davidson knowing one s own mind

7/31/2019 Davidson Knowing One s Own Mind

1/19

Knowing One's Own MindAuthor(s): Donald DavidsonSource: Proceedings and Addresses of the American Philosophical Association, Vol. 60, No. 3(Jan., 1987), pp. 441-458Published by: American Philosophical AssociationStable URL: http://www.jstor.org/stable/3131782 .Accessed: 18/10/2011 09:44

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].

American Philosophical Association is collaborating with JSTOR to digitize, preserve and extend access to

Proceedings and Addresses of the American Philosophical Association.

http://www.jstor.org
http://www.jstor.org/action/showPublisher?publisherCode=amphilosophicalhttp://www.jstor.org/stable/3131782?origin=JSTOR-pdfhttp://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/stable/3131782?origin=JSTOR-pdfhttp://www.jstor.org/action/showPublisher?publisherCode=amphilosophical


2/19

KNOWINGONE'SOWNMIND*DonaldDavidsonUniversityof California/Berkeley

Thereis no secret about the natureof the evidencewe use to decide what otherpeople think: we observe their acts, readtheir letters, study their expressions, istento their words, learntheir histories,and note their relationsto society. How we areable to assemble such material nto a convincingpictureof a mind is anothermatter;we know how to do it without necessarilyknowinghow we do it. SometimesI learnwhat I believe in much the sameway someone else does, by noticingwhat I say anddo. Theremay be times when this is my only access to my own thoughts. Accord-ing to GrahamWallas,

Thelittle girlhadthe makingof a poet in herwho, beingtold tobe sure of hermeaningbefore she spoke, said 'Howcan I know what Ithinktill I see what I say?.(1)A similarthought was expressedby Robert Motherwell: I would say that most goodpaintersdon't know what they thinkuntil they paintit.'

Gilbert Ryle was with the poet and the painter all the way in this matter; hestoutly maintainedthat we know our own minds in exactly the sameway we knowthe minds of others, by observingwhat we say, do, and paint. Ryle was wrong. Itis seldom the case that I need or appealto evidenceor observation n orderto find outwhat I believe; normally I know what I think before I speak or act. Evenwhen Ihave evidence, I seldom make use of it. I can be wrongabout my own thoughts,andso the appeal to what can be publicly determined s not irrelevant. But the possibi-lity that one may be mistakenabout one's own thoughtscannot defeatthe overridingpresumptionthat a personknows what he or she believes; n general,the belief thatone has a thought is enough to justify that belief. But though this is true, and evenobvious to most of us, the fact has, so far as I can see, no easy explanation. Whileit is clear enough, at least in outline, what we have to go on in trying to fathomthe thoughts of others, it is obscure why, in our own case, we can so often knowwhatwe think without appealto evidenceor recourse o observation.Because we usually know what we believe (and desire and doubt and intend)without needing or using evidence (even when it is available),our sincere avowalsconcerningour present states of mind are not subjectto the failingsof conclusionsbased on evidence. Thus sincere first person present-tenseclaims about thoughts,while neither infallible nor incorrigible,have an authority no second or third personclaim, or first personother-tenseclaim, can have. To recognizethis fact is not, how-ever,to explainit.

*PresidentialAddressdeliveredbefore the Sixtieth AnnualPacific DivisionMeet-ing of the American PhilosophicalAssociation in Los Angeles, California,March28,1986.441


3/19

Since Wittgensteinit has become routine to try to relieve worries about 'ourknowledge of other minds' by remarking hat it is an essentialaspect of our use ofcertain mental predicatesthat we apply them to others on the basis of behavioralevidence but to ourselveswithout benefit of such aid. The remark s true,andwhenproperly elaborated, it ought to answersomeone who wonders how we can knowthe minds of others. But as a responseto the skeptic, Wittgenstein's nsight (if it isWittgenstein's) hould give little satisfaction. For, first, it is a strange deathat claimsmade without evidential or observational support should be favored over claimswith such support. Of course, if evidence is not cited in support of a claim, theclaim cannot be impugned by questioning the truth or relevanceof the evidence.But these points hardly suffice to suggest that in generalclaims without evidentialsupport are more trustworthy than those with. The second, and chief, difficultyis this. One would normally say that what counts as evidence for the applicationof a concept helps define the concept, or at least places constraintson its identifi-cation. If two concepts regularlydepend for their applicationon different criteriaor rangesof evidentialsupport,they must be different concepts. So if what is appa-rently the same expressionis sometimescorrectly employed on the basis of a certainrangeof evidentialsupportand sometimeson the basisof anotherrangeof evidentialsupport (or none), the obvious conclusion would seem to be that the expressionisambiguous. Why then should we suppose that a predicate ike 'x believes that RasDashan is the highest mountain in Ethiopia', which is applied sometimes on thebasisof behavioralevidence and sometimesnot, is unambiguous? If it is ambiguous,then there is no reasonto suppose it has the same meaningwhen appliedto oneselfthat it has when appliedto another. If we grant(as we should) that the necessarilypublic and interpersonalcharacterof language guaranteesthat we often correctlyapply these predicatesto others, and that therefore we often do know what otherthink, then the question must be raised what groundseach of us has for thinkinghe knows what (in the same sense) he thinks. The Wittgensteinian tyle of answermay solve the problemof other minds,but it createsa correspondingproblemaboutknowledge of one's own mind. The correspondence s not quite complete, however.The original problem of other minds invited the question how one knows othershave minds at all. The problem we now face must be put this way: I know whatto look for in attributingthoughts to others. Usingquite differentcriteria or none),I apply the same predicatesto myself; so the skepticalquestion ariseswhy I shouldthink it is thoughts I am attributingto myself. But since the evidenceI use in thecase of others is open to the public, there is no reasonwhy I shouldn't attributethoughts to myself in the same way I do to others, in the mode of GrahamWallace,Robert Motherwell, and Gilbert Ryle. In other words, I don't, but I could, treatmy own mental states in the same way I do those of others. No such strategy isavailableto someone who seeks the same sort of authority with respect to thethoughts of others as he apparentlyhas in dealingwith his own thoughts. So theasymmetry between the cases remains a problem, and it is first person authoritythat createsthe problem.I have suggestedan answerto this problemin anotherpaper. (2) In that paperI arguedthat attention to how we attributethoughtsand meanings o others wouldexplain first person authority without inviting skeptical doubts. In recent years,however, some of the very facts about the attributionof attitudes on which I reliedto defend first person authority have been employed to attack that authority: it

