Royal Institute of

Throwing Away the Ladder Author(s): Cora Diamond Source: Philosophy, Vol. 63, No. 243 (Jan., 1988), pp. 5-27 Published by: Cambridge University Press on behalf of Royal Institute of Philosophy Stable URL: http://www.jstor.org/stable/3750493 Accessed: 19-11-2015 13:34 UTC

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 [email protected].

Royal Institute of Philosophy and Cambridge University Press are collaborating with JSTOR to digitize, preserve and extend access to Philosophy. http://www.jstor.org

This content downloaded from 130.88.20.147 on Thu, 19 Nov 2015 13:34:10 UTC All use subject to JSTOR Terms and Conditions ThrowingAway the Ladder

CORA DIAMOND

Whetherone is readingWittgenstein's Tractatus or his laterwritings, one must be struckby his insistencethat he is not puttingforward philosophicaldoctrines or theses;or byhis suggestion that it cannot be done,that it is onlythrough some confusion one is in aboutwhat one is doingthat one could takeoneself to be puttingforward philosophical doctrinesor theses at all. I thinkthat there is almost nothingin Wittgensteinwhich is of value and whichcan be graspedif it is pulled awayfrom that view of philosophy. But thatview of -philosophy is itself somethingthat has to be seen firstin the Tractatus if it is to be understoodin itslater forms, and inthe Tractatus it is inseparablefrom whatis centralthere, the distinction between what can be said and what can onlybe shown. Now whatabout that distinction? has writtenthat it has itssource in 'thegreat works of Frege', in Frege'sdiscussion of contrasts likethat between function and object.The differencebetween function and objectcomes out in language,but Frege, as is wellknown, held that thereare insuperableproblems in any attemptto put thatdifference properlyinto words. We cannotproperly speaking say what the differ- enceis, butit is reflectedin featuresof language; and whatholds of the differencebetween function and objectholds too of otherdistinctions oflogical category. Geach is rightthat we can bestunderstand what the Tractatusholds about sayingand showingif we go back to Fregeand thinkabout what the saying/showing distinction in itsorigin looks like there.Geach actuallymakes a strongerclaim: he saysthat 'a greatdeal of the Tractatusis best understoodas a refashioningof Frege'sfunc- tion-and-argumentanalysis in orderto remove[from it the] mistaken treatmentof sentencesas complexnames'.I That last pointof Geach's, abouthow to understandthe Tractatus, splitsinto two points if you think about it. Wittgensteinis trying to hold on to Frege'sinsight that there are distinctionsof logical category, like thatbetween functions and objects,or betweenfirst and secondlevel

I P. T. Geach, 'Sayingand Showingin Fregeand Wittgenstein',Essays on Wittgensteinin Honour of G. H. von Wright, (ed.), Acta PhilosophicaFennica 28 (1976) (Amsterdam:North-Holland), 54-55, 64.

Philosophy63 1988 5

This content downloaded from 130.88.20.147 on Thu, 19 Nov 2015 13:34:10 UTC All use subject to JSTOR Terms and Conditions Cora Diamond

functions,which cannot be put intowords but whichare reflectedin distinctionsbetween the signs for what is in one categoryand thesigns forwhat is in the other.He wantsto hold on to that,and at the same timeto getrid of the assimilation of sentences to propernames. So for Wittgensteina sentence will count as a whollydifferent sort of linguistic itemfrom a propername or any otherkind of name. But if you are holdingon to Frege'sinsight that fundamental differences in kindsof linguisticexpression are the way fundamentaldifferences in reality showthemselves, differences in realitythat cannot be putinto words- and if you are also saying,against Frege, that sentences are a wholly differentlinguistic category from any kind of name, thatwill make sense if you are also sayingthat there are featuresof realitythat can come out onlyin sentences,in theirbeing the particularkind of sign theyare, in contrastwith names. Here, then,is how Geach's pointsplits into two: (1) In the Tractatustreatment of Frege'sinsight, sentences are no longerassimilated to complexnames. (2) Makingthat break, separating sentences off that way from names, is linkedwith the possibilityof treatingthe distinctivefeatures of sentencesas reflectionsof featuresof reality,features that can onlybe reflectedin sentencesand thatcannot themselves be said to be features of reality.Such a treatmentof sentenceswould then be radically differentfrom Frege's, but could neverthelessbe said to be deeply Fregeanin spiritand inspiredby Frege. Geach himselfgives some detail of what is involvedin Wittgenstein's gettingrid of the assimilation of sentences to names;but he has rather less on what I have just been talkingabout: the applyingof Frege's insightto sentencesby taking their distinctive and essentialcharacteris- ticsto be thereflection of something in thenature of things that cannot be putinto words.2 But now to getback to whereI wasat thebeginning: ifwe wantto knowwhy Wittgenstein thinks that there cannot (in some sense)be philosophicaldoctrines, we needto see theapparent doctrines of the Tractatusas theywill look ifwe go furtherdown the road that Geach pointsout as a road. That is, we need to see whatkind of sign Wittgensteintook a sentenceto be and how,by being that kind of sign, itcan showthings that cannot be said. Butthere is somethingthat has to be donefirst. And one convenientway of doing it is togo backto Geach. I haveso farfollowed Geach in his wayof putting the Frege insight. As he puts it, variousfeatures of realitycome out in languagebut it cannotbe said in languagethat reality has thosefeatures. Geach is here followingboth Frege and Wittgensteinin an importantrespect. Witt-

2 Hans Sluga also touchesbriefly on the point; see Sluga, GottlobFrege (London: Routledgeand Kegan Paul, 1980), 144.

6

This content downloaded from 130.88.20.147 on Thu, 19 Nov 2015 13:34:10 UTC All use subject to JSTOR Terms and Conditions Throwing Away the Ladder

genstein,throughout the Tractatus, when he speaksabout what shows itselfbut cannotbe said, speaksof thesethings as featuresof reality. Thereis, he says, what cannotbe put into words. Propositionsand realityhave something in commonthat cannot be putinto words. Even the linguisticform 'what cannotbe put into words',the words 'das Unsagbare', 'das Undenkbare'-such ways of talkingrefer, or must seemto, to featuresof reality that cannot be putin wordsor captured in thought.Frege, in speakingof the distinction between first and second levelfunctions, describes it as foundeddeep in the natureof things;3 and itis evidentthat he wouldsay exactly the same about the distinction betweenfunction and object.There is a questionhow to takethis sort of talk:the use ofwords like 'reality', 'the nature of things', 'what there is' and so on, in specifyingwhat cannot be putinto words. The problemis particularlyacute in Wittgenstein,given the passage at theend of the Tractatus(6.54): 'whoeverunderstands me eventuallyrecognizes [my propositions]as nonsensical,when he has used them-as steps-to climbup beyondthem. (He must,so to speak,throw away the ladder afterhe has climbedup it.)' The problemis how seriouslywe can take thatremark, and in particularwhether it can be appliedto thepoint (in whateverway it is put) thatsome features of reality cannot be put into words. Let me illustratethe problem this way. One thingwhich according to theTractatus shows itself but cannotbe expressedin languageis what Wittgensteinspeaks of as thelogicalform of reality. So itlooks as ifthere is thiswhatever-it-is, the logical form of reality, some essential feature ofreality, which reality has all right,but which we cannotsay or think thatit has. Whatexactly is supposedto be leftof that,after we have thrownaway the ladder?Are we goingto keep the idea thatthere is somethingor otherin realitythat we gestureat, howeverbadly, when we speak of 'the logical formof reality',so that it, what we were gesturingat, is therebut cannotbe expressedin words? That is what I want to call chickeningout. What counts as not chickeningout is thenthis, roughly: to throwthe ladder away is, among otherthings, to throwaway in theend theattempt to takeseriously the languageof 'featuresof reality'.To read Wittgensteinhimself as not chickeningout is to say thatit is not,not really, his viewthat there are featuresof realitythat cannot be put intowords but showthemselves. Whatis his viewis thatthat way of talking may be usefulor evenfor a timeessential, but it is in theend to be letgo ofand honestlytaken to be realnonsense, plain nonsense, which we arenot in theend to thinkof as

I GottlobFrege, 'Function and Concept',Translations from the Philosoph- ical Writingsof GottlobFrege, P. T. Geach and Max Black (eds) (Oxford: Basil Blackwell,1952), 41.

