The Power of Russell's Criticism of Frege: 'On Denoting' pp.

48-50
Author(s): Simon Blackburn and Alan Code
Source: Analysis, Vol. 38, No. 2 (Mar., 1978), pp. 65-77
Published by: Oxford University Press on behalf of The Analysis Committee
Stable URL: http://www.jstor.org/stable/3327496 .
Accessed: 28/07/2013 10:16

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 and The Analysis Committee are collaborating with JSTOR to digitize, preserve and
extend access to Analysis.

http://www.jstor.org

This content downloaded from 129.78.72.28 on Sun, 28 Jul 2013 10:16:29 AM

All use subject to JSTOR Terms and Conditions
ANALYSIS 38.2 MARCH 1978

ANALYSIS "PROBLEM"NO. 17

THE seventeenth problem is set by Professor G. E. Hughes of the

Victoria University of Wellington, and is as follows:
Can I ever, by my subsequentactions,bring it about that something
I did on a previousoccasionwas donefroma certainmotiveratherthan
fromsomeotherone?
The word limit is 6oo words. Entries should reach the Editor of
ANALYSIS by 3' August 1978; they should not be sent to Professor
Hughes. Entries will not be acknowledged or returnedunless accom-
panied by stamps or internationalpostage coupons. Contributorsmay
submit entries under their own names or a pseudonym. Contributors
must be underthe age of thirty, or undergraduatesor graduatestudents.
A report with any winning entries will be published in volume 39
of ANALYSIS.The ANALYSIS Committee has voted a sum of ?4o which
will be awarded as a prize if Professor Hughes finds a sufficiently
deserving contribution.
The report on Problem No. i6 will appearin the next issue.

THE POWEROF RUSSELL'SCRITICISMOF FREGE:

'ON DENOTING' pp. 48-50

By SIMONBLACKBURN and ALAN CODE

i. SETTING the Scene.

In 1905 Russell first published, in 'On Denoting', the theory of
descriptions.He was conscious at the time that the theory made a pro-
found break from its predecessors,the theories of Meinong, of Frege,
and of his own work of two years earlier. Thus he included in his
articlea discussionof the weaknessesof those theories.His readershave
had no difficultyin understandingjust what his criticismsof the Mein-
ongian theoryare-indeed for a very long time it was generallyconceded
that he had successfullyunderminedit. His criticismsof Frege, and of
his own earliertheory which he regardedas similar,were not blessed
with such a fortunate reception. To the best of our belief only one
author, A. J. Ayer, apartfrom Russell himself, has ever acknowledged
in print that Russellhad an argumentwhich seriouslythreatenedFrege's
view. It is usually held that either he had no argumentat all againstany
previous theory, becausehe was hopelessly muddled, or that at best he
65

This content downloaded from 129.78.72.28 on Sun, 28 Jul 2013 10:16:29 AM

All use subject to JSTOR Terms and Conditions
66 ANALYSIS

had succeeded in undoing his previous view, expounded in The Prin-

ciplesof Mathematics [io], but totally failed to engage Frege. Dummett
thinks that Russell does, in a confusedfashion, manageto make a minor
point which Frege can easily evade. These assessmentsare quite wrong.
We do not say, on this occasion, that Russell had a decisive, or over-
whelming, objectionto Frege. But we do say that he presenteda serious
argument and even "considerations capable of determining the in-
tellect" against Frege's theory. To prove this we offer a rational
reconstructionof the trainof thought found in pp. 48-50 of 'On Denot-
ing', as numberedin the collection LogicandKnowledge, edited by Robert
Marsh[9].For conveniencewe shall letter the relevant eight paragraphs
by the italic capitals (A) to (H), and refer to them in this way. (A)
begins at the bottom of p. 48 (The relation of the meaning to .. .) and
(H) at the bottom of p. o50(That the meaning .. .).
The issues are bedevilled by problems of terminology. We shall
call both proper names and definite descriptions 'denoting phrases'.
These are the things which are in dispute. We shall talk of them having
a 'sense' and a 'reference'. The reference of the word 'Aristotle' is
Aristotle, and the word refers to Aristotle. The sense of the word is
supposed to be a third entity. It is not Aristotle, nor is it the word
'Aristotle',but insteadis some entity associatedwith the word 'wherein
the mode of presentationis contained' [6] p. 57. Part of Russell's point
will be the difficultyof referringto such a thing, so we must be pardoned
for sayingno more about it. We shall call the view that denoting phrases
have usuallyboth sense and referencethe three-entityview, for obvious
reasons. We also need a way of describing the relationshipsthat hold
between these three things. We shall say that a word 'refers to' its
referenceand 'expresses'its sense. If a sense relatesto a thing in such a
way that a word which expressesit refersto that thing, or in such a way
that a word which were to expressit would refer to that thing, we shall
say that the sense 'determines'the thing. Of course furthertheses about
all these things would be needed to describefully anyone's theory, but
for the presentwe merelyintroduce the terminology.