442 APAPROCEEDINGS


4/19

PRESIDENTIALADDRESS DAVIDSONhas been argued,on what are thought to be new grounds,that while the methods ofthe third person interpreterdeterminewhat we usually deem to be the contents ofan agent's mind, the contents so determinedmay be unknownto the agent. In thepresent paper I consider some of these arguments,and urge that they do not con-stitute a genuine threat to first person authority. The explanationI offered in myearlier paper of the asymmetrybetween first and other-personattributionsof atti-tudes seems to me if anythingto be strengthenedby the new considerations,or thoseof them that seem valid.It should be stressedagainthat the problemI am concernedwith does not requirethat our beliefs about our own contemporarystates of mind be infallible or incorri-gible. We can and do make mistakesabout what we believe, desire, approve,andintend; there is also the possibility of self-deceit. But such cases, though not infre-quent, are not and could not be standard;I do not arguefor this now, but take itas one of the facts to be explained.Settingaside, then, self-deceptionand otheranomalousor borderlinephenomena,the question is whether we can, without irrationality, nconsistency, or confusion,simply and straightforwardlyhink we have a belief we do not have, or think we donot have a belief we do have. A number of philosophersandphilosophically-mindedpsychologists have recently entertainedviews that entail or suggest that this couldeasilyhappen-indeed,that it musthappenall the time.The threatwas there in Russell's dea of propositionsthat could be known to betrue even though they contained 'ingredients'with which the mind of the knowerwas not acquainted;and as the study of the de re attitudes evolved the peril grewmoreacute.But it was HilaryPutnamwho pulled the plug. ConsiderPutnam's1975 argu-ment to show that meanings,as he put it, 'justain't in the head'. (3) Putnamarguespersuasivelythat what words mean depends on more than 'what is in the head'.He tells a number of stories the moralof which is that aspectsof the naturalhistoryof how someone learned the use of a word necessarilymake a difference to whatthe word means. It seems to follow that two people might be in physically denticalstates,andyet meandifferentthingsby the samewords.The consequences are far-reaching. For if people can (usually) express theirthoughts correctly in words, then their thoughts--theirbeliefs, desires, intentions,hopes, expectations-also must in part be identified by events and objects outsidethe person. If meaningsain't in the head, then neither, it would seem, are beliefsanddesiresand the rest.Since some of you may be a little weary of Putnam'sdoppelgangeron TwinEarth,let me tell my own science fiction story--ifthat is what it is. Mystory avoidssome irrelevantdifficulties n Putnam's tory, thoughit introducessome new problemsof its own. (4) (I'll come back to Earth,and Twin Earth,a little later.) Supposelightningstrikes a dead tree in a swamp;I am standingnearby. My body is reducedto its elements, while entirely by coincidence (and out of different molecules) thetree is turned into my physicalreplica. My replica,The Swampman,moves exactlyas I did; accordingto its nature it departsthe swamp,encountersandseemsto recog-nize my friends, and appearsto returntheir greetings n English. It moves into myhouse and seems to write articleson radical nterpretation.No one can tell the differ-ence.

443


5/19

But there is a difference. Myreplicacan'trecognizemy friends; t can'trecognizeanything, since it nevercognizedanything n the firstplace. It can'tknow my friends'names (though of course it seems to), it can't remembermy house. It can't meanwhat I do by the word 'house', for example, since the sound 'house' it makes wasnot learned in a context that would give it the right meaning--orany meaningat all.Indeed, I don't see how my replica can be said to mean anythingby the sounds itmakes,nor to haveany thoughts.Putnam might not go along with this last claim, for he says that if two people(or objects) arein relevantlysimilarphysicalstates,it is 'absurd' o think theirpsycho-logical states are 'one bit different'. (5) It would be a mistaketo be surethat Putnamand I disagreeon this point, however, since it is not yet clearhow the phrase 'psy-chologicalstate' is beingused.Putnam holds that many philosophershave wronglyassumedthat psychologicalstates like belief and knowingthe meaningof a word are both (I) 'inner' n the sensethat they do not presupposethe existence of any individualother than the subjectto whom the state is ascribed,and (II) that these are the very states which we nor-mally identify and individuateas we do beliefs and the other propositionalattitudes.Since we normally identify and individuate mental states and meanings in termspartly of relations to objects and events other than the subject, Putnam believes(I) and (II) come apart: n his opinion, no states cansatisfyboth conditions.Putnam calls psychological states satisfying condition (I) 'narrow'. He thinksof such states as solipsistic, and associatesthem with Descartes'view of the mental.Putnammay considerthese states to be the only 'true'psychologicalstates; in muchof his paperhe omits the qualifier'narrow',despite the fact that narrowpsychologi-cal states (so called) do not correspondto the propositionalattitudes as normallyidentified. Not everyone has been persuaded hat there is an intelligibledistinctionto be drawnbetween narrow(or inner, or Cartesian,or individualistic--allhese termsare current) psychological states and psychologicalstates identified (if any are) interms of external facts (social or otherwise). Thus John Searlehas claimed thatour ordinary propositional attitudes satisfy condition (I), and so there is no needof states satisfyingcondition (II), while Tyler Burgehas denied that thereare,in anyinteresting sense, propositional attitudes that satisfy condition (I). (6) But thereseemsto be universalagreement hat no statessatisfyboth conditions.The thesis of this paperis that there is no reasonto supposethat ordinarymentalstates do not satisfy both conditions (I) and (II): I think such states are 'inner', nthe sense of being identical with states of the body, and so identifiablewithout re-ference to objects or events outside the body; they are at the same time 'non-indivi-dualistic' in the sense that they can be, and usually are, identified in part by theircausal relations to events and objcts outside the subject whose states they are. Acorollary of this thesis will turn out to be that contrary to what is often assumed,first person authority can without contradictionapply to states that are regularlyidentifiedby their relationsto events andobjectsoutsidethe person.