7

This content downloaded from 130.88.20.147 on Thu, 19 Nov 2015 13:34:10 UTC All use subject to JSTOR Terms and Conditions Cora Diamond

correspondingto an ineffabletruth. To speakof featuresof realityin connectionwith what shows itself in languageis to use a veryodd kind of figurativelanguage. That goes also for'what shows itself'. This lastpoint is extremelyimportant for the issue what it is forthere supposedlynot to be philosophicaldoctrines. That is, I am contrasting twoways of taking the idea thatthere are no philosophicaldoctrines, as that idea appears in the Tractatus. You can read the Tractatus as containingnumerous doctrines which Wittgenstein holds cannot be put intowords, so theydo notreally count as doctrines:they do nothave whatcounts as senseaccording to thedoctrines in theTractatus about whathas sense. If youread the Tractatus this way, you think that, after theladder is thrownaway, you are left holding on to sometruths about reality,while at the same time denyingthat you are actuallysaying anythingabout reality.Or, in contrast,you can say thatthe notionof somethingtrue of reality but not sayably true is to be used onlywith the awarenessthat it itselfbelongs to whathas to be thrownaway. One is not leftwith it at the end, afterrecognizing what the Tractatushas aimedat gettingone to recognize. That is veryabstract and in needof some kind of illustration. Let me takethe case thatGeach was concernedwith: the distinction between functionand object. The case is fromFrege but it is usefulas an exampleof the generalpoint. We can indeedsay thatfor Frege it is a fundamentalfeature of realitythat there is that distinction;it is foundedin the natureof things.But is therea way of gettingbeyond thatway of talking? Of usingit as a ladderand thenthrowing it away? That distinctionbetween function and objectis reflectedin languagein whatFrege speaksof as the incompletenessof the signfor a function and thecompleteness of a propername, which is thesign for an object. A propername has no gap foran expressionfor an argument,whereas a signfor a functionhas gaps forone or moreexpressions for arguments of one or morekinds, one kindof argumrnt-expressionper argument place. It is not up to us to choosewhether we shallhave a languagein whichwhenever there is a function-expressionit will be incomplete.A languagemay be badlydesigned from the logical point of view, in which case the distinctionbetween the signs forfunctions and the proper nameswill not be markedin a waythat is easyto see. Suppose,though, thatwe become familiarwith a well-designedlanguage, in whichthe distinctionbetween the signs is clearlyvisible. We maycome to say,in graspingthe logical pointof the generaldistinction in thatlanguage betweenproper names and (forexample) signs for first-level functions ofone argument,that there is a distinctionbetween such functions and objectsand it is whatcomes out in thatdifference in thesigns; and yet we shallrecognize at thesame time that we cannotgo on thinkingof it as a fundamentaldistinction between functions and objects.Seeing what

8

This content downloaded from 130.88.20.147 on Thu, 19 Nov 2015 13:34:10 UTC All use subject to JSTOR Terms and Conditions Throwing Away the Ladder

itis forthat notation to be welldesigned is seeingwhat it is to talkabout a functionor functions;it is to use expressionsin a certainway. To refer to a function,that is, is to use a sign with a characteristickind of incompleteness,and to predicatesomething of a functionis touse a sign withits own furtherkind of incompleteness.And whenwe tryto say thatthere is a distinctionbetween functions and objects,we see thatwe are not thereusing language to talkabout functionsat all, because we are not thereusing signs of the distinctivesorts through which func- tions are spoken about and characterized.'There is a distinction betweenfunctions and objects,and it comesout in theclear difference betweensigns forfunctions and those forobjects in a well-designed notation':that is whatyou could call a 'transitional'remark. There is a transitionto be made,after which the word 'function' will have no place in the philosophicalvocabulary because it is not needed: thereis no workit is neededfor. Something else doeswhatever job theregenuinely is fora predicatelike 'function'to do, the somethingelse being the generallogical features of signsstanding for functions. A remarklike 'There is a fundamentaldistinction between functions and objects'is thrownout once we get the predicate'function' out of thecleaned up philosophicalvocabulary. We are leftafter the transition with a logical notationthat in a sensehas to speakfor itself. If we tryafterwards to say whyit is a good notation,we knowthat we shallfind ourselves saying thingswhich may help our listeners,but whichwe ourselvescannot regardas theexpression of any true thought, speakable or unspeakable. When we say whythe notationis a good one, whenwe explainwhat logicaldistinctions and similaritiesit makesperspicuous, we are in a sense goingbackwards, back to the stageat whichwe had been when graspingthe pointof the transition. We can then look at some of Frege's logical work as providing replacementsfor certain parts of thephilosophical vocabulary, in par- ticular, predicates like 'function','concept', 'relation'. These are replacedby featuresof a notationdesigned to makelogical similarities and differencesclear. For Wittgensteinthe provisionof replacements forterms in the philosophicalvocabulary is notan incidentalachieve- ment but a principalaim, and, more important,it is the whole philosophicalvocabulary which is to be replaced,including that of the Tractatusitself. Let me say moreabout thisdifference between Frege and Wittgen- stein. Frege thoughtthat.a contemporaryof his, Benno Kerry,was confused.Kerry had said thatthere were conceptsthat can also be objects,and thatwhen, for example, we saythat the concept horse is a concepteasily attained, the concept horse is an object,one whichfalls underthe conceptconcept easily attained. On Frege'sview, the idea thatwe can sayof a conceptthat it is a concepteasily attained, or thatit

9

This content downloaded from 130.88.20.147 on Thu, 19 Nov 2015 13:34:10 UTC All use subject to JSTOR Terms and Conditions Cora Diamond

is a concept at all, just reflectsconfusion. The predicate'concept' cannotbe predicatedof any concept: we thinkwe wantit in orderto say thingsabout concepts,but to thinkthat it will enable us to do so is confusion,confusion about what a conceptis.4 Wittgensteinthought thatthe whole philosophical vocabulary reflected confusion. We are all Benno Kerrysthrough and through.It is not thatwe are confused aboutwhat a conceptis and whatit is forus to be referringto a concept; itis thatwe areconfused about what saying something is, whatthinking somethingis. We are confusedabout the relationbetween logic and whatwe say and think;we are confusedabout whatit is forsome of whatwe say to be possibleand some necessaryand some impossible. The verywords 'possible' and 'necessary'and 'impossible'themselves are characteristicindications of lack of clarity,in just the way any attemptto use 'concept'seriously as a firstorder predicate, as Benno Kerrydid, is an indicationof lack ofclarity. What Wittgenstein wants to do is thento describea wayof writing sentences, a wayof translating ordinarysentences into a completelyperspicuous form. As partof the transitionto graspingwhat is thusmade clear,we maysay such things as thatthe possibility of a stateof affairsis notsomething that you can saybut thatit showsitself in signswith such-and-such general charac- teristics.But once thetransition is made,the analysed sentences must in a sensespeak for themselves, and we shouldnot any longer be telling ourselvesthat now we graspwhat possibility is, it is whatshows itself, what comes out, in a sentence'shaving a sense. We are leftusing ordinarysentences, and we shallgenuinely have got past the attempt to representto ourselvessomething in reality,the possibilityof what a sentencesays being so, as not sayablebut shownby thesentence. We shallgenuinely have thrownthe ladderaway. The wholeof Wittgenstein's philosophy, from before the Tractatus to his laterwork, contains different workings out ofthe kind of view of philosophyitself that I havejust sketched. I do notwant to playdown thedifferences between early and laterwork. It obviouslymarks a great changein Wittgenstein'sviews that he got ridof the idea thatyou can replace philosophicalthinking by carryingout a kind of complete analysisof sentences in whichthe essential features of sentence sense as suchare totallyvisible. But what does remainintact after that idea goes is theconviction that philosophy involves illusion of a particularkind. Recently,John McDowell, in speakingof the kind of philosophical illusionfrom which Wittgenstein in his laterwork tries to freeus, has used the phrase'the view fromsideways on', to characterizewhat we aim for,or thinkwe need to aim for,in philosophy.We have, for

4GottlobFrege, 'On Conceptand Object',Translations from the Philosoph- ical Writingsof , 42-48.

