You are on page 1of 12

Scots Philosophical Association

University of St. Andrews

Dummett on Frege
Frege: Philosophy of Language by Michael Dummett
Review by: Leslie Stevenson
The Philosophical Quarterly, Vol. 24, No. 97 (Oct., 1974), pp. 349-359
Published by: Oxford University Press on behalf of the Scots Philosophical Association and the
University of St. Andrews
Stable URL: http://www.jstor.org/stable/2217827 .
Accessed: 28/06/2014 10:10

Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at .
http://www.jstor.org/page/info/about/policies/terms.jsp

.
JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of
content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms
of scholarship. For more information about JSTOR, please contact support@jstor.org.

Oxford University Press, Scots Philosophical Association, University of St. Andrews are collaborating with
JSTOR to digitize, preserve and extend access to The Philosophical Quarterly.

http://www.jstor.org

This content downloaded from 91.220.202.121 on Sat, 28 Jun 2014 10:10:36 AM


All use subject to JSTOR Terms and Conditions
349

CRITICAL STUDY

DUMMETT ON FREGE
BY LESLIE STEVENSON

Frege: Philosophy of Language. By MICHAELDUMMETT. (London: Duck-


worth. 1973. Pp. xxv + 698. Price ?10.)

Analytic philosophers generally trace their intellectual ancestry to a


revolution in philosophy in the early years of this century, when Russell,
Moore and Wittgenstein began to make the analysis of meaning the funda-
mental philosophical method and task. One of Dummett's achievements in
this magnificent book is to demonstrate how this revolution had its real
beginning in Frege's philosophy of language, so that just as we have long
recognized Descartes as the inaugurator of a philosophical era in which
epistemology was fundamental, so we must now acknowledge Frege as the
founding father of the present era (p. 669). Russell and Wittgenstein ex-
plicitly acknowledged their debt to Frege, and were the main transmitters
of his ideas; but Moore's main achievement-the defeat of idealism by
careful analysis of meaning-was also anticipated by Frege. However, the
new understanding of the very nature of philosophical questions was more
fundamental than the change in doctrine from idealism to realism, and could
in principle have come without it (p. 683).
Dummett's book is not primarily historical. Its main content is deep
discussion of the central issues in the philosophy of language, defending,
reinterpreting, and vindicating much of Frege's approach. But Dummett's
devotion to Frege is not at all that of slavish discipleship. He clearly repudi-
ates some of Frege's famous doctrines-most notably, the treatment of
sentences as names of truth-values (pp. 183-4), the supposed analogy be-
tween the references of proper names and that of predicates (p. 243), and
the assumption that a single domain of all the objects there are can serve
as the range of every first-order quantifier (p. 476). He discusses problems
that Frege did not properly face, such as the threat to the whole notion of
objective senses posed by variation in sense between different speakers (pp.
102, 584), or was not even aware of, such as the verificationist claim that
the theory of meaning should be based on the notions of verification and
falsification rather than truth and falsity (p. 467). There are extensive and
effective criticisms of many contemporary views, such as Quine's rejection
of intensional notions (Chs. 5 & 17), Kripke's account of names (appendix to
Ch. 5), and Geach's treatment of quantification and identity (Ch. 16).
The quality and quantity of Dummett's discussion will surely make his
book one of this century's major contributions to the philosophy of language.
Having said that, let me also point to some general features which mar the
book's greatness. Firstly, its length and discursiveness is daunting even to
the determined reader. There is a certain amount of repetition, and the
argument in many chapters follows a meandering course, indulging in lengthy
digressions, coming back to already-stated positions to make qualifications,