I begin with the corollary. Why is it naturalto assumethat states that satisfycondition (II) may not be known to the personwho is in those states?Now I must talk about Putnam'sTwin Earth. He asks us to imaginetwo peopleexactly alike physicallyand (therefore)alike with respect to all 'narrow'psychologi-cal states. One of the two people, an inhabitantof Earth,has learned to use theword 'water' by being shown water, readingand hearingabout it, etc. The other,

444 APA PROCEEDINGS


6/19

PRESIDENTIALADDRESS- DAVIDSONan inhabitant of Twin Earth, has learned to use the word 'water'under conditionsnot observably different, but the substanceto which she has been exposed is notwater but a lookalike substancewe may call 'twater'. Underthe circumstances,Put-nam claims, the first speaker refers to water when she uses the word 'water';hertwin refers to twater when she uses the word 'water'. So we seem to have a casewhere 'narrow'psychologicalstates are identical, and yet the speakersmeandifferentthingsby the sameword.How about the thoughts of these two speakers? The first says to herself,whenfacinga glassof water, 'Here'sa glassof water';the second muttersexactly the samesounds to herself when facing a glass of twater. Each speaksthe truth, since theirwords mean different things. And since each is sincere,it is naturalto supposetheybelieve different things, the first believingthere is a glass of water in front of her,the second believing there is a glass of twater in front of her. But do they knowwhat they believe? If the meaningsof their words, and thus the beliefs expressedby using those words, are partly determined by external factors about which theagents are ignorant, their beliefs and meanings are not narrow in Putnam's sense.There is therefore nothing on the basis of which either speakercan tell which stateshe is in, for there is no internal or external clue to the difference available. Weought, it seems, to conclude that neither speakerknows what she means or thinks.The conclusion has been drawnexplicitly by a numberof philosophers,among themPutnam. Putnamdeclaresthat he '...totally abandons he idea that if thereis a differ-ence in meaning...then there must be some difference in our concepts (or in ourpsychologicalstate)' Whatdeterminesmeaningand extension '...is not, in general,fully known to the speaker.'(7) Here 'psychologicalstate' meansnarrowpsycholo-gical state, and it is assumed that only such states are 'fully known'. JerryFodorbelieves that ordinarypropositionalattitudes are (pretty nearly) 'in the head', buthe agreeswith Putnam hat if propositionalattitudes were partly dentifiedby factorsoutside the agent, they would not be in the head,and would not necessarilybe knownto the agent. (8) John Searlealso, though his reasons are not Fodor's, holds thatmeaningsare in the head ('there is nowhere else for them to be'), but seems to acceptthe inference that if this were not the case, first person authoritywould be lost. (9)Perhapsthe plainest statement of the position appears n AndrewWoodfield's ntro-duction to a book of essays on the objects of thought. Referring o the claim thatthe contents of the mind are often determined by facts external to and perhapsunknown to the personwhose mindit is, he says:

Because he externalrelation s not determined ubjectively, he subjectis not authoritativeabout that. A thirdpersonmightwell be in a betterpositionthan the subjectto know whichobject the subject s thinkingabout,hence be betterplacedto know whichthoughtit was. (10)Those who accept the thesis that the contents of propositional attitudes are

partly identified in terms of external factors seem to have a problemsimilarto theproblemof the skeptic who finds we may be altogethermistaken about the 'outside'world. In the presentcase, ordinaryscepticismof the senses is avoidedby supposingthe world itself more or less correctly determines the contents of thoughts aboutthe world. (The speakerwho thinks it is water is probablyright, for he learnedtheuse of the word 'water' in a watery environment; he speakerwho thinks twater is