10

This content downloaded from 130.88.20.147 on Thu, 19 Nov 2015 13:34:10 UTC All use subject to JSTOR Terms and Conditions Throwing Away the Ladder

example,the idea of ourselvesas looking,from sideways on, at the humanactivity of followinga rule, and as askingfrom that position whetherthere is or is notsomething objectively determined as whatthe rulerequires to be doneat thenext application. To thinkof the question in thatway is to tryto step outsideour ordinarysaying what a rule requires,our ordinarycriticisms of steps taken by others,our ordinary waysof judging whether someone has graspedwhat a rulerequires. We do notwant to askand answerthose ordinary questions, but to askwhat in realitythere is to justifythe answers we give when we are unselfconsciouslyinside the ordinarypractice. McDowell takesWitt- gensteinto have triedto show us how to come out of the intellectual illusionthat we are thus askinganything.5 My pointnow is thatthat image of McDowell's is usefulin characterizingWittgenstein's early viewof philosophyas well. And I wantto traceit back evenfurther to how Frege leads us to see poor old Benno Kerry,who onlywanted to lookat conceptssideways on. In theTractatus, the idea ofthe illusory view fromsideways on has a veryparticular form. When we philoso- phize we tryas it were to occupya positionin whichwe are outside logic, where logic is that throughwhich we say all the thingswe ordinarilysay, all thethings that can be said. That bringsme back to Geach's originalpoint, that we shouldsee a greatdeal ofthe Tractatus as a refashioningof Frege's insight to avoid Frege's assimilationof sentencesto complexnames. I said thatthat pointitself splits into two, and I nowwant to modifyit; I shallhave to use the 'transitional'vocabulary, the before-you-throw-away-the-lad- der mode of speaking,to do so. Wittgensteindeparts from Frege not onlyin treatingsentences as a distinctlinguistic category from names; he not onlyapplies Frege's insightin an unFregeanway by claiming thatit is in the distinctiveessential features of sentencesas signsthat certainessential features of realityshow themselves;he also triesto makeclear in hisaccount of the kind of sign a sentenceis a characteristic and unFregeanview of logic. Logic will belong to the kind of sign ordinarysentences are, and ifthat can reallybe made clear,it will be clearalso thatin speakingphilosophically, we are confusedlytrying to stationourselves outside our normalpractice of saying how things are, tryingto stationourselves 'outside the world,outside logic'. So what Wittgensteinis doingin the Tractatusshould be seen thisway: he is holdingon to Frege'sinsight, but, against Frege, taking it that sentence signsas distinctfrom other signs reflect certain features of reality; he is

I JohnMcDowell, 'Non-Cognitivismand Rule-Following',Wittgenstein: To Follow a Rule, Steven H. Holtzman and ChristoperM. Leich (eds) (London: Routledgeand Kegan Paul, 1981), 141-162passim,but especially 150.

11

This content downloaded from 130.88.20.147 on Thu, 19 Nov 2015 13:34:10 UTC All use subject to JSTOR Terms and Conditions Cora Diamond

treatingan accountof the kind of sign ordinary sentences are as includ- ing an account of logic very differentin importantrespects from Frege's,deeply critical of Frege's, and capableof being used as a footing fora fundamentalcriticism of all philosophicalthinking.

II

Whatkind of sign, then, is a sentence?Wittgenstein says that a sentence is a fact,not a name.As itstands, that is a darksaying: dark, because by calling a sentencea fact, Wittgensteinmeant that it was logically articulated.But forFrege, too, a sentence,and anycomplex name, is logicallyarticulated. Since Wittgensteinplainly intends a contrastwith Frege'sview, what exactly is the contrast? Let me putthat question alongside some others. I haveso fargiven a versionof Geach's recipe for understandingthe Tractatus. Read Frege,and readmore Frege. And reflecton Frege.Then takesentences out ofthe category of complex names, make further changes to accom- modatethat fundamental shift, and youwill thus get a greatmany of the characteristicTractatus views. Where, it mightbe asked,does Russell come in? Or is his influenceto be regardedas so secondarythat the Tractatuscan be understoodwithout one's having to graspthe relation of any of its main points to Russell's work?Again, the theoryof descriptionsseems to playa significantrole in Wittgenstein'sthought. Butwhat role? It is hardlyas ifcommentators on Wittgensteinagree on the answerto that. Here is a separatequestion. Wittgenstein says that all philosophyis a 'critiqueof language',and adds thatthe creditis due to Russell for havingshown that the apparent logical form of a sentenceneed notbe its real form.This should seem a slightlyodd remark,given that Wittgensteinhad a higheropinion of Frege'sworks than of Russell's, and it would surelyseem that Frege should be giventhe creditfor makingclear the distinctionbetween apparent logical form and real logical form. Had that not been one main purpose of Frege's Begriffsschrift?The questioncould also be put thisway. In his remark givingcredit to Russell, Wittgensteinis partiallyspecifying what it meansto call philosophya critiqueof language. But, given what Frege had accomplishedas a critiqueof language,why does Wittgenstein explainwhat kind of critiqueof languagehe has in mindby appeal to Russell? My threequestions have to be answeredtogether. To see what is meantby callinga sentencea factand nota name,we haveto see what Wittgensteingot fromRussell. To answerthe question what Wittgen- stein got fromthe theoryof ,we have to see how, as

12

This content downloaded from 130.88.20.147 on Thu, 19 Nov 2015 13:34:10 UTC All use subject to JSTOR Terms and Conditions Throwing Away the Ladder

Wittgensteinunderstood it, it involveda sharp logical distinction betweena sentenceon the one hand, and a sign thatcould standfor somethingin a sentenceon theother. And, finally, the reason Wittgen- steingives Russell the credit for showing that the apparent logical form of a sentenceneed not be its real formis thatin thetheory of descrip- tions,the revealing of logical form is tiedto the logical form of sentences as such,or whatWittgenstein saw as thatform. Frege may indeed have shownthat ordinary language conceals the logical characteristics of the thoughtsexpressed in it. But, as Wittgensteinsaw Frege'saccomplish- mentin contrastto Russell's,Frege's clarifications of logical form left in the darkthe logicalfeatures distinctive of the expression of thought as such, the logicalfeatures of sentences.Putting this last pointanother way: you can lookfor the real logical form of a sentenceonly if you are clearwhat kind of signa sentenceis. You can look forthe real logical formof a particularsentence or groupof sentencesonly by exercising yourgrasp of what the general form of sentence is. As Wittgensteinread Russell,Russell had an implicitgrasp of what was crucial;and Frege, forall his colossalaccomplishments, did not. Wittgensteinsaid thathe, likeFrege and Russell,regarded a sentence as a functionof the expressions in it-and thatpoint helps me drawmy threequestions closer still. To see whatis meantby calling a sentencea factand nota namewill be to see whatkind of function a sentenceis of theexpressions contained in it. But howdo youmake clear what kind of functionit is? You providea methodof analysisof sentences, a wayof rewritingthem. As Wittgensteinsees it,what it is forFrege and Russell respectivelyto regardsentences as functionsof the expressions in them is shownin thedifferent ways in whichthey rewrite sentences in logical notation.What Wittgenstein saw in the theoryof descriptions,then, was a methodof analysisof sentences, a wayof rewriting them, which made theirkind of functionalityclear. And because Frege's logical notation,although it is intendedto show the real logicalform of the sentenceswritten in it, reflectswhat Wittgenstein takes to be a wrong viewof the kind of function a sentenceis, he givesRussell the credit for distinguishingthe real from the apparent logical form of sentences. In whatfollows I shallnot discuss so muchwhat it is fora sentenceto be a factas whatkind of function it is in contrastwith the kind of function it is forFrege. A fullaccount of why Wittgenstein says that sentences are factswould takeme too farout of theway. And whenI talkabout the theoryof descriptions,I mean it as lookedat by Wittgenstein,not as understoodby Russell. What Russell actually does with sentences containingdefinite descriptionscan be said to be the reverseof what Frege does. Frege looks at such sentencesand regardsthem functionally. The definite descriptionitself has no argumentplace withinit, and is thussuited to

