Brandom

Embed Size (px)

Citation preview

  • 8/13/2019 Brandom

    1/23

    Philosophy and Phenomenological ResearchVol. LXIII,No. 3, November 2001

    M o d a l i t y Normativity n dIntentionalityROBERTBRANDOMUniversity of Pittsburgh

    A strikingfeatureof the contemporary hilosophical scene is the flourishingof a numberof researchprogramsaimedin one way or anotherat (as Dretskeputs it) making intentional soup out of nonintentional bones-more care-fully, specifying in a resolutely nonintentional, nonsemantic vocabulary,sufficient conditions for states of an organismor other system to qualify ascontentfulrepresentations.This is a movement with a numberof players, butfor my purposeshere, the work of Dretske, Fodor, and Millikan can serve asparadigms.The enterprise n which they arejointly engaged s not so muchone of conceptualanalysis as it has been traditionallyunderstoodas one ofconceptual engineering. That is, instead of thinking about what ordinarypeople or even sophisticatedphilosophers alreadymean by terms such as'representation',hey appealto the tools of the special sciences (for instance,informationtheory andevolutionary biology) to describeabstractly,but incriticizabledetail, how one might craft a situation in which some statearguablydeservesto be characterized s 'representationallyontentful' in vari-ous importantsenses. Insofar as the theories are good ones, they may shedlight on how human knowersactuallywork. But their immediate aim is abroaderone: to say what would count as doing the trick, rather han how wemanageto do it.It is not hard o get into a position from which intentionality1can seemphilosophically puzzling.One need not, for instance,be an atoms-in-the-voidphysicalist in order o want to understand etter the natureof the transitionfrom a worlddevoid of semanticproperties o one in which they are at leastlocally rife. One way of dispelling the sense of mystery that threatens tosurround uch a transition(indeed,where such a story is possible, the very

    Editor's note: This special symposiumderives from the 1999 BrownColloquium.In this essay, I'll speakvery roughly,and use the notions of intentionality,representation,concept-use, and semantic content, interchangeably-not because I believe that thereare no importantdistinctionsto be registered by the use of these different locutions,butbecause those differences don't matter at the level of abstraction at which I am tellingmy story.

    SYMPOSIUM587

    http://www.jstor.org/page/info/about/policies/terms.jsp
  • 8/13/2019 Brandom

    2/23

    best way to do that) is to explain in terms applicable n principle already ophenomenabeforethe transition conditionsunderwhich the new vocabularywould be applicable.This is an exciting and invigorating sort of project-not least for theprospects it holds out of doing careful, detailed, ground-levelwork thatcontributesdirectlyto the clarificationof fundamental,arge scale, longstand-ing philosophical problems. The goal of this collective undertakingistypically taken to be the productionof a naturalistic semantics: showinghow intentionalitycan be made intelligible by deploying the concepts andmethods characteristicof the natural sciences, exhibiting semantics as aspecial natural science.2 Certainly that is the way Dretske, Fodor, andMillikan thinkof it, as is evident not only from the motivations they avow,butequallyfromthe rawmaterials hey permitthemselves to employ andthecriteriaof adequacy o which they submit the resulting accounts. The projectof naturalizinghe mind and cognition has a long philosophical history. Butit is possible to see these projectsas episodes in a still broadermovement ofthought, in which naturalism s only one strand-indeed, an optional one.What I want to do in this essay is to sketch that broaderphilosophicaltradition,andto locate some of these contemporary emanticprojectswithinit.

    IIt is helpful to begin with to focus on two phases in the developmentofcontemporary hilosophical thought about intentionality:an early one and alater one. (ThoughI don't mean to pretend hat nothing else of significancefor this traditionwas going on. I'll return o some other strands n the debatelater on.) One featureof intentionality-the of-ness, about-ness, or moregenerally the representational ontentfulness characteristicof thought andtalk-that first caught semantic theorists' attention is the intensionality (orspecification relativity) of the locutions used to ascribe intentional states.Already n UberSinnund Bedeutung, Frege hadpointedout the failures ofsubstitutivity within propositional-attitudescribingcontexts that disqualifythem as extensional.Whereas n extensionalcontexts, substitutingone coref-erential term for another (or one co-truth-valuedsentence for another)preservesthe truth of the whole sentencein which they are embedded, uchsubstitutions can alter the truthvalues of sentences of the form S believes(or thinks) that (t). Again, one can believe that (t), for instance that the2 HartryField presents an interestingaccount of one way into this project, early in hisclassic essay Tarski'sTheoryof Truth Journalof Philosophy69:347-75). The differ-ence between the sorts of conceptualtools thatDretske, Fodor,and Millikan,each in hisor her own way, proposeto deploy to address this common problem,on the one hand, andthose Field appealedto, on the other, I think mark an importantdimensionof change inthe analytictraditionover the twentyfive years spanningthese projects.

    588 ROBERTBRANDOM

    http://www.jstor.org/page/info/about/policies/terms.jsp
  • 8/13/2019 Brandom

    3/23

    presentking of Franceis bald, even if t (that king) does not exist. But onecannot standin a nonintentionalrelation-say, kicking-to what does notexist. And one can want to buy a sloop, even though thereis no particularsloop one wants to buy-even though one wants only, as Quine puts it,relief from slooplessness. But one can only sail or sleep on some particularsloop.3

    By the 1950's andearly '60s, a broadlyTarskian,model-theoreticapproachto the semantics of extensional locutions (descendedrom Frege's work) waswidely agreed o providethe best availableparadigmof a successful formalsemantic theory. Frege's own approachto the semantics of intensionalcontexts, such as the propositionalattitudeascriptions by which intentionalstatesareattributed,was by contrastboth not adequately ormalizedand muchless widely acceptedphilosophically.The result was a distinctivephilosophi-cal problematic,centeredon the intensionality (non-extensionality)of inten-tionality.

    One influentialresponse, of course, was Quine's philosophical rejectionof the ultimateintelligibilityof the notion of intentionalstate, motivatedbythe recalcitranceo extensional semanticanalysis of the locutions by whichthey are attributed.Others looked for conceptualraw materials that wouldsupporta more positive response. By far the most popularcandidate eizedupon was anothersort of intensional idiom that was being philosophicallydomesticated t around he same time (from the mid '60s throughthe '70s):alethic modalities. Thus was born the fruitful and still ongoing project ofunderstanding he intentionality characteristicof mind and language bydeployinga naturalisticbutmodallyrichvocabularycenteringon the sorts ofcounterfactualsused to codify causal relationships.This explanatoryprojectforms one of the central strandsof the storyI want to tell here.The second episode in the development of contemporaryphilosophicalthoughtaboutintentionality hat I want to highlight is the heightenedappre-ciationof the normativecharacter f meaningandconceptuse aroused n partby Kripke'sdiscussion of Wittgenstein,beginningin the early '80s. The coreidea is that anything recognizable as an intentional state (for presentpurposes, we can think of these as propositional contentful states or asconceptually contentful representations)must underwritenormative assess-ments as to whether things are as they ought to be, accordingto thatstate-whether the state is correct or successful according o the standardsdeterminedby its content.Believing includescommittingoneself, undertakinga responsibilityconcerninghow things are (how they might be found to be).

    3 Significantly,from the pointof view of the storyto be told here, manyof these intensionalfeatures of intentional dioms are sharedby normativephenomena.Thus one can promiseto give someone a blue ox, even thoughthere is no blue ox, and owe someone an ox,thoughthere is no ox in particularone owes (one is obliged only to relieve his oxless-ness).

