Devitt Dummett's Antirealism PDF

Journal of Philosophy, Inc.
Dummett's Anti-Realism Author(s): Michael Devitt Source: The Journal of Philosophy, Vol. 80, No. 2 (Feb., 1983), pp. 73-99 Published by: Journal of Philosophy, Inc. Stable URL: http://www.jstor.org/stable/2026236 . Accessed: 14/02/2014 20:40
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.
Journal of Philosophy, Inc. is collaborating with JSTOR to digitize, preserve and extend access to The Journal of Philosophy.
http://www.jstor.org
This content downloaded from 75.97.192.97 on Fri, 14 Feb 2014 20:40:52 PM All use subject to JSTOR Terms and Conditions
THE JOURNAL OF PHILOSOPHY

VOLUME LXXX, NO. 2, FEBRUARY
1983
DUMMETT'S
ANTI-REALISM*
M~ICHAEL DUMMETT is a prolific,subtle, but complex in the philosophy of language and the philosophy writer LViof mathematics.It is well known that he argues against realism. In the philosophical circle centeredon Oxford the influence of this argumentis already great.' Elsewhereits influenceis 2 growing. Crispin Wright, an able and vigorous defender of Dummett,claims thatDummetthas "set up what promisesto be one of the most fruitful philosophical researchprogrammesof this century."3Yet I suspect that many philosophers are skeptical about Dummett'sargument:it smacks too much of positivismand Wittgensteinianism.4 I sympathizewith the skepticsand disagree with Wright.
li/I
* Earlier versionsof this paper weredeliveredat La Trobe University (July 1980), the 1980 Annual Conferenceof the Australasian Association of Philosophers in Sydney,the University of California,Los Angeles (January1981),the University of Michigan (February 1982), and the University of Wisconsin at Milwaukee (March 1982). I am indebted to many forcomments. Those by the following have led to changes: RogersAlbritton, JohnBigelow, TylerBurge,Joshua Cohen, Hartry Field, Philippa Foot, Ken Gemes, Karen Green,TimothyMcCarthy, PeterSnow, Kim Sterelny,and Nicholas White.I am particularly indebtedto Gregory Currieforforcing me to see that a verificationist argumentnot based on "the propositional assumption" could be abstractedfromDummett'sdiscussion. In effect the point had been made to me by Hilary Putnam in correspondence about myDesignation (New York: Columbia, 1981),but I had failed to see its significance. 'See, e.g., many of the papers in Mark Platts,ed., Reference,Truthand Reality: Essays on the Philosophy of Language (London: Routledge & Kegan Paul, 1980). 2 See, e.g., Hilary Putnam, "Realism and Reason," in Meaning and the Moral Sciences (London: Routledge & Kegan Paul, 1978): 123-138;parenthetical page referencesto Putnam will be to this book. "Critical Study:Dummettand Revisionism,"Philosophical Quarterly, XXXI, 122 (January1981): 47-67, p. 67. 4Wright was recentlyprovoked to defend a Dummettian view partly by Peter Strawson'scriticisms of thatview, "Scruton and Wright on Anti-Realismetc.," Proceedings of the AristotelianSociety(New Series), LXXVII (1976/7): 15-22,and partly by "the gratefulreceptionof [Strawson's] remarksby an audience who seemed,by and large, to thinkthatanti-realism could be nothingotherthan the Positivismof the Thirties" ["Strawson on Anti-Realism," Synthese, XL, 2 (February 1979): 283-299, p. 283].
0022-362X/83/8002/0073$02.70
1983 The Journalof Philosophy,Inc.
73
74
Dummett'sargumenthas a general formthatis claimed to cover various "realisms," e.g., about common-sense physical entities, and about the about scientific entities, about mathematicalentities, of these.My aim is to past. I shall be concernedonly with the first defendrealism about common-sensephysical entitiesfromDummett'sargument.I shall call the doctrineI defendsimplyRealism. The twentiethcenturyhas seen a "linguistic turn" in philosophy. Dummett'sargumentagainst Realism exemplifiesa strong commitment to this turn: disputes concerning The wholepointofmyapproachto [thevarious ofmeaning underlies metarealism] has beento showthatthetheory tophilosophy, I physics. IfI havemadeanyworthwhile contribution think it must lie in having raisedthisissuein these terms.5 For Dummett,"the goal of philosophy is the analysis of the structure of thought." He justifies the pre-eminencehe assigns to the philosophy of language by its bearingon thatanalysis (TOE 458).6 view of Underlyingmy criticismof Dummett is a verydifferent view. This does philosophy,what is sometimes called a "scientific" not make mycriticism out of place; forRealism is as appropriatea place as any for these rival views of philosophy to join battle. I thinkDummettwould agree (see TOE 24). in Dummett's anti-Realism: for There are three premises argument statements A. The Realism dispute is the dispute about whether truth have realist(evidence-transcendent) oronlyverificationist conditions. The statements in question hereare, of course,statements containCall ing words like 'stone', 'tree',and 'cat': "physical statements."
' Truth and Other Enigmas (Cambridge, Mass.: Harvard, 1978), p. xl; hereafter, TOE. Other works of Dummett to be cited are as follows: Frege: Philosophy of FPL; "What Is a Theoryof MeanLanguage (London: Duckworth,1973); hereafter, ing?" in Samuel Guttenplan,ed., Mind and Language (Oxford: Clarendon Press, WTM; "What Is a Theory of Meaning?(II)" in Gareth 1975): 97-138; hereafter, Evans and John McDowell, eds., Truthand Meaning: Essays in Semantics(Oxford: WTM II: Elements of Intuitionism(New Clarendon Press, 1976): 67-137; hereafter El; "Comments,"in Avishai Margalit,ed., Meaning York: Oxford,1977); hereafter, C; "Common Sense and Physand Use (Boston: Reidel, 1979): 218-225; hereafter, ics," in G. F. MacDonald, ed., Perception and Identity:Essays Presentedto A. J. CSP. Ayer,with His Replies (Ithaca, N.Y.: Cornell, 1979): 1-40; hereafter, to Dummett'swork to supportan attribuI shall oftenmake multiple references tion. In so doing I do not mean to suggestthattheseare the only places thatwould is supply the support. Indeed any attemptto be comprehensivein such references rapidlybecoming hopeless. Aside fromthe 1800odd pages in theworkscited above, of Frege's Philosophy a follow-upvolume to FPL of 621 pages, The Interpretation (Cambridge, Mass.: Harvard, 1981), appeared while the presentpaper was in the hands of this JOURNAL. And Dummett'sWilliam JamesLectures,The Logical Basis are imminent.Is thereany chance of a timeout? of Metaphysics, of this view of philosophy to Frege and sees it as definitive 6Dummett attributes "analytical philosophy" (TOE 442).
DUMMETT'S
ANTI-REALISM
75
the doctrine that these statementshave realist truthconditions Realist Truth. For DummettRealism is Realist Truth. B. The dispute about truth conditions is the dispute about is realist(evispeaker'sunderstanding whetherthecompetent or only verificationist. dence-transcendent) is only verificationist It follows fromB that if the understanding then Realist Truth is false. With A this leads to anti-Realism. is only verificationist. C. The competent speaker'sunderstanding is a more basic (A consequence of B and C is thatverification semanticnotion than truth.) There is general agreementon only one thingabout Dummett's So it may seem improbable thatit could philosophy: it is difficult. to be made "as simple as A-B-C" (as JohnBigelow nicelyremarked me). In fact B and C conceal complications: each has two distinct versions.With those complications takeninto account I do indeed with theirsupporting claim thattheabove threepremises,together constitute Dummett'sargumentagainst Realism. arguments, Dummett focuses his discussion of Realism on verificationism and premise C. The discussion of A and B is slight. Yet these two premisesare crucial to Dummett'scase against Realism. My focus will be on them. The paper is in fourparts. Part I is on premise A and the relathatA is tionship betweenRealism and Realist Truth. I argue first false (I.1), next thatDummettsubscribesto it (I.2), and finallythat his argument for it is inadequate (I.3). Nothing simply follows about Realism fromany view of Realist Truth. In Part II I distinguish two versionsof B, hence two of C, which Dummettconflates. Bi and Cl are based on theassumption thatlinguisticcompetence involves propositional knowledge of truthconditions. B2 and C2 see competence as merelya practical ability (2.1). I argue that Dummettgives the "propositional assumption" about competence no adequate support(2.2) and thatit is false (2.3). So Bi and Cl are false. Verificationist argumentsto show thatspeakersdo not know to Realist Truth. Part III takes realisttruth conditionsare irrelevant competenceas a practical ability. Interestthen centerson C2, the view that the sentencesunderstoodby the competentspeaker have truth conditions.I set out Dummett'sargument only verificationist for this (3.1) and rejectit (3.2). In Part 4 I argue that,in any case, verificationism has littlebearingon Realism. Theories of language and understanding theories of theworld (4.1). should not determine I. 1 Realism and Truth I seek first a statement of thedoctrineof Realism thatcapturesits traditional opposition to idealism and phenomenalism about
I. REALISM AND REALIST TRUTH
76