13

This content downloaded from 130.88.20.147 on Thu, 19 Nov 2015 13:34:10 UTC All use subject to JSTOR Terms and Conditions Cora Diamond

be in the argumentplace of firstorder predicates and relationterms. The truthor falsityof the whole sentences in whichit occurs must then depend on what holds of the object satisfyingthe .The functionalanalysis of sentences, in Frege'shands, is totallyincompat- ible with allowing truthor falsity,or even some invented third truth value, to sentences which contain an empty .6 Russell goes in the opposite direction. Instead of movingfrom the functionalanalysis of the sentence to its necessarilynot having a truth value if it contains an empty definite description, he goes-or so it seems-from the truthvalue the sentence has in those circumstancesto a more complex functionalanalysis in which the definitedescription itselfdisappears, does not have the role of standingfor an argument,or picking out that thing on whose propertiesthe truthor falsityof the sentence depends. So he treatsthe question what the rightfunctional analysis of the sentence is as in an importantsense secondary. You cannot, if you are Russell, answer the question whether the definite descriptiongenuinely has the role of providingan argument,unless you are able to answer the prior question whetherthe sentence has a truth value when the definite description is empty. Russell's procedure is clearestin a strikingpassage in Principia Mathematica. He thereclaims that the sentence 'The round square does not exist' is true, despite the factthat the grammaticalsubject does not exist. He goes on: 'Whenever the grammaticalsubject of a proposition can be supposed not to exist without rendering the proposition meaningless, it is plain that the grammatical subject is not a proper name, i.e. not a name directly representingsome object. Thus in all such cases, the propositionmust be capable of being so analysed that what was the grammaticalsubject shall have disappeared.7 What I am interestedin, then, is the contrastwith Frege's procedure. That is, it is truethat Frege and Russell have differentviews about what the rightanalysis is of sentences with definitedescriptions, but I am focusingon a differentdisagreement: their different mode of arrivingat theiranswers. Thinking about that differencemade me see that Part I of this paper contains something misleading. I said there, following Geach, thatwe should see as the big change Wittgensteinmade in what he took over from Frege that he gave up the idea of sentences as complex names. That is an idea of Frege's that you do not get in his

6 See Gareth Evans, The Varietiesof Reference,John McDowell (ed.) (Oxford:Clarendon Press, 1982), 11; also P. T. Geach 'Frege',Three Philos- ophers,G. E. M. Anscombeand P. T. Geach (Oxford:Basil Blackwell, 1963), 136-139. 7 AlfredNorth Whitehead and BertrandRussell, Principia Mathematica to *56 (CambridgeUniversity Press, 1962), 66.

14

This content downloaded from 130.88.20.147 on Thu, 19 Nov 2015 13:34:10 UTC All use subject to JSTOR Terms and Conditions Throwing Away the Ladder

earlywork. You getit after 1890; youdo notget it in TheFoundations of Arithmetic,in 1884. But, and thisis whatI realized,although Frege's viewschanged a greatdeal in thatperiod, they did notchange in their fundamentaldifference from Russell's. The kind of functionFrege thoughta sentencewas does not changein one fundamentalrespect overthat period, and it is that,and notwhether or notthe sentence is explicitlyregarded as a complexname, thatWittgenstein rejects. He rejects a kind of functionalview of a sentencewhen he says that sentencesare notnames, but thekind of functional view in questionis presentin Frege priorto his idea thatsentences are names,and goes deeper down in the Fregeanstructure of views. The view I mean is presentin The Foundationsof Arithmetic, in Frege's takingit thata complex designationappearing on one side of the equal sign in a mathematicalequation has the role of designatingan object,an argu- ment,and thattherefore we cannotput intosuch a place a signwhich we have not showndoes stand forsomething.8 Here we see the pro- cedurewhich goes in theopposite direction from Russell's. Frege starts withthe equal signand itsargument-places, as determiningthe logical role of the complexname; Russell argueson othergrounds that the complexname has no logicalrole of its own, and thusconcludes that evenwhen it looks as ifit must designate an object,it neednot do so for theequation or identityin whichit appearsto have a truthvalue. Wittgensteinthen sees Russell'streatment of sentences in thetheory ofdescriptions this way: ifthey are going to saysomething true or false, theirdoing so is notdependent on whetherany definite description they maycontain is satisfied.The sentence'shaving a truthvalue at all is not thekind of thing that the satisfaction or emptinessof definitedescrip- tionscan affect.That thenis puttinginto words what is betterlooked at as builtinto what Russell actually does. He treatssentences as havinga kindof functionalitydifferent from that of complexnames. That is, suppose thatthere is no such personas Beethoven'sonly half-sister, thenthere is no one who is thefather of Beethoven'sonly half-sister. And thatis reflectedin language.If the description'Beethoven's only half-sister'is empty,then complex names in whichit occursare also emptyand do notdenote anyone. So thatis thekind of functionality a complexname has. And Russell treatssentences as havinga kindof functionalityunlike that: theircapacity to say whatis trueor falsehe treatsas unlikethe capacity of a complexname to denotesomething: he treatsit as independentof the satisfaction of definite descriptions in it. I need now to turnto one furtherfeature of Russell'sanalysis of sentenceswith definite descriptions. Russell takes a sentencecontain-

8 Gottlob Frege, The Foundationsof Arithmetic,J. L. Austin (trans.) (Oxford:Basil Blackwell,1974), 90.

15

This content downloaded from 130.88.20.147 on Thu, 19 Nov 2015 13:34:10 UTC All use subject to JSTOR Terms and Conditions Cora Diamond

