Churchland Source: the Journal of Philosophy, Vol

Home , Patricia Churchland, Thomas Nagel

Journal of Philosophy, Inc.

Reduction, Qualia, and the Direct Introspection of Brain States Author(s): Paul M. Churchland Source: The Journal of Philosophy, Vol. 82, No. 1 (Jan., 1985), pp. 8-28 Published by: Journal of Philosophy, Inc. Stable URL: http://www.jstor.org/stable/2026509 Accessed: 07-08-2015 19:14 UTC

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

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

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 142.58.129.109 on Fri, 07 Aug 2015 19:14:45 UTC All use subject to JSTOR Terms and Conditions 8 THE JOURNAL OF PHILOSOPHY structureof this idiom, moreover-its embeddingof a subordinate sentence-would have been clearly dictatedby its primitiveuse in assessing children's acquisition of observationsentences. Analogi- cal extension of the idiom to other than observation sentences would follow inevitably,and the developmentof parallel idioms for other propositional attitudeswould then come naturally too, notwithstanding their opacity from a logical point of view. Naturalness is one thing, transparencyanother; familiarityone, clarityanother. W. V. QtJINE Harvard University

REDUCTION, QUALIA, AND THE DIRECT INTROSPECTION OF BRAIN STATES* DO the phenomenological or qualitative featuresof our sensations constitutea permanentbarrier to thereductive aspi- rations of any materialisticneuroscience? I here argue that theydo not. Specifically,I wish to address the recentanti-reduc- tionist argumentsposed by Thomas Nagel,' Frank Jackson,2and Howard Robinson.3 And I wish to explore the possibility of human subjectiveconsciousness within a conceptual environment constitutedby a maturedand successfulneuroscience. If we are to deal sensiblywith the issues here at stake, we must approach themwith a general theoryof scientificreduction already in hand, a theorymotivated by and adequate to the many instances and varietiesof interconceptualreduction displayed elsewherein our scientifichistory. With an independentlybased account of the nature and grounds of intertheoreticreduction, we can approach

*Research forthis paper was supportedby a grantfrom the Institutefor Advanced Study, by a research/studyleave fromthe Universityof Manitoba, and by SSHRC grant no. 451-83-3050.An earlierdraft was presentedto the Conferenceon Methods in Philosophy and the Sciences, convened in honor of Ernest Nagel, at the New School of Social Research in April, 1983. My thanks to that audience for theirin- sightfulcriticisms and helpful suggestions.Thanks also to Thomas Nagel, Frank Jackson,Daniel Dennett,Philip Hanson, Charles Marks,and Brian Loar forcritical discussions of thatearlier draft. ' "What Is It Like to Be a Bat?", Philosophical Review, ILXXXIII, 4 (October, 1974): 435-450; page referencesto Nagel are to this paper. 2"Epiphenomenal Qualia," Philosophical Quarterly, xxxii, 127 (April 1982): 127-136. 3Matterand Sense (New York: Cambridge, 1982), p.,4.

0022-362X/85/8201/0008$02.00 ? 1985 The Journalof Philosophy, Inc.

This content downloaded from 142.58.129.109 on Fri, 07 Aug 2015 19:14:45 UTC All use subject to JSTOR Terms and Conditions THE DIRECT INTROSPECTION OF BRAIN STATES 9 the specificcase of subjectivequalia, freefrom the myopia thatre- sults fromtrying to divine the proper conditions on reductionby simply staringlong and hard at the problematiccase at issue. I. INTERTHEORETIC REDUCTION We may begin by remarkingthat the classical account of intertheoretic reduction4now appears to be importantlymistaken, though the repairs necessaryare quickly and cleanly made. Suppressing niceties,we may state the original account as follows. A new and more comprehensivetheory reduces an older theoryjust in case the new theory, when conjoined with appropriate correspondence rules, logically entails the principles of the older theory.(The point of the correspondencerules, or "bridge laws," is to connect the disparate ontologies of the two theories: often these are expressed as identity statements,such as Temperature= mv2/3k.) Schematically,

TN & (CorrespondenceRules) logically entails To Difficultieswith this view begin with the observationthat most reduced theoriesturn out to be, strictlyspeaking and in a varietyof respects,false. (Real gases don't reallyobey PV = ,RT, as in classical thermodynamics;the planets don't reallymove in ellipses, as in Keplerian astronomy;the acceleration of falling bodies isn't really uniform,as in Galilean dynamics; etc.) If reductionis deduction, modus tollens would thus require that the premisesof the new reducing theories(statistical thermodynamics in the firstcase, New- tonian dynamicsin the second and third)be somehow falseas well, in contradictionto theirassumed truth. This complaint can be temporarilydeflected by pointing out that the premises of a reduction must often include not just the new reducing theorybut also some limitingassumptions or coun- terfactualboundary conditions (such as that the molecules of a gas enjoy only mechanical energy,or that the mass of the planets is negligible compared to the sun's, or thatthe distanceany body falls is negligiblydifferent from zero). Falsity in the reducing premises can thus be conceded,since it is safelyconfined to thoselimiting or counterfactualassumptions. This defensewill not deal with all cases of falsity,however, since in some cases the reduced theoryis so radically false that some or

4Ernest Nagel, The Structuresof Science (New York: Harcourt, Brace &zWorld, 1961), ch. 11.

This content downloaded from 142.58.129.109 on Fri, 07 Aug 2015 19:14:45 UTC All use subject to JSTOR Terms and Conditions 10 THE JOURNAL OF PHILOSOPHY all of its ontology must be rejectedentirely, and the "correspondence rules" connectingthat ontology to the newerontology there- foredisplay a problematicstatus. Newly conceived featurescannot be identical with,nor even nomically connectedwith, old features, if the old featuresare illusoryand uninstantiated.For example, rel- ativisticmass is not identical with Newtonian mass, nor even coextensivewith it, even at low velocities.Nevertheless, the reductionof Newtonian by Einsteinian mechanics is a paradigm of a successful reduction. For a second example, neitheris caloric-fluid-pressure identical with, nor even coextensivewith, mean molecular kinetic energy.But an overtlyfluid thermodynamics(i.e., one committed to the existenceof caloric) still findsa moderatelyimpressive reduction within statisticalthermodynamics. In sum, even theorieswith a nonexistentontology can enjoy reduction,and this fact is pro- blematic on the traditionalaccount at issue. What cases like these invite us to give up is the idea that what gets deduced in a reductionis the theoryto be reduced. A more accurate, general, and illuminating schema for intertheoreticreduction is as follows: TN & LimitingAssumptions & BoundaryConditions logically entails