    SYMPOSIUM 589

    http://www.jstor.org/page/info/about/policies/terms.jsp
  • 8/13/2019 Brandom

    4/23

    Intending also includes committing oneself, undertakinga responsibilityconcerninghow things will be (how they must be madeto be). Beliefs areessentially, andnotjust accidentally, hings appropriately ssessedas to theircorrectness in the sense of their truth. And desires and intentions areessentially,and notjust accidentally, hingsappropriatelyassessed as to theirsuccess, in the sense of their fulfillment. (Even less overtly committiveintentionalstates such as conjectures,hopes, and wishes arecontentfulonlyinsofaras they settle how things ought, accordingto them, to be.) Similarly,speech acts such as claiming and commanding essentially involve adoptingnormativestatuses, ncludingas they do the undertaking f responsibility (forhow things are)andthe assertion of authority(over how things are to be).Using a term with a determinatemeaning (using it so as to express aparticularconcept) is binding oneself to a norm that determines thecorrectnessor incorrectnessof that use (along with that of many otherpossible uses). It is this normative dimension of intentional content thatmakes it possible to distinguishtwo complementary directionsof fit thatbeliefs and intentions, or declarativesand imperativescan have, depending(intuitively) on where the fault is taken to lie if things are not as they aresupposedto be according o the norm articulated y their contents.For what an organism is doing to be intelligible as representing,theremust be room also for misrepresenting, or representationhat is incorrect.One of the hallmarksof the normativityof intentionality is that what onecommits oneself to in applying a concept outruns in principle what onetakes oneself to be committedto. The norm of correctnessone therebybindsoneself by goes beyond both the dispositions of those undertaking hosecommitmentsand whattheyconsciously envisage themselves as committingthemselvesto thereby.Because it does, a questionarises about how to under-standthe featuresof the intentionalstate or meaningfulutterance hat settlewhich determinate onceptualnormone is boundby-exactly which standardsfor the assessment of correctness or success one has implicitly put inplay-by being in that state or producing that utterance.If what one iscommittedto is not settled by what one consciously envisages (becauseonenever so envisages enough), nor by what one is disposedto accept as such(becauseone can be wrong),how is it settled?These observationscollectively give rise to anotherdistinctivephilosoph-ical problematic: fferingan accountof the distinctive normativecharacter fintentionality. Combining this new challengewith the expressive resourcesfirst called into play in responseto the recognitionof the intensionality ofrepresentational ontent-namely the vocabulary of alethic modalities, ofpossibility and necessity, of subjunctive or counterfactualconditionals-yields a satisfyingly detailedresearchprogramwith reasonablyclear criteria of adequacy. It is the project of explaining conceptualnormativity in terms of counterfactualdependencies,paradigmaticallyof590 ROBERT BRANDOM

    http://www.jstor.org/page/info/about/policies/terms.jsp
  • 8/13/2019 Brandom

    5/23

    representingevents and states of affairs on representedones. This is theprojectDretskeand Fodorhave, each in his own way, pursued n a numberofcareful and ustly influentialworksproducedover the pasttwenty years.4The prima facie difficulty that its normativity raises for attempts tounderstand intentionality directly in modal terms, by appealing tocounterfactually obust correlations,appears n a number of different, butrelatedguises. Crudelyput,one cannot take what is represented y a state orperformance o consist simply in whateverstimuli the system in question isdisposedto respondto by enteringthat state or producing hat performance.For that would leave no room for mistakenresponses, for misrepresentation.Whatever he userof a concepttakes to be correctwould count as correct.5Amore careful formulation would acknowledgethat talk of dispositions andcounterfactual ependenciesmakes senseonly as applied o types of states andevents, not for tokens. The shift to types of occurrence,understoodasindicatingothertypesof occurrence, n terms of the counterfactual eliabilityof the correlationholds out the possibility that individualtoken occurrencescould fail to correlate,andso to count as mistaken,as misrepresentations.Buteven so, the possibility of gerrymanderinghe types involved presents aproblem cognate to the original one. It is always possible in principle todraw boundariesaroundthe stimulus types (perhaps seriously disjunctiveones) so that the correlationwith the response type is perfectly reliable.Rather hanmisidentifyinga particular chidna as a porcupine,one counts ascorrectly identifying it as a porcupine-or-echidna. eaving room for thepossibility of error, of incorrect representationthen requires somehowprivileging some among all the possible types that each particular andidaterepresentation-typecorrelates with. And that seems to require furtherconceptualresources.6Another guise in which appears what is, I think,recognizablya difficultyof the same general kind, concerns the conceptualresources requiredfor picking out a distal stimulus kind as what isrepresentedby a specified response kind, holding fixed all of thecounterfactual ependencies.Forgiven any reliably covaryingcausal chain of4 It is one descriptionof thatproject.I don't meanthatthey have typically thoughtof what

    they are doing in just these terms. I'm offering a de re specification of the content oftheirproject,not a de dicto one.5 Indeed,I conjecturethat one of the sources of Kripke's own interest in the topic of thenormativityof intentionality(in the late 60's, when he wrote the prototype that waspublished only much later as Wittgensteinon Rules and Private Language [Cambridge:Harvard University Press, 1982]) was his realization that the modal apparatus thatotherwise seemed so promisingfor dealing with the intensionalityof intentionalityraisedexactly these problems.

    6 This is the point Dretske raises under the heading of misrepresentation in his well-known article of the same name,pp. 17-36 in Radu Bogdan (ed.) Belief: Form, Content,and Function,OxfordUniversityPress, 1986), and thatFodordiscusses underthe headingof the disjunctionproblem PsychosemanticsMITPress, 1987, and the title essay of ATheoryof ContentMITPress, 1990).

    SYMPOSIUM591

    http://www.jstor.org/page/info/about/policies/terms.jsp
  • 8/13/2019 Brandom

    6/23

    events leading from some stimulus to a response, later elements, moreproximal to the eventual response, will always be more reliably correlatedwith the occurrenceof a responseof that kind than will earlierones. Thus inthe chain leading from the ringing of a bell through motions in theintervening air molecules, motions in the tympanic membranesof a dog,electrochemicalchanges in its auditorynerve, and so on ultimately tostimulationof its salivaryglands,each intervening stage can also be broughtabout by some nonstandard auses, and so will less reliably elicit theresponsein questionthan its fellows downstream n the causal chain.I am not claiming that the problem of misrepresentations fatal to theenterprise.A numberof strategieshave been suggestedfor responding o thedifficulties of reconstructingthe normative dimension of concept use bymeans of modalconcepts,and no doubt there will be morein the future.(Onesociological reason for confidence hat the enterprisewill not soon be aban-doned is the mutualconceptualandmotivational supportgiven to each otherby taking subjunctivereliabilityof correlationof featuresof representingsandrepresenteds s the basis of a naturalisticsemantics, on the one hand, andexploiting such reliable correlationsas the basis of a naturalisticepistemol-ogy, on the other.7)The conclusion I am after is just that, while the use ofmodal vocabularyoffers semantic theorists a direct purchaseon the non-extensionality of intentionality, the use of that vocabularydoes not in thesame way serve directly to make its normativity intelligible. Anotherapproach s to do just that: startwith some familiar kind of normativity,andtry to exhibit the normativityof intentionalityas a species of it.Some sorts of normativitydo not provide promising raw materials forsuch an explanatory strategy. Soldier Schweik ought to march Northwestbecause his sergeantcommanded him to do that. The painting ought to behung this side up, because that is how the painter intended it. I ought todriveyou to the airport omorrow,becauseI promised to do it. These are allperfectlygood senses of 'ought', so one might try out the idea of assimilat-ing the sense in which a claim or belief ought to be true to one of them. Butbecausecommanding, ntending,andpromising are themselves all intention-ally contentful states or acts, they already nvolve the kind of normativitycharacteristic f intentionality,andso cannot serve as non-question-beggingunexplainedexplainers of it. (This is not to say that there is somethingwrong with explaining the normativity of speech acts by appeal to inten-tional states-as, say, Gricedoes--only that this is not a strategyfor under-standingthe normativedimensionof intentionalitygenerally.)A muchmore promising, andthereforemore popular, strategyis to lookto teleology, andunderstandhe norms characteristic f intentional content-

    I expresssome doubtsabout the solidityof this support n InsightsandBlindspotsof Reli-abilism, [Monist81, Number3, July 1998, pp. 371-92].