z. Previous Interpretations.
As we stand nearly alone in our insistence that Russell was at least
in the right ball-park,we would like to say a few words about the more
usual views.
Church [41 P. 302, and Butler [z] pp. 361-363 hold that Russell's
attemptto refuteFregeis vitiatedby his (characteristic)failureto observe
the use/mention distinction. On this view, once some consistent use is
made of perspicuousnotational devices distinguishing talk of expres-
sions from talk of senses of expressions, and both from talk of references
of expressions, then the alleged argument vanishes. Now Cassin [z]

This content downloaded from 129.78.72.28 on Sun, 28 Jul 2013 10:16:29 AM

All use subject to JSTOR Terms and Conditions
 RUSSELL'S CRITICISM OF FREGE 67
p. 269, hasshownthatthereis no wayof interpreting Russellif Church
andButlerareright.And on the faceof it the viewis implausible. While
it is undeniablethatRussellis quitecarelessin his notation,he wasvery
sensitiveto the underlyingpoint,which,indeed,functionscentrallyin
his own argument.Paragraph(D) is mainlydevotedto explainingit.
SinceRussellgave up his own earlierpositionbecauseof the present
argument,Churchand Butler put him in the following ridiculous
position.Thereis an argumentwhichpurportsto refuteRussell'sold
view, and which emphasizesand explainsthe use/mentiondistinction.
But the argumentdisappears oncethis distinctionis made.Nevertheless
Russellfeels forcedto abandonhis old view. Clearlywe shouldnot
acceptthisexegesisunlessabsolutelyno otheris forthcoming:it offends
againstthe cardinalprincipleof Russellianexposition,to wit: don't
makehim out a completefool if you canhelpit.
Morecommonis the view thatRusselldid have somesortof argu-
mentgoing,onlyit bearson his earlierposition,not on Frege,who only
gets draggedin becauseof Russell'snotoriousinabilityto readproperly.
This is the view expressedin Geach[7]. As he thereputs it: "readers
of 'On Denoting'will find it best simplyto ignorehis use of Frege's
name".To approachthis we must have some idea of how the two
theoriesdiffer--Geachdoesnot tell us whichfeatureof Russell'searlier
theoryrendersit, but not Frege,a possibletargetfor "OnDenoting".
There are two important differences between the theory of The
(whichwe referto as PoM) and Frege. The
of Mathematics
Principles
first, however, turns out to be basicallyterminological,and the second
cannot possibly matter to Russell's argument. The first is that for
Russell it is concepts which denote, so that it is the denoting concept
meant by a definite descriptionwhich denotes the thing referredto, or
denotation. On Frege's theory it is the definite descriptionitself-the
actual expression-which refers to the reference.Thus we cannot have
all of: sense=meaning; reference= denotation; referring= denoting;
for this does not map Russell's theory onto Frege's. In Frege one kind
of thing (an expression)refers, and in Russell a differentkind of thing
(a concept) denotes. But this is trivialby itself, since, if there is a relation
which obtains between the meaning of the description and the de-
notation, we can define in terms of it another relationholding between
the description itself (the words) and the denotation of its associated
meaning. And likewise if there is a relation which obtains between a
word, or group of words, and its reference,we may define in terms of
that relation another one holding between the sense of the expression
and that reference. Using our triangle of terms we can introduce 're-
ferring'once we have 'expressing'and 'determining',or we canintroduce
'determining' once we have 'expressing' and 'referring'. The former
order of definition, Russell's, seems preferable, since in Russell and Frege