IN [a setof theoremsof (restricted)TN] e.g.,(x) (Ax D Bx) (x) ((Bx & Cx) D Dx) which is relevantlyisomorphic with To (theolder theory) e.g.,(x) (Jx DKx) (x) ((Kx & Lx) D Mx) That is to say,a reductionconsists in thededuction, within TN, not of To itself,but ratherof a roughly equipotent image of To, an image still expressed in the vocabulary proper to TN. The correspondence rules play no part whatever in the deduction. They show up only later, and not necessarilyas material-modestatements,but as mereordered pairs: (Ax, Jx), (Bx, Kx), (Cx, Lx), (Dx, Mx). Their functionis to indicate which termsubstitutions in the image IN will yield the principles of To. The older theory,accordingly,is neverdeduced; it is just the targetof a relevantlyadequate mimicry.Construed in this way, a correspondencerule is entirely consistentwith the assumption that the older predicate it encom- passes has no extension whatever.This allows that a true theory mightreduce even a substantiallyfalse theory.

This content downloaded from 142.58.129.109 on Fri, 07 Aug 2015 19:14:45 UTC All use subject to JSTOR Terms and Conditions THE DIRECT INTROSPECTION OF BRAIN STATES 11

The point of a reduction,according to this view, is to show that the new or more comprehensivetheory contains explanatoryand predictiveresources that parallel, to a relevantdegree of exactness, the explanatoryand predictiveresources of the reduced theory.The intra-theoreticdeduction (of IN within TN), and the inter-theoretic mapping (of To into IN), constitutea fell-swoop demonstration that the older theorycan be displaced wholesale by the new, without significantexplanatory or predictiveloss.5 Material-modestatements of identitycan occasionally be made, of course. We do wish to assert that visible light = EM waves be- tween .35 ,Amand .75 ,Am,that sound = atmosphericcompression waves, that temperature= mean molecular KE, and that electric current= net motion of charged particles. But a correspondence rule does not itselfmake such a claim: at best, it recordsthe fact that the new predicate applies in all those cases where its To- doppelganger predicate was normally thought to apply. On this view, full-fledgedidentity statements are licensed by the compara- tivesmoothness of therelevant reduction (i.e., the limitingassumptions or boundaryconditions on TN are not wildly counterfactual, all or most of To's principles findclose analogues in IN, etc.). This smoothnesspermits the comfortableassimilation of the old ontology within the new and thus allows the old theoryto retain all or most of its ontological integrity.It is smooth intertheoreticreduc- tions that motivateand sustain statementsof cross-theoreticiden- tity,not the otherway around. The precedingframework allows us to framea usefulconception of reductionfor specificproperties, as opposed to entire theories, and it allows us to framea usefulconception of thecontrary notion of "emergent"properties. A propertyF, postulatedby an older theory or conceptual frameworkTo, is reduced to a propertyG in some new theoryTN just in case

(1) TN reducesTo; (2) F and G are correspondence-rulepaired in thereduction; and (3) The reductionis sufficientlysmooth to sustainthe ontology of To, and thusto sustainthe identity claim, "F-ness = G-ness." Intuitively,and in the materialmode, this means thatF-ness reduces to G-nessjust in case the "causal powers" of F-ness (as outlined in the laws of To) are a subset of the "causal powers" of G-ness (as outlined in the laws of TN).

5This sketchof intertheoreticreduction is drawn frommy ScientificRealism and the Plasticityof Mind (New York: Cambridge, 1979), section 11. For a more detailed account see CliffordA. Hooker, "Towards a General Theory of Reduction", Dia- logue, xx, 1, 2, 3 (March, June,September 1981): 38-59, 201-236, 496-529.

This content downloaded from 142.58.129.109 on Fri, 07 Aug 2015 19:14:45 UTC All use subject to JSTOR Terms and Conditions 12 THE JOURNAL OF PHILOSOPHY

Finally, a propertyF will be said to be an emergent property (relativeto TN) just in case (1) F is definitelyreal and instantiated; (2) F is co-occurrentwith some feature or complexcircumstance rec- ognizedin TN;but (3) F cannotbe reduced to any propertypostulated by or definable withinTN. Intuitively,this will happen when TN does not have the resources adequate to define a propertywith all the "causal powers" pos- sessed by F-ness. Claims about the emergenceof certainproperties are thereforeclaims about the relative povertyin the resourcesof certainaspirant theories.6Having outlined thesenotions, we shall turnto address substantivequestions of emergenceand irreducibil- ity in a few moments. Beforewe do so, several points about reductionneed to be em- phasized. The firstis that,in arguing forthe emergenceof a given propertyF relative to some theoryTN, it is not sufficientto point out that the existence or appearance of F-ness cannot be deduced fromTN. It is occasionally claimed, forexample, that the objective featuresof warmthor blueness must be irreduciblyemergent properties,since, howevermuch one bends and squeezes the molecular theoryconcerning H20, one cannot deduce fromit that waterwill be blue, but only that water will scatterelectromagnetic radiation at such-and-such wavelengths. And however much one wrings fromthe mechanics of molecular motion, one cannot deduce from it that a roaring hearth will be warm, but only that its molecules will have such-and-sucha mean kineticenergy and will collectively emit EM radiation at longish wavelengths. These premisesabout nondeducibilityare entirelytrue, but the conclusion against reducibilitydoes not follow. It is a serious mis- take even to make indirectdeducibility (i.e., deducibilitywith the

6A word of caution is perhaps in orderhere, since the expression'emergent property' is often used in two diametricallyopposed senses. In scientificcontexts, one frequentlyhears it used to apply to what might be called a "networkproperty," a propertythat appears exactlywhen the elementsof some substrateare suitably or- ganized, a propertythat consists in the elementsof thatsubstrate standing in certain relations to one another,a set of relations thatcollectively sustain the set of causal powers ascribed to the "emergent" property.In this innocent sense of 'emergent', thereare a great many emergentproperties, and quite probably the qualia of our sensations should be numbered among them. But in philosophical contexts one more often encounters a differentsense of 'emergent',one that implies that an "emergent" propertydoes not consist in any collective or organizational featuireof its substrate.The firstsense positivelyimplies reducibility;the second implies irre- ducibility.It is emergencein the second sense that is at issue in this paper.

