naturalizing the mind

7/29/2019 Naturalizing the Mind

1/193

JeanNicod CommitteeMario BorilloJean-Pierre ChangeuxClaude DebruJean- GabrielGanasciaMichellmbert

Series Foreword

The JeanNicod Lectures are delivered annually in Parisby aleading philosopher of mind or philosophically orientedcognitive scientist. The 1993inaugural lectures marked thecentenary of the birth of the French philosopher and logi -~ JeanNicod (1893- 1931). The lectures are sponsored bythe Centre National de la RechercheScientifique (CNRS) aspart of its effort to develop the interdisciplinary field of cognitivescience n France. The series hosts the texts of the lec-

PierreJacobJacquesMehlerPhilippede RoullhanDanSperber

tures or the monographs they inspire .

CNRSFran~ois RecanatiSecretaryof the JeanNicod Committeeand Editorof the Series

JacuesBouveressePresidentof the JeanNicod CommitteeAndre Holley, Secretary f the CognitiveScienceProgram


2/193

I am grateful to CNRS, The Centre National de la RechercheScientifique of France, for inviting me to give these lecturesin Paris in the spring of 1994 at the second annualConference JeanNicod de Philosophie Cognitive . Withoutthe stimulus that this 'conference provided - actually , thethreat of embarrassing myself before a French audience -these ideas would probably never have been assembled insuch a.timely way. They might not have been assembled atall .My weeks in Paris were made particularly memorable -

both socially and philosophically - by a group of philosophersand scientists , faculty and students alike (most ofthem , I think , at CREA , Centre de Recherche enEpistemologie Appliquee ), who were not only gracioushosts,. but challenging critics after each lecture . It was ahumbling experienceto find myself arguing technical issuesin the philosophy of mind with Frenchphilosophers - in themiddle of Paris no less- in English. Imagine what theywould have done to me in French. I particularly want tothank Pierre Jacob for everything he did - and there weremany things - to make my visit sopleasant.

Acknowledgments


3/193

Thanks to the Stanford Humanities Center I was able todevote almost the full year to preparing these ectures. I wasa fellow there during the year 1993- 1994 and I am deeplygrateful to the support I received from the director, WandaCorn, the associatedirector, CharlesJunkerman, and the fellowsand staff. No oneeverknocked on my door. They obviouslyknow how to run a research nstitute .

I really started working on these ectures (without realizingit ) in 1991 when I was preparing talks to give inValencia, Spain. The Valencia lectures were early stabs atthesetopics - so early, in fact, that I did not want to publishthe results. I wish to thank the philosophers in Spain, andespecially JosepCorbi, for their hospitality and helpful discussion.

Students and friends read some---and in a few casesall-of these lectures and gave me useful comments and criticism. I want to thank Sven Bemecker, and Tll11Schroederhere at Stanford, Murat Aydede now at Chicago, and EricSchwitzgebel at Berkeley for their help. They made more ofa difference than might be apparent to them . DavidRosenthal read most of the lectures and provided usefulcomments - especially on chapter 4. Berent Enc, an oldfriend from WISCOnsingave me months of grief by not lettingme ignore problems I didn ' t know how to solve Hemade me start thinking about chapter 5 two years before Igot around to writing it . Georges Rey not only read themanuscript- several times as far as I can tell- but gave mea truly astonishing number of useful suggestions and criti -cisms. Every author should be so fortunate . I am deeplygrateful to him . I tried, as best I could, to respond to his crit -icisms in final revisions, but some of our differences wentfar too deep. To keep the book a reasonableapproximationto the lectures I gave in Paris, I had to ignore many of

Acknowledgments


4/193

Georgess most perceptive and challenging questions. Sorryabout that. Wewill , I'm sure, comeback to them.Finally , I would like to thank Giiven Giizeldere . Hearrived at Stanford a few years ago and promptly got meworrying about his problems. I thought it was supposed togo the other way ar.ound. No matter. We are now both wor -ried about them. I thank him for making me realize, patientlyand .~ fu great insight , how really hard these problemsare.

Acknowledgments


5/193

Prologue

The purpose of these lectures is to promote a naturalistictheory of the mind - something I call the RepresentationalThesis. The thesis, in two parts, is that, plus or minus a bit )(1) All mentalfactsarerepresentationalfacts, and (2) All representationalfactsarefactsaboutinformationalunctions. The reasonI am interested in this thesis is that, as far as I can tell , itis the only approach to the topic of consciousness hat hasmuch to say about the baffling problems of phenomenalexperience. It does not, to be sure, remove all the mysteries.It removes enough of them, though, to justify putting one'smoney on the nose of this philosophical horse. That, at least,is where my money is. These ectures are an attempt to convinceothers that the smart money belongs there.

Representational Naturalism (as I shall call the viewdefined by the Representational Thesis) helps one understand, for example, why conscious experiences have thatpeculiar diaphanous quality - the quality of always beingpresentwhenbut never whereone looks to find them. It providesa satisfying account of the qualitative , the first -person,aspect of our sensory and affective life - distinguishing , innaturalistic terms, between what we experience(reality) andhow we experience it (appearance ). In providing this


6/193

Prologueaccount, it establishes a framework within which subjectivity

can be studied objectively. It demystifies introspection :the mind 's knowledge of itself no longer requires an internal"eye" observing the clockwork of the mind . And it providesan answer- a biologically plausible answer- to questionsabout the function or purpose of consciousnessThese benefits, and more, derive from conceiving of the mind as therepresentational face of the brain.A working premise behind the RepresentationalThesis isthat a better understanding of themind is not to be obtainedby knowledge - no matter how detailed and precise - of thebiological machinery by means of which the mind does itsjob. Information about the wetware may be useful, but it isnot enough. If you don't know what a camera is, it is of nohelp to be told about I

-stops, focal lengths, shutter speeds,and ASA numbers . What you need to know is somethingmore basic, something about pictures , something aboutwhat camerasdo. Without that you don't know what a camerais no matter how much you know about the machineryby means of which cameras do their job. Since the manipulationand use of representations is the primary job of the

mind , a deeper understanding of the nature of representationand its naturalistic basis is, perforce, a deeper understandingof the mind .The Representational Thesis is plausible enough for thepropositional attitudes - belief , thought , judgment , and thelike . I have given my own account of the propositional attitudes- particularly belief and desire - in Oretske (1988).The thesis is less plausible - some would say completelyimplausible - for sensory affairs , for the phenomenal orqualitative aspectsof our mental life. Nonetheless, in theselectures I concentrate on perceptual experience. The topic isqualia- that dimension of our conscious life that helps to

xiv


7/193

define what-it -is-like-to-be us. I focus here becausefrankly ,this is where progress is most difficult . This, then,. s whereprogress - if there is any- will be most significant .There is much that is relevant to my topic that I do notdiscuss. These lectures are already too long . I do not, forexample, examine proprioception - awareness of one's ownbodily statesand processes- even though this is the sourceof some of our most obtrusive experiences (pain , hunger,thirst , etc:). What little I do say about it occurs in chapter 4, 1. This is an omission that I think could be removed withoutfundamental alterations to the explanatory machinerydeveloped in chapter 1. There are, however, still other experiences- a general feeling of depression , forexample -about which I do not know what to say. That is the purposeof the " plus or minus a bit

" in the statement of theRepresentationalThesis.. The first four chapters of this book are the Nicod lecturesI gave in Paris in ~ ay 1994In these lectures I concentratedon the positive aspects of the Representational Thesis, onwhat it could tell us (if it were true) about self-knowledge,qualia, inverted spectra, consciousnesspoints of view, andthe possibility of knowing what it was like to be a bat or adogfish. There are, however, troublesome aspects o a representational

approach to mental phenomena, problems that Ideliberately ignored in these lectures This is especially sowhen representational ideas are given a naturalistic spin-seepart (2) of the RepresentationalThesis - and applied, asIapply them, to senseexperience. Many - perhapsmost -philosophers are convinced that this is an areawhere representationalideas are leastapplicable. Even if thought , belief,and judgment can be understood as internal representationsof external affairs, sensationsexperiencesand feelings cannot. Unlike the conceptual side of oUr mental life , experi-

Prologue


8/193

ences have a phenomenal , a what -it -is-like , quality thatdefies representational (not to mention naturalistic ) treatment. There are, furthermore , powerful intuitions - sometimeseven arguments - that support this skepticism . Icould not, therefore, publish these ectureswithout addressingthese doubts . I do so in chapter 5. This concludingchapter representsmy answer to thosewho will think , afterread.ng chapters 1- 4, that the virtues of representationalnaturalism are all well and good, but . . . well . . . it just can'tbe true. I think it can.

