arena to
the entertainment business. However, Smith and Jones are unusual in that
they are not aware of Thatchers's mutation. One thinks of the name
'Thatcher' purely as the name of the a politician. And surely, neither has
reason to believe that one person with a single name is first a politician and
then a radio personality. No one without the relevant background knowledge
could be at all confident that the same name 'Thatcher' was being used both
on the game shows and in the political histories. Therefore, Jones does not
m:iderstand Smith, and the belief which he acquires by listening to Smith is
not knowledge. Indeed, if Smith then made it clear that he was referring to
Thatcher, the radio personality, and denied any knowledge of a political
career, Jones would have strong doubts as to whether he had understood
Smith correctly in the first place.
Even a speaker who connects the name 'Margaret Thatcher' with its
bearer in exactly the same way as his audience will not be understood unless
he makes it clear which 'Thatcher' -name he is using. Suppose Smith and
Jones both recognize Margaret Thatcher by her voice -- they can discriminate
her voice from that of every other human being -- and they each associate
this voice with the name 'Thatcher'. Now, not being talented mimics, they
can't convey this method of identification to one another, so Smith can't use
his appreciation of her voice to indicate to Jones which name he is using.
All he can articulate is that Thatcher is a radio personality and this ought to
give Jones no confidence that Smith's Thatcher is the politician. Therefore,
they can't communicate by means of the name. Communication requires not
 UNDERSTANDING NAMES 147

only that two people both associate Thatcher with the name "Thatcher' but
that the hearer know which 'Thatcher' -name the speaker is employing; and
the hearer may not know this even if he attaches 'Thatcher' to its bearer in
exactly the same way that the speaker does.
To sum up, Blind and Deaf do, after all, understand one another, provided
Blind ensures that it is no coincidence that Deaf latches onto the very name
which Blind is using. Blind can do this by telling Deaf that he is speaking
of the British politician, even though the way Deaf connects the word
'Thatcher' to its bearer is quite different from the way Blind connects
'Thatcher' to its bearer. By contrast, Smith and Jones do not understand one
another because it is a coincidence that they latch onto the same name
'Thatcher' . Therefore, it is getting hold of the name that is crucial to
understanding -- it doesn't matter how you go on to connect the name with
its bearer provided you can do this somehow.
I have not made any attempt to say exactly what is required for a hearer to
know which name a speaker is using. As things are, I think most would
agree that Blind has indicated which 'Thatcher' -name he is using by telling
Deaf it is the name of the British politician, despite the fact that Blind cannot
rule out the possibility that there are other British politicians of the same
name. Similarly, Blind is said to be able to recognize Thatcher by her voice
even though he can't be absolutely certain that there are not other people
with exactly the same voice. The problem of how we can know when we
have identified a given particular is a quite general one and is neither
exacerbated not solved by claiming that named objects can be identified by
means of their name.

IDENTITY STATEMENTS

I want to conclude by saying how all this bears on the main theme of
Frege's article 'Sense and Reference', whose publication we are celebrating
with this volume. One of Frege' s motivations for speaking of the sense of a
name, or of the wayan object is designated or presented by a name, was to
explain the informativeness of identity statements. Take the statement
'Hesperus=Phosphorus'. This statement can be informative because some-
one could understand both of the names in it and yet fail to realize that they
picked out the same Object: someone could know which heavenly body
'Hesperus' referred to and could know which heavenly body 'Phosphorus'
referred to, without knowing that each referred to the very same planet.
148 DAVID OWENS

None of what I have said is inconsistent with this explanation of the

potential informativeness of identity statements. Rather, what may be under
threat is the idea, implicit in Frege's article, that informativeness is an
intrinsic property of identity statements -- that a statement is either
informative or not in virtue of its very meaning. The issue turns on the
question of what constitutes an identification of a name adequate for
communication.
For instance, suppose we thought that anyone who understands
'Hesperus' must think of it as the evening star and that anyone who
understands 'Phosphorus' must think if it as the morning star. Then, anyone
who can comprehend 'Hesperus=Phosphorus' must find it an informative
statement -- something which they couldn't know a priori. But I have been
taking the line that there is no particular way in which someone must think
of Hesperus in order to understand the name 'Hesperus'. Given this, it is
quite easy to see how 'Hesperus =Phosphorus' might be news to a speaker
but an analytical truth to his audience, even though they both understand the
statement perfectly.
Say that Mary thinks of both Hesperus and Phosphorus as the evening
star. She is informed by Joan, who is trying to tell her something which
she learnt only yesterday, that the two heavenly bodies called 'Hesperus' and
Phosphorus are identical. Mary understands Joan perfectly, since Joan has
indicated which names she is using and Mary has her own way of connecting
each name with its bearer. But since Mary connects each name with its
bearer in exactly the same way, the identity statement which Joan discovered
only yesterday is, for her, a simple consequence of the way she understands
Joan's statement. For her, there is no difference between understanding
Joan's statement and thinking that it is true. What was, for Joan, an
empirical claim is, for Mary, an a priori truth.
One might object that Mary has no real reason to suppose that the name
'Phosphorus' used by Joan is the same name as her 'Phosphorus'. True,
Joan has indicated that they are both names of heavenly bodies, but there
could quite easily be two different heavenly-body names in circulation, which
sound exactly the same, just as Smith's 'Thatcher'-name could be different
from Jones'. The fact that Joan thinks it worth uttering an apparently trivial
identity statement should make Mary suspect that she hasn't quite grasped
what Joan is saying. However, we can't assert that Mary's 'Phosphorus'-
name is a different name from Joan's just because Mary and Joan think of its
bearer in quite different ways. We have to accommodate the obvious fact that
people like Blind and Deaf, use the same name for successful communication
 UNDERSTANDING NAMES 149

even while they identify its bearer by quite independent methods. So we

must allow that Blind can indicate to Deaf which name he is using without
thereby fixing the way Deaf thinks of its bearer. And given this, we have a
recipe for constructing examples which will demonstrate that informativeness
is not an intrinsic property of identity statements.
First two names 'N' and 'N' * are singled out from all other names in
some agreed fashion (say by the use of an indefinite description of their
bearers). Then everyone proceeds to identify the bearers in his own way.
Given that everyone identifies Nand N* quite independently, it is surely
possible for at least one person to identify Nand N* in the same way while
others identify them in different ways and then we will have a situation in
which 'N'='N*' is informative to the latter but not to the former. Such
people can communicate effectively because they have means of singling out
'N' and N*' which don't fix the way they identify their bearers.
But what of Mary's puzzlement when she hears Joan Utter a trivial
identity in a sincere attempt to be helpful? Well, here Mary can reflect that
there are other instances of statements which are informative to some but
necessarily uninformative to others. For example, if Mary has been moving
around and wants to draw Joan's attention to her position, she may say' I
am here now'. This statement could not possibly express something Mary
didn't know, but it is very helpful to Jane. Mary would hardly be entitled to
doubt that Joan has understood her on the grounds that Joan found the
statement informative while she, Mary, could not. Why shouldn't it be so
with some identity statements?
NOTES

1 G. Frege, Logical Investigations (Blackwell, 1977) p. 12.

2 G. Evans, The Varieties of Reference (Oxford, 1982) p. 400.
3 G. Evans, pp. 310-11.
4 G. Evans, p. 399.
5 G. Evans, p. 384. Peter Smith has pointed out that an audience might be able to work out
which name a speaker was using by using contextual cues which were neither provided by,
nor relied on, by the speaker. Here, I think Evans would still insist that the speaker has made
no definite assertion by using the name even if his audience can work out which assertion he
intended to make from the context: they can understand what he meant to be asserting even
though they can't understand his assertion. Anyone who does not agree with Evans on this
point should simply rephrase what follows so as to allow for the possibility that the name in use
is indicated by things which the speaker makes no use of.
6 G. Evans, p. 381.

Department of Philosophy, University of Sheffield

Sheffield S10 ZTN, UK
 EROS CORAZZA & JEROME DOKIC

WHY IS FREGE'S PUZZLE STILL PUZZLING?

Gottlob Frege begins his essay Sinnund Bedeutung by stating a puzzle

(which we follow the tradition in calling "Frege's Puzzle") raised by the con-
sideration of now familiar linguistic phenomena. Our main purpose in the
present paper is to make plausible the claim that Frege's puzzle remains a
serious challenge that has not yet been dealt with in a fully satisfactory way.
In the first section, we try to formulate Frege's Puzzle in such a way that
its difficulties should be recognized by (almost) everyone, however reluctant
to accept Frege's own solution to it. In section 2, we criticize an influential
non-Fregean answer to Frege's Puzzle. In section 3, we examine some
Fregean answers to the puzzle, but we conclude that they, too, are at best in-
complete. In the final section, we try to describe what we see as the central
point of Frege's Puzzle.

1. FREGE'S PUZZLE: FOUR RELEV ANT NOTIONS

Frege's Puzzle can be stated, somewhat artificially, in the following way:

how are we to explain that a rational subject who understands and accepts as
true an identity of the form "a=b", where "a" and "b" are genuine singular
terms, may be justified thereby in extending her knowledge, while under-
standing and accepting as true some identities of the form "a=a" may not
contribute to such an extension?1
This formulation of Frege's Puzzle deserves some comments. First, by
"extending her knowledge", we mean her being in a position to draw new in-
ferences eventually leading to new actions. Consider a simple case in which
the subject has the following stock of beliefs: "Fa", "Ga", "Hb", "Ib", and in
which she suspends her belief about "a=b". As soon as she has accepted this
identity, she is justified in extending her knowledge, i.e., she has a reason for
accepting "There is a thing which is both F and H" or, if she knows that
whatever is both G and H is also J, "There is a thing which is J". Accepting
these as true can then lead to new behavior for her.2
Second, we have not supposed, as Frege does, that "a=a" (or the proposi-
tion thereby expressed) is a priori, while "a=b" (or the proposition thereby
expressed) is a posteriori. The reason is that this formulation by Frege has
been considered by some people as a petitio principii in favor of the view
that the two sentences express different propositions. Our formulation shows

151

J. BiroandP. Kotatko (eds.), Frege: Sense and Reference One Hundred Years Later, 151-168.
© 1995 Kluwer Academic Publishers.
152 EROS CORAZZA & JEROME DOKIC

that Frege's Puzzle can be stated in an interesting way without committing

one either positively or negatively on this question.3 Besides, despite our
formulation, our understanding of the puzzle does not necessarily involve
two singular terms, "a" and "b". We could also have contrasted two uses of
the same identity say "a=a", of which the one is informative, the other not.
As will become clear, it is the contrast between two uses of an identity sen-
tence which is relevant to Frege's Puzzle, not the number of singular terms
used in it.4
Third, and consequently, it is important to distinguish at least four rele-
vant notions involved in Frege's original puzzle. (i) Cognitive value. We
shall say that in typical versions of the puzzle, "a=b" is informative, unlike
"a=a", or alternatively that the cognitive value of the former is different from
the cognitive value of the latter. By compositionality, it follows that the
cognitive value of "a" (as it occurs in the aforementioned identity) is different
from that of "b".5 (ii) Cognitive content. The cognitive content of a singular
term is the body of information that the subject associates with it. In our
example, the cognitive content of "a" is made up by the information that it
is F and G, while the cognitive content of "b" is made up by the information
that it is Hand 1.6 (iii) Information content. The information content of a
sentence is what one learns to be the case about the world just by
understanding and accepting the sentence'? Intuitively, the information
content is the state of affairs that makes the sentence true. Finally: (iv)
Semantic content. The semantic content of a sentence is the entity
semantically encoded in the sentence and corresponds to what the sentence
says or expresses.
Using these different notions, we can state various problems arising from
Frege's Puzzle:

1. What gives a sentence, or a singular term, its cognitive value?

2. What is the relation between cognitive value and cognitive content? In

particular, does a difference of cognitive value entail a difference of
cognitive content?

3. What is the relation between cognitive value and information content?

In particular, given that "a=a" and "a=b" have different cognitive values,
does it follow that they have different information contents?
Alternatively, if we assume that they have different information contents,
must they have different cognitive values?
 WHY IS FREGE'S PUZZLE STILL PUZZLING? 153

4. What is the relation between information content and semantic

content? Is what a sentence says or expresses always its information
content?

Frege is traditionally interpreted as claiming that the cognitive value of a

singular term is a subset of its cognitive content (the subset that corresponds
to its Sinn), and that the semantic content of a sentence involving such a
term is identical with its information content, which itself depends on the
cognitive value of the term. In other words, the tradition sees Frege as al-
most conflating the four relevant notions involved in his puzzle. We won't
assess here the historical accuracy of this traditional interpretation. Instead,
we shall be particularly concerned with the fIrst two philosophical questions.
As for the third question, we shall follow William Taschek (1992) in
assuming that the relevant difference of cognitive value between "a=a" and
"a=b" is not to be found at the level of the information content of these sen-
tences. Frege's Puzzle is concerned with cognitive achievements which do
not require the acquisition of new information. That is, we shall assume that,
when they are true, these sentences are made true by exactly the same state of
affairs.s There are two ways of denying this assumption. The first is to claim
that "a" and "b" have different descriptive senses; the second is to question
the fact that these proper names can really be coreferential (and thus that "="
really stands for strict identity).9 Both ways recognize the significant
differences between "a=a" and "a=b" as to the conditions which must be
fulfilled for these identities to be true. The second way is not very
fashionable nowadays, as it assumes a non-standard ontology, and there are
well-known objections against the first as a general answer to Frege's
Puzzle' 10
We shall have little to say about the fourth question. The reason is that it
is closely related to very complicated problems concerning the step from
constant linguistic meaning to what is said or expressed in a particular con-
text, and the pragmatic factors that are operative in this step. In ignoring this
last question, we only hope to show that it is relatively marginal to the core
of Frege's Puzzle.
We are now ready to assess some of the most important answers to
Frege's Puzzle that have been put forth. Some of the solutions can be called
"Fregean", others "non-Fregean". By "Fregean solutions", we mean accounts
trying to solve the puzzle by introducing Sinn-like entities like senses,
modes of presentation, or guises. These entities should have the essential
154 EROS CORAZZA & JEROME DOKIC

properties Frege attributed to Sinn: they determine a unique object (even if

only relative to a context), and they account for, in some interesting yet still
metaphorical sense, a mode under which the object is given in thought. I I
Conversely, by "non-Fregean solutions" we mean accounts which deal with
Frege's puzzle without postulating Sinn-like entities.
In the following section we shall discuss a powerful position according to
which the puzzle needs a dissolution rather than a solution. I2

2. A DISSOLUTION OF FREGE' S PUZZLE?

Intuitively, Frege's Puzzle strongly suggests the possibility of knowing the

same object in different ways, and as we shall see in the next section, this
possibility is central to the neo-Fregean solution. For the moment, though,
we are going to discuss a putative dissolution of Frege's Puzzle that
recognizes the relevant possibility, but denies that it has anything to do with
the semantics of natural language. More precisely, it rests on a general criti-
cism of Frege's semantics, based on the rejection of what may be called
"Russell's Principle", roughly, the principle according to which thinking
about something entails having a cognitive fix (a representation) of the rele-
vant thing in the form of a discriminating knowledge of it. 13
Russell's Principle, it is argued, arises from a Cartesian picture of lan-
guage according to which what determines reference, viz., sense, depends on
psychological states properly correlated with words. The opposite picture is
that the significance of a word is not a function of what psychological states
are associated with it (states which are supposed to be "in the head"), but
rather of the word's being embedded in a social, linguistic practice -- the rela-
tion between a word and its bearer is somehow given in the social practice. 14
According to Howard Wettstein, this picture is un-Fregean because it
suggests that the use of language does not rest on a knowing-that ability --
and this is thought to entail that it cannot rest on discriminating knowledge
-- but depends rather on a kind of knowing-how ability.IS
How is anti-Cartesianism, with its rejection of Russell's Principle, sup-
posed to contribute to the dissolution of the puzzle? Here is Wettstein's
(1989: p. 175) answer:

Frege's puzzle (... ) is generated not by Frege's data alone, but only in conjunction
with the mental apprehension conception of reference [i.e. Cartesianism]. Is it so
obvious, I asked, that there is something so deeply puzzling about the very idea that
a speaker can be competent with two co-referring names and not know that they co-
refer? The radical change in perspective I have been encouraging makes the
dissolution of the puzzle even more dramatic. If one can refer to something without
 WHY IS FREGE'S PUZZLE STILL PUZZLING? 155

anything like a substantive cognitive fix on the referent, if the use of a name can be
virtually blind epistemically, then why should it be the slightest bit surprising that a
speaker might be competent with two co-referring names but have no inkling that
they co-refer?

And he goes further (1988: p. 157-8):

Notice that cognitive perspectives, modes of presentation, have no role in our new
form of explanation. We do not look into the speaker's head and find two different
conceptual files in terms of which we can now see why 'Cicero' and 'Tully' play
different cognitive roles for him. We rather reflect on the fact that given how little
need be in his head, his mere competence with the names puts him in no position to
decide the question of whether they co-refer. Indeed (... ) we might find identical
conceptual files for each of two cognitively distinct names.