    592 ROBERT BRANDOM

    http://www.jstor.org/page/info/about/policies/terms.jsp
  • 8/13/2019 Brandom

    7/23

    fulness as broadly unctional norms.Some care is requiredwith this strategytoo. Appealingto norms of proper unctioningof artifactsexplainedin termsof the intentions,beliefs, andpurposesof theirdesigners, will be unattractivein this methodologicalcontext because it raises circularityconcerns of thesortjust mentioned: t is hardto see how one could explain in the same waywhat counts as properly carryingout such an intention, what would makesuch a belief correct in the sense of true, how things ought to be accordingto the purposein the sense of what state of affairs would fulfill it. On theother hand, we have learned that we need not think of biologicalfunctions-for instance, the function of the heart to pump blood-in theseterms. Besides explanations of the proper functioning of subsystems ofartifacts n terms of design, there areexplanationsof the properfunctioningof subsystems of organisms in terms of selection. Understanding henormativity of intentionality on the model of norms of proper biologicalfunctioning has the structural virtue that understandingthe modelingfunctional norms does not require ven tacit appealto intentional norms ofthe sort being modeled. Understandinghe content of intentional states bylikening the way a norm of truth is integral to beliefs (a norm of successintegralto intentions, and so on) on the way a norm of blood-pumpingisintegral to hearts, involves no covert circularity. This thought generatesanother researchprogram: teleosemantics, a kind of functionalism aboutintentionality that makes its normativity intelligible as an extension andelaborationof norms of properbiological functioning. Millikan's project isby far the most sophisticatedandbest worked-out xemplarof this kind ofexplanatoryproject.Here, then, are two promising ideas: to understand he intensionality ofintentionality by appealing to counterfactualdependenciesof representingstates of affairs on representedstates of affairs,and to understandhe norma-tivity of intentionalityby appealingto norms of properfunctioningof repre-sentingsrelatingthem to representeds,modeled on norms of properfunction-ing of reproducingbiological systems that evolved by naturalselection. Inthinkingabout the relationsbetweenthese ideas, it is important,I think, tokeep in mind that the selectionalexplanationson which the second ideareliesthemselves dependessentially on the invocation of counterfactualdependen-cies that must be expressed n the sort of modally rich vocabularyon whichthe first idea relies. Maybethis is obvious. But even if it is, it is worthwhileto rehearsea couple of examples in order to make clearjust what role theappealto alethicmodalitieshas in explanationsof functional norms in selec-tional terms.

    SYMPOSIUM 593

    http://www.jstor.org/page/info/about/policies/terms.jsp
  • 8/13/2019 Brandom

    8/23

    Dretske8talks about a species of Northern Hemisphere bacteria thatcontain magnetosomes, which reliably respond differentiallyto magneticfields. The bacteria n turnreliablyresponddifferentially o the orientationoftheir magnetosomes, and move toward geomagnetic north, which in thatenvironment s also down, hence to water less rich in oxygen, which is toxicto thesecreatures.The state of these organisms'magnetosomesin this envi-ronmentreliably counterfactually orrelateswith all of geomagnetic north,deeperwater, and less oxygenatedwater. What should it be understoodasrepresenting?As Dretskepointsout:[T]his primitive sensory mechanism is...functioning perfectly well when under [a] barmagnet's influence, it leads its possessorinto a toxic environment.9Millikan counts it as a significant advantage f her selectional account overDretske's thatit privileges the latterof these in terms of its role in explana-tions of the evolution of the system.10For that responding n the way theydo to the orientationof their magnetosomesNormally leads organisms ofthis kind to less oxygen-rich waterplays an essential role in evolutionaryexplanations of the persistenceof this whole stimulus-responsesystem.'2And this essential role is articulatedby counterfactualsconcerning thedevelopmentof the system:if it had not been the case thatthe mechanism(aswe can now, retrospectively, say) successfully led the ancestors ofcontemporarybacteria o less toxic waters,then they would not now respondas they do to magnetosomes(or perhaps,have them at all). Teleosemantictheoriesof intentionalityarea specialkind of modal theory, distinguishedbythe particularform taken by their appeal to counterfactualdependenciesexpressed n a richmodalvocabulary.That this is so depends not on specific features of representationalsystems, but on the generalnature of selectional explanationsof function(andhence of functionalnorms). Consider(to use an example adapted romSober13) a device that filters marbles dependingon their size. Marbles ofassortedsizes and colors are placed in the bin at the top, and a subset isselected by findingtheirway to the bin at the bottom.If size and color of themarblesarecorrelated, t may happenthat all the marbles selectedarered, as8 In Misrepresentation, p. cit. [6], and again in ExplainingBehavior [MIT Press 1988]

    p. 63.9 In Misrepresentation, . 29.10 Retailed,for instancein both Biosemantics and Compareand ContrastDretske, Fodor,and Millikan on Teleosemantics (at pp. 93ff., and pp. 125 ff. respectively, in WhiteQueenPsychologyand OtherEssaysfor Alice [MITPress 1993]).The capitalletter indicatesMillikan's technicalsense: see LTBOC,e.g. p.96.12 The difference that in this case makes a difference between Millikan and Dretske is that

    her account allows her to look downstream,to the response a system makes to a repre-sentation,andnotjust upstream, o its antecedents.13 Elliot Sober The Natureof Selection,Universityof ChicagoPress, 1993.

    594 ROBERT BRANDOM

    http://www.jstor.org/page/info/about/policies/terms.jsp
  • 8/13/2019 Brandom

    9/23

    well as being small. The claim thatit is size, rather han color that is selectedfor, even though both are selected, must be cashed out in terms ofcounterfactuals:f a marblewere small and of some other color than red, themechanismis such thatit would be selected, while if it were largeand red, itwould not be. In the same way, Millikan can say that what themagnetosomesare or, their unction, is to indicategradientsof oxygenation,rather than of pressure or magnetic field, so that they are functioningproperly only when in fact they lead the organism to less toxic environs,because that is what was selectedfor. In fact all of Millikan's fundamentaltechnical terms from the notion of reproduction '4ith which she begins,through he definitionsof the various sorts of directproperfunctions, 15 ndof Normalexplanations '6 re (andmust be) specified in a modally richvocabulary, n termsof laws and counterfactual onditionals.In pointing out this featureof teleosemantic biofunctionalist approachesto intentionality,I by no means want to suggest that theirconceptual relianceon modal notions is in any way a defect or a drawback. point to it only tosituate them in the context of counterfactualdependency heories of inten-tional content. This kind of functionalism about the normativity of inten-tionalityshouldbe seen as addingsignificantresourcesto modal approaches,rather han as simplya differentkindof theory. It offers a functional accountof the normativity of intentionality, but seeking to understandunctionsthemselves in modal terms, explicating it by means of the sorts of counter-factuals first invoked in discussions of intentionality in response to therecognitionof its intensionality. It is against the backgroundof this broadbrush-stroke,highly selective rational reconstructionof some of the mainstrands n the traditionof recentthoughtabout intentionalitythat I want firstto considerexactly why it is appropriateo appeal to modal vocabularyinthis explanatorycontext,and then to sketch the motivations for a somewhatdifferentapproach.