This content downloaded from 142.58.129.109 on Fri, 07 Aug 2015 19:14:45 UTC All use subject to JSTOR Terms and Conditions THE DIRECT INTROSPECTION OF BRAIN STATES 13 help of correspondencerules) a requirementon successfulreduc- tion,as we saw at the beginningof this section.And thereare addi- tional reasons why it would be even more foolish to insist on the much strongercondition of directdeducibility. For example, for- mal considerationsalone guarantee that,for any predicate 'F' not already in the proprietarylexicon of the aspirant reducing theory TN, no statementswhatever involving 'F' (beyond tautologies and other trivialexceptions) will be deducible fromTN. The deducibility requirementwould thus trivialize the notion of reduction by making it impossible forany conceptual frameworkto reduce any other,distinct conceptual framework.Even temperature-thatpar- adigm of a successfullyreduced property-would be renderedir- reducible,since the term'temperature' does not appear in the lex- icon of statisticalmechanics. There is a furtherreason why the demand fordirect deducibility is too strong. It is a historical accident that we humans currently use preciselythe conceptual frameworkwe do use. We mighthave used any one of an infinitenumber of otherconceptual frameworks to describe the observable world, each one of which could have been roughlyadequate to common experienceand many of which would be roughlyisomorphic (each in its differentway) with some part of the correctaccount that a utopian theorywill eventually provide. Accordingly,we can legitimatelyask of a putativelycor- rect theoryof a given objective domain that it account for the phenomena in ( = function successfullyin) that domain. But we cannot insist that it also be able to predicthow this, that,or the other conceptually idiosyncratichuman culture is going to conceive of that domain. That would be to insist that the new theory do predictivecultural anthropologyfor us, as well as mechanics,or electromagnetictheory, or what have you. The demand that molecular theorydirectly entail our thermalor color concepts is evi- dentlythis same unreasonable demand. All we can properlyask of a reducing theoryis that it have the resources to conjure up a set of properties whose nomological powers/roles/featuresare systematicanalogues of the powers/roles/ featuresof the set of propertiespostulated by the old theory.Since both theoriespresume to describethe same empiricaldomain, these systematicnomological parallels constitutethe best grounds there can be for concluding that both theorieshave managed to latch onto the same set of objectiveproperties. The hypothesizedidentity of the propertiesat issue explains why IN and To are taxonomically and nomically parallel: theyare both at least partiallycorrect ac- counts of the verysame objective properties.IN merelyframes that

This content downloaded from 142.58.129.109 on Fri, 07 Aug 2015 19:14:45 UTC All use subject to JSTOR Terms and Conditions 14 THE JOURNAL OF PHILOSOPHY account within a much more penetratingconceptual system-that of TN. Moreover,it is to be expected that existing conceptual frameworks will eventuallybe reduced or displaced by new and better ones, and those in turnby frameworksbetter still; forwho will be so brash as to assertthat the feebleconceptual achievementsof our adolescent species comprise an exhaustive account of anythingat all? If we put aside this conceit, then the only alternativesto intertheoreticreduction are epistemic stagnation or the outrightelimination of old frameworksas wholly false and illusory. II. THEORETICAL CHANGE AND PERCEPTUAL CHANGE Esoteric propertiesand arcane theoreticalframeworks are not the only thingsthat occasionally enjoy intertheoreticreduction. Obser- vable propertiesand common-sense conceptual frameworkscan also enjoy smooth reduction. Thus, being a middle-A sound is identical with being an oscillation in air pressureat 440 hz; being red is identical with having a certain tripletof electronmagneticre- flectanceefficiencies; being warm is identical with having a certain mean level of microscopicallyembodied energies,and so forth. Moreover, the relevantreducing theoryis capable of replacing the old frameworknot just in contextsof calculation and inference. It should be appreciated that the reducing theorycan displace the old frameworkin all its observationalcontexts as well. Given the realityof the propertyidentities just listed,it is quite open to us to begin framingour spontaneous perceptualreports in the language of the more sophisticatedreducing theory.It is even desirable that we begin doing this,since the new vocabularyobserves distinctions that are in fact within the discriminatoryreach of our native perceptual systems,though those objective distinctionsgo unmarked and unnoticed fromwithin the old framework.We can thus make more penetratinguse of our native perceptual equipment. Such displacementis also desirablefor a second reason: the greaterinfer- ential or computational power of the new conceptual framework. We can thus make betterinferential use of our new perceptual judgments than we made of our old ones. It is difficultto convey in words the vastnessof such perceptual transformationsand the naturalnessof the new conceptual regime, once established.A nonscientificexample may help to get the in- itial point across. Consider the enormous increase in discriminatoryskill that spans the gap betweenan untrainedchild's auditoryapprehension of a symphonyand the same person's apprehension of the same symphonyforty years later, heard in his capacity as conductorof

This content downloaded from 142.58.129.109 on Fri, 07 Aug 2015 19:14:45 UTC All use subject to JSTOR Terms and Conditions THE DIRECT INTROSPECTION OF BRAIN STATES 15 the orchestraperforming it. What was beforea seamless voice is now a mosaic of distinguishable elements. What was before a dimly apprehended tune is now a rationallystructured sequence of distinguishable and identifiable chords supporting an approp- riatelyrelated melody line. The maturedmusician hears an entire world of structureddetail, concerning which the child is both dumb and deaf. Other modalities provide comparable examples. Consider the practiced and chemically sophisticated wine taster,for whom the "red wine" classificationused by most of us divides into a network of fifteenor twentydistinguishable elements: ethanol, glycol, fruc- tose, sucrose,tannin, acid, carbon dioxide, and so forth,whose relative concentrationshe can estimatewith accuracy. Or consider the astronomer,for whom the speckled black dome of her youth has become a visible abyss, scatteringnearby planets, yellow dwarf stars,blue and red giants, distant globular clusters, and even a remotegalaxy or two, all discriminableas such and lo- catable in three-dimensionalspace with her unaided (repeat: unaided) eye. In each of these cases, what is finally masteredis a conceptual framework-whethermusical, chemical, or astronomical-a framework that embodies far more wisdom about the relevant sensory domain than is immediatelyapparent to untutoreddiscrimination. Such frameworksare characteristicallya cultural heritage,pieced togetherover many generations,and theirmastery supplies a rich- ness and penetrationto our sensorylives that would be impossible in theirabsence.7 Our introspectivelives are already the extensivebeneficiaries of this phenomenon. The introspectivediscriminations we make are forthe most part learned; theyare acquired with practiceand expe- rience, often quite slowly. And the specific discriminationswe learn to make are those it is useful forus to make. Generally,those are the discriminationsthat others are alreadymaking, the discriminations embodied in the psychologicalvocabulary of the language we learn. The conceptual frameworkfor psychological states that is embedded in ordinarylanguage is a modestlysophisticated theo- reticalachievement in its own right,and it shapes our maturedin- trospectionprofoundly. If it embodied substantiallyless wisdom in its categoriesand connectinggeneralizations, our introspectiveapprehension of our internal states and activities would be much

'The role of theoryin perception,and the systematicenhancement of perception throughtheoretical progress, are examined at length in my ScientificRealism and the Plasticityof Mind, op. cit. secs. 1-6.