common-senseentities.There are two dimensions to this doctrine: first, a claim about what exists;second,a claim about thenatureof dimension we can say thatit is thatexistence.To capture the first common-sensephysical entities that exist. Words that frequently occur in attemptsto capture the second are 'independent','external', and 'objective'. The entitiesmust be independentof the mental; theymust be externalto the mind; theymust exist objectively anyone's opinions. We can captureboth in thattheyexist whatever thesedimensionswell enough in the followingdoctrine: Common-sense physical entities objectively exist independentlyof the mental. A lot more than this could certainlybe said to clarifyRealism.7 I fairlyobviously,even if take it, however,that thischaracterization a littlevaguely,expressesthecentralintuitionsof Realist doctrines about the external world, doctrines that have always seemed so unof Realism to be a relatively plausible. I mean this statement controversial statement such as mightbe made by any preliminary by thehistory of philosophy prior to Dummett.It person informed is, forexample, in accord with the entry"Realism" in The Encyclopedia of Philosophy.8 We are not entitledto insist that 'realism' be used in this way, ratherthan, say, forsome semanticdoctrine(as is now common), but we are entitledto wonder whetherin another use it has anything to do with traditionalmetaphysicaland epistemic disputes In the next section between realistsand idealists/phenomenalists. we shall see thatDummettcertainly does see Realism (characterized in the above way) as threatened by his arguments. What has truthto do with this doctrineof Realism? On the face of it, nothing at all. Realism says nothing about truthnor even sentences and beliefs(exceptperhaps,in about thebearersof truth, its use of 'objective', the negative point that beliefsdo not determine existence).Realism says nothingsemanticat all. Realism does not strictlyentail any doctrine of truth nor, I would claim, is thereany obviously true proposition which, togetherwith Realism, entails a (nontrivial)doctrineof truth.There in being a Realist and yettakinga thoroughly is no inconsistency I take skepticalview of theneed foran explanatorynotion of truth. it thatsuch a view is a centralpartof Quine's semanticskepticism. for a physicalisticallyminded realist to underIt is verydifficult mine thisskepticism, as StephenLeeds has pointed out in an excellent article.9
say a lot more in Realism and Truth(in preparation). R. J. Hirst, "Realism," in Paul Edwards,ed. Encyclopedia of Philosophy, vol. vii (New York: Macmillan, 1967): 77-83, p. 77. and Truth," Erkenntnis, xiii, 1 (July 1978): 111-129. 9"Theories of Reference
7 8
DUMMETT'S
ANTI-REALISM
77
So much forthe entailmentfromRealism to a doctrineof truth. It is common to link a "correentailment? What about the reverse spondence" or "realist" notion of truthto Realism. Is a person committed to Realism if he adopts such a notion for physical statements? We need to know more about this notion of truththan thatit is Some may thinkthatany notion thatcan "evidence-transcendent." are philbe defined'a la Tarski is realist.Such definitions, however, osophically neutral,as Hilary Putnam points out (2-4, 9/10): they to "a worldout there"(1). miss theintuitiveidea of correspondence is a "picturing"or "mirroring" The view thatthiscorrespondence relationship is nothing but a metaphor. What is needed, in my relationsbedependenton genuine reference view, is to make truth tween words and objective reality.We mightcapture the doctrine as follows: thatsuch a notion is appropriateforphysicalstatements objective of: (i) their are trueor falsein virtue Physicalstatements and parts between their referential relations structure; (ii) theobjective reality. of that nature reality; and (iii) theobjective Realist Truth.It does of thedoctrine, This is mycharacterization not make truthdependenton our having the evidence or on our having the capacity to get the evidence. Truth is altogetherindeThis is not to pendent of evidence; it is "evidence-transcendent." say that truthis "unknowable," whateverthat might mean, but it another.'0 is one thing,our means ofdiscovering simplythattruth Does Realist Truth entail Realism? It does not. Realism (as I have defined it), requires the objective independentexistence of common-sensephysical entities. Realist Truth concerns physical it says nothingabout the and has no such requirement: statements trueor false, exnature of the realitythat makes those statements cept that it is objective. An idealist who believed in the objective existenceof a purelymentalrealm of sense data could subscribeto are trueor Realist Truth. He could believe thatphysicalstatements false according as theydo or do not correspond to the realm of sense data, whateveranyone's opinion on the matter:we have no "incorrigibleknowledge" of sense data. (He mighteven believe in the objectiveexistenceof physical objects but thinkthemnothing but sense data.) Analogously some nominalistsin the philosophy of mathematicsaccept a doctrineof realist truthformathematical to linguisticitems.And an operstatements by takingthemto refer apparentlyabout ationist can accept realist truth for statements about observables. In unobservablesby taking them to be really will talk of not truth yieldany particularontology. sum, mere
10 See my "Realism and the Renegade Putnam: A Critical Studyof Meaning and in Nous, May 1983; sec. 5. the Moral Sciences," forthcoming
78
of is in any way constitutive I conclude thatno doctrineof truth Realism and thatpremiseA is false. This conclusion does not mean, of course, that the issues of have no bearing on each other.Indeed I thinkit Realism and truth is impossible to finda plausible epistemologyforRealism thatcan be combined with most,if not all, epistemicdoctrinesof truth.In general, we can expect a position on one of theseissues to lead by to the best explanation to a position on the other.I will inference in IV.1. However, links of this sort are consider such inferences weak and should lead to great caution in running the two issues together. to A I.2 Dummett'sCommitment idenDummettis not at all cautious. He oftenstraightforwardly that the realism/anti-realism he says Thus two issues. tifies the dispute ofthedisputed for statements appropriate concerns thenotion oftruth thekindofmeanconcerning it is a dispute class;and thismeansthat 155, have (TOE 146;see also xxx,22/23, statements ing whichthese CSP 3). 314,358/359; of RealOn this view the doctrineof Realist Truth is constitutive "possess an objective ism: it is thedoctrinethatphysicalstatements truth-value,independentlyof our means of knowing it"; their tied to thekind of evidenceforthemwe meanings "are not directly can have" (TOE 146; see also FPL 466). On the otherhand, antiRealism denies Realist Truth. This view is premiseA. SometimesDummetturges an apparentlyweakerview: the theory of meaning "underlies" the realism dispute (TOE xxx, xl) or supplies "the premisses" for various positions in the dispute (EI 382/383;TOE xxviii, 229). However, Dummettattachesno significance to the difference betweenthesetwo views: to a truthin-an adherence consists Realismrests upon-or better, for conditional semantics our language(C 218). under the term'Realism' is, Given thatwhat I have characterized fromany semanticdoctrine,it is apprima facie, quite different with Dummettover my disagreement propriateto wonderwhether premise A is "merelyverbal." Perhaps he is using the termdifferentlyand does not intend his discussion to have any bearing-at dispute. metaphysical leastnotanydirect bearing-on thetraditional This is certainlynot the case. He describes the realism that he way: with (restsupon) a semanticdoctrinein a traditional identifies thatuniverse of thephysical universe, . . . on a realistic conception ofit (El ofourknowledge an objective constitutes reality, independent 382;seealso TOE xxv,228).
DUMMETT'S
ANTI-REALISM
79
He sees the realism that is threatened by his discussion as a doctrineabout what therereally is, an ontological/metaphysical doctrine (EI 386; TOE 146/147,230) opposed to idealism (FPL 671, 681; TOE 145) and phenomenalism (TOE 147). His problem is old, but the semanticapproach is new (TOE xxxi). Indeed, as the passage quoted at the beginning of this paper shows, Dummett thinks that any worth-while contributionhe has made to philosophy lies in this approach to such metaphysicalproblems. Finally, Dummett sometimesexpresses the anti-realismthat his argument inclines him toward in ontological terms,though the expressionis briefand obscure. His anti-realism"appears a more radical repudiationof objectiverealitythanidealism" (C 223); realitycomes into existenceas we probe, though we do not createit; it is not fullydeterminate (TOE xxviii/xxix,18/19,229/30;C 221/2). So thedifference of opinion over premiseA is not merely verbal. Dummett sees his discussion as bearing directlyon Realism. He does thinkthat the particularformof his semanticversionof realism may depart somewhat fromtraditionalusage. He thinksthat usage has confused two issues: (a) the issue of whetherstatements of one kind (e.g., physical) can be reducedto thoseof another(e.g., sense-datum);(b) the issue of whetherstatements of the one kind are determinately trueor false (TOE xxxii, 157-159;CSP 1-5). His departurefromtradition, he thinks,comes in his setting(a) aside and taking(b) as the basic issue. Traditionally,as I have pointed out, therehave been two dimensions to Realism: the dimension of existenceand the dimensionof independencefromthe mental. Now the second of these has usually been an issue of reduction:Can physical objects be reducedto ideas or sense data? This issue of ontological reductionis not the same as Dummett's(a), an issue of linguisticreduction.Overlook that,for Dummettsets (a) aside. My centraldisagreement concerns (b). What has the ontological issue of what exists to do with (b), the linguistic issue of determinate truthvalues? On the face of it, nothingat all.1' III.3 Dummett'sArgument forA Dummettcompletelymisconceivesthe Realism issue. Why?The crux of the explanation is that, for Dummett,any metaphysical
" Dummettsuggeststhat the importanceof reductionto the Realism dispute has been its role as a step toward(b) (CSP 5). I suggestthatits importancecomes from the fact that anti-Realistshave typicallybeen conservative. Thus Berkeleydid not deny the veryexistenceof the common-sensephysical world, but claimed that it could be reducedto somethingmental.This conservatism was good tactics:abandon it and anti-Realism would deserverefutation by the kicked stone and the incredulous stare.
80 view is