Prologue XVI


9/193

The RepresentationalCharacter of SenseExperience

Senseexperience is the primary locus of consciousness.Marcel (1988p. 128) thinks that without it we would nothave a conceptof consciousnessat all . Whether or not this istrue, phenomenal experience - the look , sound, taste andfeel of things - dominates our mental lives. Remove it completelyand one becomes. . . what?A zombie?If , in accordance with the Representational Thesis, wethink of all mental facts asrepresentational facts, the qualityof experience, how things seem to us at the sensory level, isconstituted by the properties things are representedas having. My experience of an object is the totality of ways thatobject appearsto me, and the way an object appearsto me isthe way my sensesrepresent t .Since not all facts about representations are representationalfacts, and not all representations are mental, I beginby making some relevant distinctions among kinds of representations. Experienceis a specialkind of representation- a

nonconceptual for In of representation . The first order ofbusiness, then, is to say what kind of special representationthis isI conclude the chapter with a short preview of theexplanatory benefits of a representational approach to per-

1


10/193

1 The.Nature of Representation

Chapter

ceptual experienceIn 5 I describe he way it naturalizesintentionality and in 6 the way it provides a satisfyingaccount of why, though experiencesare in the head, onedoesnt find them there. Other explanatory benefits arereservedor laterchapters

The fundamental idea is that a system, 5, representsa property, F, if and only if 5 has the function of indicating (providinginformation about) the F of a certain domain of objects.1The way performs its function (when it performs it ) is byoccupying different statessl' s2 . . . sncorresponding to thedifferent determinate values f1, f2, . . . fn, of F. A speedometer(5) representsthe speed(F) of a car. Its job, its function , is toindicate, provide information (to the driver ) about, how fastthe car is moving (F) . When it is doing its job, its differentstates (pointer positions "24," "37," etc.) correspond to differentcar speeds(24 mph, 37 mph, etc.). Given the functionof this instrument , each of its states is supposed to carry adifferent piece of information about the speed of the car: aregistration of "37" is supposed to carry the informationthat the car is going 37 mph, "24" the information it is going24 mph , and so on. The fact that the speedometer has aspeed indicating function , and the fact that pointing at "37"means 37 mph are representational facts about the instrumentand this state of the instrument . This is what theinstrument was designed to do, what it is supposed to do,and, like any fallible system, it can fail to do what it is supposedto do. If a registration of "37" on a properly installedinstrument fails to carry information about the speed of thecar, or carries the same information that a registration of"24" carries, then it is not doing its job . The result , oftenenough, is misrepresentation.


11/193

ExperienceThe fact that the speedometer s connected to the axle by

a cable that transmits information about speed, on the otherhand, is not a representational fact about this instrument . Itis a fact about a representational system, but it is not a representationalfact. The instrument couldn 't do its job withoutthis cable, but the fact that there is such a cable does notimply that the device hasa job - let alone has the job of providinginf~rmation . For the same reason, the fact that a particularthermometer is filled with mercury, a metal whosevolume serves to indicate temperature, is a fact about a representationadevice, but it is not a representational fact. Arepresentational fact about S is a fact about what S isdesigned to do, a fact about what information it is supposedto carry. There are facts about representations - facts abouttheir color, shape, material constitution , and mode of operation- that do not tell one anything about what informationthey are supposed to supply or, indeed, whether they aresupposedto supply information at all .On the representational account of mind I intend to give,the difference between representational facts and (merel)facts about representations is the difference between themind and the brain . Neuroscientists may know a greatmany facts about the brain . These actsmay even turn out tobe facts about mental representations - facts about experiencesand thoughts. That doesnot make knowledge of suchfacts knowledge of the mind . Knowledge of the mind , ofmental facts, is, according to the Representational Thesis,knowledge of representational facts, not (merely) facts aboutmental representations . One does not understand moreabout the representational life of a system by being told itcontains mercury or has an orange pointer . Nor does oneknow what (or whether) a system represents by being toldthat it supplies information about speed. It is not whether it

The RepresentationalCharacter of Sense


12/193

supplies infonnation about speed that is important ; it iswhether it has the function of doing so.Representation s here being understood to combine tele-ological with infonnation -theoretic ideas. If the concept ofrepresentation is to do a useful job in cognitive scienceif itis to be used, in particular , to illuminate the nature ofthought and experience, it must be rich enough to allow formisreP.resentationIt must include the power to get thingswrong , the power to say that something is so when it is notso. This is what the teleology, the idea of something havingan information -carrying function , is doing in the presenttheory. It captures the normative element inherent in theidea of representation. Sincean object can retain a functioneven when it fails to perform it (think of the heart, the kidneys

, and damaged instruments ), a device can retain itsindicator function- continue to represent(i .e., misrepresent)something asgoing 34mph, for instance- even when thingsgo wrong , even when it fails to provide the information it isits job to provide . There is information without functions ,but there is no representationwithout functions.Not all events that carry information have the function ofcarrymg it . To use Matthen's (1988) example, the angle acolumnof smoke makes with the horizon carries informationabout wind speed, but that, surely, is not the function of thesmoke. The smoke cannot misrepresent wind speed. Wemay be misled by the column of smoke. We may take it toindicate something it doesn't (thus ourselves misrepresentingwind speed), but the smoke does not misrepresentwindspeed the way an anemometer can. This , incidentally , iswhy black-and-white television does not misrepresent thecolor of the sky while color television sometimes does. Thepower of the color television to misrepresentcolor lies in itscolor-depicting function . The images on the two screenscan

Chapter


13/193


be identical; yet, one misrepresents, the other does not . Thisis the difference between not representing color and misrepresentingcolor.Something like this conception of representation is, Ithink , operative in many scientific approaches to mental

representation. The terminology is not always the same, butthe intuitions and theoretical purposes are similar . AnneTreisman (1992p. 227), for example, speaks of representationsas signals or events which have been assigned the"burden " (function ?) of carrying meaning (information ?).David Marr 's (1982) well -known theory of vision assignsto

early vision the function or task (Shapiro 1993) of depictingthe spatial and chromatic properties of distal scenes.Gallistel (1990) describes he brain asrepresenting aspectsofthe environment when there is a functioning isomorphismbetween the environment and the b~ain process that"adapts" the animal 's behavior to it . Sensory organs andmechanismsare commonly described in terms of what theyare "for." The semicircular canals of the middle ear are saidto be for the detection of angular acceleration, the utricleand saccule or indication of linear acceleration, and the retinafor encoding information about light for transmission tothe brain . This way of describing perceptual mechanismsand processes is a representational way of thinking aboutthem. The sensesyield representationsof the world , not justbecause they (when working right ) deliver informationabout the world , but becausethat is their job. The sensesand the states they produce by way of performing theirfunction , are thus evaluable in terms of how well they dotheir job.Before applying these representational ideas to mentalaffairs, though , we need to know, in more precise terms,exactly what kind of representation a.mental representation


14/193

Chapteris supposed to be. If it is, indeed, the job of the various senses

to provide information about the world - thus, on thepresent account , representing that world - what is thesource of these functions, what information is it their job toprovide , and how- in representational terms - do our experiencesof objects differ from our thoughts about them ?What, furthermore , is it about certain representations thatmakes. he systemsin which they occur conscious f what isbeing represented? Good questions. I begin my answers byexplaining a number of pivotal distinctions : (1) natural vs.conventional representations; (2) representational statesvs.representational systems; and (3) represented properties vs.represented objects. Armed with these distinctions we cangive a preliminary taxonomy of mental representation andbegin the argument that consciousexperience s a speciesofnatural representation.2 Natural and Conventional RepresentationsThe function of a system (or state) is what it is designed todo - what it is, by design, supposed to do. There are different

sourcesof design. Each gives rise to a different kind offunction and, thus, a different form of representation. Oneimportant difference (for our purposes) is the differencebetween naturally acquired and conventionally assignedfunctions - hence, the difference between natural and conventiona

representations.The information -providing functions of measuringinstruments , sensors, detectors and gauges are functionsthey get from us - their makers and users. We design them.We give them a job to do. We arrange things so that certainliquids , by their placement in transparent tubes adjacentto acalibrated scale, provide us with iri formation about temper-