Wettstein's criticism of Frege's semantics follows the traditional interpreta-

tion of Frege's writings; crucially, it suggests that cognitive value is for
Frege a subset of cognitive content, i.e., a determinate conception of the ref-
erent. Now Wettstein appeals to what has been called the "Reverse Frege
Puzzle" (a puzzle that is taken very seriously now in the USA, notably on
the West coast).16 Actually, this last puzzle is a generalization of Putnam's
and Kripke's Puzzles. It has many ramifications, but the main idea is always
that it is possible that a competent subject uses two terms that have distinct
extensions although she associates with them the same body of information:
as Putnam and Kripke tell us, we can use such terms as 'elm' and 'beech',
'Gell-Man' and 'Feynmann', although we are unable to distinguish an elm
from a beech, or Gell-Man from Feynmann. It is in conjunction with the
Reverse Frege Puzzle that the traditional interpretation of Frege is thought to
be problematic. Indeed, the cognitive value of a word cannot be determined
by the body of information associated with it, because 'elm' and 'beech', for
example, may have different cognitive values even though the subject's
conception of an elm is the same as her conception of a beech.
In our view, the Reverse Frege Puzzle is a myth. More precisely, the
puzzle arises only if one ignores the rational connections among the
subject's uses of words. Once we show that the Reverse Frege Puzzle is in
no way problematic, we can see that there is logical room for a Fregean view
on Frege's Puzzle, albeit a non-traditional one, since it recognizes a
distinction between cognitive value and cognitive content.
Wettstein points out, correctly, that there is nothing problematic about
the mere idea that a rational subject who is competent with two co-referring
terms may not know that they co-refer. But it is very misleading to say that
such a subject associates exactly the same body of information with both
156 EROS CORAZZA & JEROME DOKIC

tenns. Consider the 'elm'l'beech' case, and suppose that the competent sub-
ject associates the properties F, G, and H with both terms. First of all, note
that she must further associate with elms a property that is relational with
respect to beeches; i.e., she must suppose, or (at least) be disposed to sup-
pose, of a beech, that it is a different tree from an elm, or (at least) that it
may be a different tree from an elm. Corresponding remarks apply to the
other term, so that this relational property may be thought as insufficient to
distinguish the two bodies of information, for it is presumably symmetrical.
However, if the subject is rational, she must be disposed to make at least the
supposition that some information is associated with one term but not with
the other, i.e., that an elm has some property that a beech has not (and vice
versa). Since the cognitive content of 'elm' is connected, for the subject,
with the cognitive content of 'beech' via the identity relation, her sensitivity
to elms is slightly different from her sensitivity to beeches, even if this dif-
ference is (at present) only dispositional.17
The Reverse Frege Puzzle is based on the implicit assumption that the
uses of the two terms by a single subject are unconnected, but this assump-
tion is false for all rational subjects, who must have the disposition to en-
visage, at any time, some connection between their conception of an elm and
their conception of a beech. Note that if we are right, such a connection need
not be meta-linguistic, but may be specified in the most straightforward way,
using the identity predicate. I8
Now the dissolution of the Reverse Frege Puzzle shows the way to a re-
habilitation of Frege's original puzzle. The observation that the cognitive
content of "elm" is connected with that of 'beech' for a subject who grasps
both terms, entails the possibility of a distinction between cognitive content
and cognitive value. Indeed, the circularity resulting from an attempt to de-
fine the cognitive value of 'elm' as its cognitive content for the subject of
Putnam's example, is quite obvious. For the cognitive content of 'elm'
would depend on the cognitive value of 'beech' (because it contains, for in-
stance, the information that elms are not beeches), and the cognitive value of
'beech' would depend on the cognitive content of 'elm' (because it contains,
for instance, the information that beeches are not elms). The most natural
strategy to avoid this circularity is to claim that cognitive value must be in-
dividuatedl separately from, though not independently of, cognitive content.
Note that we have criticized only Wettstein's argument for distinguishing
cognitive value from cognitive content, not the distinction itself. However,
there is an interesting difference between Wettstein's conclusion and ours. If
a difference of cognitive value entails, for the rational subject, some
 WHY IS FREGE'S PUZZLE STILL PUZZLING? 157

difference of cognitive content, cognitive value cannot be wholly external to

the mind and irrelevant to semantics -- assuming that cognitive content is "in
the head", in the relevant sense, and plays some semantic role. Indeed, this is
only the first step towards answering Wettstein's challenge; we still have to
show how to individuate cognitive value to explain its mental and semantic
roles. This is why we now turn to the question how neo-Fregeans try to
fulfill this task.

3. FREGEAN SOLUTIONS TO FREGE'S PUZZLE

We should now have an intuitive grasp of what is wrong with most non-
Fregean answers to Frege's Puzzle: it is not enough to point out (semantic
or non-semantic) differences between a trivial use of "a=a" and an informative
use of "a=b" (or between trivial and informative uses of "a=a"); we must also
show how these differences contribute to explaining why the subject is
justified, or has a reason for, an extension of her knowledge in one case but
not in the other. To put the same point in a different way, we should ask
why making two judgments of the form "Fa" and "Ga" may justify the in-
ference to "There is something which is both F and G", while making two
judgments of the form "Fa" and "Gb" may not justify such an inference even
if "a" and "b" refer to the same object.
To deal with these inferences, it is natural to turn to the common element
in the thought that Fa and Ga, i.e., to what Frege called a singular sense or
mode of presentation. Frege's Puzzle stems from the recognition that there
are subject-predicate thoughts, and a first response to the puzzle is to see
cognitive value as determined by the common singular element of a set of
possible thoughts about a given object, and cognitive content as determined
by the different predications involved in a subset of this set -- more precisely,
the subset that corresponds to the thoughts actually judged by the subject.
This picture leaves room for the fact that cognitive value (subject-ori-
ented) is separated from, albeit connected with, cognitive content (predicate-
oriented): even if the subject's conception of a beech is the same as her con-
ception of an elm, she has the capacity to grasp the thought that an elm is F
while a beech is not, for whatever F outside the relevant conception.
As we said at the beginning, descriptive senses cannot be invoked to
solve all versions of Frege's Puzzle. However, we cannot introduce any non-
descriptive senses to deal with the puzzle. To take but one example, it could
be suggested that the senses of some demonstratives in context are not de-
scriptive but are (partly) given by perceptual modes of presentation.
158 EROS CORAZZA & JEROME DOKIC

Perceptual and descriptive modes of presentation have this in common, that

they are both identified by reference to some pieces of information about the
referent (non-conceptual in the first case, conceptual in the second). Now the
advocate of this suggestion has to show that the relation between perceptual
modes of presentation and the referent is not to be understood satisfaction-
ally, i.e., following the model of the relation between a descriptive mode of
presentation and the referent; otherwise perceptual modes of presentation and
descriptive modes of presentation would stand or fall together. Indeed, we
may have the feeling that both kinds of modes of presentation lie on the
predicate-side of thought, and that we still do not know what lies on its sub-
ject-side to tie these pieces of information together and to the referent.
Admittedly, this is very vague, but, for our present purposes, it is enough to
observe that the distinction between descriptive senses and non-descriptive
ones is not really relevant to the whole issue surrounding Frege's Puzzle. l9
How are we to account, then, for the subject-side of singular thoughts? A
possible starting-point is the idea that the distinction between the predicate-
side and the subject-side of some singular thoughts is based on the distinc-
tion between information about an object and the epistemic channel through
which this information is given in thought. Channels of information, in the
epistemic sense, are tied to ways of gaining information about an object. A
non-reflective subject who is in some epistemic relation does not gain in-
formation directly from the channel (and thus indirectly from the object, that
it is so related to her), but directly from the object itself.
What we call "epistemic channels" are close to Russell's relations of ac-
quaintance, and Evans' and McDowell's writings have shown us how to lib-
eralize Russell's Principle of Acquaintance from its strong Russellian restric-
tions, in order to deal with ordinary versions of Frege' s Puzzle.
As McDowell notes, Russell equated his Principle of Acquaintance with
another principle, according to which a subject who grasps the sense of a
singular term must "know which object" the term refers to. This last princi-
ple is what we called "Russell's Principle" in section 2. Here is McDowell
(1990: p. 257) commenting on Russell's equation:

Now in part this equation on Russell's part is merely a reflection of a syntactical

insensitivity, to the difference between "know which object it is that ..." and "know
the object which ...". But we can formulate an idea of Evans's (though not in terms
that he uses himself) by saying that there is a deep point behind the syntactic
conflation. Putting Evans's point in terms of Russell's vocabulary, we can say that
the notion of acquaintance applies to a range of simple cognitive relations between
subject and object the obtaining of anyone of which suffices for satisfaction of the
 WHY IS FREGE'S PUZZLE STILL PUZZLING? 159

"know which" requirement, and thereby suffices for a distinctive variety of singular
thought..

Thus the liberalized Principle of Acquaintance is best interpreted in terms of

non-predicative (or non-propositional) knowledge, that is, it should pertain to
the French "connaitre" rather than to "savoir que".20 A subject who is ac-
quainted with an object "knows" the object, in a non-predicative sense of
"knowing". Acquaintance relations cannot be defined as just having a piece of
information, since they can be identified without mentioning some particular
piece of information or other.
Now as McDowell interprets Russell's equation, it is not the case that
any "information-link" (in Evans' words) counts as an acquaintance relation.
For example, an egocentric perceptual information-link counts as a case of
(demonstrative) acquaintance only "in the presence of a general understanding
of how egocentric placing relates to the objective order", to use McDowell's
(1990: p. 257) phrase. What this means is that only a perceiving subject
who has a conceptual scheme much like ours and is thus able to grasp
concepts of an objective world (including the concept of an objective
location) can use her perception to ground demonstrative thoughts. In its
strongest interpretation, Russell's Principle requires that the subject possess
"identifying" or "discriminating" knowledge of the referent.
Demonstrative thoughts may involve predications of two sorts. There are
predications based on the corresponding information-link, and predications
that are not so based. For example, there are perceptual predications of the
form "That is red" and "That is round", and non-perceptual predications of the
form "That is poisonous" and "That is rare". Now one may suggest that a
reference to the information-link would yield an account of some inferences
relevant to Frege's Puzzle. Consider the following example: a minimally ra-
tional subject who perceives a red tomato and makes the two judgments
"That is red" and "That is round" is immediately in a position to judge that
there is something which is both red and round. According to this sugges-
tion, the subject is justified in drawing the relevant inference because the
predications that are involved in the premisses (using the predicates "red" and
"round") are based on the same information-link. Even if the suggestion
works for perceptual predications (again, only in the presence of the general
understanding mentioned above by McDowell), it is far from obvious to us
that it works for non-perceptual predications, as Evans and McDowell
claim. 21
160 EROS CORAZZA & JEROME DOKIC

Suppose, for instance, that the subject judges the thought normally ex-
pressed by "That is poisonous" while referring to the perceived tomato. Why
is she justified in making from this and other premisses the inference to
"There is something which is red, round and poisonous"? Remember that in
our view, an answer to this question should appeal to an account of the com-
mon element between the thought "That is poisonous" and the thought "That
is red". For that reason, though, we cannot just say that there is such an ele-
ment, nor that the common element consists in the fact that the predicates
involved are based on the same perceptual act -- for the latter is just false in
this case. Evans and McDowell appeal to the "general understanding", that is,
to concepts of an objective order, in order to make sense of such predications.
Consider, though: to conceive of the tomato as an element of the objective
order, and thereby to make sense of the attributions to it not only of the
property of being red and round, but also of the non-perceptual property of
being poisonous, the subject must not only perceive the tomato "in the pres-
ence" of general concepts pertaining to an objective order; she must also be
disposed to suppose that the same tomato falls under these various concepts.
It is not enough for the subject to have the "general understanding"; she must
equally have the capacity to anchor down this understanding to particular
objects. How can she actualize this capacity without already grasping
singular thoughts whose predications involve concepts of the general
understanding? If this capacity can be actualized, Russell's Principle must
already presuppose that the subject has the capacity to grasp sets of thoughts
inferentially connected to each other, i.e., sharing the same singular mode of
presentation.
Russell's Principle requires that a given relation of acquaintance count as
both non-predicative and identifying knowledge. Now to resolve the apparent
tension between these two requirements, it is essential to suppose that the
capacity to anchor the concepts of the general understanding to the acquainted
object may not be actualized. 22 Our point, though, is independent of this
supposition. The fact remains that Russell's Principle cannot fully explain
identity and difference of sense for singular terms, because it presupposes the
notion of sharing the same singular sense. It does not follow that Russell's
Principle cannot be the consequence of the correct account of singular sense.
In our view, it is best seen as a holistic principle on complete, propositional
thoughts, according to which a given thought can be grasped only if others
(in a given set) are. As such, though, the principle plays no direct role in
Frege's Puzzle. Furthermore, it must be separated from the Principle of
Acquaintance, whatever the correct interpretation of the latter. The syntactic
 WHY IS FREGE'S PUZZLE STILL PUZZLING? 161

conflation to which McDowell refers (between "know which object it is that

... " and "knows the object which ... ") may be more puzzling than he recog-
nized.
Evans and McDowell are not the only neo-Fregean writers who counte-
nance what we have called information-free acquaintance relations. For ex-
ample, John Perry (1990: p. 22-3) has recently claimed that there are funda-
mental epistemic relations between a subject and an object which support
"special ways of knowing" the object. Corresponding to each token of such
epistemic relations, there is a repository for information about objects that
Perry calls "buffers". Buffers are only ephemeral repositories, and are the
mental correlates of indexicals. 23 Perry also introduces "notions", which are
more stable files, and are the mental correlates of less indexical singular
terms. Both buffers and notions are on the subject-side of the thought
whereas what Perry calls "ideas" (which incorporate information about the
object) are on the predicate-side.
Even if Perry's account, here grossly simplified, is on the right track, it
is not clear that it can be considered as a complete explanation of the Fregean
phenomenon. Suppose we want to know why the subject, in a given case,
treats two pieces of information as being about the same object. Even if we
concede that there is no problem when the pieces of information are grounded
in the same epistemic relation (are stored in the same buffer), what happens
when they are not so grounded? This question cannot be answered by saying
that the subject treats both pieces of information as being about the same
object because she filed them into the same notion, for we are interested
precisely in the question of why she filed them into the same notion, and
whether she was justified in doing so. In short, we are left with a similar
problem as above. To use Perry's terminology, Frege's Puzzle arises when a
given subject connects pieces of information coming from different buffers or
notions. 24

4. CONCLUSION: THE CRUX OF FREGE' S PUZZLE

If we are right, current Fregean and non-Fregean answers to Frege's Puzzle

miss its main point: in particular, the puzzle is directly concerned neither
with Cartesianism (section 2), nor with the acquaintance interpretation of
Russell's Principle (section 3). So what is Frege's Puzzle really about? We
shall try, in this concluding section, to locate the main point of the puzzle,
which concerns, as we can now see, propositional connection and, therefore,
the very possibility of the (internal) justification of inference.
162 EROS CORAZZA & JEROME DOKIC

First, it is worth noting that there are two aspects of Frege's puzzle, like
two sides of the same coin. Frege asked how it is possible that "a=b" is in-
fonnative, since "a=a" may not be, but he could also have asked how it is
possible that "a=a" is uninformative or trivial, since "a=b" may not be.
These two sides correspond to different puzzling facts. On the one hand, there
is the fact that a subject may make two judgments about the same object, at
a given time, without realizing that they are of the same object. 25 For
example, if these thoughts ascribe, respectively, the properties F and G to
the object, the question arises how the subject would not be justified to infer
(even with only minimal reflection) that there is an object which is both F
and G. On the other hand, though, there is the fact that the subject may make
two judgments ascribing, respectively, the properties F and G to the same
object and immediately be justified in making the further judgment that there
is an object which is both F and G. 26 This latter fact has surely been
neglected in favor of the former, with the following (possibly hannful)
consequelllce: usual resolutions of Frege's Puzzle that concentrate on the
fonner challenge presuppose an answer to the latter, and therefore they are at
best incomplete.
We have claimed that Frege's Puzzle is immediately connected to the
(internal) justification of some typical fonns of inferences. Consider now the
following objection. Arguably, if an inference from p to q is justified for a
rational subject, then the transition from p to q is "primitively obvious" to
her. (The tenn comes from Christopher Peacocke (1987), who has shown
that such a conception is not necessarily psychologistic, but can be part of a
nonnative account of thoughts.) Our claim can be rephrased in tenns of this
notion of primitive obviousness:

(PO) If two thoughts Tl and T2 ascribe, respectively, the properties F

and G to an object 0, and a subject grasps Tl and T2 at the same time,
then the transition from these two thoughts to the third thought "There
exists an x which is both F and G" can be primitively obvious to the
subject because, Tl and T2 share the same singular mode of presentation
of o.