This content downloaded from 142.58.129.109 on Fri, 07 Aug 2015 19:14:45 UTC All use subject to JSTOR Terms and Conditions 16 THE JOURNAL OF PHILOSOPHY diminished,though our native discriminatorymechanisms remain the same. Correlatively,if folk psychologyembodied substantially more wisdom about our inner nature than it actually does, our in- trospectivediscrimination and recognition could be very much greater than it is, though our native discriminatorymechanisms remain unchanged. This brings me to the central positive suggestionof this paper. Consider now the possibilityof learning to describe,conceive, and introspectivelyapprehend the teemingintricacies of our innerlives within the conceptual frameworkof a matured neuroscience,a neurosciencethat successfullyreduces, either smoothly or roughly, our common-sensefolk psychology. Suppose we trainedour native mechanisms to make a new and more detailed set of discriminations, a set that correspondednot to the primitivepsychological taxonomyof ordinarylanguage, but to some more penetratingtax- onomy of states drawn froma completed neuroscience.And suppose we trained ourselves to respond to that reconfigureddiscri- minative activitywith judgments that were framed,as a matterof course, in the appropriateconcepts fromneuroscience.' If the examples of the symphonyconductor (who can hear the Am7 chords), the oenologist (who can see and tastethe glycol), and the astronomer(who can see the temperatureof a blue giant star) provide a fair parallel, then the enhancementin our introspective vision could approximate a revelation. Dopamine levels in the limbic system,the spiking frequenciesin specificneural pathways, resonances in the nth layerof the occipital cortex,inhibitory feed- back to the lateral geniculate nucleus, and countless other neuro- physical nicetiescould be moved into the objectivefocus of our in- trospectivediscrimination, just as Gm7 chords and Adim chords are moved into the objective focus of a trainedmusician's auditory discrimination. We will of course have to learn the conceptual frameworkof a maturedneuroscience in orderto pull this off.And we will have to practice its noninferentialapplication. But that seems a small price to pay for the quantum leap in self-apprehension.

8 believe it was Paul K. Feyerabendand Richard Rortywho firstidentified and explored this suggestion. See Feyerabend,"Materialism and the Mind-Body Prob- lem," Review of Metaphysics, XVIII.1, 65 (September 1963); 49-66; and Rorty, "Mind-Body Identity,Privacy, and Categories", ibid. xix.1, 73 (September 1965): 24-54. This occurredin a theoreticalenvironment prepared largelyby WilfridSel- lars in "Empiricism and the Philosophy of Mind," in HerbertFeigl and Michael Scriven,eds., Minnesota Studies in the Philosophy of Mind, vol. I (Minneapolis: of Minnesota Press, 1956): secs. 45-63. The idea has been explored more recentlyin my "Eliminative Materialismand the Propositional Attitudes,"this JOURNAL, LXXVIII, 2 (February1981): 67-90.

This content downloaded from 142.58.129.109 on Fri, 07 Aug 2015 19:14:45 UTC All use subject to JSTOR Terms and Conditions THE DIRECT INTROSPECTION OF BRAIN STATES 17

All of this suggeststhat thereis no problem at all in conceiving theeventual reduction of mentalstates and propertiesto neurophys- iological states and properties.A matured and successfulneuroscience need only include, or prove able to define,a taxonomyof kinds with a set of embedding laws that faithfullymimics the tax- onomy and causal generalizationsof folk psychology.Whether fu- ture neuroscientifictheories will prove able to do this is a wholly empirical question, not to be settleda priori. The evidence for a positive answer is substantialand familiar,centering on the grow- ing explanatorysuccess of the severalneurosciences. But thereis negativeevidence as well: I have even urged some of it myself("Eliminative Materialism and the Propositional Atti- tudes," op. cit.). My negative arguments there center on the explanatoryand predictivepoverty of folkpsychology, and theyques- tion whetherit has the categorial integrityto meritthe reductive preservationof its familiarontology. That line suggestssubstantial revisionor outrightelimination as the eventual fateof our mentalistic ontology.The qualia-based argumentsof Nagel, Jackson,and Robinson, however,take a quite differentline. They find no fault with folk psychology.Their concern is with the explanatoryand descriptivepoverty of any possible neuroscience,and their line suggeststhat emergence is the correctstory for our mentalisticon- tology.Let us now examine theirarguments. III. THOMAS NAGEL'S ARGUJMENTS For Thomas Nagel, it is the phenomological featuresof our experiences, the propertiesor qualia displayed by our sensations,that constitutea problem forthe reductiveaspirations of any materialis- tic neuroscience.In his classic position paper (op. cit.) I find three distinctarguments in support of the view thatsuch propertieswill never find any plausible or adequate reductionwithin the frameworkof a maturedneuroscience. All threearguments are beguiling, but all three,I shall argue, are unsound. First Argument:What makes the proposed reduction of mental phenomena differentfrom reductions elsewhere in science, says Nagel, is that It is impossibleto excludethe phenomenological features of expe- riencefrom a reduction,in thesame way thatone excludesthe phe- nomenalfeatures of an ordinarysubstance from a physicalor chemical reductionof it-namely, by explainingthem as effectson the mindsof humanobservers (437). The reason it is impossible to exclude them,continues Nagel, is that the phenomenological featuresare essential to experienceand to the subjective point of view. But this is not what interestsme