This content downloaded from 129.78.72.28 on Sun, 28 Jul 2013 10:16:29 AM

All use subject to JSTOR Terms and Conditions
68 ANALYSIS

a word has the referenceit does as a consequence of having a certain

sense, and of this sense determining a certain thing. This point is
importantto Russelllater,as we shall see in section 3, but for the present
it is clearthatthe differenceof terminologymarksno respectin which his
own theory is less acceptablethan Frege's.
The second differenceis that for Frege the sense/referencedoctrine
applies to every singular term, whereas at the time of PoM Russell
applied his analysis (into denoting concept/object indicated) only to
definite descriptions, and not to ordinary names. Russell explicitly
notes this differencein Appendix A of PoM ? 476. If this feature made
him vulnerable rather than Frege, it could only be because in 'On
Denoting' he is arguing that the three-entityanalysisis so excellentthat
it must be appliedright acrossthe board.In fact, of course,his laterview
is that it should not be applied to anything, and therefore engulfs
Frege as well as his own more limited thesis.
So although there are these two respects in which the Russell of
PoM differsfrom Frege, neitherof them representsa weaknesswhich he
is later regretting,and it is not possible that 'On Denoting' refutes the
former but not the latter.
Authors have not failed to invent featuresof Russell's earlierview,
so that his argumentcan be seen as relevantto them, but not to Frege.
Cassin,who believes that Russell did not even intend to attack Frege,
holds that it was part of his earlierview (but not part of Frege's) that
'only terms could be denoted' but now Russell has come to realizethat
denoting concepts themselves must be denoted [2] p. 269. This is just
wrong about the earlierview. A term, in PoM, is anything which can be
a
the subjectof proposition, or an of
object thought, or a logical subject
47) and both things andconcepts are terms 48). Appendix B, ? 483
(? (?
criticizes Frege precisely for his difficulty over making concepts into
logical subjects.And in ? 476 Russell says:
If oneallows,asI do, thatconceptscanbe objectsandhavepropernames,
it seemsfairlyevidentthat theirpropernames,as a rule, will indicate
[i.e. referto] themwithouthavingany distinctmeaning[i.e. sense];but
the oppositeview, thoughit leadsto an endlessregress,does not appear
to be logicallyimpossible.
The opposite view would have a higher order sense for any denoting
phrasereferringto a given concept. In Russell'sterminologythis would
mean that when a definite descriptionindicates a denoting concept, it
itself has as its meaning another denoting concept which itself denotes
(i.e. determines)the first. This is not a possibility he rejects.
And if he hadrejectedit in PoMit would have been a simplematterto
change his mind without modifying his earlierview in any other way:
argument which shows that he shouldn't reject it would leave the
three-entity view quite intact.

This content downloaded from 129.78.72.28 on Sun, 28 Jul 2013 10:16:29 AM

All use subject to JSTOR Terms and Conditions
 RUSSELL'S CRITICISM OF FREGE 69
Cassin (p. 270) also claims that there is another bad feature of the
old view. Let us introduce the following abbreviations:
C1=some denoting conceptwhich denotes (i.e. determines)Aristotle
C2=some denoting concept which denotes (i.e. determines)C1
According to Cassinit was part of Russell's old view that:
(A) any proposition containing C, as a constituent will be about
Aristotle
We wanted it to be about C1, but the concept Cq, since it determines
Aristotle, makes the original attempt at denotation "fall through" to
Aristotle anyway. It follows that it is impossible to refer to denoting
concepts by means of expressions which have higher order denoting
concepts as their sense, or in other words it is impossible to refer to
them by descriptions, since such attempts fall through to the original
object, in this case Aristotle. It would follow that denoting concepts of
any level could at best be referredto by names,and this the laterRussell
might be finding objectionable.Searle [12] sees the argumentin essen-
tially the same way, only he regards (A) not so much as a feature of
Russell's early view, but merely as a consequenceof the conjunctionof
Frege's theory with its negation, which is all, according to him, that
Russell succeededin attacking.
There are two major problems with this view. Firstly (A) is not
stated in PoM. Nor have we found any passage into which it could
possibly be read. Secondly,had Russell surreptitiouslyheld (A) he could
hardlyhave failed to notice its effecton the regressof denoting concepts.
Yet the passagewe have quoted shows that at that time, when explicitly
discussing Frege, he thought the regress quite possible. So to interpret
Russell this way we have to suppose that he held a view which he never
stated, in spite of the thoroughness of PoM, but was unfortunately
incapableof drawingits most elementaryconsequences.It is but a short
way from here to the theory of Jager [6], who holds that Russellwas
refutinga strawmanespeciallydevelopedto be refutedin 'On Denoting'.
All these expositions offend against our cardinalprincipleof Russellian
exegesis. Clearly none of them would appear plausible but for the
difficulty of finding a sensible argument in 'On Denoting', against
either Frege or PoM. So if we can present such a thing, they may be
thankfullyabandoned.
The same stricturesdo not apply to A. J. Ayer [i], who believes
Russell to have a valid argumentagainstFrege. Ayer correctlyidentifies
the force of Russell's conclusion-that there is going to be a mystery
about identifying senses and their relationsto correspondingreferences
-but not Russell's argumentfor that conclusion. In Ayer's exposition
(p. 31i) Russell is represented as demanding that 'the first line of Gray's
Elegy' and 'the meaning of "the first line of Gray's Elegy"' should have
different meanings (senses) but the same denotation (reference). This

This content downloaded from 129.78.72.28 on Sun, 28 Jul 2013 10:16:29 AM

All use subject to JSTOR Terms and Conditions
70 ANALYSIS

would of course be quite inadmissible for Frege: it is in fact just

the demand that sense and reference be identified. But Russell's only
reason for making that demand, on Ayer's account, amounts to com-
plaining that otherwisethere will be a mysteryabout what the meaning,
or sense, is. This may be Russell'sview, but then we still await an argu-
ment for it. We think instead that Russell had a powerful and precise
argumentfor supposing that Frege is doomed to ineradicablemystery,
and it is to the presentationof this that we now turn.