15/193


ature . We call the resulting artifacts thermometers . Flagpoles and (metal) paper clips, the volumes of which are proportionalto temperature, also carry information about temperature. Their volume is also reliably correlated with temperature. That, though , is not their function . That is notwhat they are designed to do. As the names suggest, theyhave quite a different job. Though they carry the same nformation, flag poles and paper clips do not represent whatthermometers represent. They do not represent anything .We haven't given them that kind of job.When a thing 's informational functions are derived fromthe intentions and purposes of its designers, builders , andusers in this way, I call the resulting representationsconventional. Representations hat are not conventional are natural.I assume that there are naturally acquired functions and,thus , natural representations .3 I do not argue for this ; Iassume it . This view , I know, is not universally accepted(Dennett 1987especially pp . 287- 321and Searle 1992p. 52,for example, deny it ). I return to this point in chapter 5, butfor the present I follow the lead of Wright (1983, 1987),Kitcher (1993), Godfrey -Smith (1994), Millikan (1984),Neander (1991a1991b, Papineau (1993), Bennett (1976) andmany others in supposing that bodily organs and mechanismscan, in the relevant sensebe designed to do a certainjob - and, thus, have the function of doing it- without beingdesigned by anyone to do it . Philip Kitcher (1993p. 380)puts it thus: "one of Darwin 's important discoveries is thatwe can think of design without a designer." The sensesIassume, have information -providing functions , biologicalfunctions, they derive from their evolutionary history.4As aresult, perceptual (including proprioceptive ) systems produce

representations of those conditions (external or internalas the casemay be) they have the function of informing


16/193

about. The representationsthey produce by way of carryingout their informational functions have a content, somethingthey say or mean, that does not depend on the existence ofour purposes and intentions . This is why the senses- or,more precisely, the internal states(experiencesfeelings) thesensesproduce by way of performing their function- haveoriginal intentionality , something they represent , say, ormean, that they do not get from us. That is why the perceptual

representations in biological systems- unlike those inlaptop computers, speedometersand television sets - makethe systems n which they occur consciousof the objects theyrepresent.5We have, then, the following preliminary classification.

Representational systemsSystemshaving indicator flmctions

NaturalIndicator unctionsarenatural

The above classification is crude. All mental representations(not to mention nonmental natural representation s6are

Otapter1

. . . . . . . . . ." " . .. . . .

. .. . . . . .~" " ' " " " " " " " " " " " ' " " "

,

ConventionalIndicator functionsare conventionalInstrumentsLanguage Sensory systemsConcepts

Hereafter, when speaking of the RepresentationalThesis, Iwin mean the thesis that all mental statesare natural representation. This makes the thesis a form of philosophicalnaturalism .3 Representational Systemsand Representational States


17/193

classified together. Thoughts and beliefs are classified withexperiencesThis is correct, as far as it goes. Both are formsof natural representation . For our purposes, though , it isimp OrtaI1t to distinguish seeing and hearing from knowingand believing . I can seePaul playing the piano and believehe is playing the piano, but the visual experience representsthe piano playing in much different ways than does thebelief . These are different kinds of mental representation.One can seeor hear a piano being played without believinga piano is being played, and one can believe a piano is beingplayed without seeing or hearing it being played. Seeingapiano being played is constituted, in part, by a visual experience, hearing by an auditory experience. Until theseexperiencesoccur one has not seen or heard the piano .7Experiencesof piano playing do not require the conceptof apiano (at least not in the same way as a belief or judgmentrequires it ). They require no understanding of what a pianois or what it sounds like . Even mice can see and hear pianosbeing played.

Believing is something else. It requires the concept of apiano, some understanding of what a piano is. Mice whohear pianos being played do not believe pianos are beingplayed . Their understanding is, I assume, too feeble tobelieve this even though their hearing is good enough tohear it .

All representations are representations of (purported )fact, but not all such representationsare conceptual epresentations. A conceptual awarenessof facts - a belief, judgment ,or knowledge that the toast is burning - has a close tie withbehavior. For those who have language, it normally bringswith it an ability to saywhat one is aware of- that the toastis burning . This is not so with sensory awarenessOne cansee or smell (and, thus, be perceptually aware of) burning



18/193

toast while having little or no understanding of what toast isor what it means to burn . "What is that strange smell ?"might be the remark of someonewho smells the toast burningbut is ignorant of what toast is and what it means toburn . A mouse in the kitchen (the one who heard the pianobeing played) can smell, and thus have sensory awarenessof burning toast but it will not (like the cook) be aware (i .e.,believe) that toast is burning . The mouse will have little orno conceptual awarenessof this event. It will smell the toastburning but not smell that it is, not smell it as, burning . Thecook and the mouse differ in what they think about whatthey smell. They differ in how they regard the smell. Theymay also differ in how the burning toast smells to them. Butthey both smell it . They both experienceit . They both havesome kind of sensory representationof this event.

8The difference between experiences of k (as F) and

thoughts about k (that it is F)- between sensoryand conceptualrepresentations of k- is generally clear enough whenwe describe ourselves as consciousof concreteobjects (e.g.,burning toast) and events (e.g., piano playing ). When, however, we start describing ourselves as being aware ofabstract objects - differences, numbers, answers, problems,sizes, colors - an ambiguity appears. I take a moment, therefore, to remark on this ambiguity since it would otherwisemuddy the discussion to follow .When we use an abstract noun or phrase to describewhat we seehear, or feel- what we are aware or consciousof- what is being described is normally a conceptual awareness

of some (unspecified ) fact. The abstract noun phrasestands in for some factive clause (Dretske 1993). Thus, todescribe someoneas seeing (being consciousof) the differencebetween A and B is to imply that the person sees(isconscious) that they differ . To describe someone as being

Chapter


19/193

aware of the color of his (blue) shirt is to imply that he orshe is aware that the shirt is blue - thus representing thecolor in some conceptual way (asblue, the color of the sky,etc.) It would be odd to describe a person as seeing (thusbeing aware) of the color of his shirt if the person did not, atsome conceptual level , know what color the shirt was .Likewise, to be aware of the problem it isn't enough to see(experience) the thing that is the problem (e.g., a cloggeddrain). One has to see(the fact) that it is clogged. One has toconceptualize what one seesthe clogged drain , as aproblem. Until one conceptualizes it this way, one is not (as wesay) aware of the problem.These differences are important in thinking about waysof representing properties and, thus, according to a representational

theory, the properties we are consciousof. Sincethe topic is our experienceof objects, not our beliefs aboutthem, we must be careful not to describean experienceof ashirt 's color as an awareness or consciousnessof the shirt 'scolor. For this form of words implies9a conceptual representationof the shirt 's color, and this may not be present. Achild or an animal might be visually aware of the shirt 'scolor (their visual experienceof the shirt being, as they say,suffused with blueness) without their knowing or thinkingthat the shirt is blue - without sorting (or having any dispositionto sort) the shirt with other blue objects. If this is hardto imagine, think of the perceiver asan animal. Cats are notcolor blind just because they ignore differences in color.toOne can experience blue (the shirt 's color )- and, in thissensebe aware of blue - without being conceptually awarethat anything is (or looks) blue.I will try to keep thesematters straight by distinguishingbetween sensory and conceptual representations of facts,between experiences of k's blueness and beliefs or judg -



20/193

ments that k (or something ) is (or looks) blue . The word"phenomenal" is often used to describe this sensory modeof awareness, and I will sometimes use it . Phenomenalawareness s a mode of awareness that does not require -although it may, in fact be accompanied by- conceptualawarenessOne can be phenomenally conscious of a shirt 'scolor, of a piano being played, and of burning toast withoutbeing ~onscious that anything is blue, that a piano is beingplayed, or that something is burning .What, then, in representational terms, is the basis of thisordinary - and, I hope, familiar - distinction between anexperienceof color, shape, and texture, and a belief or judgmentabout color, shape, and texture? In representationalterms, what is the difference between an olfactory experienceof toast burning and a belief or judgment that toast isburning ?