inga definitedescription, say 'the present king of France', and showsus howto rewriteit. The rewrittensentence does twothings. Its havinga truthvalue is clearlyindependent of the truth or falsityof the sentences sayingthat there is a presentking of Franceand thatthere is no more thanone presentking of France. And at the same timethe rewritten sentenceclarifies the real functional character of the original sentence. The originalsentence says whateverit does say in virtueof its func- tionalcharacter, which is shownto us throughits logical relation to the analysedsentence. Its functionalcharacter can thusbe seen to leave it havinga truthvalue independentlyof the truthvalue of two other sentences,those saying that there is at leastone kingof France, and that thereis no morethan one. What sign the originalsentence is-what functionof what parts-is inseparablefrom the capacity it has (shown clearlywhen it is analysed)to keepits truth-valuedness independently of thetruth or falsityof thosetwo sentences. Further,Wittgenstein accepted a versionof Russell'saccount of the quantifiers.9He read it as enabling us to go on witha sortof analysisthat shares the essential featureof the Russellian analysisof definitedescrip- tions. In such an analysis we make clear at one and the same time two things: first,how our sentence is constructed,as what functionof what expressions, and, secondly, that our sentence is a sign that maintains truth-valuedness over the range of truth or falsity of some set of sentences. If we take the sentence 'The presentking of France is bald' in its analysed version, including quantifiers,we can go on to analyse it. We shall see it as containingsentences sayingof this individual and that one . . . thatit is a kingof France, and at the same timewe shall see that the quantified sentence retainstruth-valuedness whatever the truthor falsitymay be of any sentence saying of an individual that it is a king of France. This furtherstep then deepens our understanding of the original sentence's functionality:it shows us more clearly what sen- tence it is at the same time as it shows us that the sentence maintains truth-valuednessindependently of what the truthvalue is of any of a range (now enlarged) of sentences. What is coming out here, as seen by Wittgenstein,is thata sentence is a sort of sign such that which sign it is of that general sort is tied to its maintainingtruth-valuedness throughout any variationin truthvalues of some range of sentences. Let me put that slightlydifferently. Russell's originalsort of analysis goes some way to lettingus see what sign a sentence is, of the general sort to which sentences as such belong, by showing how it maintains truth-valuednessthroughout variations in truth value of a particular

I , Wittgensteinand the , Brian McGuin- ness (ed.) (Oxford: Basil Blackwell, 1979), 39.

16

This content downloaded from 130.88.20.147 on Thu, 19 Nov 2015 13:34:10 UTC All use subject to JSTOR Terms and Conditions Throwing Away the Ladder

rangeof sentences, whose falsity we mighthave thought deprived it of any truthvalue. We get still clearerwhat sign a sentenceis, what functionof what expressions, if we are able to writeit so thatits truth- valuednessis shownto be independentof variationsin truthvalue of someenlarged range of sentences. We shouldmake entirely clear what sentencea sentencewas, whatits functionalitywas, ifwe wereable to writeit so thatits truth-valuedness could be seento be independentof thetruth or falsityof anyother sentence. But that,as it stands,will not quite do. We need to see Russellian analysis,or its basic principle,extended in a differentdirection. The principleof Russelliananalysis, as seen by Wittgenstein,is thatthe functionalcharacter of a sentenceis graspedat thesame time as itis seen how it maintainstruth-valuedness whatever the truthor falsityis of some sentenceor rangeof sentences.Besides thinkingof Russellian analysisas goingon and on to somefinal analysis, we needto thinkof it as beginningat as itwere a zero stage;we need to thinkof it as usinga generalprinciple for reading sentences, which is actuallyapplied at one stageearlier than the one we see in thetheory of descriptions itself. Our preliminarygrasp of whatfunction of whatexpressions a sentenceis involvesseeing how by beingthat function of thoseexpressions it is truth-valued,regardless of whetherit itselfis trueor false.There will be no suchthing as a syntacticcharacterization of a sentence,such that by fixingthe meaningof the sentence'selements, it will turnout that onlyif the sentenceis truedoes it have anytruth value at all; norwill therebe such a thingas a syntacticcharacterization of a sentence,such that it can turn out that, given a certainmeaning of some of the sentence'selements, only if the sentence is falsedoes ithave a determin- atetruth value. Therewill therefore be no suchthing as a readingof the syntacticstructure of a sentencelike 'A is an object'such that, if A were not an object (whateverit may be that'object' means), the sentence itselfwould be deprivedof truthvalue. A sentencecannot require its own truthas a conditionof its own havinga determinatetruth value: thisis partof a descriptionof the kind of sign sentences are, the kind of functionalitythey have.10 I can now give a revisedversion of what I said about analysis.We makeentirely clear what sentence a sentenceis, whatits functionality is, ifwe are able to writeit so thatits truth-valuednesscan be clearly seento be independentof the truth or falsityof any other sentence and of its own truthor falsity:independent of the truthor falsityof any

10 Frege,in TheFoundations of.Arithmetic (see pp. 40 and 62), is sometimes said to hold, likeWittgenstein, a 'contrast' theory of meaning.But I believe thatWittgenstein's view is not Frege's. There is no suggestionthat Frege's viewis a view about the logicalstructure of sentences.

17

This content downloaded from 130.88.20.147 on Thu, 19 Nov 2015 13:34:10 UTC All use subject to JSTOR Terms and Conditions Cora Diamond

sentence. We should thus be clear how by being the particularsign it was, the capacity to be used to say somethingdeterminately true or false belongs to it, is internalto it. The kind of sign a sentence is can now be put two ways: if something is a sentence, it is capable of comparison with reality,yielding true or false, regardless of the truthor falsityof any sentence; and sentences in general are capable of comparison with realityregardless of whetherit is true or false. This idea of the body of sentences can be seen to need more clarifica- tion than I have given it, and to run into an immediate problem about logical truthsand logical falsehoods. Let me take that problem first.It is overwhelminglynatural to say somethinglike this: that, ifwe fixthe meaning of the sententialconnectives, and if we stickto what we have fixed, and do not equivocate, then a sentence of the form'p and not p' says something which would be true if both p and its negation were true. Because it would be true only in circumstanceswhich cannot be, it is thereforealways false, necessarilyfalse. Its truthconditions, which are perfectlydeterminate, are never met. And we shall want to say something roughly comparable about 'p or not p'. A sentence of that form has truth conditions which are in all possible circumstances fulfilled. I hope that it is clear why I say that that presents a problem for Wittgenstein.No sentence in the body of sentenceswould have a single determinatetruth value unless the truthvalue of contradictionis false; no sentence in the body of sentences would have a single determinate truth value if the tautology's truth conditions are not fulfilled. If tautologies and contradictionsare genuine sentences, the idea of sen- tences as, by the signs theyare, capable of truthor falsityregardless of the truthor falsityof any sentence-that idea has to go. And yetit seems that the overwhelminglynatural account forces us to say that if the logical signs are to be used consistently,we can form, using them, tautologies and contradictions: sentences with truth conditions fulfilledin all possible or no possible circumstances. Wittgenstein'ssolution is thatthere is no such thingas consistentuse of the logical signs: consistentuse ofthe kind we imagine. There is no explanation of conjunction, for example, or negation, independent of the character of sentences as such. The fundamentalidea is that sen- tences in the strict sense (signs so formed as to be comparable with realityregardless of the truthor falsityof any sentence) can be formed fromother sentences, but that the rules forfixing the comparison with reality of sentences so formed from other sentences also fix the con- struction in the symbolism of sentence-likestructures which merely reflectthe logical character of their component sentences, and which themselves have no comparison with reality.