3. The argument.
We do not want to discusstediouslyall of the terminologicaloptions
which paragraph(B) opens. We do not have to, since the point is to find
a sensible readingon which there is a good argumentin the offing, not
myriads of readings on which there is not. Denoting phrases are of
course the expressionsfor which the theory has to work. Their meaning
Russell calls a denoting complex. This marks a departurefrom PoM
where he would have calledit a denoting concept. It correspondsto the
Fregeansense.The thing denoted Russell calls the denotation.Denoting,
as before, is what we have called determining;it is the relationbetween
the senseandthe reference(i.e., the denotingcomplexand the denotation).
Expressions,in this terminology, do not denote. Russell is using single
quotes, here at any rate, in orderto give expressionswhich themselves
referto senses, or denoting complexes.The relationshipwhich he wants
us to consider at the end of (B) is that between sense and the reference
which it determines-i.e., determining.
His claim in (C) is that if an expression has a separate sense in
addition to its reference,then,fatally, there will be no guaranteethat
there is a logical relation between the two. Or, in Russell's words,
'. .. we cannot succeed in bothpreserving the connexion of meaning
and denotation and preventing them from being one and the same'.
It is the businessof (D)--(F) to show this, and the businessof (G)--(H)
to show how the truth of this conditionalwould renderFrege's theory
worthless. The point to be demonstratedis not that there is a difficulty
involved in the idea that a sense may itself be an object of reference,but
ratherthat there is a difficultyinvolved in specifying one in such a way
as to allow us to show that it performsa certainlogical role. Let's see
how he does this.
The firstimportantthing that he says about the determiningrelation
is thatit cannotbe 'merelylinguisticthroughthe phrase'.Whathe means
is that determiningcannot be explainedby Frege in terms of expressing
and referring.To understandthis, we must remembera little of Frege's
theory. In Frege sense is a theoreticalentity, and it is postulated that
denoting phrasescome to referto theirnormalsenseswhen such phrases
are embedded in psychological contexts. This, as is well-known, is to

This content downloaded from 129.78.72.28 on Sun, 28 Jul 2013 10:16:29 AM

All use subject to JSTOR Terms and Conditions
 RUSSELL'S CRITICISM OF FREGE 71

explainhow it is possible that 'George IV wished to know whether Scott

is the author of Waverley', and 'George IV wished to know whether
Scott is Scott' should have different truth-values. Russell finds this
explanationunsatisfactory.As pointed out above, this is not because of
any pure difficultyabout referringto senses. It is ratherbecausein the
absenceof a theoreticaldefinitionof termspurportingto referto senses,
we cannot be sure what logical role sense plays. And Russell wants to
show that there is simply no way to specify senses so as to be sure that
they play the role Frege demands. To highlight this aspect of our
interpretation,we would like to mention once again that in (C) the
difficultyis over ensuringa certainconnexion, and also point out that in
(G) it is failureof explanation,residualmystery,inextricabletangle, that
are said to menaceFrege.
The Russell of PoM was quite content to take in the notion of a
denoting concept (and likewise, Fregean sense) without requiring a
carefultheoreticalintroduction.Phrasesreferringto denoting concepts
were simply taken as indefinable.But as the recalcitranceof the Russell
paradoxbecameapparent,he realizedthat inside even the most innocent
theoreticalconstructionsthere might lurk fatal problems.Laterwe shall
see that he also had good reasonfor a change of hearttowardsdenoting
concepts.
A specificationof sense which does not guaranteeits logical role
would be to sayjust 'in orderto speakof the sense of an expression"A"
one may simplyuse the phrase"the sense of the expression'A' " ' (Frege,
p. 59. Of course, we do not dreamof implying that Frege would have
thought this to be sufficientby itself). If we postulatea sense by simply
mentioning an expressionand describingour entity as the sense of that
expression,then we can explainthe relationbetween sense and reference
only by saying: well, it is the relationwhich holds between the sense of
the term "Aristotle", and the man Aristotle, for example. This would
be to makethe relation'merelylinguistic throughthe phrase'.A forceful
way of seeing why that will not do is to considerthe pair:
(i) Aristotle taught philosophy.
(z) Jones believes that Aristotle taught philosophy.
and the pair:
(3) Aristotle, the magnate,marriedMrs. Kennedy.
(4) Aristotle, the philosopher,wrote books.
In the latter case the relationshipbetween the two things, referredto
respectively by "Aristotle", is merely linguistic through the phrase.
Now it is obvious that there must be a much closer connexionbetween
the referenceof 'Aristotle' in (i) and its referencein (2). It is not an
accidentof homonymy that the word recurs.Indeed, Russell is right in
thinking that there must be a logical connexion between (i) and (2)-
and thereforebetween the referenceof'Aristotle' in each-for they mate