Experiences (sensations) of burning toast and beliefs(thoughts) that toast is burning are representations, and allrepresentations are particular (token ) states or events .Nonetheless, token states have two different sources fortheir indicator functions . (1) A statemay derive its indicatorfunction- and , hence, its representational status - from thesystem of which it is a state. Call these systemic ndicatorfunctions (= functionss) and the representations they giverise to systemic representations(= representationss. If a system(e.g., a thermometer) is supposed to provide informationabout temperature, and B is the state (e.g., mercury atsuch-and-such level) that is supposed to carry the informationthat the temperature is, say, 32, then Bhas the systemicfunction (functionS>of indicating a temperature of 32 . StateB therefore representss he temperature as being 32 . (2) Atoken state may, on the other hand, acquire its indicatorfunction , not from the systemof which it is a state, but from

Chapter


21/193

Experiencethe typeof state of which it is a token. No matter what B systemically

represents (what it is, by design of the system,supposed to indicate), it might acquire (or be given) a special, or a different , indicator function . If we print the number"381at this point on the scale, then each and every time themercury reaches this point , it meanshat the temperature is38 . If we print the word I/DANGERI / at this (and higher)points , the mercury 's rising to this point signifies danger.Call such.functions (functions that state types are assignedor acquire independently of their - if any- systemic functions) acquiredndicator functions (= functionsa)' The indicatorfunctionsa (hencerepresentationsa) of token statesmaybe different from their systemic functions (what they represents)' As we shall seea token statemight represent in bothsystemic and acquired ways, and what it representsaneednot be what it representss. An example should make this difference clear. Supposewe have a simple speedometer mechanism that representsvehicle speed by registering the rotation of the axle. Thismechanism was designed to be used in cars equipped withdifferent sized tires. Since(at a given speed) the axle rotatesmore slowly with large tires than with small tires, the manufacturerleft the job of calibrating the face of the instrumentup to the user. If I use the instrument in a car with normaltires, I calibrate the dial in one way. If you use larger tires,you calibrate it differently . I put the number 1501at positionB of the pointer. You put the number 1601there. When ouraxles are turning at a rate corresponding to a pointer position

of B, my car is going 50 mph, your car is going 60 mph .So we calibrate differently by assigning different numbers tothis pointer position . Pointer position B, this state of the system, has the same systemic function in both cars - that ofindicating an axle rotation rate of N rpm . It nonethelessvia

The RepresentationalCharacter of Sense


22/193

different calibration, has a different acquired function in ourtwo cars. In my car stateB representsa50 mph, in your car itrepresentsa60 mph. StateB has the sameindicator functio~in all cars; what it representsa about speed), however, variesfrom car to car.

To suggest (by way of tendentious description) the waythis distinction is to be applied, we can describe individualsystems in which this "perceptual" mechanism is installedashaVing the same"experiences" (viz., of an axle rotation ofN) but as having different "belief 's" (about speed). B representss(i .e.,. is an experienceof) the samething in all cars: anaxle rotation of N . That is what this statemeans, what it rep-resentssin both your car and my car. Given the information(about axle rotation) this system has the function of supplying

, what it was designed (at the factory) to do, this particularstateof the systemhas the systemicunction of indicatingan axle rotating at N rpm . Nonetheless, although this staterepresentssthe same thing in both systems, it representsasomething different in my car than it does in your car. It rep-resentsa50 mph in my car, 60 mph in your car. Once calibrated, my speedometeras it were, "sees" an axle rotationof N rpm as a speedof 50mph . Your speedometerdifferentlycalibrated, "sees" the samething- an axle rotation of N-but sees it as 60 mphOurspeedometers have the same"experience " but the experience gives rise to different"beliefs ."Walker (1983p. 246) describes wo of Pavlov's dogs. Oneis conditioned to salivate when middle Cisplayed on anyinstrument . The other is conditioned to salivate to a clarinetplaying any note. Now imagine the two dogs hearing aclarinet

playing middle C. They hear the same sound . Theirexperience of this sound may well be the same. Theirresponse s also the same: both salivate. Yet, their responses

Chapter


23/193

are mediated by different acquired representations . As aresult of different learning, the dogs hear it differently - one(as we, not the dogs, might put it ) hears it as middle C, theother as the sound of a clarinet . The way their experiencerepresentss he sound may well be the same, but the waytheir experiencerepresentsa t is different .As my description of theseexamples is intended to suggest, experiencesare to be identified with stateswhose representationalproperties are systemicThought (conceptualstates n general), on the other hand, are states whose representationalproperties are acquiredAs a result, experienceshave their representational content fixed by the biologicalfunctions of the sensory systems of which they are states.l1How an experience representss the world is fixed by thefunctions of the system of which it is a state. The quality ofa sensory state - how things look , sound, and feel at themost basic (phenomenal) level- is thus determined phylo-genetically. Sincewe inherit our sensory systems, since theyare (at a fairly early age, anyway ) hard-wired , we cannot(not easily anyway) change the representationals characterof experience .12Through learning , I can change what Ibelieve when I see k, but I can

't much change the way klooks (phenomenally ) to me, the kind of visual experiencekproduces in me. Experiences are, for this reason, modularin Fodor 's (1983) sense. The way a belief represents theworld , on the other hand, is ontogenetically determined .We can, through learning , change our calibration . We canchange what we seesomething aswhat we, upon seeing

it , take it to be - even if we cannot, not in the same way,change what we see. This is why a representationsof k asred (a sensation of redness) is different from a representa-tiona of k as red (a belief that k is red) even though both arerepresentationsof k as red.



24/193

Another manufacturer produces a more sophisticatedspeedometer. It is designed to be used on all cars, no matterwhat size tires they use. Pointer positions are determined bytwo sources of information : not just by the rate at which theaxle is rotating, but by the height of the axle above the roadsurface. Sincethe height of the axle above the road surfaceprovides a useful measure of tire size, these two sourcesofinformation are combined to yield a reliable measure of carspeed no matter what size tires are used. The pointer positionsof these instruments - driven by two sourcesof information- thus have a speed-indicating systemic functionand are therefore calibrated at the factory. As their construction

suggests, these more sophisticated instruments aredesigned to provide information , not about how fast theaxle is rotating , but about how fast the car is going .Although the information -handling processes n thesefancyinstruments use information about axle rotation to generatea representation of speed, the final representation (pointerposition ) does not , in fact , indicate how fast the axle isrotating . The early phase in this informational processusesinformation about axle-rotation , but (by integrating it withinformation about elevation) it sacrificesthis information inorder to produce a final indication of speed. Informationabout axle rotations is thereby lost . This is an instrument

ChapterThe above example is a bit too simple-minded to reveal

the full power of the representedsrepresentedadistinctionto illuminate the sensation-cognition (experience-belief) difference. Perhaps, therefore, a minor embellishment will betterreveal its potential for capturing some of the sb"ucture ofactual sensory systems (I am thinking mainly of constancymechanisms)


25/193


whose individual states(pointer positions) representsspeed,not axle rotation . This instrument , unlike the crude onedescribed above, "experiences" speed (not axle rotation )even though the information -delivery process depends oninformation about axle rotation to generate this "experience."

But if speed is what this instrument " experiences" (=what its :variousstateshave the systemic function of indicating) what does it "believe "?This dependson what the statesof this instrument representa, what they have acquired (orbeenassigned13the job of indicating . Normally , of course, ifthe instrument was to be used as a speedometerone wouldassign functions to its various states n the way that reflectedtheir systemic functions . The state , B, having thefunctio~ of indicating a speed of 50 mph, would be giventhis functionaby printing the number "50" .at this position ofthe pointer} 4We are, however, free to assign any functionsawe pleaSeIf we aren't particularly interested in exactly howfast the car is going, we could partition the dial into severalsub-divisions and label them "slow," "medium ," and " fast."Then B (having the functions of indicating a speed of 50mph) gets the samefunctionaas the state- (having the func-tio~ of indicating a speedof 48 mph) - the functiona' namely, of indicating a mediumspeed. This, in fact, might be thefunctions these statesacquire if we used the device, not as aspeedometer, but as a mechanism to control an automatictransmission. In this casesince we want the car to be in 3dgear at both 48 mph and 50 mph, we give these two statesthe same functiona.If the various states of the device acquirethesefunctionsa' then it will end up "experiencing" 50 mphand 48 mph (these are the speeds these two states


26/193

represents)' but representinga them both in the same way(as a 3d gear speed). It experiencesa difference between 50mph and 48 mph, yes, "but it , so to speak, "sees" both thesespeedsas medium (or 3d gear) speeds. Conceptually, thissystem "abstracts" from the sensory difference between 48and 50mph .The last example is meant to be suggestive of the wayinnate sensory systems might have the function of deliveringfuformation about more or less continuous quantities(surface reflectances , speed, direction , frequency , etc.)while the cognitive system might "chunk " this informationinto parcels that are more useful to the individual system'sneeds and purposes. At the level of experience, I am sensitiveto (i .e., can discriminate ) all manner of differences inthe light , sound, pressure, temperature, and chemistry ofobjects affecting my senses I nonetheless have a limitedconceptual repertoire for categorizing these sensory differences, for making judgments about (= representationsa of)the differences I experience. I experience the differences(in the same way the speedometer described above registersa difference between 50 mph and 48 mph), but , at theconceptual level , I treat these differences as differentinstances of the same kind . At the sensory level I can discriminathundreds of different colors. At the conceptuallevel I operate with , at best, a few dozen categories for thecolors I experience. Learning calibrates me no more finelythan I need to satisfy needsand carry out purposes. This isnot to say, of course, that I could not " recalibrate " in amore discriminating way if the need arose. The informationneeded for such recalibration is already there in myexperience.The taxonomy, then, looks like this:

Chapter 1


27/193


SensoryStateswith systemicindicator functions

ExperiencesSensationsFeelings

A few clarificatory remarks are in order:(a) Experiences are representationss' but , as noted earlier,not all representationss (not even all naturalrepresentationss ) are mental - let alone experiences .Experiences are those natural representationssthat servicethe construction of representationsa' representationss thatcan be calibrated (by learning) to more effectively servIceanorganism 's needs and desires. They are the' states whosefunctions it is to supply information to a cognitive systemfor calibration and use in the control and regulation ofbehavior . SEvans (1982, chapter 7, 4) expresses a similaridea when he describes internal states whose contentdepends on their phylogenetically ancient connectionswiththe motor system. In order to qualify as conscious experiences

, Evans requires that these content-bearing statesserveas input to what he calls a "concept-exercisingand reasoningsystem." 16This is why , in our (second) speedometer

example , the early stages in the information -deliveryprocess (the ones carrying information about axle rotation ,

RepresentationsStatesavingndicatorunctions........."......... """"""""

NaturalNatural functions..."................ """""""

ConventionalConventional functions

ConceptualStateswith acquiredindicatorfunctions

ThoughtsJudgmentsBeliefs


28/193

for example) are not appropriate analogues of experience.Though theseearly processes have a functions- that of providinginformation about axle rotation (information the systemneeds to generate a representation of speed)- andtherefore represents axle rotation , this information is notavailable for calibration . The instrument uses his information

, but doesn't, so to speak, make it available to calibrational"(cognitive ) processes. There is, then , no way thisinformation can be given "conceptual" form . The informationvanish es (absorbed into information about speed)before it reaches the relevant control structures that determinebehavior (in this case, .pointer positions ). These arenatural representationss hat are not mental. Their functionis to supply information to representationsshat are.17(b) Sensory systems have phylogenetic functions and are,therefore, comparatively modular in-Fodor 's (1983) sense:hard-wired , informationally incapsulated, and so on. This isnot to say that the states (experiences) they produce areimmune to changes n their representational status. We canuse a pressure gauge (functions = delivering informationabout pressure) as an altimeter by recalibrating its face in"feet above sealevel." Pointer positions representingspressurenow (also) representa altitude . The gauge, as it were,experiences pressure as altitude . This same transformationoccurs as we learn to identify and recognize objectsand conditionswe experience. We begin by hearing (experiencing)sounds and end by hearing (recognizing ) words . We stillhear sounds, of course, but, after learning, after the kind ofcalibration occurs that is involved in language learning ,experiencesacquire an added representationaldimension} 8

Sensory adaptation may be an instance of such recalibration. Just as we can recalibrate a speedometer when we

Chapter 1


29/193


change the size of our tires (the state B acquiring the functionof indicating a speedof 60 mph, not, as it did with normaltires, a speed of 50 mph) so also can sensory systemsrecalibrate when there is a change in the information an

experience delivers . After wearing spectacles that makeeverything " look " 300 to the left , touch "educates" visionand objectsbegin to " look" to be where they can be grasped.This , at least, is one way of describing the experientialchange that accompaniesadaptation. As we age, there is agradual yellowing of the macula (spot on the retina) thatchangesthe signals sent from the retina to the brain aboutthe wavelength of light (example taken from Clark 1993p.170). Despite this constant change in the information repre-sentationalsstatescarry, there is no corresponding represen-tationalachange: we still seeblue and yellow, red and green,as blue and yellow, red and green.(c) The statesby means of which a representational systemperforms its informational functions have a structure thatenables them to acquire functions without explicitly acquiringthem. Once we assign " 12" to a certain position of theclock hands, all the remaining indicator states (positions ofthe clock hands) receive an " implied " indicator function .When there is this kind of structure among a system's indicatorstates(as there is in our experienceof color and sound,for instance), there is no reason to suppose that eachrepre-sentationalsstate has explicitly obtained its indicator functionthrough some distinguishable evolutionary process. Allthe indicator statesmay receive an implied indicator function

by one stateexplicitly acquiring it .(d) Audition provides information , not only about pitch andintensity, but about timbre and (binaurally ) the directionalproperties of sound. In the caseof vision, there seems o be


30/193

separatepathways, different systems, for infonnation aboutcolor, location, form, and movement (Fischback 1992p. 56;Zeki 1992p. 71).It would be more appropriate, then, to describe a givensensemodality , not (as I have been doing ) as having thefunction of indicating F, but as having multiple indicatorfunctions . And there is no reason experiences "associated"with a given sense modality (smell or vision ) might notexploit information from other modalities (see, e.g., Steinand Meredith 1993Dretske 1981pp . 145- 147 .(e) Finally; the subjective quality of an experience, the phenomenaappearances, are the way experience representssthings to be. If I can use the speedometeranalogy again, thefirst speedometer has axle-rotation qualia. The second hasspeed qualia. Once calibrated, though, both

"describe" (i .e.,representa) how things seemto them "in the sameway- as avehicle speed of so-and-so many miles per hour. If we aretalking about vehicle speed, things seemhe same to them.

I am, with these brief remarks, trying to indicate how Iintend to use the distinction between systemic and acquiredrepresentations. Justification for thinking of experienceandthought in this way will come later. I am quite aware thatquestionsabout qualia, about what it is like to have an experience, about the way things seem, are complex questionsthat, often enough, straddle the sensation-cognition fence.We are sometimes talking about the properties objects arerepresenteds as having (the phenomenal appearances), atother times about the ones they are representedaas having(what the phenomenal appearancesprompt us to believe, orwhat we, on the basis of the experience, are inclined to takesomething as). I return to these tangled issues n chapters 3and 5.19

Chapter


31/193


4 Represented Properties and Represented ObjectsOrdinary thennometers represent the temperature of whatevermedium they are in . Put one in your coffee and it tellsyou how hot your coffee is. Hang the samethennometer onthe wall and it tells you the temperature of the room. Put itin your mouth and it indicates something about you -whether you have a fever. It isn't the function of this instrument

to say what it is - coffee, living room, or you- whosetemperature it registers. The instrument sayshow hot this is,where this is whatever medium the thermometer is in . Itdoesnot representthe relationship (i .e., its being in this) thatmakes it this rather than that it represents the temperatureof.

The same is true of all representational devices. Pressuregaugesdo not tell you which (if any) tire it is that they representthe pressure of. If you want to know which tire (ifany) the gauge represents, which topic it is commenting on,you have to look, not at the gauge, but at the external connectionsbetween gauge and world that make it the rightfront tire, not the left rear tire, whose pressure it registers.Gauges do not supply - they do not have the function ofsupplying - infonnation about these external connections.Representationshave a sense(the properties they have thefunction of indicating ) and, often enough, a reference (anobject whose properties they represent), but the sensedoesnot determine the reference. Two representations with thesamesensecanhave different referents.

The difference between representedobject and representedproperty, between the referenceand the senseof a representation, is the same distinction Nelson Goodman (1976)was getting at in his contrast between a picture of a blackhorse ("black horse" here specifying the object the picture is


32/193

a picture of) and a black-horse picture ("black -horse" herespecifying what the picture depicts the object as). Somepicturesof black horses do not represent the black horse as ablack horse. They are not black-horse pictures. Imagine takinga picture of a black horse at a great distance. The picturedepicts the black horse as, say, a spot in the distance . Itcould be anything . Or imagine taking a picture of a blackhorse disguised to look like a brown camel. This picture of ablack .horse is a brown -camel picture . Others (e.g., Lloyd1989p. 14) make the samedistinction by contrasting explicitwith extensional content. The extensional content is whatobject it is that is being represented. The explicit content isthe way this object is being represensted- as a so-and-so.The object a representation is a representation of is notdetermined by the properties that object is represented ashaving . Nothing guarantees that black-horse pictures arepictures of black horses.