This content downloaded from 91.220.202.121 on Sat, 28 Jun 2014 10:10:36 AM


All use subject to JSTOR Terms and Conditions
350 LESLIE STEVENSON

sometimes leaving questions to be taken up later (and it is not always clear


whether and where the promises are fulfilled). Some important themes
recur in several different parts of the book, so that one needs to use the
helpful " Brief Subject Index" on p. 695 to build up Dummett's complete
account out of its parts. I thinli I shall not be untypical in needing to make
several careful readings before seeing some wood emerge from the trees.
But let me assure prospective readers that the effort is well worth it, even
if we feel that Dummett could have saved us some of it by organizing his
writing more tightly. A second general criticism is that he only rarely refers
specifically to Frege's writings. I would not have wished him to lengthen
the book still further by an elaborate apparatus of scholarly footnotes, but
surely the citing of textual support for attributions of view is something
which nobody however authoritative can be allowed to neglect. Such refer-
ences would have been very useful to all of us who are not as thoroughly
steeped in Frege's published and unpublished writings as Dummett is.
On purely historical matters, I have found only one minor point on
which to question Dummett's judgement. He claims (pp. xxiii-iv, 658) that
in 1912 Frege was totally discouraged in his work in the philosophy of
mathematics, and cites as evidence a letter to Russell in which Frege refused
an invitation to speak in Cambridge. But T. W. Bynum claims that until
1914 at least Frege still believed that the logicist programme could be carried
out, and cites a letter of 9 June 1912 (shown by Russell to Bynum) making
clear that Frege had accepted the invitation, but was forced to cancel because
his adopted son was ill.'
The proof-readers are to be congratulated. I noticed only five misprints
-p. 59, line 14 ; p. 230, line 5 ; p. 317, line 16; p. 539, line 21 ; p. 666, line 6.
It would take another book to deal with all the issues raised by Dummett.
What I will do here is to take up some main themes on which I feel his
argument can be usefully clarified and in certain respects criticized.
1. LOGICAL SYNTAX: QUANTIFIERS, PREDICATES, AND PROPER NAMES
One can have no real grasp of Frege's philosophy of language until one
has understood his logical analysis of sentences. The centrepiece of this
analysis is the notion of quantification, which Frege discovered at the out-
set of his career when he developed a logical notation in which mathematical
proofs could be rigorously expressed. Dummett reinforces the Kneales' judge-
ment2 that Frege's use of quantifiers solved the problem of multiple generality
which had blocked the progress of logic for centuries, despite the ingenuity
of the mediaeval logicians. There is a logical complexity in 'Everybody
loves somebody' which is absent from ' John loves Mary '. We must under-
stand the former as constructed in stages, by which ' loves somebody' forms
a unit to which ' everybody' is attached, whereas the latter can be regarded
as composed out of its three simple constituents in a single stage. Frege's
analysis, the basis of all modern logic, takes atomic sentences as composed
of one or more proper names with one simple predicate or relational expres-
sion, and treats all other sentences as constructed out of atomic sentences
by a sequence of operations of three kinds: the joining of sentences by
sentential connectives to form complex sentences, the omission of one or
more occurrences of a proper name from a sentence to form a predicate, and

'Frege, Conceptual Notation and Related Articles, ed. T. W. Bynum (Oxford, 1972),
p. 50.
2W. & M. Kneale, The Development of Logic (Oxford, 1962), p. 511.