than as a represenotherwise no substance a picture which has in itself of meaning(El 383; see also TOE tationof the givenconception xxviii/xxix). The metaphysicalview adds only a "metaphor" (TOE xxv/xxvi, 229).12Such a pictureor metaphoris nonetheless"natural" (TOE xxviii); it will "forceitself"on us (TOE 229/30);it is "irresistible" (TOE xxviii, 230). So although Dummettwill fromtime to time use the language of the traditionalmetaphysicaldispute,when the chips are down he talksonly of meaning. Dummett's belief in this metaphor thesis-metaphysicsbeyond meaning is mere metaphor-is central to the explanation of various puzzling aspects of his discussion which we have noted: first, of his holding premiseA; second, of his attachingno significance betweenA and the apparentlyweaker view that to the difference the dispute over truth"underlies" the Realism dispute; third,of and obscurityof his account of the ontological consethe brevity quences of his anti-Realistargument. Why does Dummett believe the metaphor thesis?Why does he thinkthe metaphysicaldispute about Realism cannot stand on its The cause is clear, but it supplies no good reason. own feet? Dummett's view of metaphysicscomes fromhis philosophy of mathematics(TOE xxiv). He thinks (i) that the critical disagreement betweenplatonists and intuitionistsis over the appropriate and that this is a disagreement formsof reasoningin mathematics over the type of over the meaning of mathematical statements, have (El 380; TOE xxvii/xxviii). truthconditions thosestatements Now thisalone does not show, of course, thatthereis not another, issue of the over the metaphysical even if less critical,disagreement nature of mathematicalobjects; it does not show that thereis not over whetherthese objects are inalso a substantivedisagreement existingabstractobjects or are creationsof thehuman dependently (ii) Dummettdoes want to claim that thereis mind. Nevertheless, no such furthersubstantive disagreement. He is impressed by "the problemis not theexistenceof matheGeorg Kreisel'sremark, matical objects, but the objectivityof mathematical statements" over (quoted, e.g., at TOE xxviii). He thinksthatthedisagreement
12 Simon Blackburn,"Truth, Realism, and the Regulation of Theory,"in PeterA. eds. MidwestStudies in French,Theodore E. Uehling Jr.,and Howard K. Wettstein, Philosophy, Volume V: Studies in Epistemology(Minneapolis: Univ. of Minnesota Press, 1980): 353-371,argues fora metaphoricalview of the Realism dispute; cf. my "Realism and Semantics: Part II of a Critical Studyof Midwest Studies in Philoin Nou4s, November1983; sec. 1. sophy V," forthcoming
DUMMETT'S
ANTI-REALISM
81
objects adds only metaphorsto the discussion, one metaphorseethe other seeing him as ing the mathematicianas an astronomer, an artist(EI 382/3; TOE xxv/xxvi,229). Suppose thisis so. It does about the not show, of course, thatall metaphysicaldisagreements ontological statusof objects have no substancebeyonda disagreement over meaning. Perhaps the disagreementbetween Realists even though that and idealists over physical objects is substantive between platonists and intuitionistsover mathematicalobjects is metaphorical. Yet, (iii) Dummett does want to extend his claim about mathematicsto all fields(EI 381-383; TOE xxix). He arrives at the metaphorthesis. over mathematicalobConsider (ii). Whydoes the disagreement over meaning?So far jects add only metaphorsto thedisagreement as I can see Dummettdoes littlemore than claim thatit adds only makes disagreement this. For Dummett,it seems,the metaphysical it is no literal sense on theface of it, and the only way to interpret to relateit to the issue of meaning. in Dummett'sdisA certainargumentdoes featureprominently over these cussion of mathematicalobjects. Take the disagreement objects literally.Dummettargues thatit does not bear on the disagreementover meaning. Assuming a platonist ontologydoes not lead to a platonist view of meaning and logic; assuming an intuitionistontology does not lead to an intuitionistview of meaning and logic (FPL 507/8; EI 382-389; TOE 230-247). Suppose Dummettwere rightabout this lack of bearing of the ontological dispute on the semantic one. (I don't thinkhe is, for reasons indicated in I.1 and consideredin IV.1. Dummett'sargument depends on his views about meaning, to be discussed in the next two parts of this paper.) The mere fact that a disagreement does not have any consequence forsemanticsdoes not show thatit is not a real disagreement, does not show thatit is only metaphorical. Most real disagreements to semantics. are irrelevant Despite the lack of argumentforit, Dummett'sclaim about the overmathematicalobjects metaphoricalnatureof thedisagreement seems to me to have some plausibility. Consider,however,how it does would strikea hard-coreplatonist. For him the disagreement not seem metaphorical;he thinkshe has a clear conception of independentlyexisting mathematicalobjects, a conception that he He froma conception of mental constructs. findsvividlydifferent will thinkDummett'sview mistaken.Furthermore, Dummett'sargument that the platonist's ontologydoes not settlehis semantics gives the platonistno reason to thinkotherwise. We see the importanceof this when we examine move (iii) in
82
Dummett'sargumentforthe metaphorthesis:the extensionof his forexto otherfields.What reason is there view about mathematics tending such sympathyas we may have for Dummett's position platonistto his position against thehard-core against thehard-core Realist about the physical world? So far as I can see Dummett offers no argumentforthis extensionbeyondclaiming thatthe argumentabout the lack of bearing of ontologyon semanticscarries over frommathematicsto otherfields(EI 381; TOE xxix). Even if this were so, and theargumentweregood, it would no more show that the metaphysicaldispute about physical objects is metaphoridisputeabout mathcal thanit earliershowed thatthemetaphysical ematical objects was metaphorical. Not only is the extension from mathematics unargued; it is betweenthe mathematicaland highly implausible. The difference physical worlds is striking.'3The platonist's conception of indemanypeople as farobjectsstrikes existingmathematical pendently fetched.What are theylike? Where do theyexist? How could we come to know about them? The intuitionist'sconception is no or How could numbersbe mentalconstructions more transparent. free creationsof thehuman mind?These conceptionsare so odd, so of the dispute is hard to grasp, thata metaphoricalinterpretation tempting(though I don't say right). In contrastthe Realist's conexistingphysical objects is the verycore ception of independently is of common sense. There certainly some vaguenessabout it, some explanation, but it is not in the least bit metaroom for further phorical. Indeed if this talk cannot be taken literally,what talk can? It must be close to a bench markof the literal. Even theidealist's conception, the traditional rival of Realism, has a certain for we are all familiarwith minds and experiences. transparency, Only a philosopher could suppose that our talk about language, one of thenewestand least developed areas of knowledge,is clearer on the face of it than our talk about ordinaryphysical things,one of the oldest and mostdeveloped areas of knowledge. In this part of the paper I have argued that Dummettsubscribes to premise A and that it is false. His reason foradopting A is the
13 See the passage from Strawson's "Scruton and Wrighton Anti-Realismetc." quoted by Dummett(TOE xxiv), which makes a similar point. In a paper fromTOE not so farcited, "Platonism" (pp. 202-214), Dummetthimselfemphasizesthedifferences betweenthemathematical and physicalworlds,drawing attentionparticularlyto the power of the physical world to affect us through observation.This seems to me verymuch along therightlines. It counts against the extension frommathematicsto physics which I have attributedto Dummett,and hence against that attribution.However, without the extensionand the metaphor thesis that is supported by it, therewould be no argumentat all in Dummettfor premiseA.
DUMMETT'S
ANTI-REALISM
83
metaphor thesis. This rests on an unsupported though possibly plausible claim about the metaphorical nature of the ontological and on an unsupported dispute in thephilosophy of mathematics, and highly implausible extension of that claim to the dispute about physical reality. Experiencesuggeststhatsome people will not accept thatDummettdoes subscribeto A, despite theevidencethathe does. Suppose he does not. What then remains from this part? The claim that nothing simply follows about Realism fromany thesisabout the truthconditions of statements.So even if an argumentis made argument is neededto establishantiagainst Realist Truth,a further Realism. And this would still leave a possibilityopen to the Realist: he mightreversethe orderof argument.He mightgive a nonsemanticargumentfor Realism and then use Realism to establish Realist Truth. These moves will be consideredin IV. 1.14 II.1. Versionsof B and C Dummett'ssecond premiseis: B. The dispute about truth conditions is the dispute about whetherthecompetentspeaker'sunderstanding is realist(evior only verificationist. dence-transcendent) of meaning This is a corollaryof Dummett's thesis that "a theory is a theory of understanding"(FPL 92; WTM 99). With A it yields the view that if understandingis only verificationist, Realism is false. On the face of it, B is odd. How could a semanticdispute about the truthconditions of sentencesbe a psychological dispute about thecompetentspeaker'sunderstanding? How could disputesabout Yet Dummettdoes such different sortsof propertybe the same?15 equate thetwodisputes(WTM II 68/69;TOE 153-155,358/9).Why? We mustdistinguishclearly,though Dummettdoes not, two versions of B, reflecting quite different assumptionsabout the nature of speaker competence. The firstand most prominentversion is based on the assumption thatcompetenceconsists (at least partly) in knowledgeof truth(falsity) conditions:16
14 I have not consideredthe details of Dummett'ssemanticversionof Realism: he makes the principle of bivalence the touchstone(TOE xxx-xxxii,149/50,155; CSP 4). The details are not centralto my argument. 15I don't mean to suggest that psychological propertiescould not enterinto an explanation of semantic properties;I am too much of a Gricean for that; see my Designation op. cit., pp. 80-86. My point is only the trivialone that the semantic and psychologicalproperties in question here are verydifferent (ibid., pp. 92/3). 16 Parenthetical additions such as thosein thissentenceshould be takenas read in future.
II. THE PROPOSITIONAL
ASSIJMPTION ABOUT COMPETENCE
84