This content downloaded from 142.58.129.109 on Fri, 07 Aug 2015 19:14:45 UTC All use subject to JSTOR Terms and Conditions 18 THE JOURNAL OF PHILOSOPHY about thisargument. What interestsme is the claim thatreductions of various substanceselsewhere in science exclude thephenomenal featuresof the substance. This is simply false, and the point is extremelyimportant. The phenomenal featuresat issue are those such as the objectiveredness of an apple, the warmthof a coffeecup, and the pitch of a sound. These propertiesare not excluded fromour reductions.Redness, an objectivephenomenal propertyof apples, is identicalwith a certain wavelength triplet of electromagnetic reflectance efficiencies. Warmth,an objective phenomenal propertyof objects, is identical with the mean level of the objects' microscopicallyembodied energies. Pitch,an objectivephenomenal propertyof a sound, is identical with its oscillatoryfrequency. These electromagneticand micro- mechanical properties, out there in the objective world, are genuine phenomenal properties.Despite widespread ignorance of their dynamical and microphysical details, it is these objective physical propertiesto which everyone'sperceptual mechanisms are keyed. The reductionswhose existenceNagel denies are in factso complete that one can already displace entirelylarge chunks of our common-sense vocabulary for observable propertiesand learn to frameone's perceptualjudgments directlyin termsof the reducing theory.The mean KE of the molecules in thisroom, forexample, is currently about . .. 6.2 x 10-21joules. The oscillatory frequency of this sound (I here whistle C one octave above middle C) is about 524 hz. And the threecritical electromagneticreflectance efficiencies (at .45, .53, and .63 MAm)of this (white) piece of paper are all above 80 per cent. These microphysicaland electromagneticprop- ertiescan be felt,heard, and seen, respectively.Our native sensory mechanisms can easily discriminate such properties,one from another, and their presence from their absence. They have been doing so for millennia. The "resolution" of these mechanisms is inadequate, of course, to reveal the microphysicaldetails and the extendedcausal roles of the propertiesthus discriminated.But they are abundantly adequate to permit the reliable discriminationof the propertiesat issue.9 On this view, the standardperceptual propertiesare not "secon- dary" propertiesat all, in the standard sense which implies that they have no real existence save inside the minds of human ob-

9See again my ScientificRealism and the Plasticityof Mind, op. cit., secs. 2-6. See also Paul and Patricia Churchland, "Functionalism, Qualia, and Intentionality", Philosophical Topics, xii, 1 (October 1981): 121-145. Reprinted in J. I. Biro and R. W. Shahan, eds., in Mind, Brain, and Function (Norman: U of Oklahoma Press, 1982): 121-145.

This content downloaded from 142.58.129.109 on Fri, 07 Aug 2015 19:14:45 UTC All use subject to JSTOR Terms and Conditions THE DIRECT INTROSPECTION OF BRAIN STATES 19 servers.On the contrary,they are as objective as you please, with a wide varietyof objectivecausal properties.Moreover, it would be a mistakeeven to tryto "kick the phenomenal propertiesinwards," since that would only postpone the problem of reckoning their place in nature. We would only confrontthem again later,as we address the place in nature of mental phenomena. And, as Nagel correctlypoints out, the relocation dodge is no longer open to us, once theproblematic properties are alreadylocated withinthe mind. Nagel concludes fromthis that subjectivequalia are unique in being immune from the sort of reductions found elsewhere in science. I draw a very differentconclusion. The objective qualia (redness,warmth, etc.) should never have been "kicked inwards to the minds of observers" in the firstplace. They should be con- frontedsquarely, and theyshould be reducedwhere they stand: outside the human observer.As we saw, this can and has in factbeen done. If objective phenomenal propertiesare so treated,then subjective qualia can be confrontedwith parallel forthrightness,and can be reduced where theystand: inside the human observer.So far then, the externaland the internal case are not different:they are parallel afterall. Second Argument:A second argumenturges the point that the in- trinsiccharacter of experiences,the qualia of sensations,are essen- tially accessible fromonly a single point of view, the subjective point of view of the experiencingsubject. The propertiesof physical brain states,by contrast,are accessible froma varietyof entirely objectivepoints of view. We cannot hope adequately to account for the former,therefore, in termsof propertiesappropriate to the lat- terdomain (cf. Nagel, 442-444). This somewhatdiffuse argument appears to be an instanceof the followingargument: (1) The qualia ofmy sensations are directlyknown by me, by introspection,as elementsof myconscious self. (2) The propertiesof mybrain states are notdirectly known by me, byintrospection, as elementsof myconscious self. ..(3) The qualia of mysensations # theproperties of mybrain states.

And perhaps there is a second argumenthere as well, a comple- ment to the first: (1) The propertiesof my brain statesare known-by-the-various- external-senses,as having such-and-such physical properties. (2) The qualia of my sensationsare not known-by-the-various- external-senses,as having such-and-such physical properties. ..(3) The qualia of mysensations # theproperties of mybrain states.

This content downloaded from 142.58.129.109 on Fri, 07 Aug 2015 19:14:45 UTC All use subject to JSTOR Terms and Conditions 20 THE JOURNAL OF PHILOSOPHY

The argumentform here is apparently (1) Fa (2) -Fb (3) a # b Given Leibniz's law and the extensional nature of the propertyF, this is a valid argumentform. But, in the examples at issue, F is obviously not an extensional property.The fallacy committedin both cases is amply illustratedin the followingparallel arguments. (1) Hitleris widelyrecognized as a massmurderer. (2) AdolfSchicklgruber is not widely recognized as a massmurderer. ..(3) Hitler# AdolfSchicklgruber. or (1) Aspirinis knownby John to be a pain reliever. (2) Acetylsalicylicacid is notknown by John to be a pain reliever. ..(3) Aspirin# acetylsalicylicacid. or, to cite an example veryclose to the case at issue, (1) Temperatureis knownby me,by tactilesensing, as a featureof materialobjects. (2) Mean molecularkinetic energy is not knownby me, by tactile sensing,as a featureof materialobjects. (3) Temperature# meanmolecular kinetic energy. The problem with all thesearguments is that the "property"ascribed in premise 1 and withheldin premise2 consistsonly in the subject item's being recogized,perceived, or known as something, under some specificdescription or other.Such apprehension is not a genuine featureof the item itself,fit for divining identities,since one and the same subject may be successfullyrecognized under one description (e.g., "qualia of my mental state"), and yet fail to be recognized under another, equally accurate, coreferentialdescrip- tion (e.g., "propertyof my brain state"). In logician's terms,the propositional function: x is known(perceived, recongized) by me, as an F is one of a large numberof intensionalcontexts whose distinguish- ing featureis that theydo not always retain the same truthvalue through substitution of a coreferentialor coextensive term for whateverholds the place of 'x'. Accordingly,that such a context (i.e., the one at issue) should show a differencein truthvalue for two terms'a' and 'b' (i.e., 'qualia of my sensations'and 'propertyof my brain-states')is thereforehardly grounds forconcluding that'a' and 'b' cannot be coreferentialor coextensiveterms! 10