This content downloaded from 91.220.202.121 on Sat, 28 Jun 2014 10:10:36 AM


All use subject to JSTOR Terms and Conditions
DUMMETT ON FREGE 351

the application of a quantifier to a predicate to form a complex sentence


(p. 16). Frege's elusive notion of "incompleteness " applies primarily to
the complex predicates, such as 'is bald and loves Mary', which are con-
structible only by the removal of names from sentences; and the notion of
such predicates is required only to explain the structure of sentences con-
taining quantifiers, such as 'Nobody is bald and loves Mary' (pp. 27-33).
Dummett suggests that the distinction between the expressions which
go to make up atomic sentences and those which, like sentential connectives
and quantifiers, form complex sentences, allows us to draw a sharp distinc-
tion between logic, as concerned with the latter, and the rest of philosophy,
which may be concerned with some expressions of the first type, e.g., ' cause '
(pp. 21-2). But surely there are many expressions of the second type, such
as 'was disappointed that', 'wonders whether', etc., the analysis of the
meaning of which we would hardly want to say is in the province of logic.
In Chapter 3 Dummett outlines the categorial grammar implicit in
Frege's analysis of sentence-structure. From the two basic types of complete
expression-sentences and proper names-derived types can be formed by
omitting expressions of any types already formed. Thus predicates can be
formed from sentences by omitting one or more occurrences of a name, and
quantifiers can be seen as what is left in a sentence when a predicate is
omitted. But Frege does not allow us to omit something from an expression
which is already incomplete-in the material mode, he does not countenance
functions whose values are functions (pp. 40-3). Frege's hierarchy of gram-
matical categories is essentially the same as Russell's simple theory of types;
it is potentially infinite, but in any particular language we shall need to
consider only a small number of types-those which contain simple signs in
the language, and those required as the complements of simple signs when
forming a sentence or complex singular term (p. 48).
This hierarchy is based on the notions of sentence and proper name.
Presumably the former can be taken as given, but Dummett feels that Frege
fails to give explicit criteria for the latter, and tries to fill the gap in Chapter
4. The ontological category of objects must be derived from the linguistic
category of proper names, rather than vice versa (pp. 55-7), so we should be
able to say what it is about the use of an expression that makes it count as
a proper name. Dummett offers criteria in terms of the validity of simple
inference-patterns involving expressions of generality (pp. 59-60), but holds
that these must be supplemented by Aristotle's insight that names do not
have contraries (p. 63), and by a means for distinguishing first-order from
higher-order generality (pp. 67-9). He then says that these purely " gram-
matical " criteria will have to be further qualified, if we are not to admit a
host of miscellaneous substantival expressions such as 'the lack of x ', 'the
whereabouts of y ', ' sliminess ', 'reciprocity', etc., as standing for objects.
At this stage he appeals to the notion of a criterion of identity, first explicitly
introduced into philosophy by Frege.3 Every proper name must have a
sense which gives a criterion of identity for the object named-a means of
settling whether the object is the same as or different from an object other-
wise identified. So a proper name can be expected to belong to a range of
names for objects in the same category (i.e., having the same criterion of
identity), and to have a certain vocabulary of predicates appropriate for
objects of that category (pp. 72-80). Thus words for colours, chemical
substances, animal species, and cardinal numbers are proper names, but the
SFrege, The Foundations of Arithmetic, trans. J. L. Austin (Oxford, 1950), ?62, p. 73.