This content downloaded from 129.78.72.28 on Sun, 28 Jul 2013 10:16:29 AM

All use subject to JSTOR Terms and Conditions
7z2 ANALYSIS

together to yield the conclusion that Jones believes something true.

So unless we are able to say something further about this connexion,
the theory of sense will be quite unintelligible:there would be nothing
but logicallyirritatingpunningto link the sentences(i) and (2). We have
thereforean "adequacy-condition"on any theory of sense: we must be
satisfiedthat on the theory this deductionis preserved.When, later, we
querywhethera properintroductionof the notion of sense can be given,
it is this standardwe have in mind. For now let us note that sense cannot
be introduced or defined "linguistically".This is also likely to be the
point of Russell'ssaying in (C) that it will be a difficultyif meaningcan
only be got at by means of, i.e., by mentioning, denoting phrases. So
what we need is a definitionof either 'the sense of "Aristotle"' or some
other phrasewhich we may suppose to refer to the sense of 'Aristotle'.
The remainderof (D)-(F) shows groundsfor supposingthatthis cannot
be done.
After this preliminary,Russellturns,in (D), to head off a false move.
Baffledat this objectionto introducingsensevia mentioning the expression,
a Fregean might make the mistake of usingit, and try to give the
requireddefinitionof 'the sense of "Aristotle"'. He might define it as
the sense of Aristotle. (Again, we do not mean to suggest that Frege
would have made such an error.) (D) shows quite decisively what is
wrong with this approach. Russell takes the case, favourable to the
Fregean,where the thing referredto is a linguisticexpressionand might
itself be supposed to have a sense. And he points out that even in that
case, its sense-the sense of the reference--couldnot be the same as the
sense of our original denoting phrase, which is the object we wish to
specify. That is made clear by the Gray'sElegy example.'Thusin order
to get the meaningwe want we must speak not of "the meaning of C"
but of "the meaning of 'C' " .. .' This, of course, takes us back to the
method of introductionwhich is merelylinguistic.
It should dawn on us at this point that Russell has discovered a
dilemmaof sorts. We ask the Fregeanto give a proper introductionto
the notion of sense, and he can eithermention expressions,or use them.
If he mentions them, then the introduction fails, partly because it is
scientifically improper just to postulate what he wants, but more
seriouslybecausethe relationbetween sense and referenceis left merely
linguistic through the phrase.If on the other hand he uses the denoting
phrase,then he refers,and you cannotintroducesense as the sense of the
thing you refer to, for even if there is such a thing, that is not what we
are after. Russell repeats the second horn at the beginning of (E):
'the moment we put the complex in a proposition, the proposition is
about the denotation; and if we make a proposition in which the subject
is "the meaning of C" then the subject is the meaning (if any) of the
denotation, which was not intended'.