18

This content downloaded from 130.88.20.147 on Thu, 19 Nov 2015 13:34:10 UTC All use subject to JSTOR Terms and Conditions Throwing Away the Ladder

On thisview, then,the rule fixingthe place in the languageof the logicalconstants (so called) is a ruledetermining the logical features of the comparisonwith realityof sentences,genuine sentences formed fromgenuine sentences.That fundamentalrule will reallybe the meaning,all rolledup intoone, as it were,of all thelogical constants. And ifwe graspthat, we shallnot be temptedto think of tautologies and contradictionsas sayingthat something or otheris the case. We shall notbe temptedreally to thinkof themas sentences. I can now finishwhat I need to sayabout the body of sentences. The idea as Wittgensteinworks it out requiresthat there be somesentences whichcan be seen directlyto havethe logical characteristic of sentence signs as such: the possibilityof comparisonto reality,comparison which yields true or false, belongs to the signs themselvesand is independentof the truth value of any sentence. The idea ofthe body of sentencesthen requires,besides such base sentences,a methodof constructionof sentencesfrom sentences, such thatif the base sen- tences have seeably got the logical characteristicof sentences,the resultswill be sentencesseeably sharing the characteristic or will see- ably be merelysentence-like constructions never comparablewith reality. Let me put thisconception slightly differently, using a termI have avoided:'entail'. I havekept clear of it becauseits meaning depends on whetheryou treatlogical relationsas Frege does or as Wittgenstein does. Whatentailment is and whichsentences entail which depend on whethera sentenceis syntacticallya sort of sign that cannot lose truth- valuednessby the truthor falsityof anysentence. Suppose itis said thatWittgenstein and everyoneelse believes that to understanda sentence,you have to knowwhat sentence it is. If my accountis correct,it looksas ifwe could neverknow what sentence a sentencewas exceptby carryingout its completeanalysis, which of course we in practicenever do. So if my accountof Wittgensteinis correct,his is a lunaticaccount. The replyis thatwe shouldthink of twothings as thesame. On the one hand thereis the fullyanalysed sentence, which would lay out clearlyin frontof us whatfunction of what expressions a sentencereally is. To get whatis goingto be on the otherhand we have to thinkof liftingup an ordinarysentence, and noticing,attached to it, likelittle wires,all thesentences which entail that it is trueor thatit is not. The ordinarysentence, together with all itslittle wires, is thesame sentence as thefully analysed one. So we can understandthe ordinary sentence eventhough we do notknow how to carryout its full analysis. The little wiresare all there,all fixedby thelogical structure of the language. As Wittgensteinputs it in section102 ofthe Philosophical Investigations: 'The strictand clearrules of the logical structure of sentences appear to

19

This content downloaded from 130.88.20.147 on Thu, 19 Nov 2015 13:34:10 UTC All use subject to JSTOR Terms and Conditions Cora Diamond

us as somethingin the background-hiddenin the mediumof the understanding.I alreadysee them(even thoughthrough the medium of the understanding)for I understandthe sign, I use it to say something'.

III

More could be said about whatkind of signa sentenceis, and whatit means to call a sentencea fact. But I need now to get back to the questionsI began with,about Wittgenstein'sview of philosophy.I spokeof an interpretationof the Tractatus,which I called chickening out. Wittgensteinsays, at theend ofthe book, that anyone who under- standshim will recognize his sentencesas nonsensicalafter he has used themto getwhere they take you. He mustthrow away the ladder after he has climbedup it. To chickenout is to pretendto throwaway the ladderwhile standing firmly, or as firmlyas one can, on it. P. M. S. Hackeris an example.He ascribesto Wittgensteinwhat you mightcall a realismof possibility.Each thinghas, internalto it and independentlyof language, fixed possibilities of occurrence in kindsof fact,possibilities shared by all membersof the categoryto whichthe thingbelongs. What we can say,what we can think,is thata thinghas (or thatit has not) one ofthe properties that, as a memberof its logical category,it can have; or thatseveral things stand (or thatthey do not stand)in one ofthe relations that as membersof their logical categories theycan standin. Whatwe cannotsay or thinkis thatthe thing belongs to such-and-sucha logicalcategory, that it has thisor theother logical propertiesof possibly being combinable in such-and-suchways. Languagemirrors these internal logical characteristics of things. They arerepresented in languageby variables, not by predicates or relational terms.Thus, forexample, being an objectis such a logicalor formal property;and, in English,if we say'An objectfell', the word 'object' is reallybeing used as a variable,and thesentence might be rewrittenin logicalnotation as '(3x) (x fell)'. We violatethe principlesof logical syntaxwhen we use a termlike 'object', a termfor a formalconcept, as a genuinepredicate, as whenwe say'A is an object'.Here we aretrying to put intowords something that shows itself in languagebut cannotbe said.1 I callthat chickening out. It involvesholding that the things we speak aboutare members of this or that logical category, really and truly, only we cannotsay so. That theyare is representedin languagein another

" P. M. S. Hacker, Insight and Illusion (Oxford: Clarendon Press, 1972), 20-24.

20

This content downloaded from 130.88.20.147 on Thu, 19 Nov 2015 13:34:10 UTC All use subject to JSTOR Terms and Conditions Throwing Away the Ladder

way.The sentencesof the Tractatus itself are taken to conveythis form ofrealism, although the doctrine itself requires that any attempt to state it as a doctrinemust fail. There are severalcharacteristic signs of this chickeningout. The firstis theidea ofa realmof necessities underlying our capacityto make sense as we do. Hacker explicitlyascribes to Wittgensteinthe view that there are ontologicalcategories, objectively fixedand independentof language, which the logical syntax of language is thenrequired to mirror.12The secondis theidea ofthere being such a thingas violatingthe principlesof logicalsyntax by using a termin what,given its syntax,goes againstwhat can be said withit. The two characteristicfeatures are thencombined in theidea thatwe violatethe laws of logicalsyntax when we tryto statethose necessary features of realitythat properly speaking show themselves in language.There is a veryclear conceptionhere of somethingyou cannot do. Or rather, perhapsit is notall thatclear, since it dissolvesinto incoherence when pushedslightly. Call itthen a seeminglyclear picture of something you cannotdo: namelyput these perfectly genuine logical features of reality intowords. There theyare, though,underlying our use of words. Wittgenstein'sphilosophy, throughout his life,is directedagainst certainways of imagining necessity. Throughout his life, his treatment of logicaims at lettingus see necessitywhere it does lie, in theuse of ordinarysentences. The troublewith chickening out, or one trouble withit, is thatit holds on to exactlythe kind of imagination of necessity, necessityimaged as fact,that Wittgenstein aimed to freeus from. Take a sentencelike 'A is an object'. If we thinkof it as statinga necessarycondition for ordinary sentences using the name 'A' to havea truthvalue, we are (it would seem) in immediatetrouble. 'A is an object'cannot be a memberof the bodyof sentences-or so it seems, sinceall membersof the body remaintruth-valued irrespective of the truthor falsityof any sentence. So thereare not,within that body, any sentencesgiving the necessaryconditions of truth-valuednessof sen- tenceswithin the body. But, now, ifwe thinkof this sentence 'A is an object',withdrawn as itwere from the body of ordinary sentences, and as statingsomething underlying the truth-valuedness of some ofthem (and hencethe inferential relations of all ofthem), we arethinking of it as itselfsaying what it does on accountof the expressionsin it. We understandit (so we think)-but whatsentence it is, whatexpressions how combined,is not separablefrom its capacityto say something truth-valuedirrespective of thetruth values of the-which?-body of sentences,standing in logical relationsto each other,to which it belongs,and includingits own negation.In so faras we takeourselves to understandit, we takeits truthand its falsityboth to be graspable.