Now it can be objected that this principle reverses the order of explanation,
for the fact that the subject finds some inferences primitively obvious is at
least partly constitutive of grasping a given singular mode of presentation. In
other words, our principle is of the fonn "p because q", when it should be of
the fonn "q because p". There is no logical gap, it is argued, between sharing
 WHY IS FREGE'S PUZZLE STILL PUZZLING? 163

a given mode of presentation and contributing to primitively obvious

inferences, so there is no real problem with Frege's Puzzle as we understand
it.
In fact, the objection does not work because the underlying account refers
to thoughts or concepts with a given content -- as Peacocke (1987: p. 156)
makes clear precisely to avoid the pitfalls of psychologism. When, for ex-
ample, the concept of conjunction is given; it is partly constitutive of its
being grasped that the subject finds the transition from "p & q" to "p" or to
"q" primitively obvious. Similarly, when a singular mode of presentation is
given, it is partly constitutive of its being entertained that the subject finds
the transition from "Fa" and "Ga" to "(x)(Fx & Gx)" primitively obvious,
but this is so because the singular mode of presentation corresponding to "a"
is what it is, and not because of the mere fact that the subject finds the rele-
vant transition primitively obvious. The crux of Frege's Puzzle is precisely
the challenge to explain why the relevant transition is primitively obvious
independently of the predicates involved.
Our discussion of Russell's Principle, in section 3, has shown that the
first step toward an explanation of the (internal) justification of the relevant
inferences is an account of propositional structure. The above principle (PO),
in the same vein, appeals to different thoughts sharing the same mode of
presentation. Briefly, then, our main claim is the following: the lesson of
Frege's Puzzle is that singular terms have two distinct functions. On the one
hand, they have the "vertical" function of referring to an object, i.e., being
semantically anchored to it. On the other hand, singular terms are not mere
Millian tags for an object (this is their "vertical" function), but they also
have the "horizontal" function of contributing to the semantic unification of
possibly many predicates which can be thereby immediately connected for the
subject through the relevant inferences. Usually the stress is put on the
vertical function to the detriment of the horizontal one. However, the de-
scription ·of the "horizontal" function of singular terms requires a sub-
ject/predicate conception of at least some thoughts and a corresponding ac-
count of propositional connection, which has not yet been given in its com-
plete form. Working on these questions should provide the beginning of an
answer to Wettstein's challenge, which sterns from a proper appreciation of
the relations between cognitive value and cognitive content.27 28
164 EROS CORAZZA & JEROME DOKIC

NOTES

1 Two remarks. First, we use "knowledge" in an extended sense, i.e., "S knows that p" does
entail "S believes that p", but not p itself. As we shall see, Frege's Puzzle may be stated in an
"internalist" way which does not depend on the truth-values of the identity statements. Second,
we shall concentrate, in the present paper, on Frege's Puzzle as it concerns singular terms.
We remain neutral on the question of the correct formulation of the puzzle at the level of, for
example, predicates.
2 Indeed, the subject may also extend her knowledge by discovering that she had
contradictory pieces of information about the same thing, and thus by adjusting her doxastic
position. Throughout this paper, however, we shall stick to the simplest case.
3 Pace Salmon (1986b), who considers the premiss that the proposition that a=a is a priori
while the proposition that a=b is a posteriori as a central element in Frege's Puzzle.
4 An identity of the form "a=a" may be informative for a subject who took the name "a" as
referring to different objects on either side of the identity predicate -- think of Pierre in
Kripke's puzzle discovering that Paderewsky (the musician) is Paderewsky (the politician) --,
and an identity of the form "a=b" may be old hat for a subject who took the different names
"a" and "b" to be synonymous -- think of "This bachelor = this unmarried man".
S Does the cognitive value of "a=a" differ from the cognitive value of "b=b"? Considering
these sentences, there seems to be no way, compositionality or not, to deduce that the
cognitive value of "a" and "b" differs. Consider, though, a subject who sincerely assents to
"a=a and b=b". This subject may not assent to "a=b". So, our argument follows: the cognitive
value of "a" differs from the cognitive value of "b".
6 Alternatively, the cognitive content of "a" for a given subject is the conception that he has
of the referent of "a".
7 Cf. Taschek (1992).
8 We think that this is true even if we distinguish the proposition expressed by "a is identical
with a" from the one expressed by "a is identical with itself' (cf. Wiggins (1976) and Salmon
(l986a». (If we reflect on the fact that identity is aformal relation, this point should be even
clearer.) So our use of the term "information content" slightly diverges from that of Salmon
(1986b).
9 This is an oversimplification of Castaneda's rather sophisticated "guise theory". (1989)
10 Following Kripke (1972), the literature on this topic is rather abundant.
11 Remember that we set aside the question of whether such entities enter the semantic
content of thoughts and utterances. So in our view, philosophers as different as Dummett,
Evans, Perry and Salmon have put forward Fregean solutions to Frege's puzzle. Salmon is
Fregean in postulating "guises" to solve epistemological problems tied to a sophisticated
version of F'rege's puzzle, although he does not think that either version requires us to depart
from the "naive" view according to which semantic content is just information content. In this
respect, Dummett and Evans are Fregean in a further sense, since they follow Frege in
claiming that information content cannot exhaust semantic content.
12 In "Is Frege's Puzzle about language?" (forthcoming) we discuss a different (non-
Fregean) answer to Frege's Puzzle, based on a meta-linguistic account. We show, or at least it
seems to us, that basically similar considerations to the ones given in this paper (viz.,
considerations having to do with the (internal) justification of inferences) may be advanced
against the meta-linguistic account.
13 Actually Kaplan argues (1989, p. 604): "Contrary to Russell, I think we succeed in thinking
about things in the world not only through the mental residue of that which we ourselves ex-
perience, but also vicariously, through the symbolic resources that come to us through our
language. It is the latter -- vocabulary power -- that gives us our apprehensive advantage over
the nonlinguistic animals. My dog, being color-blind, cannot entertain the thought that I am
 WHY IS FREGE'S PUZZLE STILL PUZZLING? 165

wearing a red shirt. But my color-blind colleague can entertain even the thought that Aristotle
wore a red shirt." (However, if neo-Fregeans are right, the possibility of exploiting the
symbolic resources of language does not necessarily conflict with Russell's Principle suitably
interpreted in terms of acquaintance). See section 3.
14 The Donnellan-Kripke picture according to which a word is linked to its referent in virtue
of a (social) chain of communication is, of course, consonant with this un-Fregean picture. It
can, however, be made to fit with a neo-Fregean picture, too.
15 There is another theme of Wettstein's dissolution of the puzzle: the rejection of a
"psychologistic" view according to which we use referring signs by associating them with
modes of presentation that are independently grasped. It is arguable, though, that this theme is
not essential to Wettstein's project. We can continue to speak of psychological states
somehow associated with the words we use, without succumbing to the problematic view.
16 Besides the traditional papers by Putnam (1975) and Kripke (1979), see Salmon (1986b),
Wettstein (1989), the eponym paper by Yagisawa (1989), Fodor (1990) and McLaughlin
(1991).
17 If we stick to the contents of the subject's judgments, the difference of cognitive content
between the two terms may be difficult to spot. However, if we vary the force of the subject's
attitudes, so that we turn from judgments to suppositions (for instance), the difference may
emerge more clearly. (This point depends on Frege's thesis about the independence of force
and content.)
18 Salmon (1986b, p.l66, n.l2), who is impressed by something like the Reverse Frege Puzzle,
is aware of the objection that the cognitive content of 'elm' is always different from the
cognitive content of 'beech', but the only objection he can think of is an objection formulated
in meta-linguistic terms: "It may be objected, for instance, that my concept of an elm tree
includes the concept of being called 'elm' in English, and perhaps even the concept of being a
different genus from the things called 'beeches' in English. If so, my concepts of elms and
beeches are significantly different after all." His counter-objection is essentially that
additional information which is supposed to distinguish 'elm' from "beech" cannot be part of
the terms' Sinn, since it is not analytical that elms are called 'elms' in English or, for that
matter, that elms are not beeches. From our point of view, this counter-objection, for what it is
worth, does not take into account the need to distinguish cognitive value from cognitive
content.
19 See McDowell (1984, p. 102): "Evans's rebuttal of Perry's 'Frege on Demonstratives' (see
'Understanding Demonstratives') turns on the lack of any basis in Perry's assumption that
Fregean senses for singular terms must be 'descriptive'. This complaint is quite correct; but it
risks being met with incomprehension as long as the framework which holds the assumption in
place is not challenged." As we shall see below, the complaint may not apply to Perry's later
work.
20 Hence, a neo-Fregean response to Wettstein would be, first, that he fails to recognize a
third case between knowing how and knowing that, i.e., knowing which (interpreted in terms
of non-predicative knowledge), and second, that the sense of singular terms can be
understood following the model of knowing which.
21 Recall Evans' "Generality Constraint" (1982, p. 100-5): "For there must be a capacity
which, when combined with a knowledge of what it is in general for an object to be F, yields
the ability to entertain the thought that a is F, or at least a knowledge of what it is, or would be,
for a to be F". Evans' account presupposes that there is an explanation of how these two
pieces of knowledge, i.e., "what it is for something to be F" and "what it is for something to be
a", can "combine" to form a complex piece of propositional knowledge. The problem we see
is that the latter piece of knowledge, unlike the former, is best reformulated as "what it is for
something to be identical with a". If we keep this formulation, though, we need to mention
another singular term to explain this combination of knowledge, which then presupposes a
166 EROS CORAZZA & JEROME DOKIC

theory about the identity and difference of sense for singular terms. Actually, the crux of
Frege's Puzzle lies in the possibility of such a theory.
22 For example, it is enough to observe that perception can "place a perceived object for the
perceiver" (McDowell (1990, p. 257», and thus count as a case of acquaintance, even if the
perceiver does not explicitly grasp concepts of the object's location (i.e., even if she does not
grasp any of the thoughts "That is there", "That is somewhere in Geneva", etc.).
Nevertheless, it must be possible to gather from perception information about the position and
the identity of the object; otherwise it becomes difficult to count perception as a case of non-
predicative identifying knowledge. In this perspective, the fact that perception is egocentric
must not be considered as a deficiency; actually, it is only on the basis of an egocentric
perception that it is possible to have (perceptual) information about the position of objects (for
example, to distinguish in perception the case where an object is on the left of another from
the qualitatively similar case where it is on its right).
23 According to Perry, buffers, being the indexical components of the thought, are context
sensitive: the "here" -buffer, for instance, links the agent in one context with a given place and
in another context with a different one.
24 Note that Perry mentions elsewhere (1988, pp. 13-15) a problem for theFregean theory of
sense that is close to the difficulties we see exposed by Frege's Puzzle. Perry offers as a
solution to this problem his theory of "mental files" (i.e., buffers or notions), but if we are
right, this theory has to be supplemented if it is to solve the puzzle. In particular, to deal with
Frege's Puzzle it is not enough to note that the subject relates different files, for we need to
know the condition under which this move is justified.
25 In this paper, we have supposed for the sake of argument that the subject draws the
relevant inferences roughly at the same time, thus ignoring problems having to do with what is
now known as "cognitive dynamics". It may be, and it is even plausible, that a complete
response to Frege's Puzzle must appeal to the dynamicity of thoughts.
26 If what was said in the preceding section is right, this fact may be less puzzling if the
predicates "F" and "G" are grounded on the same information-link. When they are not so
~rounded, however, Frege's Puzzle arises in its full force.
7 Note that it is possible to deny the premiss that singular thoughts can be of the form
subject/predicate. In our view, however, this denial would bring about major difficulties for
justifying the relevant inferences. Segal (1989), for instance, denies that singular thoughts are
of the subject/predicate form. His thesis is that they are incomplete, predicative thoughts, to be
accurately expressed by means of open sentences like "x is F". Nevertheless, if our
observations are on the right track, Segal owes us an account of the possibility of unifying
different predicates to form a body of knowledge. The mere repetition of the variable in two
different thoughts, like in "x is F" and "x is G", does not license the subject to infer the
corresponding existential generalization. Indeed, some meaning postulate has to be furnished,
to the effect that the same variable is reiterated whenever there is internal co-reference.
However, we guess that when Segal's account is properly amended, his thesis that singular
thoughts are only predicative is greatly weakened.
28 An earlier version of this paper was presented in Karlovy Vary. We express our gratitude
to Petr Kotatko, who organized the conference and forced us to think about this topic, and to
the participants who gave us their comments. We would also like to thank John Biro and
Giiven Giizeldere for insightful written comments. While working on the penultimate version,
we were visiting the Center for the Study of Language and Information at Stanford University.
Eros Corazza would like to thank the Swiss Fonds National de la Recherche Scientifique for
support (grant No. 8210-037129).
 WHY IS FREGE'S PUZZLE STILL PUZZLING? 167

REFERENCES

Almog, J., 1. Perry and H. Wettstein, (eds.): 1989, Themes from Kaplan, Oxford University
Press, Oxford.
Castaneda, H.N.: 1989, Thinking, Language, and Experience, University of Minnesota Press,
Minneapolis.
Evans, G.: 1981, "Understanding Demonstratives", in Collected Papers, 1985, pp. 291-321.
Evans, G.: 1982, The Varieties of Reference, Clarendon Press, Oxford.
Evans, G.: 1985, Collected Papers, Clarendon Press, Oxford.
Fodor, 1.: 1990, A Theory of Content and Other Essays, MIT Press, Cambridge, Mass.
Frege, G.: 1892, "Sinn und Bedeutung", Zeitschrift fUr Philosophie und Philosophische Kritik,
Vol. 100, pp. 25-50.
Kaplan, D.: 1989, "Afterthoughts", in 1.Almog.et. al. (eds.), pp. 565-614.
Kripke, S.: 1972, "Naming and Necessity", in D. Davidson and G. Harman (eds.), Semantics
of Natural Language, Reidel, Dordrecht, pp. 253-355.
Kripke, S.: 1979, "A Puzzle about Belief', in N. Salmon, and S. Soames (eds.), 1988, pp. 102-
148.
McDowell, J.: 1984, "De Re Senses", in C. Wright (ed.), 1984, pp. 98-109.
McDowell, J.: 1990, "Peacocke and Evans on Demonstrative Content", Mind, Vol. 99, No
394, pp. 255-266.
McLaughlin, B.: 1991, "Belief Individuation and Dretske on Naturalizing Content", in B.
McLaughlin (ed.), Dretske and his Critics, Blackwell, Oxford, pp. 157-179.
Peacocke, C.: 1987, "Understanding Logical Constants: A Realist's Account", The
Proceedings of the British Academy, pp. 153-200.
Perry, J.: 1980, "A Problem About Continued Belief', Pacific Philosophical Quarterly, Vol.
LXI, No.4, pp. 317-332.
Perry, J.: 1988, "Cognitive Significance and New Theories of Reference", Nous, Vol. XXII,
No.1, pp. 1-18.
Perry, J.: 1990, "Self-Notions", Philosophic Issues in Christian Perspective, Vol. 11, pp. 17-31.
Salmon, N.: 1986a, "Reflexivity", Notre Dame Journal of Formal Logic, Vol. XXVII, pp. 401-
29, reprinted in N. Salmon and S. Soames (eds.), 1988, pp. 240-274.
Salmon, N.: 1986b, Frege's Puzzle, MIT Press, Cambridge, Mass.
Salmon, N. and S. Soames (eds.): 1988, Propositions and Attitudes, Oxford University Press,
Oxford.
Segal, G.: 1989, "The Return of the Individual", Mind, Vol. XCVIII, No. 389, pp. 39-57.
Taschek, W.: 1992, "Frege's Puzzle, Sense, and Information Content", Mind, Vol. 101, No.
404, pp. 767-791.
Yagisawa, T.: 1989, "The Reverse Frege Puzzle", in J. Tomberlin (ed.), Philosophical
Perspectives 3, pp. 341-367.
Wettstein, H.: 1988, "Cognitive Significance without Cognitive Content", Mind, 97, pp. 1-28,
reprinted in Has Semantics Rested on a Mistake? and other Essays, 1991, ch. 10.
Wettstein, H.: 1989, "Turning the Table on Frege or How is that 'Hesperus is Hesperus' is
Trivial?", Philosophical Perspectives, Vol. III, pp. 317-339, reprinted in H. Wettstein,
1991, ch. 11.
Wettstein, H.: 1991, Has Semantics Rested on a Mistake? and other Essays, Stanford
University Press, Stanford.
168 EROS CORAZZA & JEROME DOKIC

Wright, C. (ed.): 1984, Frege: Tradition and Influence, B. Blackwell, Oxford.

Wiggins, D.: 1976, "Frege's Problem of the Morning Star and the Evening Star", in M. Schirn
(ed.), Studies on Frege II: Logic and the Philosophy of Language, Bad Canstatt, Stuttgart,
pp.221-255.

Dept. of Philosophy
University of Geneva
CH-1211 Geneva 4, Switzerland

Stanford University
CSLI, Ventura Hall
Stanford, CA 94305-4115
 MARTIN HAHN

THE FREGE PUZZLE ONE MORE TIME

THE PUZZLE

A plausible and common account of the Frege puzzle and the lesson we
ought to draw from it goes as follows 1: there is a theory of meaning, call it
"naive", according to which the role of the subject term of a sentence is to
pick out an object in the world so that the predicate term can ascribe
properties to it. The view has considerable pre-theoretic plausibility, since
one of the most important things we do with language, on any account, is to
describe objects. If this view were correct, however, the contribution of the
singular referring expression to what is said would be limited to the
identification of the object to be discussed, and the following puzzle would
arise:

(1) The morning star = The morning star

and

(2) The evening star = The morning star

have, according to the naive view, the same meaning. If they are construed
as predications, then they both assert of the same object that it has the
property of being The Morning Star; and if as assertions of relation, then
they each pick out the same object twice and assert its self-identity. This is
a puzzle, because while (1) is clearly true and trivial, (2) an interesting fact
of astronomy. They have, in Frege's terminology, different cognitive value
or significance. One sentence expresses, at least to some people, a
surprising fact, the other an analytic (and/or a priori) truth.
Frege's solution is to accept the result of what he regards as the
fundamental test of cognitive value and conclude that 'The morning star' and
'The evening star' differ in meaning. While the reference (Bedeutung) of the
two expressions is the same, the senses (Sinn e) differ. The identity of the
referents accounts for the fact that both the sentences are true, for the
Bedeutung of the whole is a function of the Bedeutungen of the parts. The
difference between the Sinne of 'morning star' and 'evening star' accounts for
the difference in the cognitive value of the sentences because of an analogous