This content downloaded from 142.58.129.109 on Fri, 07 Aug 2015 19:14:45 UTC All use subject to JSTOR Terms and Conditions THE DIRECT INTROSPECTION OF BRAIN STATES 21

This objection is decisive,I think,but it does not apply to a dif- ferentversion of the argument,which we must also consider. It may be urged thatone's brain statesare more than merelynot (yet) known by introspection:they are not knowable by introspection under any circumstances.In correspondence,Thomas Nagel has advised me that what he wishes to defendis the following modal- ized versionof the argument: (1) Mymental states are knowableby me byintrospection. (2) Mybrain states are notknowable by me byintrospection. ..(3) Mymental states # mybrain states. Here Nagel will insist thatbeing knowable-by-me-by-introspection is a genuine relational propertyof a thing and that this versionof the argumentis freeof the intensional fallacydiscussed above. And so it is. But now the reductionistis in a position to insist that the argumentcontains a false premise: premise2. At the very least,he can insistthat (2) begs thequestion. For if mentalstates are indeed identical with brain states,then it is really brain statesthat we have been introspectingall along, though withoutappreciating theirfine-grained nature. And if we can learn to thinkof and rec- ognize those statesunder theirfamiliar mentalistic descriptions-as all of us have-then we can certainlylearn to thinkof and recog- nize themunder theirmore pentratingneurophysiological descriptions. Brain states,that is, are indeed knowable by introspection, and Nagel's argumentcommits the same errorinstanced below. (1) Temperatureis knowableby tactile sensing. (2) Meanmolecular kinetic energy is notknowable by tactile sensing. ..(3) Temperature# meanmolecular kinetic energy. Here the conclusion is known to be false. Temperatureis indeed mean molecular kineticenergy. Since the argumentis valid, it must thereforehave a false premise. Premise 2 is clearlythe stinker.Just as one can learn to feel that the summerair is about 70?F, or 21?C, so one can learn to feel that the mean KE of its molecules is about 6.2 x 10-21 joules; for,whether we realize it or not, thatis the prop- ertyour native discriminatorymechanisms are keyedto. And if one can come to know, by feeling,the mean KE of atmosphericmole- cules, why is it unthinkable that one might come to know, by introspection,the statesof one's brain? (What would thatfeel like? It would feel exactly the same as introspectingthe states of one's

101 believe it was Richard Brandt and Jaegwon Kim who firstidentified this fallacy specificallyin connection with the identitytheory, in "The Logic of the Iden- tityTheory," thisJO' RNAL, txiv, 17 (September1967): 515-537.

This content downloaded from 142.58.129.109 on Fri, 07 Aug 2015 19:14:45 UTC All use subject to JSTOR Terms and Conditions 22 THE JOURNAL OF PHILOSOPHY mind, since theyare one and the same states. One would simply employ a differentand more penetratingconceptual frameworkin theirdescription.) One must be careful, in evaluating the plausibility of Nagel's second premise, to distinguish it from the second premise of the veryfirst version of the argument,the version thatcommits the intensional fallacy. My guess is that Nagel has profitedsomewhat fromthe ambiguityhere. For, in the firstversion, both premisesare true.And in the second version,the argumentis valid. Neitherver- sion, however,meets both conditions. The matterof introspectingone's brain states will arise once more in the final section of this paper. For now, let us move on. Third Argument:The last argument here is the one most widely associated with Nagel's paper. The leading example is the (mooted) characterof the experiencesenjoyed by an alien creature such as a bat. The claim is that, no matterhow much one knew about the bat's neurophysiologyand its interactionwith the physical world, one could still not know, nor perhaps even imagine, what it is like to be a bat. Even total knowledge of the physical details still leaves somethingout. The lesson drawn is thatthe reductiveaspirations of neurophysiologyare doomed to dash themselves, unrealized, against the impenetrablekeep of subjectivequalia (cf. Nagel, 438 ff.). This argumentis almost identical with an argumentput forward in a recentpaper by Frank Jackson." Since Jackson'sversion deals directlywith humans, I shall confrontthe problem as he formu- lates it. IV. JACKSON'S KNOWLEDGE ARGUMENT Imagine a brilliant neuroscientistnamed Mary,who has lived her entire life in a room that is rigorouslycontrolled to display only various shades of black, white,and grey.She learns about the outside world by means of a black/white television monitor, and, being brilliant, she manages to transcendthese obstacles. She be- comes the world's greatest neuroscientist,all from within this room. In particular,she comes to know everythingthere is to know about the physical structureand activityof the brain and its visual system,of its actual and possible states. But therewould still be somethingshe did not know, and could not even imagine, about the actual experiences of all the other people who live outside her black/whiteroom, and about her pos-

" "Epiphenomenal Qualia," op. cit. Howard Robinson runs a verysimilar argument in Matterand Sense, op. cit., p. 4.

This content downloaded from 142.58.129.109 on Fri, 07 Aug 2015 19:14:45 UTC All use subject to JSTOR Terms and Conditions THE DIRECT INTROSPECTION OF BRAIN STATES 23 sible experienceswere she finally to leave her room: the nature of the experienceof seeing a ripe tomato,what it is like to see red or have a sensation-of-red.Therefore, complete knowledge of the physical factsof visual perceptionand its relatedbrain activitystill leaves something out. Therefore,materialism cannot give an adequate reductionistaccount of all mental phenomena. To give a convenientlytightened version of this argument: (1) Mary knows everythingthere is to know about brain states and theirproperties. (2) It is not the case that Mary knows everythingthere is to know about sensationsand theirproperties. Therefore,by Leibniz's law, (3) Sensations and theirproperties # brain statesand theirproperties. It is tempting to insist that we here confrontjust another in- stance of the intensional fallacydiscussed earlier,but Jackson'sde- fenders" insist that 'knows about' is a perfectlytransparent, en- tirelyextensional context. Let us suppose that it is. We can, I think,find at least two othershortcomings in this sortof argument. The First Shortcoming: This defect is simplicity itself. 'Knows about' may be transparentin both premises,but it is not univocal in both premises. (David Lewis'3 and Laurence Nemirow14 have both raised this same objection, though theiranalysis of the ambi- guityat issue differsfrom mine.) Jackson'sargument is valid only if 'knows about' is univocal in both premises. But the kind of knowledge addressedin premise 1 seems prettyclearly to be different fromthe kind of knowledge addressedin (2). Knowledge in (1) seems to be a matterof having mastereda set of sentencesor propo- sitions, the kind one finds writtenin neuroscience texts,whereas knowledge in (2) seems to be a matterof having a representationof rednessin some prelinguisticor sublinguisticmedium of representation forsensory variables, or to be a matterof being able to make certainsensory discriminations, or somethingalong theselines. Lewis and Nemirow plump forthe "ability" analysis of the relevant sense of 'knows about', but they need not be so narrowly committed,and the complaint of equivocation need not be so narrowly based. As my alternativegloss illustrates,other analyses of