This content downloaded from 129.78.72.28 on Sun, 28 Jul 2013 10:16:29 AM

All use subject to JSTOR Terms and Conditions
 RUSSELL'S CRITICISM OF FREGE 73
Thus all is plain sailing to the bottom of page 49. Russell has dis-
covered his dilemmaand given, in (D), a precise statement(with only
minimalmisuse of quotationmarks)of the initialstage of the "use"horn
of it. The "mention"horn is merelyhinted at in (C), but as we have read
it thereis no solution therefor Frege. Now at the end of (E) he reminds
us of this horn: '. .. the meaninghas denotationand is a complex, and
there is not something other than the meaning,which can be called the
complex, and be said to haveboth meaning and denotation'.The point
of courseis not that Russellhas forgottenthe Fregeanview that denoting
phraseswill have both, but that he has alreadyshown that leaving the
matter there provides no explanation,or justification,of sense.
(F) is probablythe most puzzling paragraphof all. But with a little
careit becomes plain enough. The problemis still how we are going to
specifythis thing, the sense, meaning,or complex. Not, again, by using
the denoting phrase in question, because then 'what is said is not true
of the meaning,but only of the denotation,as when we say: The centre
of mass of the solarsystemis a point'. The following two sentencesthen
emphasizethat when we refer to the sense, the expressionwe use must
itself have its own sense which determinesthe sense we want to talk
about. (Note: Russellis here using 'subject'not for the topic of discourse
(the reference), but for the constituent of the proposition (Thought)
which correspondsto it, i.e., the sense, meaning, complex.) Although
these considerationsapply to any expressionpurportingto refer to the
sense we are after, Russell is here thinking primarilyof the original
defining expressionof the sense (e.g., 'the sense of "Aristotle"'). And
then comes the hammer:there is no way of identifying this secondary
sense (e.g., the sense of 'the sense of "Aristotle"') as a function of the
original sense. For quite generally, there is no way of identifying a
sense in terms of the correspondingreference:'. .. thereis no backward
roadfrom denotations to meanings,becauseeveryobjectcan be denotedby an
infinite number of denotingphrases'(our italics).
Thus by the end of (F) Russell has powerfullysharpenedthe second
horn of his dilemma. It is now not just that the phrase 'the sense of
"Aristotle"' will not do as our specificationof the object which is, in
fact, the sense of 'Aristotle'.This is what (D) pointed out. We are now
threatenedthat no denoting phrase can specify the sense of a phraseby
mentioning the referenceof that phrase,and trying to identify the sense
in terms of some fixed function of that reference.For since there is no
backwardroad,this function would be quite unexplained.So there is a
mystery as to how we define 'the sense of "Aristotle"' or any other
phrase which we may have reason to think refers to the sense of
'Aristotle'. But lacking such a definition, what guaranteeis there that
sense and reference are connected logically, as opposed to 'merely'
linguistically?

This content downloaded from 129.78.72.28 on Sun, 28 Jul 2013 10:16:29 AM

All use subject to JSTOR Terms and Conditions
74 ANALYSIS
We shall now state the case generally.Considerthe following three
expressions:
(E) 'Aristotle'
(EO)'thesenseof "Aristotle"'
(E.) 'thesenseof E,'
E2 is a phrasewhose referenceis the senseof the name'Aristotle'.
If sensesareto be properlyintroducedthentheremustbe somedenoting
phrase,suchas E2, whichwe understand, andwhosereferencewe can
thereforegrasp.The only alternativeis that we need no such phrase
because,somehow,we recognizesenses(andtheirconnexions)outright,
withoutrequiringa descriptionor a definition.We discussthis further
below,but obviouslyit is likelyto leavesensesin uninvitingobscurity.
However,the alternativethatwe recognizesensesthroughour under-
standingof somephrasesuchas E2(althoughas the "mention"hornhas
shown,E2 itselfwill not do) is opento us onlyif we aresatisfiedwith
thatunderstanding. The troubleis thatto understand E2 is to graspits
sense,andthe argumentnow repeatsitselfat the higherlevel.How are
we satisfiedthat we grasp the sense of E2? Either we recognizeit
straightoff,or thereis somedenotingphrase,suchas E3,whichrefersto
it, and whichin turnwe understand. But then, in turn,the lack of a
backwardroadwill entailthatwe needa separatestoryaboutthatpiece
of understanding, and so a regressis generated.The regresscan be
if
stopped, only we paythe cost of sayingthatthereis somelevelat
but
which we do not need an understooddescriptionor a definition,but
canrestcontentwith an outrightrecognitionof sense.'
Dummett((W)p. 267)sees Russell'sobjectionin a slightlydifferent
way. He writes that from Russell's'extremelyconfused'criticismof
Frege'we canat leastextract... a validcriticismof Frege'sdoctrineof
indirectsenseandreference'.But he also thinksthatthereis a 'simple
emendationwhich can be madeto the doctrine,whichis in harmony
with Frege'sother views, and dispels the objection'.Accordingto
Dummett,Russell,noticingthatFregeneedsanindirectsensefor a name
to expressin thosecontextsin whichit refersto its own ordinarysense,
objectsthatthereis no wayof tellingwhatthisthingis, becausethereis
no backwardroad.Dummett'semendationis to identifyindirectand
ordinarysense,so thatin suchcontextsa wordexpressesjust the same
senseasit ordinarily does:'thesenseof a wordcannotvaryfromcontext
to context,but is a propertyof the worditself,apartfromanycontext'
(p.268).
This may be satisfactoryas a doctrineabout indirectsense, and
thereforea replyto Russellas Dummettseeshim.Butit is vitalto realize
I Tom Baldwinpointedout to us that on pp. 223-227 of 'Knowledgeby Acquaintance
and Knowledgeby Description'Russellpresentsa closely relatedregressargumentagainst
Frege'sview of identitystatements.