This content downloaded from 91.220.202.121 on Sat, 28 Jun 2014 10:10:36 AM


All use subject to JSTOR Terms and Conditions
352 LESLIE STEVENSON

general run of abstract nouns like 'sliminess' and 'resemblance ' are not,
since there is no definite criterion of identity for tactual qualities or for
relations. Like Dummett, I shall return to the notion of identity after
discussing that of reference.
2. SENSE, REFERENCE, AND TRUTH-VALUE
The distinction between syntax and semantics is implicit in Frege's work,
for as well as axioms and rules of inference, he offers an account of how the
truth-values of sentences are determined. For purposes of logic alone, a
single-layered semantics is enough, assigning objects to proper names, pro-
perties to predicates, and relations to relational terms, and hence truth-
values to atomic sentences in the natural way. Dummett suggests that
Frege's notion of the reference of an expression is best approached this way,
as the assignments made in such a semantics (pp. 89-90).
Frege's famous distinction between the two semantic layers of sense and
reference is needed only when we proceed beyond formal logic to the philo-
sophy of language, when we try to say how we can establishsuch associations
between expressions and their reference. The sense of an expression is the
means by which its reference can be determined, and this is what we know
when we understand an expression. For example, if we understand the
phrase 'the tallest man in France ', we know what it would be to establish
of some man that he is the tallest in France, but we may not know which
man, if any, deserves the description. The most important notion to do
with meaning is that of knowing the meaning, and Frege's notion of sense
was first introduced in connection with the notion of cognitive value.4 So
although he is rightly famous for his cleansing of logic from psychology, it
is not true to say that he extruded all epistemology from his philosophy of
language (pp. 92-5, 679-82), as Wittgenstein did (to excess) in the Tractatus.
The sense of an expression is supposed by Frege to be public, objective,
shared by all speakers of the language. It is emphatically not a matter of
mental images or any other kind of private psychological episode, which he
assumes to be variable and incommunicable. Sense is also to be distinguished
from two other elements in the intuitive notion of meaning-from "tone ",
as in the difference between ' sweat ' and ' perspiration ', and from "force ",
as in the difference between asserting something and asking whether it is
true. The sense of an expression is that part of its meaning which contributes
to determining the truth-value of sentences in which it occurs.
But can we distinguish the respective contributions of meaning and fact ?
Is there always one standard way, common to all speakers of the language,
of establishing the reference of an expression ? Quine's scepticism on this
point must be faced. Frege's picture is not generally true of our actual use
of language; he himself realized that different users of a proper name refer-
ring to the same person may attach different senses to the name.5 However,
Dummett does a lot to ease the consciences of those of us who have continued
to employ the notion of sense despite our awareness of Quine's sceptical
arguments. In Chapter 17 he makes an extended inquiry into the coherence
of Quine's position, and concludes that it is even further from the facts than
Frege's, since an integral part of our use of language is that we enquire into
4Frege, " On Sense and Reference ", translated in Philosophical Writings of Gottlob
Frege, ed. Geach and Black (Oxford, 1952), hereinafter referred to as G & B.
6Frege, " The Thought ", translated in Philosophical Logic, ed. Strawson (Oxford,
1967), pp. 24-5.

This content downloaded from 91.220.202.121 on Sat, 28 Jun 2014 10:10:36 AM


All use subject to JSTOR Terms and Conditions
DUMMETT ON FREGE 353

the justification of statements, and thus accept certain sentences as determin-


ative of meaning. Frege's notion of sense is an ideal, but it is an ideal towards
which we strive.
Frege's notion of reference is also open to objection, for it is not at all
obvious that the semantic roles of singular terms, predicates, and sentences
all involve a relation to an extra-linguistic entity. Yet just such a conception
is implied by combining the two strands which Dummett repeatedly dis-
tinguishes in Frege's notion of reference-(1) as the semantic role of an
expression, and (2) as a relation to an extra-linguistic entity. Frege first
introduced the sense-referencedistinction for singular terms,6 but immediate-
ly extended it to the case of sentences; assuming the extensional principle
that replacement of an expression by another with the same reference must
preserve the reference of the whole sentence, he took the truth-value of the
sentence as its reference.7 (If we replace an expression by another with the
same sense, then we preserve the sense of the whole sentence, which Frege
called the thought.8) He also maintained that incomplete expressions have
references which are themselves incomplete, and called the reference of a
predicate a concept.9 The principle of extensionality leaves room for the
application of the sense/reference distinction to predicates too, for two
predicates can be said to have the same reference if they apply to just the
same objects.10 The difference between objects and concepts is not the
extensional/intensional contrast, but the complete/incomplete distinction
(derived from syntax) (pp. 208-9). It was only in later writings that Frege
explicitly talked of incomplete expressions having incomplete senses as well
as incomplete references,1 but the distinction between the senses and the
references of all expressions, including incomplete ones, is implicit in all his
philosophy of language.
Frege's two-level system of semantics can thus be summed up in the
following table:

EXPRESSIONS singular predicates and sentences


terms relational terms

SENSES complete incomplete senses thoughts


senses

REFERENCES objects concepts and truth-values


relations

(From about 1890 onwards he tended to treat sentences as a special kind of


singular term, and hence truth-values as a special kind of object.) His
theory of how the truth-value of a sentence is determined in practice can
6Frege, " On Sense and Reference " and "Function and Concept ", p. 57 and p. 29
in G & B.
7G & B pp. 62-3, and p. 31.
8G & B p. 29 and p. 62; "The Thought", p. 20.
9G & B pp. 43 ff. (in " On Concept and Object ").
10G & B p. 80 (in Frege's review of Husserl's Philosophie der Arithmetik).
"Frege, " Compound Thoughts ", translated as Appendix B in Essays on Frege,
ed. E. D. Klemke (Urbana, 1968), p. 541.