Bi. The dispute about truthconditions is the dispute about whether the competent speaker knows realist (evidenceor only verificationist transcendent) truth conditions. The knowledgein question here is propositional: it is knowledgethat and not, for example, mere knowledge-how. So on this assumption an L-speaker'sunderstanding of a sentenceof L consists in his knowing that the sentenceis true-in-Lin such and such circumstances. Now if he knows this,it must be so. Thereforea dispute over whetherthe truth conditionsthata speakerknows are, as a matterof fact,realistor only verificationist would settlewhether the truthconditions are realist or only verificationist. At least it would settle this if we assume that whatevertruthconditions a sentencehas, the speakerwill know it to have. This assumption is probablyimplicitin the "propositionalassumption" about competence.So we have establishedBi. With the propositionalassumption goes a versionof C also: Cl. The competent conspeakerknows only verificationist truth ditions. The rival to this,which the believerin Realist Truth is thoughtto be committedto, is theview thata speakerknows realist(evidencetruth conditions. transcendent) The propositional assumption about competence is a received wisdom of contemporary semantics.As HerbertHeidelbergerhas recentlypointed out, it seems to be regardedas uncontroversial, "perhaps unworthy of serious discussion." Yet it is not "obviously "17 In true. my view it is false. Hence Bi and Cl are false. I think competence is simply a set of grounded skills or abilities. I shall indicate why in II.3. The attributionof the propositional assumption, and hence of Bi, to Dummettis well based but not certain. Philosophers who make the assumption usually hedge theirbets. Dummettis no exception. It is clear thathe thinksof competenceas "a practicalability," but this alone does not count against the attribution because he mostlywritesas if the abilityconsistedin propositional knowledge of truthconditions. However, ratherthan straightforwardly identifying the ability with the knowledge, he seems to preferto say that it can be "represented"as that knowledge (FLP 461/2; WTM 105-109, 121-125; WTM II 69-71; EI 373; TOE 128/9). he thinksthatpartof thisknowledgeis only "implicit" or Further,
17"Understanding and Truth Conditions" (Midwest Studies V: 401-410), p. 402. Heidelbergergoes on to argue that though knowledgeof truthconditionsdoes not implycompetence, competence"nearlyenough" implies the knowledge.I rejectthe latterargument,"Realism and Semantics,"sec. 2.
DUMMETT'S
ANTI-REALISM
85
"tacit" (WTM II 70/1,80; EI 373/4;TOE 129).18This use of weasel words in stating the propositional assumption casts doubt on Dummett'scommitmentto it. So I shall not restmy case against Dummetton attributing thisassumption to him.'9 Abandoning the propositional assumption, and taking competenceas simplya practicalability,we get anotherversionboth of B and of C: B2. The dispute about truthconditions is the dispute about whetherthe sentencesunderstoodby the competentspeaker have realist (evidence-transcendent) or only verificationist truth conditions. C2. The sentences understood bythecompetent speakerhave only verificationist truth conditions. These differ fromBi and Cl in making no mentionof knowledge. B2 mustbe true,given the tautologicalassumptionthatthecompetentspeakerunderstands thesentences of thelanguage. Interest then focuseson C2. I shall argue in 111.2thatit is false. 11.2 An Argument for the Propositional Assumption? Meanwhile we must consider Bi and the propositional assumption on which it depends. That dependence requires that we take theassumption strictly and literally: it is no meremannerof speaking; the speakerreallydoes know thatthe truth conditionsare such and such. At first sightit may seem thatDummett'slengthy argumentsfor a verificationist and against a realisttheory of understanding (e.g., in WTM and WTM II) constitutean argumentfor the propositional assumption. For the conclusion is always presentedas: the speaker knows only verificationist truthconditions; i.e., as Cl. However this way of putting the conclusion simplyreflects Dummett'sconflationof the propositional and nonpropositionalviews of competence.There is nothingin Dummett'sargumentforverifi1 Dummettalso makes the mysterious remarkthata theory of meaning,which is forhim a theory of understanding, is not "a psychologicalhypothesis"partlyon the ground, it seems, that somethingwith "internal mechanisms" unlike ours, e.g., a Martianor a robot,mighthave therequiredimplicit knowledge(WTM II 70). This overlooks the distinctionbetweenpsychological and physical mechanisms.A Martian or a robot that is physicallydifferent fromus might be psychologicallythe same. The remarkis mysterious because if understanding is anythingit is (at least, partly)psychological. 19I know of only one place in which Dummettdoes not conflatetheview thatunderstanding is propositionalwith theview thatit is a practicalability.This is in his recent"Comments" (C) on Putnam's "Reference and Understanding"(in Meaning and the Moral Sciences and in Margalit,ed., Meaning and Use). He therecontemplates dropping the view thata theoryof truthis a theoryof understanding, and thusdropping thepropositionalassumption. In effect he restshis case on B2 and C2.
86
cationism,howeversound, thatmakes it an argumentforCl rather than C2. Despite the popularity of the propositional assumption, I can find nothing in the literature thatcould seriouslybe called an argument forit. Apparentlyit is thoughtto follow in some obvious way fromtheclaim thatspeakers"know themeaning" of sentences in theirlanguage and fromthe slogan that "the meaning of a sentence is its truthconditions." Some passages in Dummetthint at this (e.g., WTM 105-109; WTM II 68/9; TOE 153-155). Consider also thefollowingsummary of Dummett'sviewsby Colin McGinn: If[a Tarskian theory oftruth] is to serve as a theory ofmeaning forL, and ifspeakers areacknowledged to knowwhatsentences of L mean, then be a sense there in whichthetheory must or serves states, tostate, whatspeakers of L knowin knowing whatsentences of L mean.20 Remarkslike thisare common in the literature, and yettheyrepresent not so much an argumentas a play on words. EveryattemptI have made to constructan argumentaround such remarksturns into a travesty. The following attempt,suggestedby Dummett's discussion of Frege'sdistinctionbetweensense and reference (TOE 117-126; particularly 124-126),21is typical (X is a competent speaker): (1) X understands S; (2) X knows the meaning of S; (3) The meaning of S = the truthconditionsof S; (4) X knows the truth conditionsof S; (5) X knows what the truthconditionsof S are. Let TC be the truth conditionsof S. .,.(6) X knows thatthe truth conditionsof S are TC. No objection can be taken to (1). And the move to (2) is acceptable enough if (2) is taken as a mereeveryday mannerof speaking. However,if (2) is to be construedas requiringthatthereexist some entity-the meaning of S-which X knows in the sense thathe is acquainted with it, then we should resist the move. We need a strongargument,not just ordinarytalk, beforewe accept such a
20 "Truth and Use," in Platts,ed., op. cit.: 19-40, p. 20; see also Crispin Wright, "Truth-Conditionsand Criteria,"Proceedingsof theAristotelian Society,supp. vol. L (1976): 217-245,p. 221. McGinn is mostly criticalof Dummett.However,he seems to agree with the view I have quoted. In general,he concedes a greatdeal to Dummettwithoutquestion. 21 In this discussion Dummettconsidersthe view thatsense is no more than reference. He takes this to be the view that competenceis no more than knowledgeof reference. He goes on to argue thatthisknowledgeconsistsin propositionalknowledge of a certainsort.Roughly,I have constructed theargumentI claim is suggested by this passage by replacing 'reference' by 'truthconditions'.
DUMMETT'S
ANTI-REALISM
87
requirement on linguistic competence. (2) does have to be construedin this way if (4) is to follow fromit and (3). There has to exist some entitywhich is both the meaning of S and the truth conditions of S, so that if X is acquainted with the one he is acwill not go throughif quainted with the other.Thus the inference we construe(2) as just another way of saying that X knows what the meaning of S is and construe(4) similarly[i.e., construeit as contextsare opaque. [And is (2) any more ac(5)]: knowing-what ceptable construedthis way than the otherway?] to (4). It is Next consider (3), the otherpremisein the inference truth conis its sentence based on the slogan thatthemeaning of a view. Nevertheless in my one is a good ditions. As slogans go, this it would be as follows: it is only a slogan. One way of interpreting and circumstances such such and in is true a sentence because it is semantic role the special it plays in that only those circumstances in our lives thatit does play. There is no reason to suppose thata the meaning of theory guided by this slogan will posit any entity, the truthconditions of S, as (3) requires.FurtherS, or any entity, and even if (2), wereto posit such an entity more,even if the theory wereacceptable, therewould be no construedas positingan entity, would be the "theoretical"entity reason to suppose thattheformer same as the latter "ordinary"entity.Certainlythe mere fact that theywereboth called "the meaning of S" would not show thatthey were the same. but therest These are perhaps the worstaspects of theargument, of it is also bad. First,(5) does not follow from(4): a person can be is. [Of withoutknowing what theentity acquainted withan entity course, as I have noted, we could construe (4) as (5), but then it from(5) would not follow from(2) and (3).] Second, the inference to (6) is dubious, to say the least. The main problem withit is that (as Dummetthimself seems to be context-dependent knowing-what diFinally, (6) cannot be inferred notes but setsaside: TOE 126).22 rectlyfrom(4), ignoring (5): X can be acquainted with an entity withoutknowing thatit is anythingin particular. because it involves a naive The argumentseems like a travesty view firstof an ordinaryuse of the word 'meaning', second of a theoreticalslogan, and thirdof the connection between the ordinary use and the slogan. So I don't attributethe argument to Dummett.However I do claim thatif this,or somethinglike it, is not his implicit argumentforthe assumption thatcompetencerenot even conditions,thenhe has offered quires knowledgeof truth
22See also StevenE. Boer and William G. Lycan, "Knowing Who," Philosophical Studies, xxviii,5(November1975): 299-344; and my Designation,pp. 222-224.
88