This content downloaded from 129.78.72.28 on Sun, 28 Jul 2013 10:16:29 AM

All use subject to JSTOR Terms and Conditions
 RUSSELL'S CRITICISM OF FREGE 75
that it has no force whatever against the difficultyas we have put it.
The problem about the sense of a denoting phrasesuch as E, is not the
same as the problem about indirect sense. They could be confused by
supposingthat the sense of a phrasesuch as E2might itself be the indirect
sense of the name 'Aristotle'.'Dummett himself says that this is 'rather
implausible'(p. 267). In fact it is not only that, but takenin conjunction
with Dummett's identificationof indirect and ordinary sense, would
be totally lethal. For it would lead straightto the collapse of sense into
reference. We need only consider the sentence 'Jones knows that
Aristotle=Aristotle' and the sentence 'Jones knows that Aristotle=the
sense of "Aristotle"'. E2 in the context of the last sentencerefersto its
normalsense.If thatis the indirectsenseof 'Aristotle'andhence,following
Dummett, to be identifiedwith the ordinarysense of 'Aristotle', then
the phrase E2 in this context refers to the same as the name 'Aristotle'
in this context. Each denoting phrase is referring to the ordinary
sense of 'Aristotle'.Hence the two sentencesmust have the same truth-
value, by the fundamentalFregeanprinciplethat if two sentencesdiffer
only by substitutionof denoting phraseswhich, in the context of those
sentences,referto the same thing, then they cannot differin truth-value.
Hence since the first sentence is true, it follows that Jones knows that
Aristotle is the sense of 'Aristotle',in which case Aristotle is the sense
of 'Aristotle', and as Russell promised, we have failed to prevent them
from being one and the same. The argument shows that the indirect
sense of 'Aristotle',on Dummett's view of it, cannot be the same as the
sense of 'the sense of "Aristotle"'. It follows that de-mystifyingindirect
referenceis no help to Russellin his scepticismabout how we are going
to understandthe purporteddefiningexpressionof an ordinaryFregean
sense. When in (F) and (G) Russell pursues the denoting complex 'C'
which is to determinea sense, it is this which he is after,not the indirect
sense requiredby Frege.
Of course, none of this is an argumentthat we could not refer to
senses, once satisfied with their credentials.As we have stressed, the
argumentdoes not seem to be about reference.When Russellasks 'where
are we to find the denoting complex "C" which is to denote C?' his
plea is for a denoting phrasewhich we understand,the understandingof
which we can explain, and which specifiesthe ordinarysense. A phrase
such as 'the sense of "Aristotle"' fails these conditions; so does a phrase
which uses the word rather than mentioning it. If such a phrase is
apparentlyintelligibleto us ('the sense which each of us grasps when he
thinks of Aristotle', for instance),it nonetheless fails as a definitionby
presupposingthat we alreadyknow what we are talking about. It does
not tell us what we are talking about, because the lack of a backward
1 We owe our thanks to David Kaplanfor making us see this, and for his valuable
correctionsto an earlierdraft.

This content downloaded from 129.78.72.28 on Sun, 28 Jul 2013 10:16:29 AM

All use subject to JSTOR Terms and Conditions
76 ANALYSIS

road means that we would have no idea of how the word 'Aristotle'
would function in such a description.
We did notice the option of supposing that senses are named out-
right, presumablywith an attemptedexplanationof what is so named,
ratherthan introducedwith a definite descriptionreferringto them via
their definingproperties.It is worth noting at this point that (although
Russell did not know this at the time) the PoM theory of denoting makes
it impossible to directly name a sense or denoting concept. This is
because (i) when a name is used in a sentence, the thing named is a
constituent of the proposition expressed by the sentence, and (ii) if a
denoting concept is a constituentof a proposition, then the proposition
is about the object denoted and not about the denoting concept.1 If
we suppose that he had come to see this by the time he wrote 'On
Denoting', we have an explanationfor the fact that he is now insisting
that senses be introducedby means of definitedescriptions.(Notice that
although this is a problem as to how one refers to senses, if there are
such things at all, the obvious solution is not to attackFrege, but rather
to insist that his three-entityview applies to all referringexpressions.)
The heartof the matter,then, is whetherthe lack of explicitdefinition
can be tolerated. Will explanationsand indications which fall short of
this still suffice to give us confidence in the notion? In a different
passage (p. 227) Dummett writes that in saying what the reference
of a word is we showwhat its sense is, and conceding that we cannot
directly state what the sense of an expression is he offers us the con-
solation that we need only sufficientgrasp of the notion to say what it
is that someone can do when he has graspeda sense.But this consolation
is empty. Frege's theory demands that we do refer to the senses of
denotingphrases.We referto them, for instance,wheneverwe put names
in indirectcontexts: our understandingof any of these is as frail as our
understandingof preciselywhat senses are. The truth-conditionsof all
ascriptions of belief, thought, knowledge, are no more apparentthan
the propertiesof senses, nor are the relationsbetween these things and
the ordinary world of objects any more luminous than the relation
between a sense and the object which it determines.Thus consider the
common explanationin termsof a "way of findinga reference",a "mode
of presentationof a reference",or a "criterionfor identifyingan object
as being the bearerof the name". Do we understandthese things and
their relationto ordinaryobjects of talk? What is the logical connexion
between Aristotle and a mode or way of finding him? How is it that a
sentencesuch as (i) combinesin a deductionwith a sentencesuch as (2)
about this very differentthing? (Perhapswe should pauseand remember
the fate of many attemptsto illuminatethe logical relationshipbetween
criteria and things for which they are the criteria.) It is no part of our
I
Again, David Kaplan showed us this.