169
J. Biro and P. Kotatko (eds.), Frege: Sense and Reference One Hundred Years Later, 169-183.
© 1995 Kluwer Academic Publishers.
170 MARTIN HAHN

functional law for Sinne. "Meaning", then, comes in stages. A name does
not just mean its referent, as Russell would continue to insist throughout his
career. It picks it out in a certain manner, by means of a mode of
presentation -- its Sinn. The Sinn of an expression is what we have in mind
when we use it or hear it, if we are competent in the language. The referent,
say Mont Blanc with its snowy peaks, is not in any sense in our mind, but
it is determined by what is there -- a Sinn.
This, by now very familiar, story has been augmented by an almost as
familiar recent coda: the theory of direct reference needs to deal with the
puzzle anew because it rejects Frege's solution and reinstitutes the naive
theory. This has led to a reintroduction, in some quarters, of "Fregean
senses", as well as to a series of moves to preserve "Russellian singular
propositions" in others.2 My contention in this paper is that, while it is
certainly true that the Frege puzzle will force direct-reference theorists to
supplement their view to produce some account of cognitive significance,
exactly how and why it will do this has not been understood. If we accepted
direct-reference theory, as opposed to the naive view characterized above, it
would no longer be obvious why the familiar pair of sentences constitute a
puzzle at all. And when we do come to understand the challenge the puzzle
poses, it does not seem to be one that direct-reference theory, as a theory of
linguistic reference, needs to solve. What we need, whether we accept direct-
reference theory as an account of natural language or not, is an account of
intentional content, a theory of the contents of thoughts and how these
determine the thoughts' referents. I leave for another occasion the question
whether the neo-Russellian and neo-Fregean accounts in the literature in fact
provide such an account.
Among the reasons Frege's account has been such a success are
undoubtedly its economy, power and, especially, scope. It provides just as
satisfying a theory of intentionality as it does a semantics, giving a unified
account of representation, linguistic and mental. Frege's view is that he is
giving an account of the common, universal structure and content of
thought. Language is designed to express thoughts; in so far as it does not,
it is flawed. In a perfect language, the meanings of sentences would be
thoughts, composed of the Sinne of the component words. The introduction
of the notion of Sinn solves the problem Brentano posed for intentionality
theory: how it is that thoughts can represent ordinary objects, given that
thoughts are intentional and therefore subject to the vagaries that led
Meinong to postulate a special realm. The answer is that Sinne determine
Bedeutungen, but there is no road back. The answer, in more detail, is
 THE FREGE PUZZLE ONE MORE TIME 171

Frege's brilliant introduction of modern truth-conditional semantics. In other

words, intentionality theory and semantics go hand in hand, they are the
same theory, the theory of thoughts expressible in a perfect language. Sinne
are at once the meanings of the expressions of the perfect language and the
contents of thoughts.
I will call these kinds of views of reference and intentionality 'unified
semanticist theories'. Central to such theories is the contention that the
semantics of linguistic reference and the account of intentionality will be
unified. This does not necessarily mean that there is no distinction between
the contents of thoughts and the meanings of linguistic expressions, for
Frege was clearly very conscious of this distinction3. What it does mean is
that to grasp the meaning of a linguistic expression is to be in an intentional
state whose content captures the semantics of the expression. Furthermore,
even if there are mental states whose contents diverge from linguistic
meanings, the mechanism of reference is of the same kind. One might,
following a long-standing tradition started by Saul Kripke (1980, Lecture 1),
call semanticist theories "descriptivist". But it is far from clear to me that
Frege thinks all referring expressions are definite descriptions, more or less
disguised, and in any case, I want to cast my net wider. The term
'semanticist' both captures the fact that what determines the referent of an
expression or thought is a semantical mechanism and points to the view's
central role in the history of semantics. The idea is that the meaning of a
referring expression is a set of referring conditions, just as the meaning of a
sentence is its truth conditions. What, if anything, an expression refers to
depends on what satisfies (in the standard semantical sense of 'satisfaction')
these conditions in a circumstance of evaluation. This view, as applied to
the vast majority of thought contents and linguistic denoting phrases, is also
shared by Bertrand Russell. 4

DIRECT REFERENCE

According to the standard account of the alleged contrast between Frege's

(and, mutatis mutandis, any unified semanticist) view and direct-reference
theories, a succinct way to state why the naive theory got in trouble with the
puzzle is that it supposed that there was a direct relation (meaning, i.e.,
naming, in this case) between objects and expressions. But the direct-
reference theory seems to constitute just such a revival of the naive theory of
reference. It says that, for certain expressions, the only contribution made by
them to the meaning of a sentence is their referent. Proper names,
172 MARTIN HAHN

demonstratives, indexicals, and perhaps other expressions as well, pick out

their referent directly, i.e., without mediating semantical mechanisms such as
the Frege/Russell ones. 5
Frege's puzzle about identity comes back to haunt us if we accept direct
reference. Let us try it with proper names, assuming them to be just tags 6 .
Take the two sentences:

(3) Joe Lambert =Joe Lambert

(4) Joe Lambert =Karel Lambert

Now, it so happens that I was rather surprised that the person everyone
referred to as 'Joe Lambert' was the same person as the one who has written
numerous works in logic and philosophy of language and teaches at UC
Irvine, Karel Lambert. And it is also true that if accosted by a stranger in
some dark street and asked whether (3) were true, I would take it as fairly
obvious that it was. In fact, this would have been the case even before I
came to UCLA and heard anyone use 'Joe Lambert'. So (3) is trivial (a
priori and/or analytic) and (4) is a surprising fact.
It would seem, then, that direct-reference theory cannot be right. If it
were, then since 'Joe Lambert' and 'Karel Lambert' have the same referent,
they would have the same meaning, and so (3) and (4) would have the same
meaning as well. But obviously they don't, since one can know one without
knowing the other. Hence, the Frege puzzle rises again to defeat the direct-
reference view.
I want to ask, however, whether we have succeeded in setting up the
puzzle. Why did (3) seem obviously true while (4) was a surprise? When I
take (3) to be true, without knowing to whom 'Joe Lambert' refers, I am
making an important assumption -- namely that it refers to the same thing
in its two occurrences in (3). Similarly, my surprise at finding out that (4)
is true is the result of my assuming hitherto that 'Joe Lambert' and 'Karel
Lambert' refer, or at least are likely to refer, to different people.
But why assume that the names do refer to the same individual in (3), and
that they might not in (4)? After all, while it is quite common for a single
individual to have two distinct names, it is probably even more common for
a single name to name several individuals. It would not be surprising if
there were another Joe Lambert somewhere, say a real-estate salesman in
Winnipeg. The story is that I am given (3) without any special knowledge
about the referents as it is supposed to be a priori or analytically true. I
cannot tell just by looking at (3), however, whether it says something
 THE FREGE PUZZLE ONE MORE TIME 173

trivially true about a philosopher or about a real-estate salesmen or

something false about the two of them. It is even possible that it says
something true and surprising about a man with a double life. (Imagine Saul
Kripke's Peter's surprise when he finds that Paderewski is Paderewski.1)
One might claim that normally one would take (3) to be true, and
trivially s08. But why should this be relevant to what the puzzle is supposed
to show? The contention was that (3) can be seen to be true by anyone who
understands the words -- it is supposed to be a priori or analytic. So, to
make the point about normal usage relevant, we would need to show that in
the circumstances where the sentence is being used other than "normally" --
the case where 'Karel Lambert' refers to distinct people in the two
occurrences -- the words are not understood to have their standard meaning.
But that seems a tall order, given that either of the uses of the name is
perfectly standard in other circumstances.
The problem is simply that if names are just tags 9, and their form is not
a guide to their meaning, then there is only prima facie reason to regard (3)
as true and (4) as uncertain. Given the needs of normal conversation, it
would be odd to assert (3) intending the terms to refer to different individuals.
So it is likely that (3) is true. It is not a priori and/or analytic, however. By
contrast, if names pick out their referents by a semantic, conditional,
mechanism (or even conceal descriptive meaning), then (3) is a priori,
because it is analytic. Obviously, the object which has properties x,y,z is
the same object as the one which has those properties, if the definite
description succeeds at all. This is just Frege's solution. But if senses
alone do not determine reference but context is necessary as well, there is no
solution needed, for there is no more reason to take (3) to be trivially true
than (4).
Still, one might ask, if the puzzle cannot even be set up without senses
for names, how did Frege manage to do it? On the Kaplan story, the naive
theory gave rise to the puzzle because singular terms were taken to have only
the function of identifying the referent. Why was the story plausible for
'morning star' and 'evening star' if it cannot even be set up for the
'Lambert's? The answer is that not all tag views are created equal. On the
naive theory, meaning amounts to naming. The connection between a word
and its referent is taken to be the primitive semantic notion, where semantics
has, roughly, the dual function of specifying the truth conditions of
sentences and telling us what it is that a competent speaker understands. On
the view Frege is rejecting, to understand a word is to know what it refers to.
In other words, the naive theory retains the unified semanticist view except
174 MARTIN HAHN

for the particular semantical mechanism postulated. Where the unified

semanticist theory has a conditiona1link, the naive theory puts a direct one.
If we understand (3) and (4), then we should know that the terms all refer
to the same object, and thus should not find (4) any less a priori than (3).
The fact that 'Joe Lambert' may be ambiguous between meaning (i.e.,
referring to) a philosopher and a real-estate salesman is irrelevant, because to
understand a word is to know which meaning is intended in the context at
hand. For cases of ambiguity, this is uncontroversial. In order to make a
judgement whether the sentence 'His junk is garbage' is a trivial truth about
refuse or fighting words among drug dealers, we have to know how it is to
be understood. It is a reasonable demand, if accosted on a dark street and
asked about the triviality of that sentence, that we be told what sense of
'junk' and 'garbage' we are to understand. But once we do know in which
senses the expressions are being used then, if we understand the expressions
in those senses, we will be able to distinguish trivial truth from
controversial assertion. We can do this in the case of (1) and (2), but in that
of (3) and (4) we patently cannot.
On the naive theory, it is an imperfection of natural language that the
word that means the philosopher at Irvine and the one that means the
Winnipeg salesman look and sound the same. Whether we take them to have
senses or not, words in natural language are sometimes ambiguous. In the
cleaned-up logically perfect language that Russell, Frege, (early)
Wittgenstein, et. al. were striving for, context would not be needed to
disambiguate words. So, competent speakers of the perfect language should
be able to find all true identity statements equally trivial, if the naive theory
were true, since to understand a term is to know what its meaning is and the
only meaning of a singular referring expression would be its referent.
Direct-reference theory takes a very different view of context and
ambiguity. First of all, the interest of semanticists has shifted from the
construction of a perfect language to understanding the workings of natural
language. The goal is not to construct "philosophical logic" (Russell's term)
or "conceptual notation" (Frege's) in which mathematical reasoning could be
carried out with precision and assured validity. Rather, we seek an
understanding of reference in natural-language discourse. This difference
would not be significant, if we still believed, as Frege and Russell did, that
the task of semantics is independent of what used to be called pragmatics.
The old picture is that a sentence of natural language, uttered in a specific
context, expresses a proposition which is expressible more accurately by a
Quinean eternal sentence (1960, pp. 191-5). If I say 'It's cold in here', what
 THE FREGE PUZZLE ONE MORE TIME 175

I say can be put more precisely as 'It is (timelessly) cold at such-and-such a

place, at such-and-such time'). The task of semantics, on this view, is to
give an account of these eternal sentences. The connection between them and
sentences of English in a context is studied by the separate, messier,
discipline of pragmatics. This picture of intentionality, semantics and
pragmatics is necessarily rejected by the modem direct-reference view.
Suppose we say that the meaning of the term 'Aristotle' is the famous
Greek philosopher. What is it about the term that connects it to that person?
There is some initial plausibility to the view Russell and Frege seemed to
accept that words are related to their referents in virtue of their linguistic
meaning. But since a directly referential term's only function is to pick out
its referent, there is nothing about 'Aristotle' which connects it to Aristotle,
as opposed to a cat, or a Greek shipping magnate, or nothing at all. On the
naive view, what connects them is a primitive "meaning" relation, and no
explanation is needed. This is extremely implausible, as it postulates a kind
of word-magic. So, since the explanation sought cannot be semantical ex
hypothesi (the "meaning" of a proper name is the referent, nothing else),
there has to be a non-semantical account of these expressions' referential
connections. In other words, direct-reference theory denies not only that the
relation between term and referent is conditionally determined, but that it is
semantical at all. It is not a semanticist theory.
The usual direct-reference account of this relation is in terms of causal
chains 10. What connects a particular occurrence of 'Aristotle' to the Greek
philosopher is that the particular use is part of a chain of uses going back to
the Stagirite. The chain is not part of the "meaning" of 'Aristotle' in any
sense. It is not what the competent speaker understands, and it is not part of
the truth-conditions of the sentence. The connection between the linguistic
item and its referent is contextual, not semantical. Consequently, when two
individuals have the same name, it is at best misleading to talk about
ambiguity. When the same tag happens to attach to two different things, the
referent on a particular occasion is specified by a contextually determined
chain. When an expression is ambiguous, on the other hand, it is best to
regard it as two expressions which happen to be homophonic. 'Bank' and
'bank' have different semantic roles in English. 'Joe Lambert' has only one
semantic role, the same one as every other proper name, i.e., to specify
individuals by means of a contextually determined causal chain. I I Now, in
order to give an account of the truth conditions of a sentence, we need to fix
the referents of its singular terms. But these are determined only given a
particular context. In other words, there is no eternal sentence which
176 MARTIN HAHN

captures the meaning of a natural-language sentence with a proper name in it.

Since this is the case, it is clear that (3) and (4), without a context of
utterance, are equally uncertain, even if it is more likely that (3) would be
used to express a truth than that (4) would. One way to put this point is that
(3) and (4) do not have a determinate cognitive value without particular
contexts of utterance. Thus, Frege's puzzle simply cannot be set up for pairs
of sentences with proper names, once we accept direct-reference theory or any
theory on which their referent is not fully determined by what the competent
speaker grasps. Names, on such theories, simply do not have a determinate
referent without a context of utterance, and so context-less sentences
containing them cannot be classified as either surprising or a priori.

A DEEPER ANALYSIS

Let us consider an objection to my argument. I have shown at most that

Frege's puzzle cannot be set up if we take (3) and (4) to be sentences. But
the puzzle is a puzzle about propositions, or at least interpreted or understood
sentences. So one might argue that it is no objection to it that (3) and (4),
taken out of context and thus lacking a meaning for their names, cannot be
described as trivial and informative, respectively. The Frege puzzle, after all,
sets us the task of distinguishing (3) from (4) in such a way as to explain
why a competent speaker understands (3) to be trivial while interpreting (4)
as informative. If some sentences cannot be interpreted without a context,
then the puzzle needs to be set up for them as occurring in a particular
context.
So, let us take the context which is most favourable to the puzzle, i.e.,
one where one would find (3) trivially true and (4) surprising. This was the
case when I first heard 'Joe Lambert' used. It was quite clear to everyone
concerned that only one person was referred to by that name, and so had
David Kaplan presented (3) in one of his classes, I would readily have agreed
that it is trivially true (though I had at first no idea who this person was).
And, of course, the truth of (4) did surprise me. So, given the right context,
(3) and (4) can be used to set up the first step of the Frege puzzle. But this
is not enough. The puzzle for the naive theory was that according to its
predictions, (3) and (4) ought to be cognitively indistinguishable, yet they
cannot be if one is trivial and the other surprising. But the causal theory of
reference makes no such prediction. With the relevant context, I understand
all the occurrences of 'Joe Lambert' to refer to the same person (whoever he
is), and 'Karel Lambert' to refer to an Irvine philosopher. Obviously, (3)
 THE FREGE PUZZLE ONE MORE TIME 177

and (4) are cognitively very different. In the fonner I understand the names to
refer to the same person, in the latter I only understand each of them to refer
to one person. At best I am agnostic about whether they are the same one.
This is all very well, but does not the direct-reference view also predict
that (3) and (4) nonetheless "mean" the same, have the same truth-
conditions? Indeed it does. There is a clear implication from these two
results: direct-reference theory has to deny what Frege and Russell assert,
namely that there is some single item, the meaning of a tenn, which is what
a competent speaker understands (call it 'competence-meaning') and which is
the term's contribution to the truth-conditions of the sentence in which it
occurs (call this 'truth-conditional meaning'). The modern tag view is thus a
rejection of both the semanticist view and of the naive theory which tries to
hang on to unification. The naive theory made the referent into just such a
"meaning". If we add the plausible (though ultimately false, as Burge
shows 12) assumption that the competent speaker can distinguish between the
meanings of the tenns he understands, we get the clearly false conclusion
that if I know to whom each of two names refers, I know whether or not
they refer to the same thing. Any friend of Dr. Jekyll's would tell you
otherwise. The Frege puzzle brings this error to light.
To fix this problem, we can do one of two things. Frege and Russell
deny that the meaning of a natural-language singular tenn is its referent. The
direct-reference theory is an affirmation of this tag view. The latter,
therefore, has to reject the unified notion of competence and truth-conditional
meaning. If it kept the unification, the wildly implausible result would be
either that we are never competent with directly referential expressions, or
that we make infallible identifications. Russell (1956. pp. 189-216) kept the
naive theory for "genuine proper names" and was driven to the second of
these choices. Genuine proper names apply only to those things we are
acquainted with, i.e., ones about which we cannot make mistakes: present
sense data.
The result, that the notion of "meaning" has to be split in two, is
familiar from David Kaplan's work (1989). He distinguishes the content of a
demonstrative expression such as '1', from it~ character. The fonner is truth-
conditional meaning, and in the case of pure demonstratives (and other
directly referential tenns), it is just the refere~t. Character is the competence-
meaning of demonstratives. The character of '1' is, roughly, to refer to the
speaker or writer of the sentence in which it occurs. That is what the
competent speaker needs to know. The character, together with the context,
detennines the content (i.e., the referent) but is no part of the content. What
178 MARTIN HAHN