"2See, forexample, Keith Campbell, "AbstractParticulars and the Philosophy of Mind," AustralasianJournal of Philosophy, LXI, 2 (June 1983): 129-141. 13"Postscript to 'Mad Pain and Martian Pain'," Philosophical Papers, vol. I (New York: Oxford,1983). 14Review of Thomas Nagel, Mortal Questions, Philosophical Review, LXXXIX, 3 (July 1980): 473-477.

This content downloaded from 142.58.129.109 on Fri, 07 Aug 2015 19:14:45 UTC All use subject to JSTOR Terms and Conditions 24 THE JOURNAL OF PHILOSOPHY

'knowledge by acquaintance' are possible, and the charge of equivocation will be sustained so long as the typeof knowledgeinvoked in premise 1 is distinctfrom the typeinvoked in premise2. Impor- tantly,they do seem very different,even in advance of a settled analysis of the latter. In short, the differencebetween a person who knows all about the visual cortex but has never enjoyed a sensation of red, and a person who knows no neurosciencebut knows well the sensation of red, may reside not in what is respectivelyknown by each (brain statesby the former,qualia by the latter),but ratherin the different type of knowledge each has of exactly the same thing. The differ- ence is in the manner of the knowing, not in the nature of the thing(s) known. If one replaces the ambiguous occurrences of 'knows about' in Jackson'sargument with the two differentexpan- sions suggested above, the resulting argument is a clear non sequitur. (a) Maryhas masteredthe completeset of truepropositions about people'sbrain states. (b) Marydoes not have a representationofredness in herprelinguistic mediumof representationfor sensory variables. Therefore,by Leibniz's law, (c) The rednesssensation # anybrain state. Premises a and b are compossible, even on a materialistview. But theydo not entail (c). In sum, there are prettyclearly more ways of "having knowledge" than having mastereda set of sentences. And nothing in materialism precludes this. The materialistcan freelyadmit that one has "knowledge" of one's sensationsin a way thatis independ- ent of the scientifictheories one has learned. This does not mean that sensations are beyond the reach of physical science. It just means that the brain uses more modes and media of representation than the simple storageof sentences.And this proposition is pretty obviously true: almost certainlythe brain uses a considerablevar- iety of modes and media of representation,perhaps hundreds of them.Jackson's argument, and Nagel's, exploit this varietyillegit- imately:both argumentsequivocate on 'knows about'. This criticismis supported by the observationthat, if Jackson's formof argumentwere sound, it would prove far too much. Sup- pose that Jackson were arguing, not against materialism, but against dualism: against the view that thereexists a nonmaterial substance-call it "ectoplasm"-whose hidden constitution and nomic intricaciesground all mental phenomena. Let our cloistered

This content downloaded from 142.58.129.109 on Fri, 07 Aug 2015 19:14:45 UTC All use subject to JSTOR Terms and Conditions THE DIRECT INTROSPECTION OF BRAIN STATES 25

Mary be an "ectoplasmologist" this time,and let her know1every- thing thereis to know about the ectoplasmic processesunderlying vision. There would still be somethingshe did not know2:what it is like to see red. Dualism is thereforeinadequate to account forall mental phenomena! This argumentis as plausible as Jackson's,and forthe same reason: it exploits the same equivocation. But the truthis, such arguments show nothing, one way or the other, about how mental phenomena mightbe accounted for. The Second Shortcoming: There is a furthershortcoming with Jackson's argument,one of profound importancefor understand- ing one of the most exciting consequences to be expected froma successfulneuroscientific account of mind. I draw yourattention to the assumption that even a utopian knowledge of neuroscience must leave Mary hopelesslyin the dark about the subjectivequalit- ative nature of sensationsnot-yet-enjoyed. It is true,of course, that no sentenceof the form"x is a sensation-of-red"will be deducible frompremises restricted to the language of neuroscience.But thisis no point against the reducibilityof phenomenological properties. As we saw in section I, directdeducibility is an intolerablystrong demand on reduction,and if this is all the objection comes to, then thereis no objection worthaddressing. What the defenderof emergent qualia must have in mind here,I think,is theclaim thatMary could not even imagine what the relevantexperience would be like, despite her exhaustive neuroscientificknowledge, and hence must still be missing certaincrucial information. This claim, however,is simply false. Given the truthof premise 1, premise2 seems plausible to Jackson,Nagel, and Robinson only because none of thesephilosophers has adequately consideredhow much one mightknow if,as premise 1 asserts,one knew everything thereis to know about the physical brain and nervous system.In particular,none of these philosophers has even begun to consider the changes in our introspectiveapprehension of our internalstates that could follow upon a wholesale revision in our conceptual frameworkfor our internalstates. The factis, we can indeed imagine how neuroscientificinforma- tion would give Marydetailed informationabout the qualia of various sensations. Recall our earlierdiscussion of the transformation of perception through the systematicreconceptualization of the relevantperceptual domain. In particular,suppose that Mary has learned to conceptualize her inner life, even in introspection,in termsof the completedneuroscience we are to imagine. So she does