This content downloaded from 129.78.72.28 on Sun, 28 Jul 2013 10:16:29 AM

All use subject to JSTOR Terms and Conditions
 RUSSELL'SCRITICISMOF FREGE 77
argumentthat these questionsare unanswerable,but it is our contention
that Russell's argument is strong in proportion as their answers are
obscure.Logic may be ineffablebut it is a pity to build its ineffabilityinto
the very bottom of our theory about the relationbetween thoughts and
things. Perhapsa theorist can plead some licence for intuitive notions,
and not everything can be defined--but is it too exacting a critic who
complains when the undefined construct has indefinablebut firm and
clear logical relations?
Russell'ssolution is to give sentencescontainingdefinitedescriptions
a quantifier-plus-propositional-function interpretation,thus leaving the
object referredto no place in either our understandingof them or our
account of what makes them true. It has been suggested by Kaplanthat
since this is availablewe may simplyidentify the Russellianpropositional
function with Fregeansense, thus meeting any Russellianqualmsabout
the ontological and logical status of senses, although not satisfyinghis
distrustof a mixed theory and the logical (senseto reference)connexions
it must have. Whether or not such a move will in the end prove
satisfactory(will it provide a sense for phrases such as 'the sense of
"Aristotle"' which Russell does not contextuallydefine, for instance?),
we must still keep it in mind that prior to his developmentof the theory
of descriptionsit could not have been made at all. Hence the objections
to Frege could not have been met. It is hard to take away from Russell
the credit for just the respectin which his theory was shatteringlynew.
It may be that for some or many denoting phrasesFrege's theory can
be resurrected,but we are certain that in 1905 Russell knew of diffi-
cultiesfor it which, in the intervening seventy years, have seldom been
noticed, and never laid to rest.
( SIMONBLACKBURNAND ALAN CODE 1978
PembrokeCollege,Oxfordand University
of California--Berkeley
[i] A. J. Ayer, Russell and Moore: TheAnalyticalHeritage,Macmillan (London, x971), 30-32.
[2] R. J. Butler,'The Scaffoldingof Russell'sTheoryof Descriptions',Philosophical
Review,
LXIII(i954),350-64.
[3] ChrystineE. Cassin,'Russell'sDiscussionof Meaningand Denotation:A Re-examina-
tion', EssaysonBertrand
Russell,ed. E. D. Klemke, Universityof Illinois Press (I970),
256-72.
[4] Alonzo Church,'Carnap'sIntroductionto Semantics',Philosophical
Review,LII (I943),
298-304.
[5] M. Dummett,Frege:Philosophy Duckworth(London,
of Language, i973).
[6] Gottlob Frege, 'On Sense and Reference',Tratnlations
from thePhilosophical
Writingsof
GottlobFrege,ed. P. T. Geachand M. Black,BasilBlackwell(Oxford,1952), 56-78.
[71P. T. Geach,'Russellon Meaningand Denoting',Analysis,XIX (1959),69-72.
[8] R. Jager,Russell'sDenoting Complex',Analysis,XX (1959),53-62.
[9] BertrandRussell,'On Denoting',LogicandKnowledge, ed. R. C. Marsh,George Allen &
Unwin Ltd. (London, 1956), 41-56; reprinted from Mind, n.s., XIV (1905), 479-93-
[io] BertrandRussell, ThePrinciples 2nd ed., George Allen & Unwin Ltd.
of Mathematics,
(London, 1937; first ed. 1903).
[x1] BertrandRussell, 'Knowledge by Acquaintanceand Knowledge by Description',
Mysticismand Logic, George Allen & Unwin Ltd. (London 1917) pp. 223-227.
[12] J. R. Searle,'Russell'sObjectionsto Frege'sTheoryof Senseand Reference',Analysis,
XVIII (1958),
137-43.

This content downloaded from 129.78.72.28 on Sun, 28 Jul 2013 10:16:29 AM

All use subject to JSTOR Terms and Conditions