the competent speaker needs to grasp in order to understand a sentence with

'1' in it is not the pronoun's contribution to the content (truth conditions) of
the sentence, that is, its referent.
What, then, is the competence-meaning or character of a proper name?
On the purest form of direct-reference theory we have been considering, not
much, obviously. If proper names are tags, the only thing a competent
speaker needs to understand to use a name is that it is a name and that its
referent is determined contextually by means of a causal chain. That is why
the Frege-puzzle cannot be set up with sentences that lack a context of
utterance. Being fully competent with (3) and (4) does not make one
conclude that one must be trivial and the other might be informative.
Let us return to the objection. It can now be put this way. We grant
you that, as far as linguistic competence is concerned, the puzzle cannot be
set up because, on the direct-reference view, (3) and (4) do not have any
content (truth-conditional meaning) without a context. But clearly Frege
meant the puzzle to concern items with truth-conditional meaning. These,
on your view, are sentences-in-context. Let us then run the puzzle that way.
Now, my response was that in any given context, and certainly one where (4)
is surprising and (3) is not, the two are cognitively quite different. I
understand 'Joe Lambert' to refer to a single person at the end of the single
chain terminating in David Kaplan at my end, and 'Karel Lambert' to refer to
some philosopher at the end of quite a different chain. Moreover, nothing in
direct-reference theory predicts that what the speaker understands in such a
context will be the same for the two sentences. All that direct-reference
theory predicts is that the truth-conditional meanings (contents) of the two
sentences will be identical. Hence, as I said, it follows that a distinction
needs to be made between the two senses of 'meaning' .
Direct-reference theory does distinguish between the truth-conditional
meaning (content) and the competence-meaning (character) of expressions.
But this distinction will be of no help to us here. What is needed is a notion
of "meaning" on which (3) and (4) differ even though they have identical
truth-conditional meanings. In the case at hand, however, the characters as
well as the contents of the two expressions are identical. The character is the
conventional linguistic meaning that the competent speaker grasps. It stays
the same from context to context. On the radical view that names are
nothing but tags, which has been our running example throughout, the
character of all proper names is identical. And on less radical views, it is not
obvious that one will be able to argue successfully that 'Joe Lambert' and
'Karel Lambert' have different characters. But suppose that such an argument
 THE FREGE PUZZLE ONE MORE TIME 179

can be mounted and it turns out that an orthographic difference between two
names, that is, a difference in the ordered sets of letters composing the
names, ipso facto yields a difference in character. Then we still have the case
of Peter who, in the different contexts, takes 'Paderewski = Paderewski' to be
trivial (as asserting the self-identity of a pianist or a politician), or surprising
(as asserting the identity of the pianist with the politician)13. Character, or
competence-meaning, will not provide the relevant difference between (3) and
(4).14
There must be yet a third sense of 'meaning' to be distinguished, if direct-
reference theory is to be saved. I shall call it 'intentional content'. Russell
and Frege did not distinguish between intentional content and meaning in
either of our two other senses. The meaning of a sentence was to them what
the competent speaker knew, the thought he had when he understood the
sentence and the truth-conditions as well. We have distinguished between
what the competent speaker understands and truth-conditions. But the former
is ambiguous between: 'what one needs to know in order to use the sentence
competently' and 'the content one has in mind when one uses it'. So, to use
Frege's own example (1968), the competence-meaning of 'I' is its character
and it is identical for all normal uses of the word in English, making it
possible to translate 'I' into either 'je' or 'moi' in French, etc. It is the
general rule that the referent is the producer of the utterance. The truth-
conditional meaning of 'I' varies from context of utterance to context of
utterance and can be identical to the truth-conditional meaning of 'he' (in a
different context) or 'Alain Chapel', etc. It is the referent itself, according to
direct-reference theory. The intentional content of the first-person thought I
have when I say 'I am hungry' is yet a third thing, and Frege might well be
right that no two people have identical thoughts when they use '1'.
Similarly for proper names. The character of 'Paderewski' is whatever
conventional linguistic meaning all competent uses of the name have in
common. I have assumed, for the purposes of argument, that all proper
names have identical competence-meaning, but that is not a necessary part of
the direct-reference view. The content, once again, is just the referent
determined by the character along with the context of utterance. But the
intentional content is yet a third thing. Where most of us have but one
thought which we might express by 'Paderewski is a great pianist', Peter has
two. One he takes to be false, the other true.
The unified semanticist theory identifies mental contents with the
meanings of linguistic expressions (at least in a perfect language) and then
gives the same answer to the central questions of intentionality and of
180 MARTIN HAHN

reference. Mental states and linguistic expressions pick out their objects by
specifying a set of conditions an object must satisfy in order to be the
referent. The elegance of the solution is evident, the success well
documented. The essence of the direct-reference view is to reject the
Frege/Russell unified semanticist solution for linguistic reference. Linguistic
(competence) meaning is distinguished from what determines reference (the
context along with competence-meaning), which in turn is distinct from the
contribution to truth-conditions the singular term makes (it supplies the
referent itself). The rejection of semanticism thus leads to the rejection of
part of the unified picture: where a Fregean Sinn performed three functions,
we now have three separate accounts.
But, if I am right, it is no part of the essence of direct-reference theory to
keep the other parts of the unified view. In fact, the Frege puzzle shows that
such unification must be rejected. If we try to keep the identification of
cognitive significance and truth-conditional meaning, we get into trouble, as
the last version of the puzzle shows. In the jargon of many direct-reference
theorists, we are forced to the conclusion that propositions (truth-conditional
meanings) cannot be identical to thought-contents. It seems that thought
contents are sometimes individuated more finely than propositions 15 . In the
case at hand, while (3) and (4) express the same proposition, they have
obviously different cognitive values.
The lessons of the Frege puzzle for the direct-reference view thus come in
two stages. If we try to set up the puzzle as being about sentences, we are
forced to recognize that competence-meaning and truth-conditional meaning
cannot be identified. But this is a result that David Kaplan has recognized
and written about extensively. The direct-reference view is a two-stage view
of meaning, truth-conditional meaning only emerges at the second stage with
input from the context of utterance. If, on the other hand, we try to set up
the puzzle at the level where there is truth-conditional meaning, i.e., as
involving sentences in a particular context, we are forced to distinguish
between truth-conditional meaning and intentional content. This distinction
is more controversial, partly because it has been obscured by the blunt term
'proposition' in the literature 16 . Nonetheless, I believe direct-reference
theorists must make it.
Once the distinction is drawn, however, we see that direct-reference theory
has much less to deal with than is often thought. Semantics has the twin
tasks of providing the truth-conditions of all the sentences of a language and,
for natural languages, explaining what it is that a competent speaker knows.
In other words, direct-reference theory has the task of accounting for
 THE FREGE PUZZLE ONE MORE TIME 181

competence-meaning (the character of referring expressions) and truth-

conditional meaning (content). Once direct-reference theory rejects the unified
semanticist view, as any theory that takes context to be essential to
reference-determination must, if I am right, we see not only that the two
tasks of semantics are separate, but also that the further task of accounting
for intentional content is, at least in principle, independent of the two
semantical tasks. 17

NOTES

The account is basically due to Kaplan and has been widely accepted by direct-reference
theorists and their critics. It figures, for example, in Salmon (1986). The account, as I present
it, uses 'meaning' in a pre-theoretical sense, not making many distinction I will want to make,
and some that Frege himself was keenly aware of.
2 The neo-Fregeans are represented prominently by Evans, McDowell, Peacocke and others,
the neo-Russellians by Kaplan, Salmon, Perry and others. Recanati (1993) has argued that the
differences between the two approaches are not as significant as it might seem.
3 Cf. Burge (1979a, 1990) for two arguments I find quite convincing on this matter.
4 Russell wants to maintain something like the naive theory for logically proper names, of
course. But this does not alter the fact that he holds the unified semanticist view for the vast
majority of denoting phrases and corresponding thoughts.
5 The three classic places to find arguments for the view are in Donnellan (1970), Kaplan
~1989), and Kripke (1980). I shall not commit myself for or against the view.
I will run the argument of this paper with proper names, taking their function to be solely to
introduce the referent into the truth-conditions of the sentence in which they occur. This
involves two simplifications. One is that direct-reference theory has prima facie difficulties
with the Frege puzzle across the board, not just for names. The arguments I am making could
be made, mutatis mutandis for examples such as 'This = this' and 'That is this'. Indeed, the
argument of the present section would likely encounter less resistance with such an example.
The second one is that proper names may, consistently with direct-reference theory, have
richer linguistic meanings than the sparse tag view suggests. More on t:tis below.
7 I am referring to an example found in Kripke. (1979, p. 265) Peter believes that
Paderewski, the pianist, had musical talent but that Paderewski, the Polish statesman, had none.
8 I doubt that this is true. "Normally" we do not utter analytic sentences any more than we
"normally" use the same name to refer to two distinct individuals in the same seMence. As far
as normal usage goes, we have a stand-off here.
9 Or, for that matter, if names have an indexical as well as a conceptual element, as they do
on, say, Burge's view. The point is just that if names have any token-reflexive or contextual or
demonstrative component to their reference-determ~nation, then mere competence with the
name does not suffice for knowing whether N=N is true on a particular occasion of use.
10 Actually, neither Donnellan nor Kripke are particiularly happy with the characterization of
these chains as 'causal'. The idea behind the charact~rization is, I believe, precisely to signal
a contrast with 'semantical' or any other a priori G:onnection between name and referent.
What the determining chains of uses are, precisely, ~nd whether they are ultimately causal or
not, fortunately need not concern us here.
11 This, of course, is an extreme version of the direct-reference account. There is nothing
to preclude a much richer conception of the linguistic meaning of proper names so that
different names have different meanings. But however much detail one adds, the contextual
182 MARTIN HAHN

element will remain -- it is the very essence of a direct-reference account. To see this, we
can think about demonstratives. 'He' and 'she' are clearly different in meaning, but that will
not capture the difference in the truth conditions of the sentence 'He's one of the leading
chefs of this century' when uttered in a context where Ronald Reagan is the referent, on the
one hand, and where it is Alain Chapel, on the other.
12 (l979b) and (1982). I should add here that the relationship between Burge's points and the
sort of "exterualism" advocated by direct-reference theory is much more tenuous than is often
sfPosed.
1 The avenue of insisting that 'Paderewski' is, in fact, two names here with two different
linguistic competence-meanings strikes me as deeply unpromising. I will not argue for my
view here, except to note that such a move would lead to the fragmentation on every name
into a (potentially) infinite number of competence-meanings -- one for every fact (or rumour)
about the person named. ("Oh, so the Martin Hahn who took the bus yesterday is the same
Martin Hahn who promised to get his grades in on time!?").
14 At this point I might wish I had chosen to run the entire argument with demonstratives. No
one, I take it, would want to argue that every occurrence of 'this' entails a different
competence-meaning and so that the relevant difference between 'This = this' understood as
trivial and as informative is to be found in the characters of 'this' and 'this'.
15 It is equally easy to find cases where intentional content is individuated more coarsely than
propositions. Take the case where someone thinks that Fred Smith is identical to Fred Smith
when, in fact, there are two people with the same name that are talked about in the various
contexts in which the person hears the name. Then the single thought "Fred Smith makes a
mean souffle" will point indiscriminately to two different truth-functional meanings, mirroring
the way a single proposition such as "Paderewski is a great musician" is linked to two thoughts
in Kripke's Peter (1979).
16 Of course, there are some arguments for such an identification. These often proceed
from some facts about propositional-attitude ascriptions. I believe such arguments fail. I
discuss them, and related issues, in (1993).
17 Just how independent the two projects are is a very interesting question. Clearly, one
cannot have a semantical theory on which competence is psychologically impossible. But it is
not at all clear just what this means for the account of intentional content. It may even be
possible to have a fully descriptivist account of intentional content along with a direct account
of semanticaJ. reference. So, for example, the semantics of the first-person pronoun could be
as direct-reference theory states: it has the linguistic function of referring to the
speaker/writer, and the truth-conditional contribution it makes is just the referent. But it may
tum out that all my thoughts about myself are fully conceptualized. Now, I happen to think
that this account is false and that first-person thoughts are essentially indexical or de re, but I
do not think that direct-reference theory drives us to this conclusion. The arguments for it are
related but separate. (Cf. Hahn 1993) The point of the present paper is that, even if
intentional and semantical content both essentially have indexical elements, we ought not to
conclude that the two will get identical accounts. They may, indeed, will, differ both in the
individuation of the indexical elements (so that, for example, two intentional "names" might
correspond to the single 'Paderewski') and in how the referents are determined (so that, for
example, while the intentional referent of my particular use of 'this' will be perceptually
determined, the linguistic referent may well be determined by conventional rules of
ostension).
An interesting case of the possibility of radical divergence between intentional content
and thought ought to be familiar from the work of Kripke. In "Semantic Reference and
Speaker's Reference", he imagines what he calls a 'strong Russell language' in which all
definite descriptions work as Russell analyses them in "On Denoting". In that language, he
claims, people would still use descriptions referentially. His own reasons for saying this are, I
 THE FREGE PUZZLE ONE MORE TIME 183

believe, not relevant to the Donnellan's distinction or de re thoughts -- he points to the fact that
people would still make mistakes. Nonetheless, he is right. If our language contained no
indexical expressions at all, no names and no demonstratives (this is the strong Quine
language, actually), we would still have de re intentional contents.

Simon Fraser University

LB 7400
Burnaby Be V5A 156
Canada
 JOHN BIRO

THE NEO-FREGEAN ARGUMENT

Is a name a mere meaningless tag, whose function is simply to denote its

bearer, as Mill maintained? Or does it, as Frege thought, have, in addition to
its reference, a sense, one its user can associate with it, understand, or
"grasp"? An argument in favour of the second, Fregean, alternative many
philosophers have found well-nigh irresistible is one that takes as its starting
point the fact that co-denoting names are apparently not always substitutable
salva veritate in sentences ascribing propositional attitudes. Only the
postulation of senses, this argument -- which I shall call neo-Fregean --
goes on to claim, can explain this: only a difference in the senses of the
names could account for (what is seen as) their different semantic
contributions to the sentences in which they occur. 1
Many writers endorse, or at least flirt with, some version of the neo-
Fregean argument. 2 There are differences in the way the argument is formu-
lated, as well as in the precise nature of the conclusion drawn. In particular,
some regard the semantic property whose existence the argument putatively
proves as equivalent to, and expressible by, a (set of) description(s), others
think that it is sui generis and something we can get at only indirectly.
Some are content to use Frege's term 'sense,' others speak of "non-
descriptive connotations" (Ackerman) or of the names' expressing an
"individual essence" (Plantinga). These differences are interesting and
important, but they will not figure in the present discussion. What I want to
do is to ask whether arguments of this sort do not face some quite general
difficulties, regardless of the particular formulation the anti-Mill ian
conclusion receives.

II

I should begin by giving a sketch of what I take to be the most general form
of arguments of this neo-Fregean kind. Their target is a view commonly
attributed to Mill (hence 'Millianism'), according to which names have only
one semantic property or function: to stand:for -- stand in for -- their bearers.
As Mill himself put it, names are "mere m¢aningless tags": they are " ... not
connotative; they denote the individuals wijo are called by them, but they do
not indicate or imply an attribute as belongtng to those individuals."3

185
J. Biro and P. Kotatko (eds.), Frege: Sense and Reference One Hundred Years Later, 185-206.
© 1995 Kluwer Academic Publishers.
186 JOHN BIRO

But this picture is at odds with what some see as unchallengeable facts
about our use of names in ascriptions of propositional attitudes. It seems
that it is possible -- indeed, not uncommon -- for someone to believe that a
is F and not to believe that b is F, even though a is b and thus the names 'a'
and 'b' denote the same thing.4 After all, it is undeniable that people
sometimes assent to a sentence of the form 'Fa' but not to one of the form
'Fb' when 'a' and 'b' name the same thing; and their doing so is, on the face
of it, evidence that they have beliefs whose content can be specified by using
these very sentences in our attributions. However, if the Millian view were
the right one, and the semantics of a name were exhausted by its denotation,
there would be only one thing expressed by two sentences in which the same
predicate was attached to 'a' and to 'b', respectively. Since it is difficult to
swallow the idea that it is common for people both to believe and not to
believe the same thing, we must, it is suggested by the neo-Fregean,
conclude that the two sentences must not, after all, be saying the same thing.
That people accept one and not the other, then, is neither surprising nor
problematic 5.
Note that if we want to argue in this way, we must take the fact that
someone exhibits a certain pattern of verbal responses -- assenting to and
dissenting from sentences -- as evidence that he accepts or rejects what they
express. Doing so makes sense, of course, only on the assumption that he
understands the sentences in question, that is, knows what they express.
Given this assumption, we must take his differing responses to different
sentences as evidence that these express different things. But if there is to be
a difference in what the two sentences express, so that one can rationally as-
sent to one and not to the other, we have to ask what that difference can
consist in. What other candidate is there, the neo-Fregean asks, but a differ-
ence in the senses of the names in the respective sentences, given that by
hypothesis both the names' reference, and everything else about the two sen-
tences, is the same?
I emphasize the role of verbal behaviour in order to highlight the fact that
it is the ground-level phenomenon from which the neo-Fregean argument
must start and that thus the so-called "phenomenon" of non-substitutivity is
already a result of interpretation based on non-trivial theoretical
commitments about the evidential relation between the verbal behaviour we
observe and the beliefs we attribute. We will see that a failure to appreciate
this can lead to confusion about what one's account needs to explain and thus
about what sort of fact one needs to aver to if one is to explain it.
 THE NEO-FREGEAN ARGUMENT 187

The way to avoid the unwelcome consequence mentioned above, of at-

tributing contradictory beliefs to rational subjects, is, according to the neo-
Fregean, to recognize that there are, in fact, two things expressed by the two
sentences (let us call them two propositions). There must be two propo-
sitions, for no one proposition can have two incompatible properties such as
being -- even possibly -- believed and being -- even possibly -- not believed
(by one person at one time). If there are two propositions, the names must
have, in addition to possessing a reference or denotation, some other
semantic property that is responsible for the difference in the propositions
expressed. (As just noted, by hypothesis, they do not differ in reference and
there are no other differences of any kind between the two sentences in which
they occur). It is natural to think of this additional semantic property of a
name as the property of having (or expressing) a sense (or whatever we
choose to call the property of its carrying some information, in virtue of
which property it names the thing it names), and to think of the name as
contributing that sense to the Fregean sense or thought expressed by a sen-
tence in which the name occurs.
Having distinguished the propositions expressed by two sentences differ-
ing only in containing the co-denoting names 'a' and 'b', we can, according
to the neo-Fregean, now explain our originally puzzling phenomenon in a
perfectly straightforward way. The recognition that the objects of the differ-
ing attitudes are different propositions removes the feeling of paradox: of
course one can believe that p without believing that q, when p and q are
different! The price of this natural explanation is accepting a proliferation of
semantic properties, but it is a price we must be prepared to pay. So, names
must have senses, and 'a' and 'b' must have different senses, if 'Fa' and 'Fb'
are to express different propositions. When 'Fa' and 'Fb' are then used to
specify the contents of attitudes we are attributing, they will specify different
contents: the joint truth of'S believes that Fa' and'S does not believe that
Fb' remains quite un-mysterious even if a is identical with b, and even if we
know this. 6