the glimmerof an argumentfortheassumption. If theassumption is false,so also are Bi and Cl. 11.3 Arguments against the Propositional Assumption In my view competencein a language does not consist in any semantic propositional knowledge at all. It is a set of grounded skills or abilities. It consistsin being able to do thingswith a lana language guage, not in having thoughtsabout it. Understanding no more involves having propositional knowledge of a semantic sortabout the language than being able to ride a bicycleinvolves having propositional knowledgeabout mechanics,or being able to digest food involves having propositional knowledge about digestion.23 Gilbert Harman has raised a verygood objection to theproposito the speakerretional assumption.24 The knowledgeit attributes to himselfthecondiquires thathe have some way of representing tions thatwould make sentencestrue.But whatdoes competencein the representing language consist in? Eitherthe same problem has reappearedor we are facedwith an equal one.25 I have raised a related objection.26Briefly, a person could not have semanticpropositionalknowledgewithouthaving thesemantic vocabulary of some language. That vocabulary is an isolable part of a language, just as is the biological or economic vocabulary. A person could be competentin the nonsemanticpart of a language without being competentin its semanticpart or in the semanticpart of any otherlanguage. So competencein the nonsemantic part does not consistin semanticpropositional knowledge. So competencein the language as a whole does not either. This is not the place to attemptthe large task of giving a theory of competenceand its relationto semanticproperties. In thecourse
23 Perhaps this overstates thecase a little(see also Designation, p. 107). It is plaufora termlike 'bachelor' is in terms sible to thinkthatthe explanation of reference of 'adult', 'unmarried','human', and 'male' (ibid., pp. 202/3). If so of the reference it may be the case that understandingthis termconsistsin knowing thatit means adult unmarriedhuman male. This could be the sort of termof which Dummett speakerhas explicitpropositional knowledge(EI 373). What I thinksthecompetent and in general, am most intenton denyingis thatcompetencerequires primarily, implicitor explicit. semanticpropositionalknowledge,whether 24 "Language, Thought, and Communication," in Keith Gunderson, ed., Language, Mind, and Knowledge: Minnesota Studies in the Philosophy of Science, Vol. 7 (Minneapolis: Univ. of Minnesota Press, 1975): 270-298,p. 286. 25 This objection indicatesan important factabout competencein L: thiscompetencemightcover many competencesincluding, e.g., competencein understanding competencein thinking spoken L, competencein writingL and, mostimportantly, in L. In thispaper I mostlyfollow the usual, rathermisleading,practiceof conflat(spoken?) L, and ing competencein speaking L with competencein understanding of writingas if thesecompetenceswere the only ones thatconcernedus. 26 Designation, pp. 97-100.
DUMMETT'S
ANTI-REALISM
89
of offering a semantictheory I have elsewheremade some preliminaryremarks towardthattask.27 These indicatehow we mightview competenceas a set of groundedskills or abilities. I summarize. I take the usual line of explaining meaning largelyin termsof truthconditions. Truth conditions are to be explained ultimately in termsof reference under the guidance of Tarski, as interpreted and developed by HartryField.28I look to causal theoriesfor the ultimateexplanation of reference. A causal theory explains the natureof reference in termsof a certainsortof causal chain, a chain with threekinds of link: groundings of word in object; reference in a person; borrowingswhen a word is passed on or reinforced and abilities with a word gained and sustainedby groundingsand reference borrowings. Implicit in this semanticsis a view of competence.Thus, to have the ability we all have with 'cat', to understandthe English word 'cat', is to be appropriatelylinked to the networkof causal chains for 'cat', a networkinvolving other people's abilities as well as groundingsand reference borrowings. To have thisabilitya person must be able to combine 'cat' appropriatelywith other words to formsentences.He mustbe able to have thoughtswhich those sentencesexpress. Furthermore, these thoughtsmust be grounded in cats. A Twin-Earthian,who in otherrespectshas the same ability with 'cat' that we have, does not have our understandingof the termbecause his ability is grounded not in cats but in apparently similar but really quite different animals, Twin-Earth-cats. However,having our abilitydoes not require knowingthat'cat' has any particularsemanticor syntactic property, nor does it requirebeing able to recognizecats. The theory has littleto say about the property of understanding a sentence,forexample, the sentence'The cat is on the mat'. That understanding involves,of course, having abilities with contained wordslike 'cat'. Beyondthatit involveshaving thesyntactic skill of combining words of those typesinto sentencesof thatstructure. It does not involve knowing that the truth conditionsof the sentence are such and such; it does not involve knowing that the sentence has any particularsemanticproperties. On thisview of linguisticcompetence, any propositional knowl27Ibid.,pp. 101-110,129-133,196-199. "Tarski's Theory of Truth," in Platts,op. cit.: 83-110. Field's views have been criticizedby Donald Davidson, "Reality without Reference," ibid.: 131-140 (cf. my Designation 118-124); by John McDowell, "Physicalismand PrimitiveDenotation: Field on Tarski," ibid.: 111-130(cf.myRealism and Truth);and by Hilary Putnam, Meaning and the Moral Sciences [cf. my "Critical Notice" of that book, Australasian Journalof Philosophy,LVIII, 4 (December 1980): 395-404].
28
90
overand above his edge ofa language thata personhas is something about thelanguage. gained fromtheorizing competence,something In the previoussectionI argued thatDummettgives thepropositional assumption no adequate support. In this section I have argued that the assumption is false and have suggestedan alternativeview of competence.If I am rightthen Bi and Cl are false. It is worthhighlightingthe consequence of this for Dummett's argumentagainst Realism. If the propositional asverificationist sumption is false then a person committed to Realistic Truth sentences should hold only to theview thatthespeakerunderstands that have realisttruth conditions. So it is quite beside the point to argue against him that speakersdo not know realist truthcondiarof Dummett'sverificationist tions. Yet thatis the centralthrust gument against Davidson (WTM, WTM jl.29 Davidson is open to the argument,of course, because he accepts the propositional assumption). Establishing merelythat speakersdo not know realist truthconditionscasts no doubt on Realist Truth. Even less does it cast doubt on Realism. 111. 1. Dummett'sArgument for Verificationism Consider now the alternativeversionsof B and C. B2 is trivial. Interest settleson C2: speakerhaveonly understood bythecompetent C2. The sentences verificationist truth conditions. To establish this Dummettneeds to argue thata speakercould not understanda sentencethatas a matterof facthad realist(evidencetruth conditions. transcendent) In theiroriginal formDummett's verificationist argumentsare The presenttaskis permeatedby talk of the speaker'sknowledge.30 compethat treats an argument such talk and abstract to ignore all tenceas simplya practicalability.The task is hard, forthe precise at abstraction argumentis elusive. What followsis mybestattempt (fromFPL 467/8,WTM 115-123; WTM II 70-111; El 4-6, 373-380; TOE xxxii-xl, 16-18,23/4, 132/3,153-155).3"
29 See also Wright,"Truth-Conditionsand Criteria," and "Strawson on AntiRealism." 30The "Comments" (C) mentionedin fn 19, are an exception. However, the argumentthereis (i) very brief, (ii) unclear (to me, at least),and (iii) aimed specifically at Putnam's theory of truth, which does not make use of an explanatorynotion of reference and hence is not what I have herecalled "Realist Truth" (see my "Critical Notice," op. cit., fn 28). 31 See also Dag Prawitz,"Meaning and Proofs:On the ConflictbetweenClassical and IntuitionisticLogic," Theoria, XLIII, 1 (1977): 2-40, pp. 4-7; Wright, "TruthConditionsand Criteria,"pp. 224-228,and "Strawson'sAnti-Realism,"pp. 284-286.
III. COMPETENCE
AS A PRACTICAL ABILITY
DUMMETT'S
ANTI-REALISM
91
of a sentenceS (1) The competentspeaker's understanding is a practicalability; a particular (2) This practicalabilityis an abilityto manifest sortof behavior; the speak(3) The only sortof behavior thatcould manifest of S is thatbehaviorwhichbringshim er's understanding into the position in which, if the condition obtains that conclusivelyjustifies the assertionof S, he recognizesit as so doing;32 of S is his ability to mani(4) The speaker's understanding festbehaviorthatbringshim into theposition in which, if the condition obtains which conclusivelyjustifiesthe assertionof S, he recognizesit as so doing; (5) The recognizableconditionsof S's conclusivelyjustified truth conditions. assertionare its verificationist and abbreviateby using the phrase Let us put (4) and (5) together 'associates recognitionally'. (6) The speaker's understandingof S associates S recogtruth conditions. nitionallywith verificationist is verificationist This conclusion establishesthatall understanding but not thatall truthis. The 'only' in C2 requiresthatS not have ones: that it not any truthconditions other than verificationist the recognizableconditions of have truthconditions transcending condiconclusivelyjustifiedassertion;thatit not have realisttruth tions. Dummett has to rule out the possibilitythat S has realist truthconditions retruthconditions as well as the verificationist quired forunderstanding. conditionsotherthanthoseassociatedwith (7) S has no truth by the speaker'sunderstanding; it recognitionally conditions. truth (8) S has only verificationist S stands in here for any sentence understood by the competent speaker,and so C2 followsimmediately.33 two tie unThe threekeypremisesare (2), (3), and (7). The first S truth conditions.The thirdprevents derstanding to verificationist conditionsthatare not so tied. fromhaving any truth