This content downloaded from 91.220.202.121 on Sat, 28 Jun 2014 10:10:36 AM


All use subject to JSTOR Terms and Conditions
354 LESLIE STEVENSON

be illustrated by the following " flow-path" for the simplest case of an


atomic sentence:12

singular term predicate sentence


l l
sense sense thought
I I t
object concept truth-value
I I t

When hearing or reading a sentence we must be aware of its components


and of the syntactic structure by which they are put together; then know-
ledge of the senses of these components gives us understanding of the mean-
ing of the whole sentence (p. 152). Frege's theory is that the sense of each
component determines its reference, and that these references then determine
the truth-value of the whole (p. 159). This is what Dummett calls the
" direct means " of establishing truth-values (p. 237), that which corresponds,
step by step, with the structure of the sentence. In some cases there may
be a short-cut via the thought, for instance in 'The tallest Frenchman is a
man ', where obviously we do not need to start taking tape-measures across
the Channel, since we can see from the sense of the whole sentence that it is
true. But in general, the path must go via the references of the components.
How adequate is this as an account of how we understand sentences and
determine whether they are true or false ? At the level of sense, it is a spelling
out of the obvious " compositional " conception of the meaning of a sentence
as determined by the meaning of its constituents and the syntactical struc-
ture of their combination. It is hard to imagine how any other account
could be given of the familiar fact that we can understand sentences which
we have never heard before, provided we know the grammar of the language
and the meaning of the words. As Dummett puts it (p. 4), in the order of
recognition, the sense of a word is primary; although when we want to give
a general account of what it is for a word to have a sense, we have to say
that its sense consists in the contribution it makes to determining the sense
of sentences in which it occurs.
But at the level of reference, Frege's account looks highly questionable.
Are the relationships between predicate and concept, and between sentence
and truth-value, the same as, or even analogous to, that between singular
term and object ? Indeed, can we even say that there are such things as
concepts and truth-values ? Does anything important hang on maintaining
the neatness of the above diagram by putting in concepts and truth-values
alongside objects in the " realm of reference" ?
Dummett reserves his strongest condemnation for Frege's assimilation of
sentences to proper names, calling it a " ludicrous deviation ", a " gratuitous
blunder " (p. 184). The motivation for it seems to be the resulting simplifica-
tion in ontology, whereby truth-values become just a special case of objects,
and concepts a special case of functions (p. 50). But is there even an analogy
between sentences and proper names ? Well, sentences can form parts of
12My diagram is an adaptation of that used by Wiggins in " Identity-Statements"
in Analytical Philosophy, second series, ed. R. J. Butler (Oxford, 1965), p. 57.