III

A question one may be inclined to raise about the argument right away is,
just what makes it the case, according to the proponent of the neo-Fregean
argument, that there are two propositions expressed by the two otherwise
identical sentences in which different but co-denoting names occur? Whatever
propositions are, it is reasonable to think that they have their truth-
188 JOHN BIRO

conditions essentially. Having such-and-such a truth-condition is an intrinsic

property of a proposition. By contrast, relational properties are, in general,
best seen as extrinsic, and it is not clear why they should be seen differently
in this case. Why should we regard relational properties, such as being
assented to, or being believed by, someone, as an intrinsic, or essential, or
individuating property of a proposition?7 But suppose we did: is it not then
puzzling to find the neo-Fregean claiming that what must constitute the
difference in propositions is a difference in the senses contributed by the
names that occur in the sentences (or clauses) used to express them? Surely,
one might protest, he has already offered us one explanation of what makes
the propositions different: a difference in their relational properties (that is, in
one's being believed and the other's not being believed). Either the fact that
there are (]possibly) different attitudes is sufficient for there being different
propositions, or it is not. If it is, we need not advert to any other. If it is
not, we need not accept the claim that there are different propositions merely
on the basis of a (possible) difference in attitudes. But then it is difficult to
see what could support the next step in the neo-Fregean argument, for that
step is to infer a difference in the senses of the names from a difference in the
propositions expressed by the sentences in which they occur.
There is, of course, a ready answer to this objection. Even if it were be-
ing the object of different attitudes that made the two propositions different,
that difference would entail a difference in -- and thus, of course, the existence
of -- senses. What the appeal to senses is supposed to explain is the fact that
there can be different attitudes in such cases, by way of explaining how the
holder of the attitudes can understand the sentences with which he is
confronted as expressing different propositions. It is his having different atti-
tudes to the propositions that makes them different, but it is because he un-
derstands them to be different that he has those attitudes. He could understand
the propositions expressed to different only if they were different in some
respect; since, by hypothesis, everything about the sentences, including the
reference of the names, is the same, that difference must lie in some other
semantic property the names have: their sense.
The first thing that should strike one as a bit odd about this reply is that
it is by no means obvious that one can think that two sentences express dif-
ferent propositions only if they in fact do so. What is there about this par-
ticular kind of case that precludes one's making a mistake and thinking that
there are two propositions where there is only one? Granted, to make such a
mistake, one must think two propositions. But it does not follow that of the
two sentences involved, one expresses one of these, the other, the other, even
 THE NEO-FREGEAN ARGUMENT 189

if the thinker so believes. However, as I have already argued, the Millian

need not go along with the assumption that someone's responses to
sentences provide grounds for an immediate inference to what he believes.
Hence he need not be drawn into an argument over whether someone's be-
lieving that two sentences express two different propositions does or does not
entail that they do.

IV

The main question I want to discuss in this paper is whether what I have
characterized as the ground-level phenomenon -- a person's responding differ-
ently to two sentences differing only in containing different but co-denoting
names -- is best explained by the postulation of senses for the names.
The fact that people sometimes display a pattern of responses of this sort
certainly needs an explanation. Would the fact, if it were one, that names
have senses provide one? Not in and of itself, it seems. If we are to have an
explanation, we must find something that can be seen as causing the differ-
ence, when there is one, in our subject's responses to the two sentences con-
fronting him. This can only be something in the subject's mind, something
to do with his grasp or understanding of the putative difference in sense.
This, we just noted, was the thought behind the neo-Fregean's own proffered
explanation: he wanted there to be different things the subject could be
thought to understand in understanding the two sentences. But even if we
conceded that the propositions expressed by the sentences are different, that
they are different in virtue of the embedded sentences' having different mean-
ings, and that these sentences have different meanings in virtue of a difference
in the senses of the names occurring in them -- even if we accepted all this,
we would still not have said enough to explain why our subject has the
attitudes he has. We would have said something that makes it possible for
him to have those different attitudes, true enough. And, as we will see later,
the central question in judging the success of the neo-Fregean argument is
whether it offers the only account of this ,possibility. But my present point
is that the mere existence of different sen~es would not, by itself, suffice as
an explanation of the difference in respon!ses, when it shows up, unless we
thought of the responder as believing that there is such a difference. His be-
lief need not, of course, be explicit. But it is one he must have, or we have
no explanation of his responses even if the senses do differ, and even if we,
the attributors, believe that they do. With senses, no less than with denota-
tions, a person using the expressions whqse denotations or senses they are
190 JOHN BIRO

may have an incomplete or erroneOus understanding of what they are. In par-

ticular, the user of two expressions may not know whether they share a de-
notation or a sense. (All this holds equally of extension of common nouns.
It is also no less true of mental representations than of expressions in a pub-
lic language. I shall return to both these points below.)8
While the neo-Fregean argument requires only the possibility of some-
one's exhibiting different responses to the two sentences, rather than actually
doing so, this does not alter the point. Our interest is in explaining what
facts make it possible for a particular person to do so when he does. That
possibility still requires the possibility of that person's thinking that there
are different propositions, something he may fail to do even when there in
fact are. He may· fail to recognize the difference in sense even with expres-
sions with respect to which no-one would dispute that there is one. He may
believe that 'red pepper' and 'paprika' have the same sense (and thus the same
extension) and even put chili pepper in his paprikas or paprika in his chili.
And so, too, with names, should they turnout to have senses. I might mis-
takenly believe that two names have the same sense when in fact they do
not. If I did, it would not be possible for me to have different attitudes when
confronted by two otherwise identical sentences containing the names. 9 The
fact that it is still possible for someone (else) to have different attitudes when
confronted by the same sentences only reinforces the point I am making: that
to explain why he and I differ, one must appeal to a psychological difference
between uS,not (merely) to a semantic difference between the expressions
involved~ One must pitch one's explanation at a psychological level,
between whatever objects and properties one's semantic theory countenances
and the propositional attitudes to be explained, in order to have something
with a fine enough grain to do the explaining.
So, what we need, if we are to have an explanation of our basic phe-
nomenon, is not only a difference in semantic properties (the senses of the
names), even if there is one, but a difference in psychological properties (the
understanding or grasping of these senses by particular users of the names).
But then we should ask, do we need a difference in semantic properties at all?
We must, of course, concede the obvious point that if there are to be
graspings of senses, there must be senses to be grasped. And it may seem
that this is enough for the neo-Fregean to secure his conclusion; the expla-
nation of (the possibility of) different responses by an appeal to the thinker's
psychology still seems to point to a conclusion about the semantics of
names. However, the appearance is misleading and the conclusion does not
follow. Before showing this, though, it may be useful to spend a little time
 THE NEO-FREGEAN ARGUMENT 191

on the question, what kind of property is the neo-Fregean's putative semantic

property?

v
The question need not, for our purposes, be construed as one about the
ontological nature and status of these putative properties. That there are dif-
ficulties in that direction, as well as in the direction of accounting for our
contact with them, is well known. 10 But here what matters is only what
senses, whatever they are, are supposed to do in a semantic theory. Thus
even if we would, perhaps, like to have a clearer characterization them than
some neo-Fregeans give (such as Ackerman's "non-descriptive connotations"
or Plantinga's "individual essences"), the question I want to ask is the more
modest one of what, from the point of view of semantic theory, does it mean
to say that an expression has a sense. What does a sense, whatever it is, do?
How does it earn its keep in a theory of language and language use?
A natural thing to say is that for an expression of some language (public
or, if such are possible, private) to have a sense is for it to carry some in-
formation about whatever it is that it (purportedly) denotes; perhaps, also, it
is in virtue of carrying that information that it denotes what it does. The
property of carrying such information is a property of the expression, and the
fact that an expression has that property is a fact about the language of which
it is an expression -- even if that language has only one speaker. Such facts
about languages should be distinguished from facts about the speakers of
languages, including from psychological facts such as those involved in their
knowledge of their language generally and, in particular, in their beliefs
about the semantics of the expressions of the language, whether the latter are
beliefs about denotation or sense.
A sense, even so distinguished from properties of speakers, should not
be thought of as a property of an expression. A sense (or a body of informa-
tion) can no more be a property of a linguistic item that can a referent. (The
sense of the name 'Venus' is no more a property of it than is the planet it
denotes.) Having a sense could be a property of expressions, as can having a
(certain) reference, and we can ask what having such a property would consist
in. It seems that it must consist in some potential of the expressions to be
understood in a certain way. Thus we are back to the point made in a different
way in the last section: even if two expressions have the potential to be
understood in different ways, their having that potential neither determines
nor explains how they are understood by a particular speaker. That a person
192 JOHN BIRO

understands an expression in one way or another (when there are, in the

language in question, different ways in which the expression may be un-
derstood) is a psychological fact about that person. And we have seen that it
is such facts that must be invoked to explain the patterns of response the
neo-Fregean claims require, and receive, an explanation through the postula-
tion of senses.
Why is it necessary to insist on these distinctions, in themselves rather
obvious, even, one might say, platitudinous? The reason is that much of the
appeal of the neo-Fregean argument depends on a conflation of three very
different things: there being a (certain) sense (possessed by, or associated
with, an expression), the expression's (having the property) of having that
sense, and a person's (having the property) of attaching a sense to an
expression. Keeping these distinctions in mind, and keeping our focus on the
right explanandum -- the fact that a person exhibits patterns of response
different from the expected, will enable us to zoom in on the question I think
is central to the dispute between the neo-Fregean and the Millian.ll

VI

Why should we assume, with the neo-Fregean, that the psychological fact
that explains how someone can have different responses to the sentences
involved in these "puzzle" cases, the grasping or understanding we have seen
was necessary to have an explanation of these responses, must be a grasping
of senses?
A defender of Millianism may have more room to manoeuvre here than
his opponent allows. He can offer his own Millian explanation of the phe-
nomenon, one in terms of facts about the responder's beliefs about the deno-
tations of the names involved. Such an explanation would say simply that
someone with divergent responses lacks the belief that the two names in
question are co-denoting. It is, if you like, his meta-linguistic beliefs, his
beliefs about the reference of the names, that explain why he responds the
way he does (and appears, thus, to have different attitudes) in spite of the
names' actual co-referentiality.
Meta-linguistic explanations of the sort of phenomenon on which the
neo-Fregean argument is based are often viewed with suspicion. In the first
place, it is insisted, the phenomenon is not essentially linguistic, as is
shown by both the possibility of describing it without mentioning any lin-
guistic expressions and the intuitive plausibility of its extending to non-lin-
guistic creatures. Second, it is alleged that on any meta-linguistic account,
 THE NEO-FREGEAN ARGUMENT 193

the contents of the attitudes attributed will come out as differing from the
intuitively correct ones.
On the first score, the neo-Fregean may complain that it is a mis-repre-
sentation of his argument to focus on sentence pairs of the form 'Fa' and
'Fb' and responses to them, when it is more properly couched in terms of the
possible truth of attributions of the form'S believes that Fa' and'S does not
believe that Fb'. Put this way, there is nothing obviously linguistic
involved. In making such attributions, we are neither talking about names
ourselves nor attributing beliefs about names to our subject. (We are, of
course, using names in making true and, hence, on a Millian account of
names, paradoxical, attributions. But that does not imply that the beliefs we
are attributing have anything to do with names.) When the phenomenon is
described in this way, the possession of a language seems inessential to it:
such attributions could, on the face of it, be true of infants and animals.
Hence a meta-linguistic account cannot be right.
Three things may be said about this first objection. In the first place, it
ignores the fact, insisted upon earlier, that attributions of attitude already
amount to an interpretation of the observed behaviour which is the ground-
level.phenomenon in which the argument must be rooted. The neo-Fregean's
argument for his conclusion about the semantics of names will work only if
it is the only -- or, at least, the best -- explanation of that phenomenon. So,
even if it looks as if the neo-Fregean argument can be stated in a non-lin-
guistic way, that is only an appearance. Second, the Millian account is ar-
guably no more meta-linguistic than the neo-Fregean's own. For the latter,
what was doing the explanatory work was the subject's "attaching" different
senses to the names, "grasping" the names's different senses, or "understand-
ing" the names as different. It is not at all clear what is meant by these locu-
tions. A simple way of understanding them would be to see them as being
ways of talking about the subject's beliefs about what senses are associated
with the names. Or, if this is thought too strong, one may think that these
attachings, graspings, and understandings at least entail the subject's having
such beliefs. If so, we have something as,meta-linguistic with these as with
the beliefs about the names' referents to which I have suggested the Millian
can appeal. Third, while the Millian account naturally focuses on linguistic
expressions -- after all, it is about whether names have senses that we have
been arguing -- it is readily extended to the non-linguistic cases the neo-
Fregean is concerned about. What I have been calling the ground-level phe-
nomenon may take the form of behaviour involving differential responses to
non-linguistic -- say, perceptual -- representations. Such behaviour would
194 JOHN BIRO

require interpretation before it could be seen as puzzling, but, as we have

seen, this is no less the case with linguistic behaviour. The only difference is
that the evidence for the interpretation is different: linguistic behaviour in the
one case and non-linguistic behaviour in the other. While this difference does
require a difference in methods of interpretation, it is not seem to be a
fundamental difference vis-a-vis the issues at stake between the Millian and
the neo-Fregean. Thus the Millian's account of the phenomenon involving
names should be seen as a special case of what might be called a meta-
representational account of similar phenomena involving representations of
all kinds.
I have just urged that if there are sound objections to any account that in-
corporates meta-representational elements, the neo-Fregean's own will be as
vulnerable to them as the Millian's. Still, what about the force of these ob-
jection? The chief one has always been that meta-linguistic accounts tum our
attributions into ones whose content is meta-linguistic. In doing so, in
addition to distorting the nature of the phenomenon, they run afoul of the so-
called translation test. 'A believes that the bearer of "N" is F' is clearly not
synonymous with 'A believes that N is F'; but, surely, if the former is
proffered as an analysis of the latter, they should be. Furthermore, the atti-
tude the former attributes can be seen to be different from that attributed by
the latter by noticing that they sustain different inferences. 12
But this objection involves a non sequitur. Suppose that what explains
why someone responds to sentences the way he does is, at least in part, what
he believes about the reference of the names in those sentences. It simply
does not follow that the attitudes we attribute to him on the basis of those
responses (using the best methods of interpretation we have at hand) are
about the lllames. While the way I express my beliefs depends on my beliefs
about what names refer to what things, my beliefs are nonetheless about the
things to which I believe my names to refer. The meta-linguistic belief to
which the Millian explanation of my verbal behaviour appeals is a different
belief from the belief (if any) that we attribute on the basis of that behaviour.
If I believe two names to refer to two different things, any beliefs that
may be attributable to me (using appropriate methods of interpretation) on
the basis of my use of the names will be about two putative things, and the
fact that this is so must be reflected in the contents ofthe attitudes attributed.
If I come to believe that the names refer to just one thing, the attributions
must be adjusted accordingly. What makes matters so slippery is that in the
former sort of case an attributor may believe that the names refer to the same
thing, and his use of the names will, of course, reflect this belief. 13 But this
 THE NEO-FREGEAN ARGUMENT 195

kind of link between meta-linguistic beliefs and object beliefs neither

constitutes nor requires an identity of content. While the link is both causal
and conceptual, the beliefs linked, and their contents, are distinct. Thus the
chief objection to meta-linguistic accounts of the phenomenon we are dealing
with fails. A meta-linguistic explanation of someone's responses to
sentences does not require a meta-linguistic construal of the attitudes he
would express by using those sentences.