32 forwhich he thinksour knowledge Dummettmakes thisclaim only of sentences beof truth conditions is implicit. Wheretheknowledgeis explicit themanifesting beof thoseconditions. He thinksthatthe basic manifesting havior is thestatement conditions.See also fn23. wherewe could not statetruth havior mustbe forsentences 33 Dummettsometimesleans towardpreferring a semanticsbased on falsification would require obto one based on verification (WTM II 127-137). This preference between vious adjustmentsto my version of Dummett'sargument.The difference falsificationism and verificationism can be overlookedforour purposes.
92
111.2 The Rejection of Dummett's Verificationism34 Dummett'sverificationist argumentraises too many issues to be discussed in detail here. My strategy is firstto make objections to the argumentfromthe perspective of severalcontemporary theories; second, to rejectthe considerationsDummettoffers in support of some of his premises. A thorough rebuttal of the argument would include a worked-out theory supportingRealist Truth. Such a theory is alluded to in IV. 1. Step 4 stands opposed to holism in psychologyand epistemology. The stand on psychologycomes mainly frompremise2. This requires,forunderstanding at least,a kind of behaviorism:to have an ability is to manifest a particularsortof behaviorin the appropriatecircumstances. In myview,recentworkin thephilosophyof mind decisively favorsfunctionalism overthiskind of behaviorism. The difference can be put like this. Behaviorismsees each mental stateas a simple input-outputfunction:to be in a mental stateis simply to be apt to yield certain behavioras output given certain stimulias input. Accordingto functionalism a mentalstateis not a simple input-outputfunction:it is relatedto input and output by causal relations to other mental states,usually complicated relations. Competencewitha wordor sentence can no morebe tiedto a particularmanifestation than can pain, love, belief,or bravery." Dummettdoes argue against a holistic view of understanding, particularly in discussing Davidson and Quine (FPL 592-601; WTM 115-138; TOE 134-140, 301-309). However, so far as I can see, theargumentdepends on accepting the propositional assumption about competence. Furthermore, from functionalismdiffers the holism of Davidson or Quine. Dummett'scommitment to anti-holistepistemologycomes with premise3. It requires thattherebe a particularrecognizablecondition in which the beliefexpressedby S is conclusivelyjustified.In my view the best recentwork in epistemologyand the philosophy of scienceshows this to be an impossiblerequirement. The relation between worldly conditions and a justified belief is much more complicated than is presupposed by the requirement.Many differ34For some objections to Dummett related to some of mine see Alan Millar, Cur"Truth and Understanding,"Mind, LXXXVI, 343 (July1977): 405-416; Gregory "Knowledge of Meaning," Nouts(forthcoming). rie and PeterEggenberger, 3'Dummett concludes a discussion of the realist view that a person who died withouteverbeing put in dangereitherwas braveor was not with thefollowingastounding statement:"it is evident that only a philosophically quite naive person would adopt [the] realistview" (TOE 150). In my view, he was braveif he realized the appropriatefunctionalstateand not braveif he did not. His quiet life is beside the point.
DUMMETr'S
ANTI-REALISM
93
ent worldlyconditions can produce the same sensorystimulation. past experiencesand presentbeliefs,the same stimGiven different otherbeexperiences.Given different ulation can produce different "observational" beliefs,the same experiencecan lead to different otherbeliefs,the same "observational" belief liefs. Given different can lead to different "theoretical"beliefs.No beliefis conclusively justified. Each belief is tied loosely to a range of conditions in which, relative to other beliefs,it is justifiedin varyingdegrees, positivist epistemology. Premise3 requiresan unreconstructed fromthe epistemic There are two signs of Dummett'sretreating extremismof (3). First,though Dummettusually talks of conclu(e.g., FPL 148, sive justificationin presentinghis verificationism 467, 514, 586; WTM 123; WTM II 111, 132; EI 375), he now seems to thinkthis is a mistake(TOE xxxviii). Second, he is sympathetic to the epistemic holism of Quine's "Two Dogmas" (FPL 591; WTM 11 1; TOE 297/8). Despite thishe does still want to hold, it seems, that some beliefs-the peripheralbeliefsI have called "obcondijustifiedin certainspecifiable -are conclusively servational" of observationI tions. This is at odds with the theory-ladenness have described:the retreat has not gone farenough. is that it However, the most importantpoint about this retreat threatens thecollapse of Dummett'sposition. How can (3), (4), and manyrec(5) be revised?The problemis thatthereare indefinitely ognizable conditions that could give a beliefsome degreeof support. And the one condition could give some degreeof support to many beliefs.Are we to modify(3) so thatthe speaker indefinitely is to be able to recognizeone of these conditions? Or a few? Or many? Or all? Whateverthe answer, thereseems to be no way to modify(5). We can pick out no condition fromS's set of possible conditruth confirming conditionsand make it the verificationist tions of S in particular. particuThe retreatleads straightto a holistic verificationism, larly when it is accompanied, as it should be, by a move to a funcis not, of tionalist theoryof the mind. Holistic verificationism course, any comfort to the believer in Realist Truth (see IV.1 below). However, it is certainlynot what Dummett wants. His "molecular" verificationism psychologyand requiresa behaviorist a positivistepistemology. The objections so farare freeof any semantic presuppositions. Not so the one that follows. The theoryof understandingbriefly of (3) describedin 11.3is in directconflictwith the verificationism and (4). The basic disagreement is thatbetweencausal and description theories of reference. According to a description theory,a
94
competent speaker of a proper name or natural-kindtermassothereferent. to identify ciates with the terma descriptionsufficient elements,a By allowing the descriptionto include demonstrative recognitionalcapacdescriptiontheory can covera straightforward to otherpeople's references ity.By allowing thedescriptionto refer cathe theorycan cover a much more attenuatedidentificational pacity. All of this fitsnicely with (3). Causal theoriesof reference were born out of the rejection of descriptiontheories.A speaker can use a termto referthough almost entirelyignorantabout its referent. He may not be able to describeit, recognizeit, or know view accordhow to trackit down. Causal theorieshave a different ing to which reference is fixed by an appropriate causal link to reality.Judgingwhich part of realityis so linked is a job for the not theessenceof what everyspeakercan do. The opinions experts, and of expertsdepend on theirtheories, and so opinions maydiffer remainsconstant. change over timeeven though reference in detail. In my This is not the place to air this disagreement is leadopinion-some might say a biased one-the causal theory have not producedany ing heavilyon points. Descriptiontheorists (save to thered herring, effective response to thedetailed criticisms "rigid designation").36Criticismsof causal theoriesby Dummett (FPL 135-151;TOE 140-144,420-430) and othersmayappear more effective than theyreallyare because of the undevelopedstateof the a development theoriesat which theyare aimed. I have attempted (in Designation). It is worthemphasizing that this disagreement over understandor not "we would ordinarilysay" that ing is not settledby whether a person had "fully grasped the meaning" of a termin this and thatcircumstance. Even if Dummettwererightin his claims about such matters(which I thinkhe mostlyisn't) thatwould only show is claimWhat thecausal theorist somethingabout our folktheory. is all thatis ing is that his more austereconcept of understanding so differs, needed to explain the behaviorof speakers.If folk theory much the worseforfolk theory (ibid., 8, 87-90, 99/100,198). If the causal theoryof understanding is right then the verificationisttheory is wrong. How is understanding accordmanifested ing to the causal theory? In a multitudeof ways, many of them having nothingto do with verification. My objections so farhave been to (2) and (3). The causal theory
36 Rigid designationis a red herring against because the best Kripkeanarguments descriptiontheoriesmake no appeal to the modal intuitionson which thatnotion depends (Designation,pp. 13-23).
DUMMETT'S
ANTI-REALISM
95
is also opposed to (7). It is a consequence of thattheory thatwords have referents, and hence sentenceshave truthconditions,which are not associated with them recognitionally by the speaker's understanding. I have not yetconsideredDummett'sargumentsfor (3) and (7). The slogan for theseargumentsis an appealing one which DummetttakesfromWittgenstein: "meaning is use." It is importantto see thata good deal of this slogan's appeal comes fromits suggesting an indubitablefact:it is what people do withwordsthatmakes them mean what theydo; in particular,takinga centralaspect of meaning, it is what people do thatmakes it the case thata certain object is the referent of a word. Similarly,it is what a person does that makes it the case thata certainobject is his child. It no more follows fromthe first fact thata person must be able to recognize the object in question as the referent (let alone know that it is the referent) than it follows fromthe second factthata person must be able to recognizethe object in question as his child (know thatit is his child). Our actions can relate us to objects and conditions withoutour having thecapacityto recognizetheobjectsand conditions as so relatedto us (know that theyare so related). The causal theoristcan embrace the indubitable fact suggested by the slogan with as much enthusiasm as the verificationist. It alone gives no support to (3) or (7). Wittgenstein's slogan has to be construedin a Wittgensteinian way to give therequiredsupport: 'use' mustbe takento mean "recognizable conditions of conclusivelyjustified use." That is how Dummett does construeit. The step fromthe indubitable fact to this construalis a giant step. What justifiesit? It is clear that Dummett thinks that the slogan, and hence (3) and (7), are justifiedby the nature of communication and by the way language is learnedand taught. A view of communication and a view of understanding are so closely relatedas to stand or fall together. For communicatingsuccessfullywith S normallyrequires speakerand audience to understandS in thesame way. Dummett'sview,in myterminology, is that theyassociate S recognitionally with the same verificationist truth conditions. All the objections to his view of understanding apply equally to this view. If communicationis to be possible it must require speakerand audience to associate S withsomething otherthan its recognizableconditionsof conclusivelyjustifiedassertion. Dummett's briefremarkson language learning do not yield a view thatis both plausible and supportiveof (3) and (7). Oftenthe