This content downloaded from 142.58.129.109 on Fri, 07 Aug 2015 19:14:45 UTC All use subject to JSTOR Terms and Conditions 26 THE JOURNAL OF PHILOSOPHY not identifyher visual sensationscrudely as "a sensation-of-black", "a sensation-of-grey",or "a sensation-of-white";rather she identi- fies them more revealinglyas various spiking frequenciesin the nth layerof the occipital cortex(or whatever).If Mary has the relevant neuroscientificconcepts forthe sensational statesat issue (viz., sensations-of-red),but has never yet been in those states,she may well be able to imagine being in the relevantcortical state,and imagine it with substantial success, even in advance of receivingex- ternalstimuli thatwould actually produce it. One testof her ability in this regardwould be to give her a sti- mulus thatwould (finally)produce in her the relevantstate (viz., a spiking frequencyof 90 hz in the gamma network:a "sensation-of- red" to us), and see whethershe can identifyit correctlyon intro- spectivegrounds alone, as "a spiking frequencyof 90 hz: the kind a tomato would cause." It does not seem to me to be impossible that she should succeed in this,and do so regularlyon similar testsfor other states, conceptualized clearly by her, but not previously enjoyed. This may seem to some an outlandish suggestion,but the following will show that it is not. Musical chordsare auditoryphen- omena that the young and unpracticed ear hears as undivided wholes, discriminableone fromanother, but without elementsor internal structure.A musical education changes this, and one comes to hear chordsas groups of discriminablenotes. If one is suf- ficientlypracticed to have absolute pitch, one can even name the notes of an apprehendedchord. And the revereseis also true:if a set of notes is specifiedverbally, a trainedpianist or guitaristcan iden- tifythe chord and recall its sound in auditoryimagination. More- over,a reallyskilled individual can construct,in auditoryimagination, the sound of a chord he may neverhave heard before,and cer- tainly does not remember.Specify for him a relativelyunusual one-an F#9thaddl3thfor example-and let him brood for a bit. Then play forhim threeor fourchords, one of which is the target, and see whetherhe can pick it out as the sound that meets the description. Skilled musicians can do this. Why is a similar skill beyond all possibilityfor Mary? "Ah," it is temptingto reply, "musicians can do this only because chords are audibly structuredsets of elements. Sensations of color are not." But neither did chords seem, initially, to be structuredsets of elements. They also seemed to be undifferentiatedwholes. Why should it be unthinkablethat sensations of color possess a comparable internal structure,unnoticed so far, but awaiting our deter-

This content downloaded from 142.58.129.109 on Fri, 07 Aug 2015 19:14:45 UTC All use subject to JSTOR Terms and Conditions THE DIRECT INTROSPECTION OF BRAIN STATES 27 mined and informedinspection? Jackson's argument, to be successful, must rule this possibilityout, and it is difficultto see how he can do this a priori. Especially since therehas recentlyemerged ex- cellent empirical evidence to suggest that our sensations of color are indeed structuredsets of elements. The retinex theoryof color vision recentlyproposed by Edwin Land"5 representsany color apprehendable by the human visual systemas being uniquely specifiedby its joint position along three vertices-its reflectanceefficiences at three critical wavelengths, those wavelengthsto which the retina's triunecone systemis selec- tivelyresponsive. Since colors are apprehended by us, it is a good hypothesisthat those threeparameters are representedin our visual systemsand thatour sensations of color are in some directway de- terminedby them. Sensations of color may turnout literallyto be three-elementchords in some neural medium! In the face of all this, I do not see why it is even brieflyplausible to insist that it is utterlyimpossible fora conceptuallysophisticated Mary accurately to imagine, and subsequentlyto pick out, color sensationsshe has not previouslyenjoyed. We can already foreseehow it mightactu- ally be done. The preceding argument does not collapse the distinction(be- tween knowledge-by-descriptionand knowledge-by-acquaintance) urged earlier in the discussion of equivocation. But it does show thatthe "taxonomies" thatreside in our prelinguisticmedia of rep- resentationcan be profoundlyshaped by the taxonomiesthat reside in the lingustic medium, especially if one has had long practice at the observational discrimination of items that answer to thoselinguistically embodied categories.This is just a furtherillus- trationof the plasticityof human perception. I do not mean to suggest,of course, that therewill be no limits to what Mary can imagine. Her brain is finite,and its specific anatomy will have specific limitations. For example, if a bat's brain includes computational machinery that the human brain simply lacks (which seems likely), then the subjectivecharacter of some of the bat's internalstates may well be beyondhuman imagi- nation. Clearly,however, the elusivenessof the bat's inner life here stemsnot fromthe metaphysical"emergence" of its internalqualia, but only fromthe finitecapacities of our idiosyncraticallyhuman brains. Within those sheerlystructural limitations, our imagina- tions may soar far beyond what Jackson, Nagel, and Robinson

""The Retinex Theory of Color Vision," ScientificAmerican (December 1977): 108-128.

This content downloaded from 142.58.129.109 on Fri, 07 Aug 2015 19:14:45 UTC All use subject to JSTOR Terms and Conditions 28 THE JOURNAL OF PHILOSOPHY suspect, if we possess a neuroscientificconceptual frameworkthat is at last adequate to the intricatephenomena at issue. I suggest then,that those of us who prize the flux and contentof our subjectivephenomenological experienceneed not view the advance of materialisticneuroscience with fearand foreboding.Quite the contrary.The genuine arrival of a materialistkinematics and dynamicsfor psychological statesand cognitiveprocesses will constitute not a gloom in which our inner life is suppressed or eclipsed, but rathera dawning, in which its marvelous intricacies are finallyrevealed-most notably,if we apply ourselves,in direct self-consciousintrospection. PAUL M. CHURCHLAND Universityof California/San Diego

BOOK REVIEWS

Perception: A Representative Theory. FRANK JACKSON. New York: Cambridge UniversityPress, 1977. x, 180 p. $27.50. Frank Jackson's admirable book forcefullyargues that all visual experiences involve seeing sense data, that sense data are mental objects,and that seeing a material object consists in seeing a sense datum that,in a sense Jacksondefines, "belongs to" that material object. These claims constitute,according to Jackson,the core of a representativetheory of visual perception. Jackson argues in the course of chapter III that,like visual sense data, bodily sensations are mental objects; otherwise he confines his attention to visual perception.But his argumentswould verylikely adapt successfully to othersensory modalities. Jackson's conclusions will strikemany readersas somewhat ex- travagant.Nonetheless, his argumentsfor these conclusions are in- variablymodels of clarity,conciseness, and cogency. He builds his case for sense data and representationalismwith meticulous care, and helpfullydirects attention to the connectionsamong the various steps in his over-allargument. Accordingly, I firstdescribe the structureof that argument. I then turn to several steps that seem especially open to objection. Finally, I sketch a way one might enjoy the advantages of Jackson'saccount while avoiding its more extravagantaspects.

0022-362X/85/8201/0028$01.40 ? 1985 The Journalof Philosophy, Inc.

This content downloaded from 142.58.129.109 on Fri, 07 Aug 2015 19:14:45 UTC All use subject to JSTOR Terms and Conditions