445


7/19

probablyright, for he learned he word 'water' n a twateryenvironment.) Butskepti-cism is not defeated; it is only displacedonto knowledge of our own minds. Ourordinarybeliefs about the external world are (on this view) directedonto the world,but we don't know what we believe.There is, of course, a difference between water and twater, and it can be dis-covered by normal means, whether it is discoveredor not. So a personmight findout what he believes by discoveringthe difference between water and twater, andfinding out enough about his own relationsto both to determine which one his talkand beliefs are about. The skeptical conclusion we seem to have reached concernsthe extent of first person authority: it is far more limited than we supposed. Ourbeliefs about the world are mostly true, but we may easily be wrong about whatwe think. It is a transposed mageof Cartesian kepticism.Those who hold that the contents of our thoughts and the meaningsof ourwords are often fixed by factors of which we are ignoranthave not been much con-cerned with the apparentconsequenceof their views which I have been emphasizing.They have, of course, realized that if they were right, the Cartesian dea that theone thing we can be certain of is the contents of our own minds,and the Fregeannotion of meaningsfully 'grasped',must be wrong. But they have not made muchof an attempt, so far as I know, to resolvethe seemingconflict between their viewsandthe strong ntuition that first personauthorityexists.One reason for the lack of concernmay be that some seem to see the problemas confined to a fairly limited rangeof cases, cases where concepts or words latchonto objects that are picked out or referred to using propernames, indexicals, andwords for natural kinds. Others, though, argue that the ties between languageandthought on the one hand and external affairs on the other are so pervasive hat noaspect of thought as usually conceived is untouched. In this vein Daniel Dennettremarksthat '...one must be richly informed about, intimately connected with, theworld at large, its occupants and properties,in order to be said with any proprietyto have beliefs'. (11) He goes on to claim that the identification of all beliefs isinfected by the outside, non-subjectivefactors that are recognized to operate inthe sort of case we have been discussing. Burgealso emphasizes he extent to whichour beliefs are affected by external factors, though for reasonshe does not explain,he apparentlydoes not view this as a threatto firstpersonauthority.(12)The subject has taken a disquietingturn. At one time behaviorismwas invokedto show how it was possible for one person to know what was in another'smind;behaviorismwas then rejected in part because it could not explain one of the mostobvious aspectsof mentalstates: the fact that they arein generalknownto the personwho has them without appealto behavioristicevidence. The recent fashion, thoughnot strictly behavioristic,once more identifiesmental states partlyin terms of socialand other external factors, thus makingthem to that extent publicly discoverable.But at the sametime it reinstates he problemof accounting or firstpersonauthority.Those who are convincedof the externaldimensionof the contents of thoughtsas ordinarily identified and individuatedhave reacted in different ways. One re-sponse has been to make a distinction betweenthe contents of the mindas subjective-ly and internally determined, on the one hand, and ordinarybeliefs, desires, andintentions, as we normally attribute them on the basis of social and other outwardconnections, on the other. This is clearlythe trend of Putnam'sargument(althoughthe word 'water'has differentmeanings,and is used to expressdifferentbeliefs when

446 APAPROCEEDINGS


8/19

PRESIDENTIALADDRESS DAVIDSONit is used to refer to water and to twater, people using the word for these differentpurposesmay be in 'the same psychologicalstate'). JerryFodor accepts the distinc-tion for certain purposes, but argues that psychology should adopt the stance of'methodological solipsism' (Putnam'sphrase)-that is, it should deal exclusivelywithinner states, the truly subjective psychological states which owe nothing to theirrelations o the outsideworld. (13)Steven Stich makes essentially the same distinction, but draws a sternermoral:where Fodor thinks we merely need to tinker a bit with propositionalattitudes asusually conceived to separate out the purely subjective element, Stich holds thatpsychological states as we now think of them belong to a crude and confused 'folkpsychology'which must be replacedby a yet to be invented 'cognitivescience'. Thesubtitleof his recentbook is 'TheCaseAgainstBelief. (14)Clearlythose who drawsuch a distinction have insured that the problemof firstperson authority, at least as I have posed it, cannot be solved. For the problemIhave set is how to explain the asymmetrybetween the way in which a personknowsabout his contemporarymental states and the way in which others know about them.The mental states in question are beliefs, desires,intentions, and so on, as ordinarilyconceived. Those who accept somethinglike Putnam'sdistinction do not eventry toexplain first person authority with respect to these states; if there is first personauthorityat all, it attaches to quite differentstates. (In Stich'scase, it is not obviousthat it canattachto anything.)I think Putnam, Burge, Dennett, Fodor, Stich, and others are right in callingattention to the fact that ordinarymental states, at least the propositionalattitudes,are partlyidentified by relations to society and the restof the environment,relationswhich may in some respects not be known to the personin those states. They arealso right, in my opinion, in holding that for this reason(if for no other), the con-cepts of 'folk psychology' cannot be incorporated nto a coherent and comprehen-sive system of laws of the sort for which physics strives. These concepts are part ofa common-sensetheory for describing, nterpreting,and explaininghuman behaviorwhich is a bit freestyle, but (so I think) indispensable. I can imaginea science con-cerned with people and purged of 'folk psychology', but I cannot think in whatits interestwould consist. Thisis not, however,the topic of this paper.I am here concernedwith the puzzlingdiscovery hat we apparentlydo not knowwhat we think--at east in the way we think we do. This is a real puzzle if, like me,you believe it is true that external factors partly determine the contents of thoughts,and also believe that in generalwe do know, and in a way others do not, what wethink. The problem arisesbecause admitting the identifying and individuatingroleof external factors seems to lead to the conclusion that our thoughts may not beknownto us.But does this conclusion follow? The answer depends, I believe, on the wayin which one thinks the identificationof mentalcontents dependson externalfactors.The conclusion does follow, for example, for any theory which holds that pro-positional attitudes are identified by objects (such as propositions, tokens of pro-positions, or representations)which are in or 'before' the mind, and which containor incorporate(as 'ingredients')objects or events outside the agent; for it is obviousthat everyoneis ignorantof endless features of everyexternalobject. That the con-clusion follows from these assumptions is generallyconceded. (15) However, forreasonsI shall mention below, I reject the assumptionson which the conclusion is inthis casebased.