    IIWe can ask a very general programmaticor methodological question aboutthe explanatory trategyof using modal notionsto account for the non-exten-sionality andnormativityof intentionality. Suppose we bracket he questionof whether these conceptualraw materials are sufficient to account for thephenomena n question-even supplementedand articulatedn the way thebest teleosemantic accounts do. What makes us think that using theseresources s so much as legitimate?The questionbecomes more pointed inhistoricalperspective.For at least the first half of this century,philosophers14 Her condition(3) is explained(LTBOC,p. 20): Roughly,the law in situ implies that hadA been different...B wouldhave differedaccordingly [italics in original].15 Beginningat LTBOCpp. 27ff.16 BeginningatLTBOCpp. 33ff.

    SYMPOSIUM595

    http://www.jstor.org/page/info/about/policies/terms.jsp
  • 8/13/2019 Brandom

    10/23

    (especially English speaking ones) were extremely suspicious of modalidioms. What such talk expressed seemed at best obscure, and at worstperhaps n principleunintelligible.Duringthis periodone of the outstandingphilosophical problems andprojectswas precisely to renderrespectabletheuse of counterfactuals ndappealsto what was possibleor necessary. Concep-tually fastidiousphilosophersin the broadly empiricist traditionfelt them-selves obliged either to demonstrate he legitimacy of these notions or toshow how to do without them. Thus Peirce worriedabout what it meant tosay of a diamondthatspentits entireexistence, from initial crystallization ofinal incineration, waddled n cottonwool, thatit was hardor fragile-that isthat it would have scratched lass or shatteredhad it hada differentcareer.17Such claims seemed to have a statuswholly differentfrom those attributingacertainshape, size, or velocity to the diamond. Modal vocabularydoes notseem to be (andin fact is not) reducibleto ordinary,nonmodal, descriptivevocabulary.We know (we think)what it is for the cat to be on the mat. Butwhat is it for it to be possible for the cat not to be on the mat, butnecessary that she is a mammal or that she is gravitationallyattracted o themat in inverseproportion o the squareof the distancebetween them and indirectproportionto the productof their masses? What sort of properties,relations,or arrangements f objectsconstitutethese facts, or the fact thatif amouse came within rangeshe would leap on it? What are we committingourselves to when we make this sort of modal claim, and what counts assufficient evidence for them?The centralrole played by the notion of habit inpragmatistthought, as a naturalphenomenondiscerniblein the inorganicworldno less than the organic, but that could nonetheless be deployedtoaccount for discursive and intentional phenomena, is in no small part aresponse to this issue. Humean concernsabout the ultimate intelligibility ofnecessaryconnections,and so of the idea of causation itself arepervasiveinthe empiricist tradition,and aregiven addedmpetus by the developmentofextensional formal languagesfor the codificationof empirical claims. Firstorder quantificational languages could express regularities andgeneralizations with hitherto undreamed f precision. But for philosopherssuch as Russell andCarnap, hat ust made all the more urgentto explain orexplain away laws and lawlike generalizations, whose content somehowextendedbeyondwhatcould be capturedwiththeseexpressiveresources.I think it is worthwhileremindingourselves how surprisedphilosopherswho lived and moved and had their being in this milieu would have been todiscoverthatby the end of thecentury,when their successorsfound the inten-sional andnormativecharacter f intentional idioms problematic, their firstimpulse and dominantstrategy would be to appeal to modal notions toexplain them.Just how did that which seemedmost in need of philosophical17 Sober,E. The Natureof Selection.Cambridge:MITPress, 1984.

    596 ROBERTBRANDOM

    http://www.jstor.org/page/info/about/policies/terms.jsp
  • 8/13/2019 Brandom

    11/23

    explanationanddefense become transformed o as to be unproblematicallyavailableto explainotherpuzzlingphenomena?The most cursory familiaritywith intellectualhistoryaccustoms us to the spectacleof explanandabecom-ing explanans.But in this particular ase, what did we come to understandnthe interimthat made the transition egitimate,ratherthan its being merely aproductof changeof fashion, fatigue, or amnesia?Quine has kept the oldquestions alive. He finds the intensionality even of modal vocabularyultimately andin principleunintelligible, and has continuedto argue force-fully for the wisdom of eschewing all such nonextensionaltropes. The worldspirit has not moved his way, but why not? Why are we right not to bebotheredby his scruples,hallowed as they areby tradition?Here are three answers.I think the first two, which areprobablythe mostpopular responses, will not do as they stand. They contain elements of amoreadequateanalysis,butcan be understoodproperlyonly in the context ofthe third,which I take to be the deepestof the three.The first answer is thatwhat happened s a revolution in modal logic and semantics. Here's a quicksketch. C.I. Lewis began a process that yielded formally adequateaxiomaticways of expressing various senses of 'necessary'and 'possible', and thenKripkedevelopeda complete model theoretic semantics for those systems.With the extension of Kripke's semantics from modal logics to modallanguagesmoregenerally,the processwas complete.In the light of the tradi-tion I havejust been talking about, it has seemedparticularly ignificant tomanythatKripkeshowed us how to do semantics for modal languages usinga first orderextensional metalanguage.18Modal operatorsare interpreted yfirst orderquantifiersover possible worlds. Given a structure of possibleworlds and accessibility relations among them, all we need to do is todescribe that structure n the ordinary,nonmodal language of first orderlogic, in order o understandhe use of distinctively modal vocabulary.Sohere was an answer to the question, what sort of facts are modal facts? Andthe answer was extensional,usingonly ordinarydescriptivevocabulary.Thusmodal vocabularybecame demystified and respectable. Now it no longerseemed to present a problem that laws and the dispositions expressed bysubjunctive or counterfactualconditionals loom so large in the naturalsciences. And for this reason modal vocabulary becomes available tophilosophersto do explanatoryworkin its own right-not just as an objectof explanation, but as a tool of explanation. So when certain potentiallypuzzling featuresof semantic vocabularyarenoticed-for instance, that thevocabulary n which we describeintentionality s intensional-it is natural oappealto the newly respectablemodal vocabulary o explain them. This is18 Whathas been called California emantics s second order (in both the version due to

    Kaplanand thatdue to Montague), in quantifyingover functions, predicates, and so on.But in anothersense it is still extensional,even in its treatmentof intensions.

    SYMPOSIUM 597

    http://www.jstor.org/page/info/about/policies/terms.jsp
  • 8/13/2019 Brandom

    12/23

    one of the conceptual inheritancesof our generation of philosophers: theshiny new toolkit of modally rich philosophical metalanguages.So it hasbeen one of our cardinal asks to use suchmetalanguages o reduceor explainwhat is expressed by specifically semantic vocabulary: talk of whatrepresents what, of what expresses what content, of what somethingmeans.

    There is obviously something importantly right about this line ofthought,but it is important o be carefulaboutjust what it is. It cannotbethat,faced with the choice betweengiving up the use of suspect modal locu-tions andshowingthem to be respectableby reducingthem to ordinary xten-sional nonmodalones, we did the latter. The Kripkesemantics is not even acandidate orprovidingsuch a reduction,becauseit owes its extensionalchar-acter to the introductionof new primitivenotions, possible worlds andacces-sibility relations (and in the case of quantifiedmodal languages, furtherapparatus ermitting e-identification f individualsandacrossworlds)that arethemselves richly modal (andwhose deploymentturns out to require urthermetaphysicallynontrivial commitments concerningabout what is essentialand what accidental).Any legitimate reasons to doubt the legitimacy of talkof necessity andpossibility arejust going to be passed off and transformedinto corresponding easons to doubtthe legitimacy of appealto such primi-tives. Thusan appealto advances n formal semantics for modal logic is notreally responsive to the original conceptual challenge. The new semanticsgives us much greatercontrol over our use of modal vocabulary,helping usto get clearerabout what we commit ourselves to by using it, how differentmodal notions relatedto one anotherand to some nonmodalones, articulatingthe fine structure f modalconcepts. But this clarification s resolutely inter-nal to a language deploying the modalconcepts, and does not addressn anydirectway more global worries about the legitimacy in principleof all suchlanguages.Here is a second answer to the questionof whatjustifies the radicalshiftin attitudefrom treatingmodal notions as the paradigmsof philosophicallyproblematicandsuspect conceptseminently n needof explanationto treatingthem as primeraw materialsto employ in the explanationof otherpuzzlingconcepts. It, too, makes an importantpoint, but one that is prone to mis-understanding.ts basis is the very observation that made the analysis andjustificationof modal concepts seem so urgentto the empiricist tradition nthe first place: their ubiquityin scientific theory andpractice.If the naturalsciences generally help themselves generously to modal notions such asdispositions, if they distinguish what is possible from what is not, treatsome but not all regularitiesas lawful, and endorse counterfactual onclu-sions, why should not a scientific semantics be entitled to do the same? Whatjustifies puttingmore restrictiverequirements n the scientific explanationofmeaningthan one putsfor instance on the scientific explanationof the distri-598 ROBERT BRANDOM