96
following argument seems to be suggested (FPL 467/8; EI 4-6, 375-380; TOE 16-18, 188-190).37 A person learns to understandS by being taughtto associate it with therecognizablecondition that in factconclusivelyjustifiesits assertion.He can learn to associate no other semantic propertywith S. This supports (3). So S can have no othersemanticproperty. This supports(7). This argumentis too crude to be Dummett's actual argument, but it is quite unclear how Dummettwould want it modifiedand how such modifications could remedy its failings. First,theview of language learningis obviouslyat odds with the following relativelyuncontroversialfacts. (i) Most sentences we understandwe neverhear uttered.(ii) We mostlylearn a language not frombeing taught it but fromobservingits use. (iii) Hardly are of sentencesin the presenceof condiany of theseobservations tions that are close to being candidates forconclusivelyjustifying are of false or unjustithe sentences.(iv) Many of the observations fiedassertions which,nevertheless, manifest linguisticcompetence.38 Next,all our earlierobjections count against Dummett'sview of language learning.Indeed theperspective of thoseobjectionsseems much more likelyto providean explanation of the uncontroversial So refactsof language learning than Dummett'sverificationism. flectionon language learning seems likely to provide a good argumentagainst verificationism. In particular,it suggeststhatverificationism is too passive. If we are to learn anything in the linguistic situations that we actually experience we must go way "beyond theevidence." I suggestwe learn a language in thecourse of veryactive theorizing about the world in generaland about people in particular.And thatwhat we primarilylearn with sounds is to associate themwith the thoughtsthatare conventionallyrelated to them.But that'sanotherstory(see Designation 75-86). Finally, the support the crude argumentis alleged to give to (7) is spurious. Even if it were the case that speakers recognitionally associate with S only the condition of its conclusivelyjustifiedassertion,it would not follow thatS had no semanticproperty other than association with that condition: the act of association can conferon S properties we don't recognize.See above on parenthood. In this section I have attemptedto expose the elementsof Dummett's verificationism and indicate their opposition to well-sup"See also Prawitz,op. cit.,pp. 3-6, 10. of "correctuse" encourages a confusion here. An assertioncan be correct in the sense of manifesting competence,but incorrect in thesense thatit expressesa mistakenview of the world. Mistakenviews are one thing,linguisticincompetence another.
38 Talk
DUMMETT'S
ANTI-REALISM
97
and sein psychology, epistemology, theories portedcontemporary is right,Dummett'sviewis in severe mantics.Ifany of thosetheories thereis littlein DumI thinktheyare all right.Further, difficulty. mettto cast doubt on them.C2 is false. Realist Truthis unscathed. and Realism IV.1 Verificationism Dummett'sargumentagainst Realism restson threepremises.I as thetrivhave argued thateach of these(except B when construed ial B2) is false. If I am right,Dummett'sargumentis bad frombeginning to end. In this section I shall consider the general question of the relaThis is importantin tion between Realism and verificationism. attemptsto undermine order to guard against otherverificationist Realism (and against theinevitablecharge thatDummett'sreal arto him). gumentis otherthan the argumentI have attributed What sort of connection might therebe between Realism and My argument shows that the issues of Realism, verificationism? truth,and understandingare distinct.So therewill be no entailmentrelationsbetweena position in one area and a position in the other. What we can expect are inferencesto the best explanation the case forRealholding betweenpositions. Withsuch inferences ism looks good. startsfrom I think that the most promising of such inferences and about theproperties with some observations Realism, together relationsof theobjects the Realist believesin, and argues forRealThe doctrineof understanding. ist Truth and a nonverificationist part of such an argument.The sketchedin II.3 is, in effect, theory major problem for this inferenceto the best explanation is in showing that the Realist needs an explanatorynotion of truthat all.39 However I thinkwe can hope to show this need in order to Showing thiswould complete the explain learningand teaching.40 case against verificationism. supposing that it Now the strikingthing about this inference, fromRealism. Those impressed can be made good, is thatit starts with Dummettwill object, claiming thatwe should startfromthe of understanding and see what we can inferfromthatabout theory Realism. For example, suppose we can establish a verificationist
39See 1.1 above and Leeds, op. cit. For some briefsuggestionsalong these lines see HartryField, "Mental RepreXIII, 1 (July 1978): 9-61, pp. 47/8,and my Designation,pp. sentation,"Erkenntnis, 68/9. In my "Critical Notice" of Putnam's Meaning and the Moral Sciences,p. 403, I agreed with Putnam thatwe needed truthto explain the success of beliefs.I now thinkthisa mistake:see Realism and Truth.
40
IV. VERIFICATIONISM
AND REALISM
98
thenit may be claimed thatwe can infer theory of understanding, fromthis an epistemicdoctrineof truth.Then, given the earliermentionedimpossibilityof findinga plausible Realist epistemology to combine with thatdoctrineof truth(I.1), we can inferantiRealism. I have threecommentson this. of understanding is the wrongplace to start.The (a) The theory of theory task is to put together the most plausible comprehensive the phenomena thatconfront us. Theories of language and undertheoriesthatmustbe standingare only two among many scientific fittedinto the comprehensivepicture. Realism is an overarching theory or principle. It is initiallyplausible. It empirical (scientific) can be supportedby argumentsthatmake no appeal to theoriesof e.g., it is the only plausible explanalanguage or understanding: tion of the way thingsseem; it accords with our best science; criticisms of it fail; it is supported by naturalizedepistemology;rival overarchingprinciples such as those of idealism all fail. What in firmer place could therebe to stand than Realism, as we theorize such undeveloped areas as those of language and understanding? thepoor stateof theoriesin thoseareas, whether verifiIn contrast, cationistor not, makes thema bad place fromwhich to starttheorizing, particularlyin determiningoverarchingprinciples about the nature of reality.To thinkotherwiseis to put the cart before the horse. (b) Suppose, however,thata good argumentcould be produced for a verificationist theory, making us feel inclined to waive our as a starting place. objection in principle to using understanding we could inferanti-Realism thatfromthat theory Suppose further in the way suggested.What should not be overlookedis that,however good that inferencewas, it alone would not underminethe earlier "promising" inference fromRealism to Realist Truth. Our choice betweenthesetwo inferences should be guided by our view of which starting plausibility.I suggest assumptionhad thegreater that the argumentsfor Realism are verystrongand that it would fora verificatake an argumentfar strongerthan any yetoffered of understanding tionisttheory to make it the moreplausible starting assumption. to the bestexplanation froma (c) Finally, I doubt thatinference of understanding verificationist would yieldanti-Realism.If theory not, thenit does not matterto the Realist how stronga case could be made forverificationism. Consider the austere physicalism described by Leeds (op. cit.), drawing on Quine. This position is unequivocally committedto So faras I Realism, but skepticalabout semanticnotions like truth.
DUMMETT'S
ANTI-REALISM
99
can see it could accommodate a verificationist theoryof understanding (of a Quinean sort,forexample). Given the independent plausibilityof Realism, it seems likelythatthebestexplanation of the world fora verificationist should be not that of anti-Realism but thatof Leeds-Quine. In thesecommentsI have taken theoriesof language and understanding to be ordinaryscientifictheories.I give them no special role in settlingour comprehensive world view. I thinkwe should resistDummettianattemptsto use themas the basis fora "bornagain first philosophy." IV.2 Conclusion In its more prominent version, Dummett's argument against Realism restson the theorythat linguisticcompetenceconsistsin knowledge of truthconditions. In general it restson a linguistic theory of metaphysics, a behavioristtheory of themind,a positivist epistemology, and a descriptiontheory of reference. In myview all these theoriesare false. In part I have argued this here; in part I have relied on what has been done elsewhere. Moving away fromDummett'sactual argumentto consider the general bearing of verificationism on Realism, I have found little for the Realist to worry about. Realism is too stronga doctrineto be overthrown bycurrent speculationsabout understanding. Whateverone makes of thosespeculations,thebest theory of theworldis Realist. I thinkthe best theory should include Realist Truth and a of understanding theory of thesortsketchedhere. However,even if I am wrongabout this,the Realist should not despair: thereis still the Leeds-Quine position. Betterto throwall the semanticplanks overboardthan abandon the Realist boat. University of Sydney
MICHAEL DEVITT