This content downloaded from 91.220.202.121 on Sat, 28 Jun 2014 10:10:36 AM


All use subject to JSTOR Terms and Conditions
DUMMETT ON FREGE 355

larger sentences, and if the connectives are extensional, it is only the truth-
values of the constituent sentences which matter for the truth-value of the
whole. (When the connective is not extensional, Frege maintains the general
principle of extensionality by saying that the reference in such cases must
be what is ordinarily the sense:13this is his doctrine of " indirect reference ",
to which Dummett devotes an important chapter (9) which I bypass here.)
But what we must realize is that there are considerable disanalogies between
sentences and proper names: firstly, truth-values are not identifiable con-
stituents of reality in the way that material objects are (pp. 406-8); and
secondly, although complex singular terms must lack reference if one of
their constituents lacks reference, there is no obstacle to our ascribing truth-
values to sentences containing names without bearers (p. 412). This then is
one of the points where the two strands in Frege's notion of reference come
apart.
The question of the reference of incomplete expressions calls for a subtly
different revision of Frege's theory. Dummett's position appears to be that
although there can be no doubt about the existence of concepts and relations,
provided we can use higher-order quantification (p. 245), as he thinks we
can (p. 541), there is still not a strong enough analogy between proper names
and predicates for Frege to apply his two-stranded notion of reference to
the latter (pp. 243, 428). When first introducing Frege's notion of reference,
Dummett says that it coincides with the modern semantics for first-order
logic (p. 90). But that semantics assigns sets of objects to predicates, whereas
Frege ascribes incomplete entities (concepts) to them. It is clear that the
normal " direct means " of establishing the truth-value of an atomic sentence
does not proceed via the set of objects of which the predicate is true. We do
not establish that Fido is brown by first identifying the set of all brown
things and then seeing whether Fido is among them, for clearly it must
already be discovered whether Fido is brown before we could say whether
he is a member of the set of brown things. In this respect at least, the
standard set-theoretic semantics, however useful for proving metatheorems
about completeness, etc., gives no understanding of the actual working of
language. (It is worth noting that a similar remark applies to the possible-
world semantics for modal logic-for we do not establish whether p is possible
by looking through the set of possible worlds to see whether p is true in at
least one of them; rather we would first have to show whether p is possible
before we could say whether it is true in some possible world-cf. Dummett,
pp. 284, 290.) But is it even true that the establishing of truth-value pro-
ceeds via the identification of a concept(or property, p. 173) as the reference
of the predicate ? That is, do we first bring into play our knowledge of the
sense of the predicate ' is brown ', then establish what property this predicate
stands for, and finally establish whether the object Fido has that property ?
It is not clear that there is any real distinction between the first two stages
here, as there undoubtedly is in the case of singular terms. Should we not
just say that to know the sense of a predicate is to have a criterion for
deciding, for any given object, whether the predicate applies to that object ?
Dummett suggests just this as a " first approximation " (p. 229), and the
qualifications he makes do not seem to affect the matter at issue here, so
he concludes that it is virtually impossible to make any sense of the concep-
tion of identifying a concept as the reference of a predicate (p. 241).
It would seem then that we should amend our flow-path diagram by
13G& B p. 59, pp. 66 ff.

This content downloaded from 91.220.202.121 on Sat, 28 Jun 2014 10:10:36 AM


All use subject to JSTOR Terms and Conditions
356 LESLIE STEVENSON

simply omitting concepts as a separate stage in the process, and by not


attempting to put truth-values on a level with objects. So it can now be
drawn as follows:
-I---- -+ predicate--sense- ---

sentence truth-value

I ~~--singular--sense >object--->
term

There is a further aspect of the matter which we have ignored until now
-the dependence of an expression on the context in which it is uttered or
written. For instance, the reference of a pronoun may depend on the linguis-
tic context, and of a demonstrative expression on the non-linguistic context,
and the truth-value of a sentence may depend on the time and place of its
use. To do justice to these well-known " token-reflexive " aspects of language-
use, we must distinguish between the tokensof linguistic expressions (partic-
ular sounds and marks) and the types (the recurring patterns which sounds
and marks can instantiate). Frege's interests were primarily in mathematical
and logical uses of language, so he did not even consider token-reflexiveness
until his late essay " The Thought ". There he maintained that it is thoughts
that primarily are true or false, and that thoughts have a timeless existence
(in the realm of sense) so that they logically cannot change their truth-values.
So he had to modify his definition of a thought as the sense of a sentence,
and admit that the context often determines what thought is expressed by
a particular use of a sentence. Dummett shows that Frege's argument to
the timeless existence of thoughts has no force (p. 370), and, after a discussion
of the phenomenon of tense as the most important kind of token-reflexiveness,
concludes that the notion of a thought is a secondary construct, since our
utterances themselves are the primary bearers of truth-value (p. 400). This
being so, we can amend our flow-path to make clear that that of which we
establish the truth-value is a sentence-token used in a particular context:

------ >predicate---sense >

sentence- -- sentence- ->truth-


token type value

-I -->singular -- sense --object -


term t
context I -

This is as far as I shall pursue this flow,path method of analysis here, but
it may be worth remarking on some of the extra complications which would
be required to make the picture more realistic. Firstly, I have been treating
only atomic sentences so far: complex sentences would require their connec-
tives, and the structure of the whole sentence, to figure in the analysis.
Secondly, I have not attempted to distinguish the different ways in which
linguistic context, non-linguistic context, and the actual facts or states of
affairs in the world, make their contributions to the determination of truth-
value. The last diagram above does however make clear that it is expression-

This content downloaded from 91.220.202.121 on Sat, 28 Jun 2014 10:10:36 AM


All use subject to JSTOR Terms and Conditions
DUMMETT ON FREGE 357

types which primarily have sense, but tokens of those types which have
reference. (According to Frege's doctrine of indirect reference in intensional
contexts, the sentential context can affect the sense as well as the reference
of an expression, but Dummett claims that there is no need for a theory of
indirect sense (p. 268).) Thirdly, in cases of word-ambiguity, the sense of a
word-token may be determined by the context. Fourthly, in cases of syn-
tactic ambiguity, even the syntactic structure of the sentence may depend
on the context.
3. ONTOLOGICAL COMMITMENT, EXISTENCE AND IDENTITY
Dummett explains in Chapter 14 and elsewhere how Frege's logical
syntax makes a complete break from the traditional approach to ontological
questions, still exemplified in Strawson's Individuals. In that tradition, the
questions of ontology are what particulars there are, and what universals
(if any) there are, the assumption being that a universal can be alluded to
both by a predicate like 'is red' and an abstract singular term like 'red-
ness'. For Frege this is a logical mistake, since it is senseless to suppose
that one entity could be referred to both by an incomplete and by a complete
expression. Objects (the referents of proper names) and concepts (the refer-
ents of predicates) are radically different, and Frege marks the ontological
gulf by his insistence that concepts themselves are incomplete.14 So for him
the ontological questions are what objects there are, and what concepts and
relations, etc., there are (p. 473).
Quine's well-known criterion of ontological commitment, that to be is to
be the value of a variable, i.e., that one is committed to the existence of all
those objects which must be in the range of values of one's bound variables
of quantification if what one says is true, is derived from Frege's under-
standing of objects as the domain over which first-order quantifiers range.
But Dummett maintains that Quine has over-simplified ontological ques-
tions by reducing them to questions about objects only, dispensing with
higher-order quantification (pp. 223-6, 479). Quine draws a sharp distinction
between bound variables and schematic letters: the former are explained as
ranging over a set of entities, the latter only as showing places for linguistic
expressions. More recently, a similar distinction has been drawn between
two ways of interpreting quantification itself: objectually, in terms of extra-
linguistic entities; and substitutionally, in terms of linguistic expressions.
This has led many writers, including myself,15 to say that it is only when
objectually interpreted that quantification carries ontological commitment,
and thus higher-order quantifiers need not commit one to the existence of
concepts or properties as the references of predicates. However, Dummett
claims that there is no ultimate distinction to be drawn here,16since we are
bound to ascribe reference to any expression which finctions as a significant
unit of the language (pp. 526-8).
Now this is one crucial place where I find that Dummett's position does
not fully cohere with what he has said earlier in the book. For the upshot
of his extended discussion of Frege's notion of reference seemed to be that
the two strands in the notion of reference-as semantic role, and as relation
to an extra-linguistic entity-pull apart when the notion is extended from
14G& B pp. 43 ff.
5Stevenson, " Frege's Two Definitions of Quantification ", PQ, 23 (1973), 207-23.
16Dummett's position here is supported by Bostock in his recent book Logic and
Arithmetic (Oxford, 1974), Ch. 3.