    http://www.jstor.org/page/info/about/policies/terms.jsp
  • 8/13/2019 Brandom

    13/23

    butionof different sortsof animals, plants, or mineralsover the earth?Oncethe demandthatexplanations n semanticsbe a priori and availablefrom thearmchairs recognizedas anunwarrantedelic of explodedphilosophical prej-udices, they can take theirplace alongsideother scientific explanations, andhelp themselves to the same sorts of conceptualtools.The troublewith this argumentis that semantics is not just one morespecial science. It is (also) a philosophical discipline. Among its tasks ispreciselytheunderstandingndif need be criticism of the concepts employedby other disciplines. Part of its business is adjudicatingdisputes over thelegitimacy of varioussortsof attemptsto mean something. It is accordinglymethodologically requiredo be much more self-conscious and critical aboutthe concepts it itself relies upon than are the disciplines whose concepts itstudies. The geologist andthe evolutionarybiologist are entitled to ignorechallenges to the ultimate coherence and legitimacy of talk of dispositionsandlaws, andjust get on with the business of talking about the movementsof tectonicplatesandtheadaptiveadvantagesof homeothermy.The semantictheorist is in a differentposition, subjectto differentexplanatoryobligations,and cannotjustify the same attitudesimply by appealto their example. Atmost the indispensabilityof modal locutions for the special sciences wouldshow that there must be a philosophically satisfactoryresponseto the chal-lenge to their legitimacy forwarded y empiricism in both its traditionalandits distinctivelylogical twentiethcenturyforms. It does not by itself provideone.

    Now, I thinkit is entirely legitimate for the contemporary emantic theo-rist to employ counterfactuals,appeal to dispositions, and in general todeploy a richlymodalvocabulary n formulating heories about intentionalityandmeaning.But I also think it is important o understandwhy, andexactlywhat role the two sorts of considerationust sketchedplay in justifying ourcurrent evel of comfort with the use of modal vocabulary.With the wisdomof hindsight, I think we should see that there were two different sorts ofphilosophical challenge regardingmodalities. One of them was legitimate,but is adequately espondedto by the developments in formal logic andsemantics of the last half century. Those developments do not provide aresponsiveanswer to the other sort of challenge-but thatchallenge is illegit-imate,for reasons that aregesturedat, though not madeexplicit, by appealsto the indispensabilityof modal locutions in science.The empiricist tradition that stretches from Hume to Quine offers thefriends of modalitya starkchoice: eithershow how to explain modalities innonmodal termsor learnto live without them.But properlyunderstanding hepresuppositionsof this challenge gives us good reasons to reject it. For it ispredicated n the ideathat the legitimacy of modal vocabularyturns on itsreducibilityto nonmodalvocabulary.And that presupposesa level of inde-pendentlyandantecedentlyntelligiblediscoursethat is purely descriptiveand

    SYMPOSIUM 599

    http://www.jstor.org/page/info/about/policies/terms.jsp
  • 8/13/2019 Brandom

    14/23

    nonmodal,as a background nd modelwithrespectto which the credentialsofmodaldiscoursecan then be invidiouslycompared.An argumentdue to Kantand revivifiedby Sellarsurges that this idea is a chimera. The ability to useordinary empirical descriptive terms such as 'green', 'rigid', and 'mass'alreadypresupposes grasp of the propertiesand relations made explicit bymodalvocabulary.Sellarssummedup the claim admirably n the title of oneof his early papers: Conceptsas InvolvingLaws, andInconceivableWithoutThem. '9The incompatibility of the propertyof being green with that ofbeing red, andso the impossibilityof one color patch being at once greenandred, is partof the content of the conceptgreen. To call something 'rigid' isto commitoneself to variousconsequences concerningwhat would happenifone applieda potentially deforming orce. That an object has a finite massmeans that the applicationof force is necessary to accelerate t. Apartfromtheirinvolvements in materialincompatibilities, counterfactual ependencies,and necessitations of the sorts these examples illustrate, bits of ordinarydescriptiveempiricalvocabulary annothave themeaningstheydo.20

    Quine, consistently with his anti-modal stance, insists that a maturescience must not appeal to 'dispositional properties'.He holds out as anexampleof how this revisionaryprojectmight proceedaccounts of the solu-bility of sugarin terms of its microstructure.Theremightbe some go to thisthought if it is limited to overtly dispositionalterms such as 'soluble' and'rigid'. But if it is extended o include all terms whose use, like that of'green'and 'mass', essentially, albeit implicitly, brings with it commitmentto counterfactualonsequences,then it is incoherent.For 'structural' ppealsto propertiessuch as valence and featuressuch as molecular orbitals explainphysical or chemical behavioronly in virtue of the modally robust conse-quencesthey have, e.g. for what would happen if two molecules having avalence of +1 combine with one molecule having a valence of -2. One canperhapshave anempirical anguage n use that lacks the expressivepower tomakeexplicit modal relationsamongits terms andpredicates.But one cannothave anempirical anguagethe use of whose terms is intelligible apart romprohibitionsandrequirements oncerning heirrelationswhich would be madeexplicit by the use of modal vocabulary.This is one of the centralpointsKantwas arguingfor in insistingthat commitmentto the applicabilityof thecategories(thepure conceptsof the understanding),which underwrite ecessi-ties, is implicit in the applicabilityof all empirical concepts. Bracketing he'9 Reprintedat pp. 87-124 in J. Sicha (ed.) Pure Pragmaticsand Possible Worlds:The EarlyEssays of WilfridSellars [Ridgeview PublishingCompany,Atascadero, CA, 1980]-here-afterPPPW.20 Semantic theories of the sort I call strongly inferentialist (in Making It Explicit

    [Cambridge:HarvardUniversity Press, 1994, pp. 131-32] treat this sort of relation assufficientto determine the meaning of the broadly inferentially related terms. But thecurrent, weaker, claim depends only on the necessity of standingin such relations forhaving the meaningsthey do.