VII

We have yet to establish, though, that an explanation appealing to beliefs

about reference is fully adequate as an explanation of the ground-level
phenomenon. Suppose it established, however. This would not show, in and
of itself, that the explanation offered by the neo-Fregean cannot be also ade-
quate. We would need to ask whether there are any considerations militating
in favour of one or the other of these rival explanations beyond their ability
to explain this particular phenomenon. In this section, I shall propose what
seems to me to be a plausible candidate for a test of adequacy, one I shall ar-
gue the Millian explanation satisfies at least as well-- indeed, arguably, bet-
ter than -- the neo-Fregean one. In the next, I shall, assuming for the sake of
a comparison that the latter is also an adequate explanation, suggest some
considerations which may, even so, tilt the balance in favour of the former.
Our explanandum, as I have argued, must be the fact that a person does,
or may, exhibit a certain pattern of behaviour (usually, though not always,
verbal): a pattern of assent and dissent to pairs of otherwise identical sen-
tences containing different but co-denoting names. We are asking, what facts
can explain these facts, in the sense of making it (causing it to be) the case
that they obtain. We have seen that these explaining facts must be facts
about our subject's psychology, in particular, facts about his beliefs about
the reference of his names. (I have suggested that we can, harmlessly, call
these meta-linguistic beliefs. But nothing hangs on this.) We have distin-
guished these facts from facts about the objects of these meta-linguistic be-
liefs. For the moment, let it be an open question whether the latter include
facts about senses, as the neo-Fregean claims. I have argued that even if they
do, we still need to advert to additional facts in explaining what causes the
different responses when these are present.
With these reminders, we are ready to consider the test of adequacy I want
to propose. We have seen that on both accounts it is one's beliefs about
one's names that is thought to determine the pattern of one's assent and dis-
196 JOHN BIRO

sent to the sentences in which they occur: on the one, beliefs about their
reference, on the other, beliefs about ("graspings" of) their sense. The obvi-
ous way to tell whether a belief of a certain sort is really doing the work it is
alleged to be doing must be, one is inclined to say, to see whether a change
in that belief leads to a change in the responses of the person holding it. In
our case, the test would be this: if a change in my beliefs about the co-
referentiality of a pair of names results in a re-alignment of my responses to
sentences containing them, that is good reason to think that my earlier
beliefs were responsible for my earlier responses. To apply the test, we just
need to imagine me as being given the information about co-referentiality.lf,
as a result, I come to (be disposed to) assent and dissent identically to pairs
of otherwise identical sentence containing the respective names, my lacking
the information (and thus the belief a rational, linguistically competent and
undistracted person can be presumed to base on it) can be reasonably seen as
an adequate explanation of my earlier disposition to assent and dissent
differently.
By such a test, the Millian explanation (as I have presented it) seems to
fare at least as well as the neo-Fregean one. It does seem that one's acquiring
the belief, when one did not have it before, that two names are co-denoting,
is sufficient, by itself, to re-align one's responses. One need only be told --
and be told only -- that 'Cicero' and 'Tully' are co-denoting names for all dif-
ferences in one's response to sentences differing only in containing one or
the other, respectively, to disappear. 14 This will happen even if one has no
beliefs about what the names' actual (single) referent is and, even more im-
portantly in the present context, even if one attributes to them no properties
other than their co-referentiality.
By contrast, the neo-Fregean explanation seems to fail the test I am
proposing. According to that explanation, it will be recalled, it was my be-
lief that the two names had different senses that accounted for my different
responses to the otherwise identical sentences containing them. But while it
seems reasonable to think that if I came to believe that two names had
(exactly) the same sense, I would eo ipso come to believe that they have the
same referent, it does not seem as clear that coming to think that two names
have different senses guarantees that I must think of them as having different
referents. (Presumably neo-Fregeans themselves believe that 'Cicero' and
'Tully' have different senses but the same referent.) It seems clear that a per-
son's thinking that two names have different senses need not cause him to
have different responses to two otherwise identical sentences in which they,
respectively, occur. But thinking that they have different referents will, it
 THE NEO-FREGEAN ARGUMENT 197

seems, inevitably do so. (Actually, all that is necessary is that one not think
that they have the same referent. As long as that is the case, the possibility
of different responses seems to remain. But coming to think that there is
only one referent seems to foreclose the possibility.)
The neo-Fregean may try to appeal to the principle that, at least for a
particular person, beliefs about sense determine beliefs about reference. He
could argue that while two different senses can, of course, determine the same
referent, in those cases where he needs to appeal to the principle, that is,
where a difference in the senses one attaches to two co-denoting names is
supposed to explain the difference in one's responses to the (otherwise iden-
tical) sentences containing them, thinking that there is a difference in sense
will cause one to think at least that there may be a difference in reference.
But it is clear that this line of defence is suicidal, for it amounts to accepting
that even if there were senses, and even if we attributed beliefs about these to
our subject as an explanation of his responses, that explanation would work
only on condition that we also attributed to him beliefs about reference .. This
concedes the necessity of bringing in beliefs about reference, and it leaves
open the possibility that doing so is, as the Millian urges, sufficient by
itself to yield the explanation we seek. If that is so, the neo-Fregean has not
provided either the only, or the most economical, account of the
phenomenon to be explained.
Still, pairs of sentences in which co-denoting names occur in an opaque
position, such as those used in attributing attitudes, may seem to some to
present counter-examples to my claims on behalf of this test. Suppose I have
reason to assent to both 'My nephew wants a book by Mark Twain for his
birth-day' and 'My nephew does not want a book by Samuel Clemens for his
birth-day.' I can have such reason, based on my interpretation of my
nephew's pattern of responses to other sentences (among them, perhaps, an-
swers to questions like 'Would you like a book by Mark Twain/Samuel
Clemens for your birth-day?) whether I myself believe that the names 'Mark
Twain' and 'Samuel Clemens' refer to one person or two. Suppose I think
the latter. My coming to learn that I am wrong will not -- and, of course,
should not -- make me change my responses to the two attribution-sentences.
What matters is what my nephew believes, and -- we are assuming -- I have
no reason to believe that that has changed.
So, while I stated the test in a completely general way, we actually need
to restrict its application to cases where the responses in question are to sen-
tences which are not themselves attitude-attributing ones. Such a restriction
should not, however, be seen as an ad hoc one, given our goal of isolating
198 JOHN BIRO

the beliefs that cause a person's pattern of responses to be what it is. A sen-
tence attributing an attitude to someone is, as I have argued, already an in-
terpretation ofthat person's pattern ofresponses. The only things that can be
the cause of that pattern are beliefs (among them, meta-linguistic beliefs) of
that person. The only beliefs of an attributor that are relevant are his beliefs
about his subject's beliefs.I 5 His own meta-linguistic beliefs cannot be
assigned a causal role vis-a-vis his subject's responses. True, the attributor's
own responses to the attribution-sentences remained unaltered in spite of his
change of belief about co-referentiality. But that is because his meta-linguis-
tic beliefs played no causal role in his coming to assent to the two attribu-
tion-sentences. (As I noted, an attributor. could have good reason to assent to
these whether he himself believed the names to be co-referring or not.)
At least by this test, it seems that one can continue to be a Millian, in-
sisting that we should think of someone's different responses as resulting not
from his thinking of the names as having different senses, but from his
understanding of them as having different referents. It is beliefs about tbe
reference of names that explain the relevant explanandum, not beliefs about
their sense. And if such an explanation works at least as well as its neo-
Fregean rival --let alone if it works better -- we are not forced to postulate
senses for the explanatory beliefs to be about. The conclusion must be that
the way of arguing for names' having senses that we have been examining is
blocked.

VIII

In spite of the considerations just mooted, which may cast doubt upon the
ability of the neo-Fregean account even to pass what seems to be a rea-
sonable the test of adequacy, I now want to proceed as if in this respect there
were no difference between it and its Millian rival. Let us assume that both
explanations pass the test, so that we can ask what reason there might be to
favour one over the other. The obvious candidate that comes to mind is our
preference, other things being equal, for the simplest explanation we can find
for what we want explained. Here the Millian seems, on the face of it, to
have the edge. His explanation seems more economical, in that it appeals to
fewer semantic objects and properties. But whether it is really so can be
finally judged only after seeing what success it has with the broader range of
questions to which a semantic theory must give an answer. Before conclud-
ing that the Millian account is more economical overall, one would need at
 THE NEO-FREGEAN ARGUMENT 199

least to consider how it fares as an explanation of some other, closely related,

phenomena involving expressions other than names.
It seems reasonable to expect a unified account of similar phenomena, so
we would like one that gives the same explanation of the different responses,
whenever such occur, to pairs of sentences differing only in their subject
term, whether that term is a proper name, a description (definite or indefinite)
or a common noun. It is an unquestionable virtue of the neo-Fregean account
that it offers such a unified explanation. Can the Millian account match this
virtue?
Let us allow that common nouns and, thus, descriptions and the like have
senses. After all, it is uncontroversial that, most words other than names
have meanings. Let us take common nouns. Can their having meaning
explain someone's having different responses to a pair of sentences differing
only in containing two such nouns with different meanings but with the
same extension? If someone were to (appear to) say that a fortnight is longer
than two weeks, would the fact that the two expressions differ in meaning be
the best -- let alone, the only -- way to explain his odd utterance?I6 It seems
that considerations parallel to those discussed in connexion with proper
names apply. The mere existence of a difference in meaning, even when there
is one, could not suffice: here, too, the subject must believe that there is a
difference. I7 But even that will not do, if we apply a test of the sort I have
proposed, for he may correctly believe that there is a difference in the
meaning of the two expressions without believing that there is a difference in
their extensions. I8 So, if a subject really does respond differently to
sentences containing such a pair of expressions, it seems that the only thing
that can provide an adequate explanation of his doing so is his belief that the
extensions of the expressions are different. Even when such a belief is false --
as here -- it seems to fare better on the r~-alignment test proposed than a
belief about a difference in meaning. A change in one's beliefs about
extensions will alter one's pattern of responses, where a mere change in
one's beliefs about meanings will not. One's lacking the belief that two
expressions are co-extensive will explain how one can have different
responses to sentences containing those expressions (or, if we accept the
interpretation of this basic phenomenon a~. evidence for a difference in atti-
tudes, how one can have such different attitudes), just as one's lacking the
'.

belief that two names are co-referential didiwhen these were involved. Having
that belief will, in a rational person, blqck (the possibility of) differing
responses in the same way. This will be so even if the expressions have
i
200 JOHN BIRO

different senses and even if the person in question knows that they do.
(Recall also our earlier example from the kitchen.)
To evaluate the merits of the rival neo-Fregean and Millian accounts
fully, we would need to pay attention, in addition to the apparent failures of
substitutivity in propositional-attitude contexts, to at least two other phe-
nomena that seem to some to threaten the Millian position on names: the
presence in natural languages of non-denoting (sometimes called "empty" or
"vacuous") names, and the possibility of making true statements, including
denials of existence, with sentences containing such names. 19 As I said at the
outset, I shall not attempt to deal with these challenges in this paper. But it
should be conceded here that if the Millian account cannot deal satisfactorily
with these other problems, it cannot claim to have the overall advantage. My
goal here has been the limited one of considering whether the argument from
(alleged) non-substitutivity by itself suffices to establish the neo-Fregean
conclusion. A verdict on this question can be given independently of a more
comprehensive evaluation of the merits of Millianism. 20
There is, however, one more consideration the neo-Fregean can offer that
I need to take up briefly, since it may be thought to provide the kind of
knock-out blow to Millianism that would render such larger comparisons
unnecessary. The neo-Fregean may make an appeal to the compositionality
of meaning and claim that if we accept that, we cannot, in the end, escape his
conclusion. He will insist that (what he takes to be) failures of substitutivity
in propositional-attitude contexts show that the content-specifying sentences
which differ only in containing different (though co-denoting) names must
differ in meaning, for if they did not, it would not be rational to assent to
one and not to the other, as people sometimes do. But since the meanings of
sentences are a function of the meanings of their component parts, and since
the only parts that could make a differential contribution to the meanings of
the two sentences are the different names they contain, these parts must have
(different) meanings. Call what these parts -- the names -- have meanings or
senses or anything you like, having them is a different property from having
reference. So, Millianism cannot be right.
What this line of argument assumes, however, is that from (the possibil-
ity of) different responses to two sentences we can infer that the sentences
have different meanings. This is an unwarranted inference, as a moment's re-
flection on the phenomenon of inter-language synonymy will show. The fact
that someone ignorant of a foreign language does not accept (or assent to) the
translation of a sentence he does accept in his own obviously cannot be taken
to show that the second sentence is not, after all, a translation of the first. In
 THE NEO-FREGEAN ARGUMENT 201

the case of different responses to a pair of otherwise identical sentences

containing different but co-denoting names, analogous ignorance about
reference is needed to explain them: if one does not believe that the names
denote the same thing, one may assent to the one sentence and not the other,
even if one is fully rational and linguistically competent. Ignorance of this
sort is not irrationality, obviously. How far (ideal) linguistic competence
includes complete mastery of the denotations of all the names in the
language one speaks is an interesting question. (My view is, not very far.)
But it is not a question we need to take up in order to see that even if the
lack of such complete mastery constituted lack of (full) competence, no
conclusion about names' having meanings (or senses) would be thereby
forced on us.
Such an appeal to compositionality is, then, in the present context, ques-
tion-begging. Whether or not two sentences must be thought of as differing
in meaning (or sense, or in what thought or proposition they express) just
because someone (may) accept one and not the other is one of the things at
issue here. If (the possibility) of someone's doing so can be explained in an-
other way, than we have no argument from that possibility to the names'
having to make different semantic contributions and hence no argument for
there having to be different semantic objects or properties (other than their
having a, by hypothesis common, referent) associated with them. And the
obvious way to explain someone's accepting one sentence and not another,
where these would be naturally thought to be translations of each other were
it not for this possibility, is to focus on what that someone takes them to
mean. It is, I hope, obvious that someone can take two sentences to differ in
meaning when in fact they do not. But then no semantic conclusion can be
inferred from people's acceptance of one and non-acceptance of the other of
two sentences, in particular, no conclusion about their non-synonymy.21

IX

I have claimed that the well-worn argument of neo-Fregeans from the

(alleged) phenomenon of non-substitutivity in propositional-attitude contexts
fails. The chief reason is that the neo-Fregean has not made out a case that
the postulation of senses is the only way to explain the ground-level
phenomenon, namely, puzzling patterns of (usually verbal) behaviour. We
should recognize that to have an explanatiqn of that phenomenon we need to
appeal to a psychological property of the ~esponder in addition to whatever
semantic properties we may think different kinds of terms have. Suppose that
202 JOHN BIRO

such a psychological property -- we may call it, innocently, I think, a meta-

linguistic belief, since it is a belief about the semantic properties of one's
terms -- will explain what we want explained, whether we take it to be a
belief about the sense of one's terms or about their reference. Then a com-
parison of the two alternatives may show that the Millian one is the more
economical one, since it can make do with fewer semantic objects or proper-
ties than its rival. (Such a verdict, however, has to wait on a further investi-
gation of the Millian account's ability to deal with, and deal equally
economically with, other challenges to its adequacy as an account of the
semantics of names.)
I have also suggested a test for the adequacy of any proposed explanation
of the phenomenon: does learning about the facts which, according to that
explanation, account for one's differing responses, suffice to re-align those
responses? Judged by this test, the Millian explanation seemed, contrary to
the assumption I made for the sake of the comparison in economy, to have
the edge.
The overall conclusion must be that the case of the neo-Fregean has not
been made out. This does not mean that the Millian must be right. There
may be better arguments than the one I have examined for a Fregean view of
the semantics of names. On the other hand, the fact that this particular way
of arguing for their existence fails in the way it does may suggest a salutary
general lesson about the relation between semantics and psychology. One
reason for the appeal of the neo-Fregean explanation of the phenomenon is
that it emphasizes the cognitive significance of a term for an individual who
uses it. But, first, as we have seen, insofar as the explanation requires an
appeal to cognitive significance, sense itself is not fine-grained enough (any
more than is reference, of course) to provide these. Second, and more impor-
tant, that a term has this or that cognitive significance for an individual is a
psychological fact about that individual, a fact of a different order from facts
about the role that term plays in a language, in other words, from semantic
facts. One must be careful not to conflate, as the neo-Fregean argument does,
these radically different sorts of facts.
It is ironic that writers taking their inspiration on these matters from
Frege, the champion of the liberation of semantics from psychology, should
fall into the trap of confiating the two through basing a claim about what
semantic properties expressions in a language have on considerations of the
cognitive significance of those expressions. Whether Frege himself in fact
avoided this trap is not clear. But he would have wanted to; and so should his
followers. 22
 THE NEO-FREGEAN ARGUMENT 203

NOTES

1 John Stuart Mill (1843); Gottlob Frege (1892) The relation of the argument 1 shall discuss to
Frege's own is a complicated matter. 1 use the label 'neo-Fregean' both to suggest the
argument's obvious inspiration and to indicate its possible distinctness.
2 Examples: Ackerman (1979), (1979a), (1979b), Plantinga (1978). Flirters include, arguably,
Russell, Dummett, Searle, Castaneda and Evans.
Proponents of the neo-Fregean argument usually find other reasons as well for espousing
a neo-Fregean position on senses. There are, as is well known, two other problems to which
the postulation of senses is thought to provide a natural, and perhaps the only, solution: the first
concerns sentences containing empty names, the second, true negative existential sentences.
Since my chief interest in this paper is in what support the argument from non-substitutivity
provides for the neo-Fregean conclusion, I want to treat it in isolation from other arguments
~urporting to lead to that conclusion. (I discuss these other matters in Biro (forthcoming».
Mill, p.2l.
4 I say 'that is' (that is, treat the fact that a is b as equivalent to the fact that 'a' and 'b' denote
the same thing) for two reasons. One is that it is difficult to see what else an assertion of the
former could come to beyond a claim of self-identity, which would hardly be worth making.
(Recall Frege's remarks on the relative informativeness of 'a=b' and 'a=b'.) Second, as 1
shall be emphasizing, any claim that names are non-substitutable salva veritate in attributions
of propositional attitude must be grounded in some behavioural evidence on the part of the
subject of those attributions, and the relevant facts indicated by such behaviour are, in the first
(and, 1 shall argue, the last) place, facts about beliefs about the denotation of the names.
5 1 have put things here in a way that does not depend on any particular view of what sorts of
things sentences express, let alone any particular view of the nature of those things. Having
done so, and having thus shown that the neo-Fregean argument can be stated in a way that
makes no primaJacie theoretical commitments on the nature of propositions (other than that
whatever they are, they are what well-formed indicative sentences express), I shall, in what
follows, fall in at times with the talk of propositions common in statements and discussions of
the argument.
6 Note, however, that the subject cannot be thought of as knowing (or even believing) that a is
identical with b. Suppose that names do have senses and that he attaches different senses to
the names 'a' and 'b', respectively. He might still believe that they have a single referent. (If
'Hesperus' and 'Phosphorus' have senses, and if these are, as would be plausible to think,
different, most of us are, with respect to these names, in just this position.) It seems clear that
his responses to sentences in which the names occur would still be uniform. (I shall discuss this
point more fully in Section VII. See also fn. 9.)
7 Sometimes the proponent of the neo-Fregean argument makes an explicit appeal to
Leibniz's Law in arguing for the claim that there must be two propositions in the cases we are
considering. (See, for example, Ackerman, op. cit.). For misgivings about this strategy, see
Bertolet (1984).
8 I sometimes put things in terms of knowing or not knowing such facts, for brevity's sake. But
what actually matters is not knowledge of such things, only belief concerning them. Even the
mistaken belief that two expressions have the same referent (or the same sense) is enough to
ensure that there will be no difference in attitude. (See below)
9 Suppose names have sense and suppose that, contrary to what most of us think, 'Cicero' and
'Tully,' express different senses. If I were unaware of this, or believed the contrary, I would
exhibit a pattern of responses to sentences containing these names different from the normal
one. The Millian account I defend below explains this by noting that my believing that the
names share a sense, together with the fact -- not in dispute in the present debate -- that
sameness of sense guarantees sameness of reference, entails my believing that they share a
reference.
204 JOHN BIRO