Experiencesare like pictures in this regard; they are de remodes of representation (Burge 1977; Bach 1986, 1987;Recanati1993). Although token experiencesmay be individuatedby being experiencesof a particular object, k, tokensof the same experience type can occur with a different k orwith no object at all (hallucinations, dreams, imagination ).Although my present black-horse experience is an experienceof a black horse (I am, that is, seeing a black horse), Ican nonetheless have exactly the sametype of experience, ablack-horse experience, without its being an experienceof ablack horse or, indeed, an experience of any object at all .20What determines the referencefor a de re mode of representation(the object it is a representationof) is not how t is represente, but a certain external causal or contextual relationI will designateas C.21There is nothing in the content of therepresentation , nothing the representation says, which

Chapter


33/193


makes it about this object rather than that objector no objectat all . De re modes of representation have their referencedetermined contextually, by the relation I am here calling C.5ince the veridicality of an experience depends on its reference- on what object (if any) it is an experience of- theveridicality of experienceis determined, in part, by context(C). C makes it the case that k is the object 5 representsasblue. C ~ ereby helps to determine whether the representation(that k is blue) is veridical or not. Nonetheless, the factthat it is k (rather than some other object or no object at all)that stands in relation C to the representation is not what therepresentation represents. Representations do not (indeed,cannot) represent context. They represent k as being blue,but they do so without representing it to be k that they represent

to be blue.The speedometer n my car is connected to the axle of mycar, not your car. It therefore represents(or misrepresents, asthe casemay be) the speed of my car, not your car. It is invirtue of this special relation , C, to my car that it can dowhat other speedometers cannot do - viz ., say something(whether truly or falsely) about my car. Other speedometerscannot even say something false about my car. C is the relationsuch that, when a representational system5 is functioning

properly, and k stands in C to 5, then 5 will indicate theF of k. If k is the objectwhose F is representedat time t, thenwe can say that, at time t, 5 representsk. At other times 5may be deployed differently - thus, at different times, representingdifferent objects. Changing the way a system isdeployed, changing the object that stands in relation C to 5,changeswhat 5 represents(what objectt represents), but notnecessarilywhat 5 says(represents) about it .As I am using terms, then, the fact that 5 representsk isnot- at least not a pure - representational fact about S. It is a


34/193

Chapterfact about a representation - an important fact for manypractical purposes - but it is not a fact that has to do exclusivelywith what the systemhas the function of indicating . 5has the function of indicating the F of those objects whichstand in C to it , but it does not have the job of indicating -does not therefore represent - which objects - or evenwhether there is an objectthat stands in C to it . When wedescribe 5 (whose function it is to indicate F) asrepresentingk, thiS implies that, for some F, 5 represents (possibly misreprese) the F of k. That 5 representsk, therefore, impliesa representational factthat, for some F, 5 representsthe Fof k. But it also implies something that is not a representationalfact- viz ., that k stands in relation C to 5. 50 factsabout the object of representation are hybrid facts - part representati, part not . This will be important when we discuss(in chapter 2) introspective knowledge . For what isknown by introspection are mental- hence, (according tothe RepresentationalThesis) representationalfacts. Hybridrepresentational facts (that 5 represents k) are mixtures ofrepresentational facts and facts about representations. Thisis why one cannot know, at least not by introspection, whatobject (or whether there is an object) one is experiencing.22Clearly, then, my use of the word "something " in thedescription of 5 as representing something as being blue isnot an existential quantifier . It may turn out that 5 is misrepresentin

something to be F when there is something (in thenext room, say) that is F. The fact that there is something inthe world that is going 25 mph does not mean that aspeedometer registering "25 mph " says something trueabout the world . For what the speedometer "says" is notthat there is something in the world going 25 mph, but thatthis (whatever it is that stands in C to the instrument ) isgoing 25 mph . H representational system 5 says anything at


35/193

all when it represents color- and, thus, represents somethingasbeing (say) blue---it is that this is blue where this iswhatever object (if there is one) to which S stands in the C-relation. If there is no object of representation, then, S represents(i .e., misrepresents) this asbeing blue when there is nothis. If we take this to mean that, owing to a failure of presupposition, what S represents to be so (the content of the

represen~ tion) is neither true nor false, this merely showssomething about the nature of sense experience that weknew all along: viz ., that misrepresentation takes two forms.An experience can misrepresent by (1) saying somethingfalse, by saying that this is blue when it (the object of representation), s not blue;and (2) by saying what is neither truenor false - that this is blue when there is no "this" that is notblue. When there is no object of representation, nothing thatS representsto be blue, I shall continue to describeS as representingsomething to be blue . It should be understood ,however, that when there is no object, "something" standsin for a failed indexical .

Wethus have two different kinds of misrepresentation:

Object 01representationThere is aI) objectSrepresents o be blue

MisrepresentationThe objectSrepresentsto beblue is not blue


SrepresentscolorSrepresentssomething to be (say) blue........../ """'-.......,...No object of representationThere is no objectSrepresentsto be blue

MisrepresentationSrepresentssomethingto be blue, but there isno object that Srepresentsto be blue

/ ' "VeridicalThe objectSrepresentstobe blue is blue


36/193

Brentano (1874) conjectured that a mark of the mental wasintentionality . Whatever, exactly, Brentano meant by inten-tionality , and whether or not he was right about its being afeature of all , and only, mental events, most philosopherstake intentional characteristics(variously understood) to bedistinctive of a great many mental phenomena. What followsis a brief catalog of those aspectsof intentionality thathave figured most prominently in the recent literature . Ineach case we find that a representational account of themind provides a satisfying explanation of intentionality . Byconceiving of mental facts - and, in particular , those aboutsenseexperience- aspart of the natural order, asmanifestationsof overall biological and developmental design, onecan seewhere intentionality comesfrom and why it is there.In each caseintentionality is real e~ough, but it turns out,asFOdor(1987 suggests t must, to be really something eIse.1. ThePower to MisrepresentChisholm (1957 describes thisas the first mark of intentionality . Beliefs and experienceshave the power to "say" or "mean" that k is F when k is notF. Indeed, they have the power to say this (seeabove) evenwhen there is no k. If we understand representations asstateshaving indicator functions (systemic or acquired), wecan understand where they get this power. Unlike symbols,diagrams, gauges, and instruments, moreover, it is a powerthat natural representationsdo not get from us. They arestates that exhibit the first mark of intentionality in an original(nonderived ) form .2. AboutnessThis, of course, is the power or capacity of onestate of affairs to refer to or be about anotherS seesTom,hears him , has thoughts about him and desires for him .These are thingsS cannot do unlessS occupies states that

Chapter5 Intentionality


37/193

have Tom as their object, aswhat they refer to, aswhat theyare thoughts aboutexperiencesof, and desiresfor . In seeingTom, S's experiencesof Tom may not depict who (or what )it is they are experiences of. S can, after all, seeTom on afoggy night and mistake him for someoneelse. But if Tomstands in the right causal relationship to these experiencesthey are experiencesof him .. Echoing Goodman, S's experienceneed not be a Tom-experience n order to be an experience

ofTom.We not only experienceand think about objects, we alsoexperienceand think about their properties. I seea golf balland, in seeing it , I experience its color, shape, movement,and dimpled texture . My experience is of or about theseproperties asmuch as it is of the golf ball . I could be experiencingthese properties (during a dream or hallucination )when there was no golf ball, no object that had the propertiesI experience. Whether or not there is an object, the experienceis still an experienceof whiteness, roundness, movement, and texture.Even the simplest measuring instruments exhibit thesedimensions of aboutness albeit in a derived way. In representing

the pressurein an intake manifold , a pressuregauge"says" something about the manifold . It is not only aboutthe manifold (an object), it is about the pressure in it (aproperty ) and, therefore, about the manifold 's having thatpressure (a condition or state). If pressuregaugeswere conscious, if their intentionality was original rather than conventional, the manifold (the one standing in relation C tothe gauge) would be the object the gauge was consciousof,having a pressureof 14psi would be the property the gaugeperceived it to have, and its having a pressure of 14 psiwould be the condition , state of affairs, or fact the gaugewas aware of.



38/193

Though aboutnesss here being understood in terms ofthe reference of a representational state, and reference, inturn , is fixed by contextual relation C, this should not betaken to mean that this dimension of intentionality is beingequated with a causal or informational relation (whateverthe relation C comes down to) between two objects. Whatmakes 5 about he objectk is not simply the fact that k standsin relation C to 5. 5tanding in relation C to 5 is necessarybut not sufficient, to make 5 about k. The system to which kstands in this relation must be a representationalystem, onehaving thefunction of indicating the F of thoseobjectswhichstand in C to it . If 5 did not have this function , if there was-n't something 5 was supposedo indicate about k, then evenif k stood in relation C to 5, 5 would not be about k. A playgroundteeter-totter 's position stands in the samerelation tothe children playing on it that a beam balancestands to theobjects whose weights it has the function to compare. Thebehavior of the balance "says" or represents somethingabout the objectson it . It "says" that one is heavier than theother. The teeter-totter does not. This is not to say that wecould not learn something (exactly what we learn from thebalance) about the relative weight of the children from thebehavior of the teeter-totter. This, though, doesn't make theteeter-totter 's stateabout he children in any relevant senseof"about." A drunk 's behavior is not abouthis blood chemistryjust becausewe can use his behavior to tell something aboutthe chemistry of his blood.3. AspectualShapen thinking about a ball I think about it inone way rather than another- as red not blue, as round notsquare, as stationary not moving . These are the aspectsunder which I think of the ball . I can desire an apple, yes,but in desiring an apple I desire to eat it , taste it, throw it ,hold it , look at it , or simply have it . These are the aspects