    600 ROBERT BRANDOM

    http://www.jstor.org/page/info/about/policies/terms.jsp
  • 8/13/2019 Brandom

    15/23

    particularsof his account, it is an argumentwe should endorsetoday. Theappealto the fact that even the special sciences that are not much in thebusiness of formulating necessary laws nonetheless routinely employmodallyrichcounterfactual-supportingdioms has force in justifying a simi-larpolicy for semantics, insofar as it does have force, only as backedup byan argumentof this sort.Now I have only sketched the conclusion of this argument,not done whatwould be requiredo establish it. But if and insofar as thereis a good argu-ment that the applicabilityof ordinaryempirical descriptivenonmodal vocab-ularyimplicitlypresupposesthe applicabilityalso of modal vocabularysuchas 'possibility', 'necessity', and 'counterfactual ependence', he empiricistdichotomy-either reducemodalconcepts to nonmodal ones or resolve to dowithout the modal ones-has no bite. For one can neverbe in the supposedposition from which the challenge is issued. One can neverbe in a positionin which the use of ordinarynonmodal terms such as 'green', 'rigid', and'mass', 'cat', and 'mat', is in principlelegitimate andunproblematic,whilethe use of modal terms is in principle questionableandproblematic.For wecannot make sense of what is expressedby nonmodalvocabularyapart romall considerationof counterfactually obust consequencesandimplications. Ifthat is so, if modalityis already mplicitly intricated n the very meanings ofnonmodalterms, then the only legitimate challenge is to make sense of thewhole complex idiom that includes both modal and nonmodal descriptivevocabulary.So the only remainingtask is to be sure that we know what wemean by using the modal vocabulary,wherethat is something to be under-stood in terms of having inferentialcontrol over it in relation both to thebasic descriptive vocabulary,and to other modal vocabulary. That is theenterprisewithin which I thinkwe should understandhe conceptualprogressthat has been made by the developmentof model theoretic semantics formodalformallanguages.It contributes o ourunderstanding f the functionofmodal vocabularywithin the whole idiom that comprises both modal andnonmodalvocabulary. thelps articulate herelationbetweenordinarydescrip-tive predicatesandexplicitly modal ones, without pretending o reducethecontent of the latterto the former,or explainthe one in termsof the other.This is what I meant when I said that we should distinguish two differentphilosophical challenges regardingmodalities. The legitimate challenge, toclarifythe expressive role of modal vocabularyandestablish adequate ware-ness of and control over the inferential consequenceswe are committingourselves to by its use, is adequately esponded o by the developments informal logic andsemantics of the last half century.Those developmentsdonot providea responsiveanswer to the other sort of challenge-the challengeto the in-principle egitimacy of modal discoursetout court. But that chal-lenge is itself illegitimate, because predicatedon an ultimately untenable

    SYMPOSIUM 601

    http://www.jstor.org/page/info/about/policies/terms.jsp
  • 8/13/2019 Brandom

    16/23

    commitment to the autonomous intelligibility of what is expressed bynonmodaldiscourse,withrespectto whatis expressedby modal discourse.III

    I have indicatedwhat I take to be the real reasonswhy it is entirely in orderfor semantic theoriststo appealto alethicmodalities,typically in the form ofcounterfactual ependencies,n explaining semantic phenomenasuch as theintensionalityor the normativity of intentionality. But I think the story Ihavegesturedat hasfurther essons for the way we think about normativity,and hence about intentionality.For I believe we should tell an analogue ofthat story, mutatis mutandis, with normative vocabularyplaying the roleoriginally played in it by modal vocabulary.For we might ask why thereis aphilosophical problemabout normativity(whether n the context of thoughtabout intentionalityor not), and what sort of problem it is. Guidedby thepreviousdiscussion,we can see that one sort of problemtakes the form of achallenge to its legitimacy: eitherexplain the use of normative vocabularyreductively n terms of the use of nonnormative(including classical modal)vocabulary,or undertake o do without t (for instance, in accounts of seman-tic content or intentionality).On the other hand, if the legitimacy and in-principle intelligibility of normativevocabulary s not in this way globallychallenged,a verydifferent ortof enterprisecan still be in order:explicatingandclarifyingtheexpressiveroleplayed by normativevocabulary n a contextthat ncludes bothnormativeandnonnormative ocabulary.I alluded above to Kant'sargument or the conclusion that(to put it in mywords)what is madeexplicit by modal vocabularycan be seen already o beimplicit in the use of ordinary,nonmodalvocabulary.In fact, I think hisargument s much broader,and concerns normativity generally, ratherthanjust modality. The scope of his claims is easy to overlook for two reasons.First,he couches it in terms of 'necessity' (Notwendigkeit),which we auto-matically but I think anachronistically eadback in terms of alethic modali-ties. Second, we tend to think of Kant(at least in this context), as respondingto Hume's worriesabout induction, causation, and necessaryconnection intheoretical philosophy. But on the first point, Kant uses the same term,'Notwendigkeit', to talk about moral necessity. Indeed,for him, naturalnecessityand moralnecessityarespecies of a genus-and the genus clearly isnot alethicmodality.And on the secondpoint,Kant was equallyconcerned orespondto Hume in practical philosophy.He sees the issues about the rela-tion between must be and is as of a piece with those about the relationbetween ought and is.2121 Kripke implicitly rediscovers this connection in modeling his skeptic's worries about

    'ought' on Hume's inductive skeptic's worries about 'necessarily'. In fact in English(and German) we are quite comfortable using 'must' to express both Kantian naturalnecessity and Kantianmoralnecessity.

    602 ROBERT BRANDOM

    http://www.jstor.org/page/info/about/policies/terms.jsp
  • 8/13/2019 Brandom

    17/23

    What ties the two together for Kant is precisely normativity. For him'necessary' ust means according to a rule (his jurisprudentiallynfluencedterm for the form of norms).Objectivelybinding rules he calls 'laws'-onthe practical,as well as the theoreticalside. Specifically moral norms andthose invoked in a scientific setting arejust two species of this genus. Themost basic sortof rule he calls 'concepts', andhis overarching oncern is tounderstand he bindingness or validity (Giltigkeit) of concepts. PerhapsKant's most basic insight is that what it is thatjudgments andactions-thesubject mattersof theoreticalandpracticalphilosophy respectively-have incommon thatdistinguishesthem from the responsesof mere animals is thatthey arethingswe arein a distinctive way responsible for, that they expresscommitments of ours. Central to what we are responsible for (doing) isgiving (or at least having) reasons for them. And what counts as a reason fora givenjudgmentor action is determinedby the rules we govern ourjudgingor acting-that is, the concepts that we therebycount as applying.22 So forKant the discursiverealm('discursive'for himjust meant 'of or pertainingtoconcept use'), that is, the intentional, is pickedout by its being articulatedaccordingto a certainkindof norm:normsgoverningwhat is a good reason,what else one counts as committing oneself to by a certainjudgment oraction,and whatwould count asjustifying those commitments. Graspingtheconceptualor intentionalcontent of a state or performance equiresmasteringthose inferentiallyarticulatednorms thatdeterminewhen it would be correct,and whatelse one has obliged oneself to thereby.Kantgoes on to drawconclusions about the specific forms of judgmentsin which concepts must be able to participate n order o play this normativerole. Leavingaside thedetailsof his argument,what connection can be recov-ered between this generalview about the inferentiallyarticulatednormativityof intentionality as such and discerning alethic modal involvements asimplicitin the use of ordinarydescriptiveconcepts?In an early paper,Sellarssays I shall be interpretingour judgements to the effect that A causallynecessitatesB as the expressionof a rule governingour use of the terms 'A'and 'B'. 23Elsewhere, he develops this idea, (following out a thought ofCarnap's24),ccordingto which thelanguageof modalities is interpreted s a'transposed' anguageof norms. The transposition s from the formal to thematerialmode, from talk abouttalkingaboutthings to talk aboutthings. Theidea is thatendorsingthe claim thatA necessitatesB is endorsingthe propri-ety of a certainkind of inference rom (A) to (B). Of course what the claimthat A necessitates B says is not that this inference is good. It doesn'tmentionexpressions, proprieties,or inferences.But understandingthe claim22 In the lattercase, this is bringingthe performanceundera maxim.23 Language,Rules, and Behavior ootnote2 to p. 136/296 in PPPW, op. cit.24 Inferenceand Meaning, pp. 279/331-284/336 in PPPW, op. cit.