447


9/19

Tyler Burge has suggestedthat there is another way in which external factorsenter into the determination of the contents of speech and thought. One of his'thought experiments'happenspretty well to fit me. Until recentlyI believedarthri-tis was an inflammationof the joints causedby calciumdeposits;I didnot know thatany inflammationof the joints, for example gout, also counted as arthritis. So whena doctor told me (falsely as it turnedout) that I had gout, I believed I had gout butI did not believe I had arthritis. At this point Burgeasks us to imaginea world inwhich I was physicallythe same but in which the word 'arthritis'happenedactuallyto apply only to inflammation of the joints caused by calciumdeposits. Then thesentence 'Gout is not a form of arthritis'would have been true, not false, and thebelief that I expressed by this sentence would not have been the false belief thatgout is not a form of arthritisbut a true belief about some diseaseother than arthri-tis. Yet in the imaginedworld all my physicalstates, my 'internalqualitativeexperi-ences', my behaviorand dispositionsto behave,are the sameas they are in this world.My belief would have changed, but I would have no reasonto supposethat it had,andso could not be saidto know what I believed.

Burgestresses he fact that his argumentdependson...the possibilityof someone'shavinga propositionalattitudedespitean incompletemasteryof some notion in its content...if the thoughtexperiment s to work, one mustat some stagefind the subjectbe-lieving(orhavingsomeattitude characterized y) a content, despitean incompleteunderstanding r misapplication. 16)

It seems to follow that if Burgeis right, whenever a person is wrong, confused, orpartiallymisinformedabout the meaningof a word, he is wrong,confused,or partial-ly misinformedabout any of his beliefs that is (or would be?) expressedby usingthatword. Since such 'partialunderstanding' s 'commonor even normalin the case ofa large number of expressionsin our vocabularies'accordingto Burge,it must beequally common or normalfor us to be wrongabout whatwe believe(and,of course,fear,hope for, wishwere the case, doubt,and so on).Burgeapparentlyaccepts this conclusion;at least so I interpret his denialthat'...full understandingof a content is in generala necessarycondition for believingthe content'. He explicitly rejects '...the old modelaccording o whicha personmustbe directly acquainted with, or must immediately apprehend, the contents of histhoughts...aperson's thought content is not fixed by whatgoes on in him, or by whatis accessible o him simplyby carefulreflection.'(17)I am uncertainhow to understand hese claims,since I amuncertainhow seriouslyto take the talk of 'direct acquaintance'with, and of 'immediatelyapprehending',a content. But in any case I am convincedthat if what we mean and think is deter-mined by the linguistic habits of those around us in the way Burgebelieves theyare, then first personauthorityis very seriously compromised. Since the degreeandcharacterof the compromiseseem to me incompatiblewith what we know aboutthe kind of knowledge we have of our own minds, I must reject some premise ofBurge's. I agree that what I mean and think is not 'fixed' (exclusively) by whatgoes on in me, so what I must rejectis Burge'saccountof how social andother exter-nal factorscontrol the contentsof a person'smind.

448 APA PROCEEDINGS


10/19

PRESIDENTIALADDRESS DAVIDSONFor a number of reasons, I am inclined to discount the importanceof the fea-tures of our attributions of attitudes to which Burge points. Suppose that I, whothink the word 'arthritis'applies to inflammation of the joints only if caused by

calcium deposits, and my friend Arthur,who knows better, both sincerelyutter toSmith the words 'Carlhas arthritis'. Accordingto Burge,if other things are moreor less equal (Arthurand I are both generallycompetent speakersof English,bothhave often applied the word 'arthritis' o genuine cases of arthritis,etc.) then ourwords on this occasion mean the same thing, Arthurand I mean the same thing byour words, and we express the samebelief. My error about the dictionarymeaningof the word (or about what arthritis s) makes no difference to what I meant orthought on this occasion. Burge'sevidence for this claim seems to rest on his con-viction that this is whatanyone (unspoiledby philosophy)would reportabout Arthurand me. I doubt that Burgeis right about this, but even if he is, I don't think itproves his claim. Ordinaryattributions of meaningsand attitudes rest on vast andvague assumptionsabout what is and is not shared (linguisticallyand otherwise)by the attributer,the person to whom the attributionis made, and the attributer'sintended audience. Whensome of these assumptionsprove false, we may alter thewords we use to make the report, often in substantialways. Whennothing muchhinges on it, we tend to choose the lazy way: we take someone at his word, evenif this does not quite reflect some aspect of the speaker's hought or meaning. Butthis is not because we are bound (outside of a law court, anyway) to be legalisticabout it. And often we aren't. If Smith (unspoiled by philosophy) reportsto stillanother party (perhaps a distant doctor attempting a diagnosis on the basis of atelephone report) that Arthurand I both have said, and believe, that Carlhas arthri-tis, he may actively mislead his hearer. If this dangerwere to arise,Smith, alert tothe facts, would not simply say 'Arthurand Davidson both beleive Carlhas arthri-tis'; he would add something like, 'But Davidsonthinks arthritismust be causedbycalcium deposits'. The need to make this addition I take to show that the simpleattributionwas not quite right;there was a relevantdifference n the thoughtsArthurand I expressedwhen we said 'Carlhas arthritis'. Burgedoes not have to be budgedby this argument,of course, since he can insist that the report is literally correct,but could, like any report,be misleading. I think, on the other hand, that this replywould overlook the extent to which the contents of one belief necessarilydependon the contents of others. Thoughtsare not independent atoms, and so there canbe no simple,rigid,rulefor the correctattributionof a singlethought. (18)Though I reject Burge's nsistence that we are bound to give a person'swordsthe meaningthey have in his linguisticcommunity,and to interprethis propositionalattitudes on the samebasis, I thinkthereis a somewhatdifferent,but veryimportant,sense in which social factors do control what a speakercan mean by his words. Ifa speakerwishes to be understood,he must intend his words to be interpreted n acertain way, and so must intend to provide his audience with the clues they needto arriveat the intendedinterpretation. Thisholdswhether the hearer s sophisticatedin the use of a language he speakerknows or is the learnerof a first language. It isthe requirementof learnability,interpretability, hat providesthe irreduciblesocialfactor, and that shows why someone can't mean something by his words that can'tbe correctly decipheredby another. (Burge seems to make this point himself in alaterpaper.)(19)