Chapter 1


39/193


(the language of "aspects" is taken from Searle1992) underwhich I desire the apple. Our mental statesnot only have areferencean aboutness, an object (or purported object) thatforms their topic ; they represent that object in one wayrather than another. When an object is represented, there isalways an aspectunder which it is represented. Even whenthere is no object, the aspectemains.The same s true of the objectswe experience. The objectsI see looKa certain way to me - red, round, and stationary.The objects I touch feel a certain way to me - rough andcold. Experiencesare of objects, yes, but one cannot experiencean object without experiencing it under some aspect.This follows immediately from the fact that the experiencedobject is simply that object, if there is one, whose determinable

properties the sensory system represents. One cannot, therefore, seesmell, or taste an object without experiencingit ashaving determinate properties."Aspects come finely individuated and this, too, is part ofwhat is meant in speaking of aspectualshape. If we supposethat, in accordancewith gas laws, the pressureof the gas ina (constant volume) tank cannot be 14 psi without the gas

having a temperature of (say) 38C, an instrument cannothelp delivering information about temperature (that the gasis 38C) in delivering information about pressure (that it is14 psi). It can, nonethelessrepresenthe one without representingthe other. An instrument can have a pressure-indicatingfunction without having a temperature -indicatingfunction even when it cannot deliver information aboutpressurewithout delivering information about temperature.A pressuregauge is not a thermometer just because t deliversinformation about temperature (though we could certainlyuse it as, and therefore make it , one). Doing X doesn'tautomatically give you the function of doing X, and having


40/193

Chapter 1

the function of providing temperature information is whatis needed to represent temperature . It is the specificity ofindicator functions to this property rather than that property(even when the properties are nomically correlated ) thatexplains the aspectual profile of representationsand, therefore(according to the RepresentationalThesis) the aspectualcharacter of experience and thoUght.23It is the explanationfor why experiencing a certain pressure is different fromexperiencing a certain temperature even when theseaspectscannot be separated n the objectone experiences4. DirectednessThe objective reference of an experience -what object the experience s an experienceof- is arelational, a contextual, property of the experience. Two experiencescan be subjectively indistinguishable , both be black-horseexperiencesand, yet, one be an experienceof a black horse,the other not. Directedness, on the other hand, is supposedto be a quality of experiencethat is supposed to be intrinsicto the experience. It is supposed to be part of what it is liketo have that experience. H I understand it (and I'm not sure Ido) it is the quality of experienceby meansof which experiencepurports to refer to someobjector fact. Experiencesaresupposed to exhibit this kind of intentionality - directed -ness- whether or not they have an objective referent ,whether or not they are about something in the sensedescribed above. Miller (1984) explains Husserl's technicalterm "noema" in this way :Sinceanact hasdirectednessegardless f whetheror not it hasanobject, something other than the object must be that whichaccountsor the actsdirectednessThis"something" accordingoHusserlis the acts noema(16)There is, however , a connection between the noema of anexperience and its reference . Following Fo:llesdal (1969),


41/193

Experiencehe RepresentationalCharacter of Sense

Miller (p. 17) takes the noema to be that which determineswhich object (if there is one) is the object of the experience.For an experience to be directed is, simply, for it to have anoema (p. 31). To say that experienceshave this quality ofdirectedness, then, is to say that they have a subjectivelyaccessiblequality that is such that if they have an objectivereference (some object they are o.f), this quality determineswhich object it is.If we "understand directedness in this way, however, itsounds as though there is supposed to be some subjectivequality of experiencethat is such that if I am (in fact) seeinga black horse, it determines which black horse I am seeing.If this is what directedness s supposed to be, it seems clearthat experiencesdo not have it . When I am experiencing anobject, nothing in my experience of it determines whichobject I' m experiencing anymore than "here is somethingabout a gauge's representationof a tire's pressurethat determineswhich tire it is registering the pressure of .Representationsarenot like that. Neither areexperiencesI' m not sure, though , that I understand noema and thedirectedness they are supposed to confer on experiences.Miller (1984p. 69) says that the part of the noema (or noematic sinn: the reference determining aspect of the noema)which determines the purported objectof the act (in abstractionfrom properties) is indexical (also Christensen 1993p.758). If this means, as it seems o mean, that what object, ifany, an experience is directed at is determined, not withinthe subjective "accessiblecharacter of the experienceitself,but by the contextual or relational properties of the experience(what I called relation "C" in 3), then, once again,directedness comes out the same as aboutness as alreadydefined. It does not represent an additional aspectof inten-tionality .


42/193

Chapter 1

Sometimes, of course, the properties being representedapply uniquely to only one object. This is especially true invision where experience representsobjectsas being in particular

places, places that (given the way we representobjects) can only be occupied by one object. Something isrepresented as being here and something is represented asbeing there. It does not make much sense to suppose thatthe Qbject being represented as here is really being representedas there I can, at a given time, represent more thanone objectas red so that an object'sbeing representedas redwill not determine which object (if any) I am representing.But if I representan object asbeing here (where it is understoodthat only one object can be representedasbeing here),the representation of it as being here determines whichobject (if there is one) is being representedthat way. It is theone being represented as here. If this is what directednesscomes down to, then directednesshas to do, not with object-reference(if any) of the experience, but with certain specialproperties that this purported object is represented as having. Even simple instruments can have this kind of directed-ness (in a derived way ). Radar is "directed " at whateverobject it representsasbeing at coordinates [x,y,z]. If there isan object there, the representation of it as being there, at[x,y,z], determines which object it is that is being representedasbeing there. It is the objectthat is there - at [x,y,z].6 Mind and Brain: The Whereabouts of ExperienceInsofar as a representational account of sense experiencedepicts experience as representational, it likens experienceto something like a story. In thinking about a story we canbe thinking about : (1) the words that tell the story or (2)what those words mean, the story they tell . Call the first the


43/193

Whenever one talks about representations, there is thiskind of ambiguity, the ambiguity between representationalvehicle and representational content . Our talk aboutthoughts and experiencesexhibits this kind of ambiguity . Inthe sameway that stories are in books, thoughts and experiencesare in heads. What is in the head, of course, are theexperience

-vehicles, the physical states hat have a representational-content, the states that tell (express) a story aboutthe world - that say or mean, for example , that there issomething red over here and something triangular overthere. Thoughts and experienceshave to be in the head (orat least somewhere in the body ) if our having them is(sometimes) to explain why we act the way we do. I go tothe closet because I want to get dressed and think that iswhere my clothes are. If my thought (that my clothes are inthe closet) and desire (to get dressed) were not in me- atleast states or conditions of me, how could they possiblyexplain my trip to the closet? The itch I feel has to be in meto explain why I scratch. If mental states are to have anexplanatory role, if we sometimes do what we do becauseofwhat we think and experience, then beliefs and experienceshave to be in the systems whose behavior they explain .Aside, though, from their role in explaining behavior (somephilosophers are scepticalof this), experienceshave to be inus in order to explain why, for instance, I cease o experience


story-vehicle and the second the story-content. Stories (i .e.,story vehicles) are in books, but what happens in the story(content) does not happen in a book . It need not happenanywhere. The maiden was rescuedby a knight in shiningarmor- that episodewas certainly in the story- but , thoughthis story is in a book, no one expects to find a maiden, aknight , or a daring rescue in (i .e., between the covers of) abook.


44/193

a room full of people when I closemy eyes. The room full ofpeople doesnot vanish, but something ceases o exist whenI close my eyes. And this something has to be in me or astate of me. Why else would its existence be so utterlydependent on the position of my eyelids?But though the thoughts and experienceshave to be inus, we all know what we will find when we look inside theskuU. 1f there are sensationsand feelings in there, they seemto be well camouflaged. When Wilder Penfield (19571959)looked into the brain of his patient, a patient who was (as aresult of electrical stimulation by Penfield) experiencing avariety of sensations, he did not observe the colors andsounds the patient reported experiencing . This is exactlywhat one would expect on a representational theory ofexperience. Justas what makes a story the story it is is whatit is a story about (i .e., the story-content), so what makes amental state- and, in particular, an experience- the experienceit is is what it is an experienceof. We can have storiesabout blue dogs (blue-dog stories) and blue -dog experiences. Stories about blue dogs - the blue dog vehicles - areneither blue nor doglike . Just look in the book. It is all blackand white . Just so, what we find by looking in the brain of aperson experiencing blue dogs is neither blue nor doglike .We do not find the content of experience, the properties thatmake the experience the kind of experience it is. What wefind , instead, are electrical and chemical activity in gray,cheesy brain matter. We find the experience vehicles (seeLycan 1990, p. 111Tye 1991, p . 118, for the same point ).What we find , in other words, are the same sort of thing wefind when we look in books: representational vehicles thatin no way resemble what theserepresentationsrepresent.