12 Hacker, 23.

21

This content downloaded from 130.88.20.147 on Thu, 19 Nov 2015 13:34:10 UTC All use subject to JSTOR Terms and Conditions Cora Diamond

Even in thinkingof it as truein all possibleworlds, in thinkingof it as somethingwhose truth underlies ordinary being so and notbeing so, we thinkof it as itselfthe case; our thoughtcontrasts it withas it werea differentset of necessities. Our ordinarypossibilities have the character ofpossibility, given that these underlying necessities are as theyare, not someother way. This wayof taking our sentence 'A is an object'has got built into it a way of thinkingof what is necessary,where what is necessaryhas got its own logic distinctfrom the logicof our ordinary descriptionsof what is thecase. Fromthe perspective we nowseem to ourselvesto occupy,the logical rules governing ordinary sentences, the logical'scaffolding' internal to the originalbody of sentences,will be thoughtof as needingto match,to be determinedby, which necessities do hold.-From theperspective we nowseem to occupy;but Wittgen- stein'saim is to letus recognizeit to be onlythe illusion of a perspective. The contrastI want is the contrastbetween saying that that is the illusionof perspectiveand sayingthat it is the correctphilosophical perspective,only you cannotput intowords what is seen fromthere. The philosophicalperspective is fine,but you just need to shutup. On that second, chickeningout, interpretation,Wittgenstein's general accountof sentencehoodrules out theexpression, by anysentence, of theview from the philosophical perspective. What is seenfrom there is representableonly in internalfeatures of thebody of sentencesitself. But if thingswere differentas seen fromthat perspective, different necessitieswould thusbe representedin internalfeatures of language; thesystem of possibilitiesin languagewould be different.So we have twocrucial features of (whatI am suggestingis) thewrong interpreta- tionof Wittgenstein. There is stillinsistence on viewingpossibility and necessityas fixedsome particular way rather than some other; they are stillreally being conceived in a space. Whatis possiblein thecontingent world,what is thinkable,what is sayable,is so because of the way ontologicalcategories are fixed.And withthis there is the idea that sentencesattempting to expressany of thesethings are illegitimate, countas nonsenseby the doctrinesof the Tractatus. We need to go back to the apparentlyinnocent way I led us intoall this:by showing, or seemingto, that, given Wittgenstein's view of what sentencesare, theremust be deep troublewith the sentence'A is an object'.The argumentwas thatno sentencegiving necessary conditions forthe truth-valuednessof a sentencecan belongto the body of sen- tences.So 'A is an object'was pushedoutside, and thenwe had more trouble. But whatWittgenstein says is thatthere is nothingwrong with any possiblecombination of signsinto a sentence.He says at 5.4733 that any possiblesentence is, as faras itsconstruction goes, legitimately put together,and, ifit has no sense,this can onlybe becausewe havefailed

22

This content downloaded from 130.88.20.147 on Thu, 19 Nov 2015 13:34:10 UTC All use subject to JSTOR Terms and Conditions Throwing Away the Ladder

to give a meaningto some of itsconstituents, even ifwe thinkthat we havedone so. Thus thereason why 'Socrates is identical'says nothing is thatwe have notgiven any adjectival meaningto theword 'identical'. The word'identical' as it occursin (e.g.) 'The morningstar is identical withthe eveningstar' is, syntactically,a totally different symbol from what we have in 'Socrates is identical'.So the sentence'Socrates is identical'is legitimatelyput together, in thesense in which'Socrates is frabble'is, as faras its structuregoes, legitimatelyput together.Both containwhat are syntacticallyadjectives; all theyneed is for some adjectivalmeaning to be fixedfor them. What I am emphasizingis that on Wittgenstein'sview, the only thing wrong with 'Socrates is identical' is the absenceof an adjectivalmeaning for 'identical', where the need for a meaningmay be hidden fromus by the fact that the word 'identical'has otheruses in which it is meaningful.A good logical notationwould mark syntactical differences by the use ofunconfusable signs,signs with visibly different structure. And thenthere would be no such thingas givinga word like 'identical'a new syntacticalrole, whereits alreadyhaving a differentrole, in whichit was meaningful, hid its meaninglessnessin the new role. We needto applythis to 'A is an object'.What is wrongwith it, on the viewof Wittgenstein which I am attacking,is thatit is an attemptto put intowords what really does underliethe intelligibilityof whatwe say trulyor falselyof A in ordinarysentences. That A is an object does underliethe intelligibility of ordinary sentences, but it is a violationof logicalsyntax to put that intowords. Against this, here is how I am suggestingwe interpretWittgenstein. The veryidea of thephilosoph- ical perspectivefrom which we consideras sayableor as unsayable necessitiesthat underlieordinary being so, or possibilitiesas them- selves objectivefeatures of reality,sayably or unsayably:that very perspectiveitself is the illusion,created by sentenceslike 'A is an object',which we do notsee to be simplynonsense, plain nonsense. 'A is an object' is no morethan an innocentlymeaningless sentence like 'Socratesis frabble';it merelycontains a wordto which,in its use as predicatenoun, no meaninghas beengiven. But we inflateit, we blowit up into somethingmore, we thinkof ourselvesas meaningby it somethingwhich lies beyondwhat Wittgenstein allows to be sayable. We thinkit has to be rejectedby himbecause of that. We thinkof there beinga contentfor it, which according to hisdoctrines, no sentencecan have. Butthis conception of what we cannotsay is an illusioncreated by our takingthe word'object', which works in meaningfulEnglish sen- tences essentiallyas a variable,and puttingit into othersentences whereit has a whollydifferent grammatical function. When Wittgen- steinsays thatwe cannotsay 'There are objects',he does not mean 'There are, all right,only that there are has to get expressedanother

23

This content downloaded from 130.88.20.147 on Thu, 19 Nov 2015 13:34:10 UTC All use subject to JSTOR Terms and Conditions Cora Diamond

way'. That thesentence means nothing at all, and is notillegitimate for any otherreason, we do not see. We are so convincedthat we under- standwhat we are tryingto saythat we see onlythe two possibilities: it is sayable,it is notsayable. But Wittgenstein'saim is to allowus to see thatthere is no 'it'. The philosophicalinsight he wantsto conveywill come whenyou understandthat you wantto makeuse ofa syntactical construction'A is a such-and-such',and thatyou are freeto fixthe meaningof the predicatenoun in any way you choose, but thatno assignmentof meaning to itwill satisfy you. Thereis notsome meaning you cannotgive it; but no meaning,of thosewithout limit which you can giveit, will do; and so yousee thatthere is no coherentunderstand- ing to be reachedof what you wantedto say. It dissolves:you are left withthe sentence-structure 'A is an object',standing there, as it were, innocentlymeaning nothing at all, notany longer thought of as illegiti- mate because of a violationof the principlesof whatcan be put into wordsand whatgoes beyondthem. Really to graspthat what you were tryingto say shows itselfin languageis to cease to thinkof it as an inexpressiblecontent: that which you weretrying to say. Take Wittgenstein's remarkthat there is only logical necessity (6.375). It is a wonderfulremark. Logical necessityis that of tautologies.It is notthat they are true because their truth conditions are met in all possibleworlds, but because theyhave none. 'True in all possibleworlds' does not describeone specialcase oftruth conditions being met but specifiesthe logicalcharacter of certainsentence-like constructionsformulable from sentences. But theremark that there is onlylogical necessity is itselfironically self-destructive. It has the form, the syntacticform, of 'There is onlythis sort of thing',i.e. it uses the linguisticforms in whichwe say thatthere are onlythises rather than thisesand thats.It belongsto itssyntax that it itself says something the otherside ofwhich can be representedtoo. If thereis onlysquiggledy wiggle,the language allows wiggles that are notsquiggledy as well. But whateverthe sentenceaims to do forus, it is not to place the kindof necessitythere is as thissort rather than that. It does notconvey to us the philosophicalbut unsayablefact that there is onlytautology not genuinelysubstantial necessity. In so faras we graspwhat Wittgenstein aims at, we see thatthe sentence-formhe uses comes apartfrom his philosophicalaim. If he succeeds,we shall not imaginenecessities as statesof affairsat all. We throwaway the sentencesabout necessity; theyreally are, at theend, entirelyempty. But we shallbe awareat the end thatwhen we go in forphilosophical thinking, the characteristic formof such thoughtis preciselythat the sentence-forms we use come apartfrom what we have takento be our aims. Not because we have chosenthe wrong forms.