449


11/19

Now I would like to return to Putnam'sTwin Earth example, which does notdepend on the idea that social linguistic usagedictates (undermore or less standardconditions) what speakersmean by their words, nor, of course,what their (narrow)psychological states are. I am, as I said, persuadedthat Putnamis right;what ourwords mean is fixed in part by the circumstancesn which we learned,and used, thewords. Putnam'ssingle example (water) is not enough, perhaps,to nail down thispoint, since it is possible to insist that 'water' doesn't apply just to stuff with thesame molecular structureas water but also to stuff enough like water in structureto be odorless, potable, to support swimmingand sailing, etc. (I realize that thisremark, ike many others in this piece, may show that I don't know a rigid designatorwhen I see one. ( don't.) The issue does not depend on such special cases nor onhow we do or should resolvethem. The issue depends simply on how the basiccon-nection between words and things, or thoughts and things, is established. I hold,along with Burgeand Putnam if I understandthem, that it is establishedby causalinteractions between people and parts and aspects of the world. The dispositionsto react differentially to objects and events thus set up are central to the correctinterpretationof a person'sthoughtsand speech. If this werenot the case,we wouldhave no way of discoveringwhat others think, or what they mean by their words.The principleis as simpleand obvious as this: a sentencesomeone is inspired caused)to hold true by and only by sightingsof the moon is apt to mean something like'There'sthe moon'; the thought expressedis apt to be that the moon is there; thethought inspiredby and only by sightingsof the moon is apt to be the thought thatthe moon is there. Apt to be, allowing for intelligible error, second hand reports,and so on. Not that all words and sentences are this directlyconditioned to whatthey are about; we can perfectly well learn to use the word 'moon' without everseeing it. The claim is that all thought and languagemust have a foundation in suchdirect historical connections, and these connections constrainthe interpretationofthoughts and speech. PerhapsI should stress that the argumentsfor this claim donot rest on intuitions concerningwhat we would say if certaincounterfactualsweretrue. No science fiction or thought experimentsarerequired. 20)I agreewith Putnamand Burge, hen, that

...the intentionalcontent of ordinarypropositionalattitudes...cannotbe accountedfor in termsof physical,phenomenal,causal-functional,computational,or syntacticalstatesor processes hat arespecifiednonintentionallyand aredefinedpurelyon the individual n isolationfrom his physicalandsocialenvironment. (21)The question remainswhether this fact is a threat to first personauthority, as Burgeseems to think, and Putnamand otherscertainlythink. I haverejectedone of Burge'sargumentswhich, if it were right,would pose such a threat. But thereis the positiondescribed in the previous paragraph,and which I hold whether or not others do,since I think this much 'externalism'is required to explain how languagecan belearned,andhow wordsandattitudescan be identifiedby aninterpreter.Why does Putnam think that if the referenceof a word is (sometimes)fixed bythe naturalhistory of how the word was acquired,a user of the word may lose firstperson authority? Putnam claims (correctly, in my view) that two people can be inall relevant physical (chemical, physiological, etc.) respects the same and yet mean

450 APAPROCEEDINGS


12/19

PRESIDENTIALADDRESS- DAVIDSONdifferent things by their words and have different propositionalattitudes (as theseare normallyidentified). The differences are due to environmentaldifferencesaboutwhich the two agentsmay, in some respects,be ignorant. Why,underthese circum-stances, should we suppose these agents may not know what they mean and think?Talkingwith them will not easily show this. As we have noted, each, when facedwith a glass of water or twater says honestly, 'Here's a glassof water'. If they arein their home environments,each is right;if they have switchedearths,eachis wrong.If we ask each one what he means by the word 'water', he gives the right answer,using the same words, of course. If we ask each one what he believes,he gives theright answer. These answers are right because though verbally dentical, they mustbe interpreteddifferently. Andwhat is it that they do not know (in the usualauthori-tative way) about their own states? As we have seen, Putnamdistinguisheshe stateswe have just been discussingfrom 'narrow'psychologicalstates which do not pre-supposethe existence of any individualother than the subjectin that state. Wemaynow start to wonder why Putnam is interested in narrowpsychologicalstates. Partof the answeris, of course, that it is these states that he thinks have the 'Cartesian'propertyof being known in a special way by the personwho is in them. (The otherpart of the answerhas to do with constructinga 'scientific psychology'; this doesnot concern us here.)The reasoningdepends,I think, on two largelyunquestionedassumptions. Theseare:

(1) If a thoughtis identifiedby a relationto somethingoutside the head,it isn't wholly in the head. (It ain't in the head.)(2) If a thoughtisn't wholly in the head,it can't be 'grasped'by themind in the way requiredby firstpersonauthority.That this is Putnam'sreasoning s suggestedby his claim that if two heads arethe same, narrowpsychologicalstates must be the same. Thus if we suppose twopeople are 'molecule for molecule' the same ('in the sense in which two necktiescan be "identical"';you may add, if you wish, that each of the two people 'thinksthe sameverbalized houghts..., has the samesense data, the samedispositions,etc.'),then 'it is absurd to think [one] psychologicalstate is one bit different from' theother. These are, of course, narrowpsychologicalstates, not the ones we normallyattribute,whichain't in the head. (22)It is not easy to say in exactly what way the verbalizedthoughts, sense data,and dispositionscan be identical without revertingto the neckties, so let us revert.Then the idea is this: the narrowpsychological states of two people are identicalwhen their physical states cannot be distinguished. There would be no point indisputingthis, since narrowpsychologicalstates are Putnam's o define; what I wishto question is assumption(1) above which led to the conclusionthat ordinarypro-positionalattitudes aren't n the head,andthat thereforefirstperson authoritydoesn't

applyto them.It should be clearthat it doesn't follow, simply from the fact that meaningsareidentified in part by relations to objects outside the head, that meaningsaren't inthe head. To suppose this would be as bad as to arguethat becausemy being sun-burned presupposesthe existence of the sun, my sunburnisn't a condition of myskin. My sunburnedskin may be indistinguishable rom someone else's skin that

451


13/19

APAPROCEEDINGSachieved its burnby other means (our skins may be identical in 'the necktie sense');yet one of us is really sunburnedand the other not. This is enough to show that anappreciationof the external factors that enter into our common ways of identifyingmental states does not discredit an identity theory of the mental and the physical.AndrewWoodfieldseems to thinkit does. Hewrites:

No de re state about an object that is externalto the person'sbraincanpossiblybe identical with a stateof that brain,sinceno brainstatepresupposes he existence of an externalobject. (23)Individual states and events don't conceptually presupposeanything in themselves;some of their descriptionsmay, however. My paternalgrandfather idn'tpresupposeme, but if someone can be described as my paternalgrandfather, everalpeople be-sidesmy grandfather, ncludingme, must exist.Burgemay makea similarmistake n the followingpassage:

...no occurrenceof a thought...couldhave a differentcontent and bethe very sametoken event... [T]hen...aperson's houghtevent is notidenticalwith any event in him that is describedby physiology,biology,chemistry,or physics. For let b be any givenevent described n termsof one of the physicalsciencesthat occurs n the subjectwhile hethinksthe relevant hought. Let 'b'be suchthat it denotesthe samephysicalevent occurringn the subject n our counterfactual ituation...b need not be affected by counterfactualdifferences[that do notchangethe contents of the thoughtevent]. Thus...b [the physicalevent]is not identical with the subject'soccurrent hought. (24)Burgedoes not claim to have establishedthe premise of this argument,and so notits conclusion. But he holds that the denial of the premiseis 'intuitivelyvery im-plausible'. He goes on, '...materialist dentity theories have schooled the imaginationto picture the content of a mental event as varyingwhile the event remains fixed.But whether such imaginingsare possible fact or just philosophical ancy is a separatequestion'. It is because he thinks the denial of the premiseto be very improbablethat he holds that 'materialist dentity theories' are themselves'rendered mplausibleby the non-individualistichought experiments'. 25)I accept Burge's premise;I think its denial not merely implausiblebut absurd.If two mental events have different contents they are surely different events. WhatI take Burge'sand Putnam's maginedcasesto show (andwhat I think The Swampmanexampleshows more directly) is that people who are in all relevantphysical respectssimilar (or 'identical' in the necktie sense) can differ in what they mean or think,just as they can differ in being grandfathers r being sunburned. But of coursethereis something different about them, even in the physicalworld;their causalhistoriesaredifferent.I conclude that the merefact that ordinarymentalstatesand events areindividua-ted in terms of relations to the outside world has no tendency to discreditmental-physical identity theories as such. In conjunctionwith a numberof further(plausi-ble) assumptions, he 'externalism'of certainmental states and events can be used, Ithink, to discredittype-type identity theories;but if anything t supportstoken-token

452


14/19

PRESIDENTIALADDRESS- DAVIDSONidentity theories. (I see no good reason for callingall identity theories 'materialist';if some mental events are physical events, this makes them no more physical thanmental. Identityis a symmetricalrelation.)