    SYMPOSIUM603

    http://www.jstor.org/page/info/about/policies/terms.jsp
  • 8/13/2019 Brandom

    18/23

    requiresbeing able to distinguish he inferencesone would become committedto by endorsingit. Those normativeconsequencesfor what inferences onewho applies the concept of necessaryconsequence herebybecomes commit-ted to are in that sense implicit in the concept. Making them explicit (thereversetranspositionrom material o formalmode) requiresconcepts pertain-ing to the use of expressionsthat are not made explicit by that use-conceptssuch as expression, inference, and (most importantlyfor our purposeshere)normativeconcepts such as propriety, commitmentandentitlement, obliga-tion and permission.25 That a given bit of vocabularyexpresses causalnecessitation, an alethic modality, is a matter of the sort of inference itimplicitly involves (thatone implicitly endorsesby applying the concept).The inferences its use commits us to are counterfactually obust. All thecoins in my pocket arecopper.It follows thatif I had heatedone of them to1082?C., it would not have melted, but if I had heated it to 1084? C., itwouldhave, (thoughI didn't andit didn't).But it does not follow that if thisnickel had been in my pocket, it would have been copper.I said in the previoussection that Kant and Sellars think thereis an argu-ment that the use of any ordinary mpiricaldescriptiveconceptcommits us tothe correctnessof at least some counterfactually obust inferences(though Ididn'tpretend o presentsuch an argument).Thus the use of any such conceptpresupposes the applicability in principle of modal vocabulary, whoseexpressive role it is to make explicit the differencebetweenthe two sorts ofinferences llustratedby those concerningthe coins. So if what is expressedby the use of ordinary mpiricaldescriptivepredicates s intelligible in prin-ciple, then so is what is expressedby modal vocabulary.My present point isthat the backgroundof thatargument s a more generalone, to the effect thatmasteringthe use of ordinaryempirical descriptive predicates,which is inpractice understandingheir content or the meaning they express, requiresbeing able to distinguishsome uses and inferencesas correct, and others asincorrect. It requiresknowing (in the practicalsense of being able to distin-guish, a kindof know-how)at least somethingabout what one is committingoneself to by applying the concept, and what would entitle one to do that.Treatingone descriptivepredicateas applicable in a particularcase obligesone to consider others (suitably inferentially related to it) as applicable,(normatively)precludes one fromapplyingothers,and licenses one to applyothers. Since this essential dimension of the use of even ordinary,descriptive,nonnormativeconcepts (in belief andjudgment no less than inlinguistic assertion) is what is made explicit by normative vocabulary, itcannot be that ordinary empirical descriptive concepts are coherent andintelligible in principle, but normative concepts are incoherent and25 Sellarsgesturesat this pointby his somewhatdarkdistinctionbetweenwhat one asserts

    by using a sentence,and what one merely conveys by using it. Ibid.p. 281/333.

    604 ROBERTBRANDOM

    http://www.jstor.org/page/info/about/policies/terms.jsp
  • 8/13/2019 Brandom

    19/23

    unintelligible in principle. The discriminations that are made explicit bynormativevocabularyareimplicitalready n the ability to use andunderstandnonnormativevocabulary.One may not yet have the words to make themexplicit, but just by using the nonnormativevocabularyone alreadyhas allthe abilities needed to understand he basic normative concepts. Thisargument s strictlyparallelto thatfor the implicit,practicalmasteryof whatis madeexplicit by modal vocabularybeing partof the backgroundagainstwhich alone masteryof the use of nonmodalvocabulary s intelligible.This means that no concept user can be in a position of being skepticalabout the in-principle ntelligibilityof specifically normativeconcepts, with-out thereby being skeptical about the in-principle intelligibility of conceptuse (andso meaning,content,andintentionality)generally.It is a version ofthis thought that sets the stage for Kripke's Wittgenstein's skeptic's argu-ment for precisely that paradoxical ttitude:a reductio ad absurdam of thecoherence of conceptualnormativity,and so of the use of concepts, includingthose appealedto in mountingthe argument.The form of that argumentwasa challenge:to say what it is about how a concept is or has actually beenappliedin the past that settles how it ought to be appliedin the future. Theforce of the argument urnscritically uponwhatvocabularyone is allowed touse in specifying the past use. Kripke's skeptic is willing to countenancemodal vocabulary-for instance the invocation of dispositions.Now the rulesfor determining ust what vocabularyone is allowed to use, and their ratio-nale, are neverexplicitly discussed. But I think it is clear that Kripke'sskep-tic is not willing to countenancethe use of normative vocabulary n describ-ing the past use. For if one acknowledges hat applying a concept correctlyor incorrectly is also something we actually do and have done, then theproblemtakes on a verydifferentshape.It becomes theproblemof projectingfutureproprietiesfrom past proprieties,of saying what it is for us to beboundnow by a prior decision about what was correctand incorrect,abouthow an expression ought to be used. Such a questioninvites certainkinds ofphilosophicalexplication,but does not threaten o yield a paradox.In askinghow the actual use of concepts could determinehow they oughtto be used, how it would be correct o use them, Kripke's skeptic implicitlyrestricts he vocabularyused to specify that use to nonnormativevocabulary.He thus assumes that normativespecificationsof proprietiesof concept usearein principle intelligible only if they can be reducedwithout remainder ospecifications of nonnormativeproperties of concept use. But the moregeneralKantianargumentwe have been considering nvalidatesthis assump-tion. We cannot be in a position where the specification of the use ofconcepts in nonnormativevocabulary s unproblematic,but the applicabilityof richernormativevocabulary n specifying thatuse is in principlequestion-able. For graspingthe nonnormativedescriptionsrequiresalreadymaking thediscriminationsthat are expressed by normative vocabulary. The implicit

    SYMPOSIUM 605

    http://www.jstor.org/page/info/about/policies/terms.jsp
  • 8/13/2019 Brandom

    20/23

    restriction o nonnormative pecificationsof the actual use of concepts begsthe question, in the context of acknowledgments f the essentially normativecharacterof concept use. The credentialsof normativeconcepts should nomore be held hostage to their reducibilityto nonnormativeones than thecredentialsof modal concepts should be held hostage to theirreducibilitytononmodalones, and for the same reasons.If that is right, then there is methodologicalroom for another sort ofexplanatory trategyfor addressingphilosophical issues about intentionality,besides the one consideredn my opening remarks.For just as it is legiti-mate, wherehelpful, to use modal vocabulary n elucidatingand explainingvarious featuresof intentionality,so it is in principlelegitimate to appealtonormative vocabulary o do so. It is not circularor otherwiseobjectionableto appealto non-extensionalmodalconcepts to make sense of the kind ofintensionality peculiarto intentional phenomena. And it is not circularorotherwiseobjectionable o appealto normativeconceptsto make sense of thespecifically conceptual normativitydistinctiveof intentionalphenomena.Inshort, I think there is room to lay normativeapproaches o intentionalityalongside the modal ones made familiar by Dretske, Fodor, and others(including,in her distinctiveway, Millikan).I haveclaimed that we shouldreject global skeptical challengesto the in-principle intelligibility of normativeconcepts, as we should reject corre-spondingchallenges to the intelligibility of modal ones. They proceed romanddependon an ultimately unsustainablepictureof the meanings of exten-sional, descriptive,empirical predicatesas autonomously intelligible apartfromcounterfactual-supportingonnectionsto otherpredicates,andnormativefeaturesof their use. But to rejectthatpicture,and with it the global skepticalchallenges it sustains, is not to rejectmore local demands or explication oftheexpressiveand inferential olesplayedby vocabularyof the various sorts,andof interrelationsamong them-the sort of enterprise o which possibleworld semantics makes such a signal contribution in the case of modalvocabulary.An explicative responsedoes not require he reductionof norma-tive to nonnormativevocabulary,but its clarificationas a component in awhole that comprisesboth, as well as modal vocabulary.One can imaginemany ways of approaching his undertaking.Two that have been workedoutin some detail are explications of the norms internal to intentionality inDarwinian or developmental terms, as Millikan elaborates in Language,Thought,and OtherBiological Categories, andexplications of those normsin Wittgensteinianor social terms,as I do in MakingIt Explicit.