The reverse case is, of course, equally possible. I may believe that there is a difference
of sense where there is not. My doing so would then lead to differences in my responses
where the rest of my community thinks there should not be any -- as long as I also believed
that each sense determined a different referent. But this second belief is what is really doing
the work here: if one merely believes that two names have different senses (or attaches
different senses to two names -- see fn. 17 below) without believing that they have different
referents _.. as most of us do with 'Cicero' and 'Tully' -- doing so does not affect one's re-
sponses in the way believing that there is a difference in referents would.
10 See, for example, Swoyer (1979).
11 Again, this somewhat overstates things. It is the point at dispute as far as this particular neo-
Fregean argument, from non-substitutivity in contexts of propositional attitude, is concerned.
As noted before, there are other arguments that may tilt the overall case against the Millian.
12 For the translation test, see Church (1950). For a discussion of some related objections
against "going meta-linguistic," see Corazza and Dokic, "Is Frege's puzzle about language?"
(forthcoming) (See also Corazza and Dokic, "Why Is Frege's Puzzle Still Puzzling?", this
volume.) An unorthodox account of the role of meta-linguistic considerations in semantic
theory is presented in Roberts (1985).
13 It is not always noticed that the reverse is also possible. An attributor may believe that two
names refer to two different things, even if his subject does not. Much confusion in discussions
of our topic comes from the fact that we do not notice that we could always be in either of
these positions, a possibility obscured by the fact that in our attributions we necessarily rely in
on what we take to be the correct meta-linguistic beliefs. (For detailed discussions of this
foint, see the references in fn. 16.)
4 Assuming, of course, as I have been all along, that we are discussing cases involving the
responses and attitudes of rational beings. This idealization is necessary for both sides to the
argument, so I shall not dwell on it.
It should also be noted that my claim about responses being re-aligned as a result of
learning about co-referentiality does not, of course, extend to sentences in which the names
are mentioned, rather than used. The sentences '''Cicero' has six letters" and '''Tully' has six
letters" will continue be the objects of different responses by a rational and linguistically
competent person.
15 Of course, when the subject's belief are themselves about what someone else believes, the
attributor must take account of the first subject's beliefs about the second subject's meta-
linguistic beliefs, and so on. The possibility of such iteration does not affect the point being
made here.
16 I include the parenthetical 'appear to' as a reminder that I am, all along, speaking of
observable verbal behaviour, as yet un-interpreted. This, as I have insisted, is the phenomenon
from which the neo-Fregean argument must start and the one which any account must explain.
The relation between what we may call the thin (un-interpreted) sense of 'say' and its thick
(interpreted) sense is what is in question between those who accept and those who deny the
genuineness of the alleged phenomenon of non-substitutivity: the former do, while the latter
hesitate to, take thin sayings (or assentings) as evidence for attributing thick sayings and
propositional attitudes. For a discussion of some of the principles involved in basing
interpretations on verbal behaviour, see Biro (1984 and 1992).
17 As with the "graspings" of the putative senses of names discussed earlier, the
psychological fact required seems best described as a belief about a difference in the
meaning of the expressions. Kent Bach has suggested (in conversation) that the locution that
seems most natural here, namely, that 'the two expressions mean different things to the
subject', should not be construed as implying the existence of beliefs about the expressions'
meanings. Perhaps not -- but then we are owed some account of what else such locutions
should be taken to mean. It does seem intuitively right to say at least that in a case where a
 THE NEO-FREGEAN ARGUMENT 205

person responds differently to two sentences, he must be thought of as taking them to mean
different things, however 'taking to mean' is to be un-packed.
18 Someone may believe that there is a metaphysically possible difference in the extensions of
two terms whose actual extensions are the same. In such a case, he will respond differently to
modal sentences containing the two terms. The test I have proposed can be extended to deal
with such cases in an obvious way.
19 See, for example, Plantinga, (1978). And there is, of course, the fact that identity
statements involving distinct but co-denoting names can be informative. It should be clear that
the phenomenon so described is, at bottom, the same as that of the failures of substitutivity I
have been discussing.
20 I discuss some other phenomena allegedly posing a problem for Millianism in Biro
(forthcoming).
21 A reminder: I am talking about synonymy in a public language, not in a person's idiolect or
mental language. One may, of course, be skeptical about one or the other of these different
notions. But it seems clear both that the neo-Fregean argument's intended conclusion is that
names as used in a public natural language have senses and that it is in this that the argument's
interest lies.
22 This paper was a long time in the writing. Among the many helpful critics over that time,
Felicia Ackerman, Kent Bach, Kit Fine, Kirk Ludwig, William Lycan, Reinhard Muskens,
Greg Ray, Georges Rey, Chris Swoyer, and Howard Wettstein are owed special
hanks.

REFERENCES

Ackerman, F.: 1979, "Proper Names, Propositional Attitudes and Non-Descriptive

Connotations,"Philosophical Studies, vol. 35.
Ackerman, F.: 1979a, "Recent Work in the Theory of Reference," American Philosophical
Quarterly.
Ackerman, F.: 1979b, "Proper Names, Essences and Intuitive Beliefs," Theory and Decision,
vol. 11.
Bertolet, R.: 1984, "Ackerman on Propositional Identity," The Philosophical Quarterly, vol. 34.
Biro, J.: 1984, "What's in a Belief?," Logique et Analyse, vol. 107.
Biro, J.: 1991, "Individualism and Interpretation," Acta Analytica, vol. 6.
Biro, J.: 1992, "In defence of social content," Philosophical Studies, vol. 67.
Biro, J.: 1995, "What is right with meta-linguistic approaches to puzzles about reference?"
(forthcoming) .
Corazza, E. and J. DokiC: 1995, "Why Is Frege's Puzzle Still Puzzling?" (this volume).
Corazza, E. and J. DokiC: 1995, "Is Frege's Puzzle about Language?" (forthcoming).
Church, A.: 1950, "On Carnap's Analysis of Statements of Assertion and Belief," Analysis,
vol. 10, no. 5.
Frege, G.: 1892, "Uber Sinn und Bedeutung," Zeitschrift fiir Philosophie und philosophische
Kritik, vol. 100; English translation as "On Sense and Reference," in Frege' s
Philosophical Writings, E. Anscombe and P. Geach (eds.), Oxford, 1952.
Mill, J. S.: 1843, A System of Logic, London.
Plantinga, A.: 1978, "The Boethian Compromise," American Philosophical Quarterly, vol. XV.
206 JOHN BIRO

Roberts, L.: 1985, "Problems about material and formal modes in the necessity of identity,"
The Journal of Philosophy, vol. 82.
Swoyer, C.: 1979, "Sense and Nonsense," Canadian Journal of Philosophy, vol. IX.

University of Florida
Department of Philosophy
330 Griffin-Floyd Hall
PO Box 118545
Gainesville, FL 32611-8545
 INDEX

Ackerman, D. (F.)185, 191

Anscombe, E. 60
Aristotle 59
Austin, J.L. 63
Benveniste, E. 100
BonJour, L. 109
buffers 161
causal theory 176
character 10-12, 23, 29, 33, 36,43-44,48,78,89,177-181
Chomsky, N. 18-19,36,41-42
communication 28, 47-51, 55-56, 86-89, 95-98, 100, 143-144, 146-148
concept 113-115
content 2,6-7, 10, 16,23,32-36,43-44,51-53,63, 75, 87,91,97,99-101,
111,130-137,143,152-153,156-157,163,171,177-179, 181, 185-186,
194
convention 33-34, 55-56
Davidson, D. 9-10, 13, 16-18
direct reference 170-172
dossiers of information 99
Dummett, M. 1-4,7-13, 15,27-29,32,50,60,65,88,96,98,110
emotions 121-122
evaluative concept 113, 115, 119-124
evaluative judgements 117
evaluative terms 114-118, 123-124
Evans, G. 35, 98, 134-135, 158-161
fIrst person 95-101
Fodor, J. 27, 31
force 5-8, 48, 53-54
Geach, P. 5, 10,60
Goodman, N. 62
grasping 25,31-33,65-66,68, 71, 81-83, 90, 160-163, 190-193, 196
Husserl, E. 65, 67, 129-131
indexicals 23, 61, 79, 96, 99-101,161,172
interpretive theories 16-18, 21-23
judgement 18, 21, 28-30, 44
Kaplan, D. 23, 173, 176-178, 180
knowledge of meaning 32, 103

207
208 INDEX

~pke,S.64, 76-77,155,171,173
Leibniz, O. 72
Lewis, D. 13
Locke,1. 59
Lotze, R.H. 1
Lycan, W.20
Mackie,1. 124
McDowell,1. 35, 121, 158, 159, 160, 161
meaning 11-13, 16,47-50,53-56,59-64,70-71,75-76,79-80,83,85,90,92,
96,101,104,107-111,115,132-137,153,169-181,189-201
mention 34
meta-linguistic account 193-195
metalinguistic theory 133, 135
Mill, 1.S. 59, 185
mode of presentation 26, 49, 65-66, 80-81, 84, 95-96, 98-101,107,157-158,
160, 163, 170
natural kind 59-62, 64, 67-69, 75-76, 78-80, 82-83, 85-88,90-92,107
Peacocke,C. 162, 163
Perry, 1. 98-99, 161
Pettit, P. 120, 123, 125
Planting a, A. 185, 191
predicate 8, 11, 15,38,62,64-69,80-82,88-89, 107-108, 156-161, 163, 169
property 3, 8, 10, 13,27-28, 144, 148-149, 156, 160, 169, 187-188, 191-192,
200,202
propositional attitudes 55, 104, 118, 185-186, 190
Putnam, H. 59-64, 67-72, 75-79, 81-86
radical interpretation 13
Recanati, F. 96, 101
reference 2-3, 6-8,11, 15,28,32-39,43-44,50,59-61,64-69,75-77,80-90
Russell, B. 154, 158-160, 162-163, 170-172, 174-175, 177, 179
Searle, 1. 53
self-knowledge 103, 109
semantic content 97-101, 152-153
semantic properties 28, 42-43, 185-188
semantic theory 11, 18, 190-191, 199
semanticist theory 174, 175, 180
singular propositions 98, 170
T-theory 15-19,21-22
Tarski, A. 15-16
 INDEX 209

Taschek, W. 153
theory of knowledge 17,25
theory of meaning 9, 16-23,32-34,47,50,54-56,169
theory oftruth 1,4, 12-13
third person 98, 100-101
thought 1,4-5,8-11,84,111
thought-content 27
truth conditions 23, 66, 171-175, 178
truth-theory 15, 17-18,54
understanding 17-21,27-28,31-36,42,55-56,59-61,64,68-70,80,82-84,
88-90,96-98, 141-143, 147-148, 151-152, 159-160, 174, 189-193, 198
Wettstein, H. 154-157, 164
Wiggins, D. 75, 78-91,111,113-116,125
Wright, C. 120
 PHILOSOPHICAL STUDIES SERIES
Founded by Wilfrid S. Sellars and Keith Lehrer

Editor
KEITII LEHRER, University ofArizona, Tucson

Associate Editor
STEWART COHEN, Arizona State University, Tempe

Board of Consulting Editors

Lynne Rudder Baker, Allan Gibbard, Denise Meyerson, Ronald D. Milo,
Fran~ois Recanati, Stuart Silvers and Nicholas D. Smith

1. JAY F. ROSENBERG, Linguistic Representation, 1974.

2. WILFRID SELLARS, Essays in Philosophy and Its History, 1974.
3. DICKINSON S. MILLER, Philosophical Analysis and Human Welfare. Selected
Essays and Chapters from Six Decades. Edited with an Introduction by Lloyd D.
Easton, 1975.
4. KEITII LEHRER (ed.), Analysis and Metaphysics. Essays in Honor of R. M. Chisholm.
1975.
5. CARL GINET, Knowledge, Perception, and Memory, 1975.
6. PETER H. HARE and EDWARD H. MADDEN, Causing, Perceiving and Believing.
An Examination of the Philosophy of C. J. Ducasse, 1975.
7. HECTOR-NERI CASTANEDA, Thinking and Doing. The Philosophical Foundations
of Institutions, 1975.
8. JOHN L. POLLOCK, Subjunctive Reasoning, 1976.
9. BRUCE AUNE, Reason and Action, 1977.
10. GEORGE SCHLESINGER, Religion and Scientific Method, 1977.
11. YIRMIAHU YOVEL (ed.), Philosophy of History and Action. Papers presented at the
first Jerusalem Philosophical Encounter, December 1974,1978.
12. JOSEPH C. pm, The Philosophy ofWilfrid Sellars: Queries and Extensions, 1978.
13. ALVIN 1. GOLDMAN and JAEGWON KIM, Values and Morals. Essays in Honor of
William Frankena, Charles Stevenson, and Richard Brandt, 1978.
14. MICHAEL J. LOUX, Substance and Attribute. A Study in Ontology, 1978.
15. ERNEST SOSA (ed.), The Philosophy of Nicholas Rescher: Discussion and Replies,
1979.
16. JEFFRIE G. MURPHY, Retribution, Justice, and Therapy. Essays in the Philosophy of
Law, 1979.
17. GEORGE S. PAPPAS, Justification and Knowledge: New Studies in Epistemology,
1979.
18. JAMES W. CORNMAN, Skepticism, Justification, and Explanation, 1980.
19. PETER VAN INWAGEN, Time and Cause. Essays presented to Richard Taylor, 1980.
20. DONALD NUTE, Topics in Conditional Logic, 1980.
21. RISTO HILPINEN (ed.), Rationality in Science, 1980.
22. GEORGES DICKER, Perceptual Knowledge, 1980.
23. JAY F. ROSENBERG, One World and Our Knowledge of It, 1980.
24. KEITH LEHRER and CARL WAGNER, Rational Consensus in Science and Society,
1981.
25. DAVID O'CONNOR, The Metaphysics ofG. E. Moore, 1982.
26. JOHN D. HODSON, The Ethics ofLegal Coercion, 1983.
27. ROBERT 1. RICHMAN, God, Free Will, and Morality, 1983.
28. TERENCE PENELHUM, God and Skepticism, 1983.
29. JAMES BOGEN and JAMES E. McGUIRE (eds.), How Things Are, Studies in
Predication and the History of Philosophy of Science, 1985.
30. CLEMENT DORE, Theism, 1984.
31. THOMAS L. CARSON, The Status of Morality, 1984.
32. MICHAEL 1. WHITE, Agency and Integrality, 1985.
33. DONALD F. GUSTAFSON, Intention and Agency, 1986.
34. PAUL K. MOSER, EmpiricalJustijication, 1985.
35. FRED FELDMAN, Doing the Best We Can, 1986.
36. G. W. FITCH, Naming and Believing, 1987.
37. TERRY PENNER, The AscentJrom Nominalism. Some Existence Arguments in Plato's
Middle Dialogues, 1987.
38. ROBERT G. MEYERS, The Likelihood of Knowledge, 1988.
39. DAVID F. AUSTIN, Philosophical Analysis. A Defense by Example, 1988.
40. STUART SILVERS, Rerepresentation. Essays in the Philosophy of Mental Rerepresen-
tation, 1988.
41. MICHAEL P. LEVINE, Hume and the Problem of Miracles. A Solution, 1979.
42. MELVIN DALGARNO and ERIC MATTHEWS, The Philosophy of Thomas Reid,
1989.
43. KENNETH R. WESTPHAL, Hegel's Epistemological Realism. A Study of the Aim
and Method of Hegel's Phenomenology of Spirit, 1989.
44. JOHN W. BENDER, The Current State of the Coherence Theory. Critical Essays on the
Epistemic Theories of Keith Lehrer and Laurence Bonjour, with Replies, 1989.
45. ROGER D. GALLlE, Thomas Reid and 'The Way of Ideas' ,1989.
46. J-C. SMITH (ed.), Historical Foundations of Cognitive Science, 1990.
47. JOHN HElL (ed.), Cause, Mind, and Reality. Essays Honoring C.B. Martin, 1990.
48. MICHAEL D. ROTH and GLENN ROSS (eds.), Doubting. Contemporary Perspectives
on Skepticism, 1990.
49. ROD BERTOLET, What is Said. A Theory of Indirect Speech Reports, 1990
50. BRUCE RUSSELL (ed.), Freedom, Rights and Pornography. A Collection of Papers
by Fred R. Berger, 1990
51. KEVIN MULLIGAN (ed.), Language, Truth and Ontology, 1992
52. JESUS EZQUERRO and JESUS M. LARRAZABAL (eds.), Cognition, Semantics and
Philosophy. Proceedings of the First International Colloquium on Cognitive Science,
1992
53. O.H. GREEN, The Emotions. A Philosophical Theory, 1992
54. JEFFRIE G. MURPHY, Retribution Reconsidered. More Essays in the Philosophy of
Law, 1992
55. PHILLIP MONTAGUE, In the Interests of Others. An Essay in Moral Philosophy,
1992
56. 1.- P. DUBUCS (ed.), Philosophy of Probability. 1993
57. G.S. ROSENKRANTZ, Haecceity: An Ontological Essay. 1993
58. C. LANDESMAN, The Eye and the Mind. Reflections on Perception and the Problem
of Knowledge. 1993
59. P. WEINGARTNER (ed.), Scientific and Religious Belief 1994