This content downloaded from 91.220.202.121 on Sat, 28 Jun 2014 10:10:36 AM


All use subject to JSTOR Terms and Conditions
358 LESLIE STEVENSON

the paradigm case of singular terms to the supposedly analogous cases of


predicates and sentences. Indeed, he says on p. 427 that the word ' reference '
itself ceases to be useful to us, for just this reason. Yet here he is, 100 pages
later, still using it in a situation where its two strands of meaning must surely
be distinguished. For it is tautologous to say that every significant unit of
language has a semantic role, but highly controversial to hold that in every
case that role must consist in a relationship to an extra-linguistic entity.
He suggests that Quine has no alternative account to offer for the semantic
role of predicates (p. 525), but surely Dummett himself has implicitly offered
such an alternative in the account of predicates outlined in the third diagram
above (cf. pp. 229-44).
If there is no such thing as the identification of a property or concept
as the referent of a predicate (p. 241), or the identification of a truth-value
as the referent of a sentence (pp. 412-3), surely we should conclude that the
use of predicates and sentences does not commit us to the existence of con-
cepts and truth-values ? One of the recurring themes in Dummett's book
is that if we are to say that a certain entity exists, we should be capable of
identifying that entity in some way other than as the referent of one type
of linguistic expression (pp. 73 ff., 232, 407, 488-9). On the question of the
existence of abstract objects, he requires in the end some analogue of observa-
tion (p. 511). So it seems inconsistent for him to maintain that there is no
room for doubt about the existence of concepts (pp. 245, 541).
The notions of identification and of a criterion of identity play a fund-
amental role both in syntax (as we have seen in section 1 above) and in
semantics, for we have just seen how the existence of extra-linguistic entities
is bound up with the means of identification available for them. Thus one of
the main lessons to be learnt from Frege is, as Dummett puts it on p. 503, that
" what objects we recognizethe world as containing depends upon the structure
of our language. Our ability to discriminate, within reality, objects of any
particular kind results from our having learned to use expressions, names
or general terms, with which are associated a criterionof identity which yields
segments of reality of just that shape ".
So Quine's ontological question " What is there ? "17 must be glossed as
"What objects are there ? ", where the notion of object is understood in
Frege's way. Dummett's work deepens our understanding of Frege's
insight that different kinds of object have different criteria of identity. So
we must put a further gloss on the ontological question, and rephrase it as
" What categories of object are there ? ", for after all the ontologist is not
asking for a complete list of objects, nor even for a list of all kinds of object,
if a " kind " is understood as given by any general term at all. The notion
of category needed here is that of a class of objects all of which have the
same criterion of identity.18 Dummett claims that there must always be a
categorial term corresponding to a criterion of identity (pp. 75-6), although
he displays some hesitancy about how to identify such terms, and later
suggests that it is only normally the case (but not universally ?) that there
is such a term (p. 546). Further philosophical analysis is needed here, for
if we are to talk of two terms supplying the same, or a different, criterion
of identity, then clearly we are assuming that we have a criterion of identity

17Quine, " On What There Is ", in From A Logical Point of View (Cambridge, Mass.,
1953), p. 1.
18The notion is closely akin to Wiggins' notion of an ultimate sortal, in Identity and
Spatio-Temporal Continuity (Oxford, 1967), pp. 32-3 and footnote 40.

This content downloaded from 91.220.202.121 on Sat, 28 Jun 2014 10:10:36 AM


All use subject to JSTOR Terms and Conditions
DUMMETT ON FREGE 359

for criteria of identity themselves! This assumption is vulnerable to


scepticism.
There are many other deep issues raised by Dummett (particularly that
of verificationism against realism in the theory of meaning) which I have
not had space to discuss here. Let me end by recommending all who call
themselves analytical philosophers to ponder deeply the works of Frege and
Dummett on Frege. For this is a book worthy of the genius of its hero, and
there can hardly be higher praise than that.

University of St. Andrews

This content downloaded from 91.220.202.121 on Sat, 28 Jun 2014 10:10:36 AM


All use subject to JSTOR Terms and Conditions

You might also like