Devitt Dummett's Antirealism PDF

Enviado por

Dados do documento

Descrição original:

Direitos autorais

Formatos disponíveis

Compartilhar este documento

Compartilhar ou incorporar documento

Opções de compartilhamento

Você considera este documento útil?

Este conteúdo é inapropriado?

Direitos autorais:

Formatos disponíveis

Devitt Dummett's Antirealism PDF

Enviado por

Direitos autorais:

Formatos disponíveis

Journal of Philosophy, Inc.

THE JOURNAL OF PHILOSOPHY

1983 The Journalof Philosophy,Inc.

THE JOURNAL OF PHILOSOPHY

THE JOURNAL OF PHILOSOPHY

THE JOURNAL OF PHILOSOPHY

THE JOURNAL OF PHILOSOPHY

THE JOURNAL OF PHILOSOPHY

II. THE PROPOSITIONAL

ASSIJMPTION ABOUT COMPETENCE

THE JOURNAL OF PHILOSOPHY

THE JOURNAL OF PHILOSOPHY

THE JOURNAL OF PHILOSOPHY

THE JOURNAL OF PHILOSOPHY

THE JOURNAL OF PHILOSOPHY

THE JOURNAL OF PHILOSOPHY

THE JOURNAL OF PHILOSOPHY

THE JOURNAL OF PHILOSOPHY

Você também pode gostar

Devitt Dummett&#39;s Antirealism PDF

Enviado por

Dados do documento

Descrição original:

Direitos autorais

Formatos disponíveis

Compartilhar este documento

Compartilhar ou incorporar documento

Opções de compartilhamento

Você considera este documento útil?

Este conteúdo é inapropriado?

Direitos autorais:

Formatos disponíveis

Devitt Dummett&#39;s Antirealism PDF

Enviado por

Direitos autorais:

Formatos disponíveis

Journal of Philosophy, Inc.

THE JOURNAL OF PHILOSOPHY

1983 The Journalof Philosophy,Inc.

THE JOURNAL OF PHILOSOPHY

THE JOURNAL OF PHILOSOPHY

THE JOURNAL OF PHILOSOPHY

THE JOURNAL OF PHILOSOPHY

THE JOURNAL OF PHILOSOPHY

II. THE PROPOSITIONAL

ASSIJMPTION ABOUT COMPETENCE

THE JOURNAL OF PHILOSOPHY

THE JOURNAL OF PHILOSOPHY

THE JOURNAL OF PHILOSOPHY

THE JOURNAL OF PHILOSOPHY

THE JOURNAL OF PHILOSOPHY

THE JOURNAL OF PHILOSOPHY

THE JOURNAL OF PHILOSOPHY

THE JOURNAL OF PHILOSOPHY

Você também pode gostar

Devitt Dummett's Antirealism PDF

Devitt Dummett's Antirealism PDF