According to the Representational Thesis, the facts thatmake what is in the head mental, the facts that convert elec-

Chapter


45/193


trical and chemical activity in the cortex into blue- dog experiences, are facts that are not identifiable by looking , exclusively, at what is in, the head. What makes a certain patternof electrical activity in the cortex into a blue-dog experienceis a fact about what this activity representswhat it has thefunction of indicating . What something represents, what ithas the function of indicating , are not facts one can discoverby loo~ g in the cortex, at the representations. One could aswell hope to figure out what, if anything , a voltage differencebetween two electrical contacts in a computer means(represents, signifies ) by taking ever more exacting measurementsof the voltage. One could as well hope to figureout what , if anything , ancient scratchings on a cave wallmeant by peering at them with a magnifying glass. Thatisn 't going to tell you what anything means or, indeed ,whether there is meaning- content - there at all . Accordingto the Representational Thesis, that is why looking at thebrain of a person having an experience isn't going to tellyou what that person is experiencing. The experience-vehicleis there, but the experiencecontent isn't there, and, justlike stories, it is the content that makes the experience (the

story) the experience(the story) it is.What sometimeshelps to obscure the difference betweenrepresentation- vehicle and representation-content is a confusion(fostered by conventional measuring instruments )about how one goes about determmmg what a representationrepresents. Instruments come with their representational" faces" calibrated in degrees Fahrenheit , pounds persquare inch, ohms, miles per hour, revolutions per minute ,and so on. This enables users to tell , not only what theinstrument is designed to measure, but how it representsthat quantity. One merely looks at the representation (e.g., apointer pointing at the numeral "900" on the face of an


46/193

instrument labeled TACHOMETER and calibrated in rpm 's)to discover what the instrument is " saying, " what it represents, what story it is telling about the engine. Looking atthe representation -vehicle , the pointer pointing at "900,"tells you what the representation-content is - that the engineis turning at 900rpm . With everything neatly labeled in thisway, one can, as it were, seewhat the instrument is "seeing," Seewhat its "experience

" is like, by looking in (or at)the instrument . This obscures the distinction between therepresentation itself , something (the vehicle) that is in theinstrument , something (the vehicle) that tells a story aboutthe engine, on the one hand, and, on the other hand, what isbeing represented (= content), the story being told , somethingthat is not in the instrument . An animal'sbrain doesn'tcomeneatly labeled in this way. That is why you cannot tell ,by looking at (or in) it , what the animal whose brain it is isexperiencing. You couldn 't do it with an instrument eitherwithout the help of labels. The representationsare there, buttheir content is not. In this sensethe mind isn't in the headany more than stories (i .e., story contents) are in books.

Chapter 1


47/193

Introspection

I think today is Wednesday. How do I know I think this?From this angle A looks longer than B. How do I know itdoes?Theseare not questions about how I know that today isWednesday or that A is longer than B. .They are questionsabout how I know something about myself- of what I thinkand how things look to me. They are questions about whatJohn Locke called reflection the notice that the mind takesof its own operations and states. Natsoulas (1983) definesone form of consciousness- reflective consciousness- as a

privileged ability to be noninferentially aware of (all orsomeof) one's current mental occurrencesWe seemto havethis ability . In telling you what I believe I do not have to figurethis out (as you might have to) from what I say or do.There is nothing from which I infer that A looks longer thanB. It just does. That is somehow given - knowledge themind has of itself . I will call the mind 's direct knowledge ofitself introspective knowledge . Introspection is the processby means of which we comeby suchknowledge .That we have introspective knowledge is obvious .Philosophers may disagree about the mode of accesswehave to our own thoughts and feelings and whether the


48/193

authority we enjoy is that of infallibility or something less,but most are willing to concedethat there is a dramatic differencebetween the way you know how I feel and the way Iknow it . The problem is to explain how we come by suchknowledge and what gives us this first -person authority .If the idea of the mind turning its attention onto itselfstrikes .one, as it struck Wittgenstein (1974, 412) as " thequeere.st hing there could be

" a representational theory dispelsthe mystery. It gives an account of introspective knowledgethat is compellingly simple and plausible. There is noappeal to---no longer a need forinternalS CaImersmonitoringthe clockwork of the mind . What one comesto know byintrospection are, to be sure, facts about one's mental life -thus (on a representational theory ) representational facts.These facts are facts, if you will , about internal representations. The objects and events one perceives to learn thesefacts, however, are seldom internal and never mental.1Onebecomesaware of representational factsby an awareness ofphysical objects. One learns that A looks longer than B, notby an awareness of the experience that represents A aslonger than B, but by an awareness of A and B, the objectsthe experience is an experience of. On a representationaltheory of the mind , introspection becomesan instanceof displaced

perception knowledge of internal (mental) facts viaan awarenessof external (physical) objects.2 .A representational theory of the mind is an externalisttheory of the mind and, therefore, facesthe sameproblemsthat all externalist theories confront (Boghossian 1989). Ifmental facts are constituted, not by the intrinsic characterofthe events occurring inside, but by the relations hese internalevents bear to external affairs (e.g., by their indicatorfunction ) then how is it possible to know , by " lookinginward ," by introspection, what IS going on in the mind ?

Otapter2


49/193

That would be like trying to determine , with the aid of mirrors, whether one was a father or a citizen of the UnitedStates . This " problem " evaporates once one understandsthat introspection is not a process in which one looksinward . Even if we could look inside , there would be no

point in doing so. The mental facts one seeks to .discover byintrospection are not there .

1 DisplacedPerceptionMany of the facts we come to know through ordinary senseperception are facts about objectswe do not perceive.I see how much I weigh by looking at the bathroom scaleon which I stand. The object I see s the bathroom scale. Thefact I learn is a fact about me-othat I weigh 170pounds. Thispattern- perceptual object in one place, perceptual fact inanother- is familiar . One looks at a gaugeon the dashboardin order to see how much gas remains in the gas tank- afact about an object, a tank located beneath the car, that onedoesnot see. One listens for the sound of a timer to find outwhen a cake in the oven is done and reads the newspapersor watches television to seewhat is happening on the otherside of the world . During instrument landing all the importantfacts are outside, but the instruments and gauges- theobjectsthe pilot actually sees- are inside. In such casesonecomes to know that k is F- sees that k is F- by seeing andhearing, not k itself, but h. The perceptual fact is displacedfrom the perceptual object.3

Perceptualdisplacement- seeing that k is F by seeing notk, but some other object, h, occurswhen there is conceptual,but no corresponding sensoryrepresentation of k4 In seeingthat I have gained five pounds by looking at the bathroomscale, there is a conceptual representation of meas having

Inb' ospection


50/193

gained five pounds , but a sensory representation of thescale, not me. The properties sensuously represented (i .e.,representeds) are properties of the scale (its color, orientation, size, shape, etc.), not properties of me. The propertyconceptually represented (i .e., representeda )- havinggained 5 pounds - is a property of me, not the scale.5Perceptual displacement requires belief (knowledge? jus-tified ~elief?6) that an appropriate connection exists betweenthe properties of the perceived object (h) and the propertiesof the target object (k)- the object the perceptual belief is abelief about. One must believe that h would not be G unlessk were (probably) F. Thosewho do not believe this will notseethat k is F when they seethat h is G. If you don't knowthat the object you are standing on is a scale, don't (therefore

) believe that the positien of its pointer indicates theweight of the person standing on it , you will not see howmuch you weigh by looking at it . Those who perceive hwithout an appropriate connecting belief (as I will call it ) - abelief about the relationship between h and k- will perceivethe sameobjectsbut not the same acts.Perceptualdisplacement enlargesthe number of factsone

perceiveswithout a corresponding enlargement in the numberof objectsone perceives. To expand one's observationalpowers , one does not need microscopes and telescopes.They help (because hey increase the number and kind ofobjectsone can see), but they aren't necessaryOne can alsoseemore facts, not by seeingmore objects, but by expandingone's knowledge of what the objectsone can already seesignify

about the objectsone cannot see. This is what connectingbeliefs (e.g., well -confirmed theories) provide .Before turning to the demonstration that, on a representationaltheory of perception, introspective knowledge (atleast of certain mental states) is a form of displaced percep-

Chapter 2


51/193

tion, a word about the facts that introspective knowledge isknowledge of. Introspective knowledge is knowledge of themind - i .e., mental facts. Since mental facts (according tothe thesis I am promoting ) are representational facts, introspective

knowledge is a (conceptual ) representation of arepresentation- of the fact that something (else) is a representationor has a certain representational content. It is

naturalizing the mind

Documents