    606 ROBERT BRANDOM

    http://www.jstor.org/page/info/about/policies/terms.jsp
  • 8/13/2019 Brandom

    21/23

    I am not claiming that these alternativesare exhaustive, or even exclu-sive.26Nor will I attempthere to assess their relative merits. I want rather oconcludethe abstractmethodologicaldiscussion I have presentedhere with afew considerationsabout the explanatoryecological niche that such accountsmight be seen to occupy. One reasonto want a story of the genus of whichthese two arespecies is that it is important hat the advent of intentionalitynot be left seeming magicalor mysterious.There was a time when there wasno conceptuallyarticulated ctivity-nothing boundby or subject to assess-mentaccordingto conceptualnorms. Later,there was. An importantpartofunderstanding urselves is understandinghe natureof this transition. Thefinal point I want to make concernsthe criteriaof adequacyt is appropriateto imposeon candidatecontributions o suchunderstanding.

    For it is, I think, in the first instancein connection with these considera-tions that it comes to seem right and properto insist on a naturalisticaccount of intentionality n all its aspects. Now if the alternative s to allowsupernaturalexplanations, it is hard to see how to object to such acommitment. But, grantingthat, I think it is useful to keep in mind thatthere are still options to be considered.One of them, to be sure (perhaps heone thatcomes most easily to mind) is to offer an account that is naturalisticin the sense of being reductive-showing how conceptualnormativitytakesits place as a species of a genus that is intelligible entirely in terms ofvocabularyalready n use in the natural ciences. Millikan'sstoryis a tour deforce of this kind. But there are other possibilities. One might be anonnaturalistabout conceptualnormativity simply in the sense of beingnonreductionistaboutit. On the approach have recommended,or instance,normativity of the sort that articulates discursive practice (and sointentionality n the sense I address) s seen as essentially a productof socialinteractions: here were no commitments andentitlements, no authorityandresponsibility, until creatures started in practice taking each other to becommitted and entitled, recognizing or attributing authority, holding eachotherresponsible.The normativestatuses that areproductsof such essentiallysocial interactionsare the subject matter,not of the natural sciences, but atmost of the social sciences. Explicatingthem may requirea vocabulary hatis in principle richer than that employed by the natural sciences27-for

    26 I am very interested,for instance,in what I see as Hegel's way of combining social anddevelopmental (though not, for pardonablechronological reasons, WittgensteinianandDarwinian)considerations n his account of the natureand content of conceptualnorms.27 In spite of the remark above, while Millikanbegins her story with concepts availablealreadyto the naturalsciences, it is much less clear thatby the end she has not introducedsocial considerations of a sort that go substantiallybeyond those countenanced by, say,naturalhistoryor populationbiology. Insofar as her project depends merely on the acontinuousdevelopment of concepts rooted in biology, there may be less differencebetween our approachesthaninitially appears.

    SYMPOSIUM607

    http://www.jstor.org/page/info/about/policies/terms.jsp
  • 8/13/2019 Brandom

    22/23

    instance,a normativevocabulary.Nonnaturalism f this nonreductivekind isstill safely removed fromany sort of supernaturalism.In fact it appears hatone can mix and match the two kinds of approach oconceptualnorms with naturalismor nonnaturalism.Thus-picking a fewprominentexamples-I take it that a broadlysocial approach o conceptualnormativityis pursued n a nonnaturalistic orm by Wittgenstein, and in anaturalistic orm by Crispin Wright. A broadly developmentalapproach sdevelopedin nonnaturalisticform by Hegel, and in naturalistic form byMillikan. Within Millikan's own account, it is a worthwhile enterprisetodistinguish those features that are consequencesof her general naturalisticcommitments, those that are consequencesof the particulardevelopmental-evolutionary way she follows out her program,and those that are featuressimply of taking conceptualnormativity seriously.28Prominent among thelatter, for instance, is her denialof meaning rationalism:the claim that thecontents of our thoughts andour concepts is transparento us. Part of thesignificancealreadyof Kant's displacementof Cartesiancertaintyas the cen-tral phenomenonin concept use in favor of the rule-governedness e called'necessity' was a shift from concernwith our grip on concepts to concernwith theirgripon us. So he asks not Are they clearand distinct to us? butArethey valid for or binding on us? 29Accordingalso to Wittgenstein'snonnaturalized ocial account of conceptual norms (or that of Making ItExplicit), the commitmentsone undertakesby making a move in a languagegame may well outrunwhat the one undertakinghe commitmentappreciates.Again, any theorist who thinks of applying a concept primarilyas undertak-ing a commitmentor bindingoneself by a norm ought to find the undertak-ing of incompatibleconceptualcommitments (and so having an incoherentconstellation of commitments)no more mysterious in principlethan is thepossibility of binding oneself by incompatible promises. White Queenpsychology too has nonnaturalistic arieties.

    28 There are many other separablethreads in Millikan's rich tapestry,whose responsibilityfor variousattractive eaturesand consequencesit would be interestingto investigate: theexplanatorypriorityaccorded to propositional representations, looking downstream toconsumers of representationsas well as upstreamto their producers, the fact that herfunctionalismis not only normative (rather than simply causal), but also social (bothdiachronicallyand synchronically), rather than individualistic,and so on. (For instance,the superiorityof her account of the magnetosomecontainingbacteria over Dretske'sderives from her considerationof the consumers of representations. See pp. 125ff ofWhiteQueen Psychology.)

    29 This feature of the normativeconception of conceptual content is of the utmostimpor-tance in making palatableholismabout such contents.For the problemof how it is possi-ble to share such contents(andso to communicate) akeson a completely different shapewhen one understands t as a questionof how individualswith different beliefs can cometo be boundby a sharedset of norms.

    608 ROBERTBRANDOM

    http://www.jstor.org/page/info/about/policies/terms.jsp
  • 8/13/2019 Brandom

    23/23

    A prominentandwelcome featureof thephilosophical andscapeof our timesis a set of researchprogramsoffering reconstructionsof notions of intention-alityandrepresentationhat arevastlymore detailedand well worked-out hanany previously available.They are standardly nd properly seen as develop-ments in two traditions.First, as regards ts aim, the common enterpriseepitomizedby the workof Dretske, Fodor, andMillikan seeks a naturalistictreatmentof the centralphenomenaof intentionality and semantics. Thisaspiration s recognizableas a version of the unity of science programthatmotivatedTarski,Carnap,andQuine. Second, as regards ts explanatoryrawmaterials,this constellation of approachess distinguishedby its relianceonand comfort with the use of alethic modal vocabulary:appeals to disposi-tions, counterfactuals,and so on. These are idioms that their naturalistprede-cessors regarded s at best themselves too much in need of explanation forthem to be eligible to serve in explainingintentionalor semantic idioms, andat worst as in principle defective and incoherent.I think the contemporaryattitude owards he legitimacy of modal locutions is entirely in order.But Ialso think thatunderstandingwhyit is in orderopens up a broader ntellectualcontext within which it is illuminatingto locate these semanticprojects. SoI have said next to nothing about explanatoryvirtues and vices of particularundertakingsn this tradition.My concernin this essay has ratherbeen to tryto understand etter the philosophical landscapethey occupy, and so theirrelationto other approaches o intentional and semantic phenomena, whichshare with them neither their motivating naturalismnor the central method-ological placethey accordto recentlyrehabilitatedmodal notions.

    SYMPOSIUM609