24

This content downloaded from 130.88.20.147 on Thu, 19 Nov 2015 13:34:10 UTC All use subject to JSTOR Terms and Conditions Throwing Away the Ladder

Back now at last to Wittgenstein'srelation to Frege. The analogies couldbe speltout in detail,and so couldthe differences. What I shalldo insteadis workthrough one particularanalogy to makeclear at thesame timethe deepestdifference between them. For Frege,you referto a conceptby using a termpredicatively, that is, byusing a termwhich makes a characteristickind of contribution to thetruth or falsityof the sentences in whichit occurs. You cannotfirst predicatehorse of Bucephalus,say, and thenas it wereturn round on yourselfand grabhold of what you referred to bythe predicative use of 'horse'and catchit by usinga termas a logicalsubject. It was Benno Kerry'sidea thathe could do just thatthat Frege criticizedhim for. Frege'scriticism would also applyto Russell,13 whoseconfusion (confu- sion,that is, as we mayimagine it seen from Frege's point of view) is ofa particularlyhelpful sort. Russell believes that the logical subject humanityis actuallyone and thesame thingas theconcept ascribed to Socrates when we say that he is human (they are 'exactly and numerically'the same); and he adds that the differencebetween humanitywhen it is spokenof by a termwith the logical features of a propername and whenit is predicatedof somethingis in theexternal relationsof the concept humanity and notin theintrinsic nature of the thingwe are talkingabout. This idea thatthe logicalcharacter of the expressionsyou use in talkingabout somethingis irrelevantto what kindof thing it in itselfis is thenreflected in that'rounding on oneself' thatI spokeof, that attempt to catchat thevery thing you had referred to bya predicativeuse ofan expression,and to holdit up in frontof you for philosophicalconsideration. I cannot here tryto show what is involvedin Frege'srejection of any view like Russell's as incoherent."4 But itis importantto see thatRussell's view depends on a notion(or on imaginingthat he has a notion)of the identityof a thing,a notionof identityuntied from the substitutabilitysalva venitateof expressions referringto it. There is, he supposes,a position,a perspective,from whichto thinkabout things-with-their-intrinsic-nature, outside of the use of propernames and functionalterms, with their different logical character,in sentences. Justas, for Frege, you referto a conceptby using a termpre- dicatively,and it is confusedto thinkyou can roundon yourselfand grabhold ofwhat you thusreferred to, so forWittgenstein, something thatmay be thecase getssaid to be thecase by a sentence,and it is confusedto thinkyou can roundon yourselfand grabhold of what you

13 BertrandRussell, Principles of Mathematics (London: Allenand Unwin, 1937), 46. 14 I have discussedthose argumentsof Frege's in 'What does a Concept- scriptdo?', PhilosophicalQuarterly 34 (1984), 343-368.

25

This content downloaded from 130.88.20.147 on Thu, 19 Nov 2015 13:34:10 UTC All use subject to JSTOR Terms and Conditions Cora Diamond

thussaid to be so, and treatit as a logicalsubject. For Frege,the sign for a conceptis an expressionwith the logical character of a predicate,with one or anotherkind of incompletenessinternal to it, carryingwith it a particularkind of rule forsubstitution salva veritate,different from the rule forsubstitution characteristic of propernames. So, forWitt- genstein,the sign for what is thecase (or is notthe case) is thesentence, a signto whosefunctional character it belongsthat no sentence'struth or falsitycan rob it ofits capacity for comparison with reality; and that logical characteris tied to a particularkind of substitutionrule, differentfrom that for any referring sign. Possibilityand necessityget expressedin the use of ordinarysentences, in inferencesfrom these sentencesto moreof thesesentences. It is a mistaketo thinkthat you can in thoughtcatch hold of,mean, that possibility that is reflectedin theordinary sentence you use, and, forexample, consider what under- lies itsbeing possible, as ifthat were a characteristicof something. The mistakeis ofthe same general character as theattempt to ignorethe way a conceptis referredto, the attemptto mean it, bare of the logical accoutrementswith which language covers it-which was whatRussell explictlysays we can do. I wantnow to saywhere my present argument will wind up, beforeI get it to go there.There is the Russellconfusion, in whichyou think thatyou can graspthe identityof a thingabstracted from the use of a termfor it in sentences,and againstit there is theidea runningthrough Frege's thought,that logical characteristicsof the expressionsthat standfor a thingbelong to whatit is we weretalking about. In Frege's hands,that principle is directedagainst Benno Kerry, and wouldhave been directedagainst the remarksof Russell'sI quoted. From Witt- genstein'spoint of view, that very confusion (the confusion of thinking thatyou can graspwhat you are talkingabout, pulled awayfrom the logical featuresof any expressionfor it) is present(without Frege's seeingit) in Fregehimself, and in anyonewho thinks of the laws of logic as true.The turningof the Frege principleagainst Frege depends on thispoint: ifRussell's confusion is theattempt to thinkof something, abstractingfrom the logical featuresof how we speak about it, then whatyou will take to be the same confusionas Russell'sdepends on whatlogical features you thinkthere are, internalto how we referto things. For Frege, the distinctionbetween functions and objects, betweenfunctions of differentlogic levels, and betweenfunctions withdifferent numbers of arguments:these distinctions all belongto the logicalfeatures of the referringexpressions, the names,of every language.For Wittgensteinas forFrege, the logicalcharacteristics of referringexpressions are tied to the role of such expressionsin sen- tences,where a sentenceis a functionof the expressions it contains. But theydisagree very deeply about whatkind of functionof the expres-

26

This content downloaded from 130.88.20.147 on Thu, 19 Nov 2015 13:34:10 UTC All use subject to JSTOR Terms and Conditions ThrowingAway the Ladder

sions in it a sentenceis, and how you tellthat an expressionis one of those that a sentenceis a functionof. For Wittgensteinbut not for Frege,you can tellwhat the sentenceis, whatexpressions occur in it, onlyrelative to thelogical character of the sentence itself as a signwhose capacityfor truth or falsityis independentof the truth or falsityof any sentence.The logicalrelations of sentences to each otherenter the way we tellwhat sentence our sentence is, whatexpressions how combined. The wholeof logic is internalto thelogical character of every referring expression. Here is a roughsketch of another way of making the contrast between Fregeand Wittgenstein.For Fregethe fundamentallogical relation is thatbetween ;they are 'made for each other'.15That relationis reflectedin languagein the relationbetween predicate and propername. Wittgenstein substitutes for Frege's fundamental logical relationbetween concept and objectthat between object and situation: theyare made for each other.And thatfundamental relation is reflected in languagein the relationbetween two sortsof sign: one, thatstands forsomething, and theother, that in whichthere is standingfor; where thelatter kind of sign,the sentence, has as itscharacteristic feature its capacityto occur as argumentof truth-functionaloperations. The truth-functionalcalculus, within which sentences have their identity as signs,is whatgoes withany referring expression. (Frege and Wittgen- steincould not differ about what sort of functions sentences are without differingabout what sort of argument expressions in themare, without differingabout the logicalcharacter of argumenthood.) If you thinkthat the whole of logic is internalto any referring expression,you will see the Russellconfusion wherever anyone treats anypart of logic as externalto whatwe aretalking about. Anyone who, likeFrege, treats logical laws as holdingof objects and functionswill be imagininga kindof referenceto objectsand functionswhich (on your view) is an illusion:such a criticismis analogousto thatwhich Frege could have directedat Russell. Given Wittgensteins account of the characterof sentences, it willappear that anyone who thinksof logical truthsas genuinelytrue, anyone who thinksof logicaltruths as true because theirtruth conditions are met,will be in a confusionof the same essentialcharacter as Russell's: he will be supposinghimself to have access to whathe is talkingabout, eventhough he is abstracting fromthe logical character of the signs he uses to sayanything. The idea ofa scienceof logic is, on Wittgenstein'saccount, nothing but illusion.

Universityof Virginia

15 GottlobFrege, Posthumous Writings, Hans Hermeset al. (eds) (Oxford: Basil Blackwell,1979), 118, 178.

27

This content downloaded from 130.88.20.147 on Thu, 19 Nov 2015 13:34:10 UTC All use subject to JSTOR Terms and Conditions