Putnam and Woodfieldare wrong, then, in claimingthat it is 'absurd' o thinktwo people could be physicallyidentical (in the 'necktie'sense)andyet differin theirordinarypsychological states. Burge,unless he is willing to make far strongerplaythan he has with essentialistassumptions, s wrong n thinkinghe has shownall identi-ty theories mplausible.Wearetherefore ree to hold that people can be in all relevantphysical respects identical while differing psychologically:this is in fact the positionof 'anomalousmonism' for which I havearguedelsewhere. 26)One obstacle to non-evidentialknowledge of our own ordinary propositionalattitudes has now been removed. For if ordinarybeliefs and the other attitudes canbe 'in the head' even though they are identified as the attitudes they are partly interms of what is not in the head, then the threat to first person authority cannotcome simply from the fact that external factors are relevant to the identificationof the attitudes.But an apparent difficulty remains. True, my sunburn,though describableassuch only in relation to the sun, is identical with a condition of my skin which can(I assume) be described without reference to such 'external' factors. Still, if, as ascientist skilled in all the phjsical sciences, I have access only to my skin, and amdenied knowledge of the history of its condition, then by hypothesis there is noway for me to tell that I am sunburned. Perhaps,then, someone has first personauthority with respect to the contents of his mind only as those contents can bedescribedor discoveredwithout reference to external factors. In so far as the con-tents are identified in terms of external factors, first person authority necessarilylapses. I can tell by examiningmy skin what my privateor 'narrow'condition is,but nothing I can learn in this restricted realm will tell me that I am sunburned.The difference between referring o and thinkingof water andreferring o and think-ing of twater is like the difference between being sunburnedand one's skin beingin exactly the same condition through another cause. The semantic differenceliesin the outside world, beyond the reachof subjectiveor sublunarknowledge. So theargumentmightrun.This analogy, between the limited view of the skin doctor and the tunnel visionof the mind's eye, is fundamentallyflawed. It depends for its appeal on a faultypicture of the mind, a picture which those who have been attackingthe subjectivecharacterof ordinarypsychologicalstates sharewith those they attack. If we canbringourselvesto give up this picture, first personauthoritywill no longerbeen seenas a problem;indeed, it will turn out that first person authority is dependent on,and explained by, the social and public factors that were supposed to underminethat authority.Thereis a pictureof the mind whichhas become so ingrainedn our philosophicaltradition that it is almost impossible to escape its influence even when its worstfaults are recognizedand repudiated. In one crude, but familiar,version,it goes likethis: the mind is a theater in which the conscious self watches a passingshow (theshadows on the wall). The show consists of 'appearances', ense data, qualia,whatis givenin experience. Whatappearon the stage are not the ordinaryobjects of theworld that the outer eye registersand the heart loves, but theirpurportedrepresenta-tives. Whateverwe know about the worldoutsidedependson what we cangleanfromthe innerclues.

453


15/19

The difficulty that has been apparentfrom the start with this descriptionof themental is to see how it is possible to beat a track from the inside to the outside.Another conspicuous, though perhaps less appreciated, difficulty is to locate theself in the picture. For the self seems on the one hand to include theater, stage,actors, and audience; on the other hand, what is known and registeredpertainstothe audience alone. This second problemcould be as well stated as the problemof the location of the objects of the mind: are they in the mind, or simply viewedby it?I am not now concernedwith such (now largely disavowed)objects of the mindas sense-data,but with theirjudgmentalcousins, the supposed objectsof the proposi-tional attitudes, whether thought of as propositions,tokens of propositions, repre-sentations, or fragmentsof 'mentalese'. The central idea I wish to attack is thatthese are entities that the mind can 'entertain', 'grasp','have before it', or be 'ac-quainted' with. (These metaphors are probably instructive: voyeurs merely wantto have representationsbefore the mind'seye, while the more aggressivegraspthem;the Englishmay be merely acquaintedwith the contents of the mind, while morefriendlytypes will actuallyentertainthem.)It is easy to see how the discovery that external facts enter into the individua-tion of states of mind disturbsthe pictureof the mind I have been describing. Forif to be in a state of mind is for the mind to be in some relation like grasping o anobject, then whateverhelps determine what object it is must equally be grasped fthe mind is to know what state it is in. This is particularlyevident if an externalobject is an 'ingredient' n the object before the mind. But in eithercase, the personwho is in the state of mindmay not know what state of mindhe is in.It is at this point that the concept of the subjective--ofa state of mind-seemsto come apart. Onthe one hand,thereare the true innerstates,with respectto whichthe mind retains its authority; on the other hand there are the ordinarystates ofbelief, desire, intention and meaning, which are polluted by their necessary con-nectionswith the social andpublicworld.In analogy, there is the problem of the sunburnexpert who cannot tell by in-spectingthe skin whether it is a caseof sunburnor merelyan identical condition withanother cause. Wecan solve the sunburnproblemby distinguishing etween sunburnand sunnishburn; unnishburn s just like sunburnexcept that the sun need not beinvolved. The expert can spot a case of sunnishburnust by looking, but not a caseof sunburn. This solution works becauseskin conditions, unlikeobjectsof the mind,are not requiredto be such that there be a special someone who can tell, just bylooking,whetheror not the conditionobtains.The solution in the case of mental states is different, and simpler;it is to getrid of the metaphor of objects before the mind. Most of us long ago gave up theidea of perceptions,sense data, the flow of experience,as things 'given'to the mind;we should treat propositionalobjects in the sameway. Of coursepeoplehavebeliefs,wishes, doubts, and so forth; but to allow this is not to suggestthat beliefs, wishesand doubts are entities in or before the mind, or that being in such states requiresthereto be correspondingmentalobjects.This has been said before, in varioustones of voice, but for different reasons.Ontologicalscruples, for example, are no part of my interest. We will always needan infinite supply of objects to help describeand identify attitudes like belief; I amnot suggestingfor a moment that belief sentences, and sentencesthat attributethe

454 APAPROCEEDINGS


16/19

PRESIDENTIALADDRESS- DAVIDSONother attitudes, are not relational n nature. WhatI am suggesting s that the objectsto which we relate people in order to describe their attitudes need not in any sensebe psychologicalobjects, objects to be grasped,known, or entertainedby the personwhose attitudes are described.This point, too, is familiar;Quine makes it when he suggeststhat we may useour own sentences to keep track of the thoughts of people who do not know ourlanguage. Quine's interest is semantical,and he says nothing in this context aboutthe epistemologicaland psychological aspects of the attitudes. We need to bringthese various concerns together. Sentences about the attitudes are relational;forsemantic reasons there must therefore be objects to which to relate those who haveattitudes. But having an attitude is not havingan entity before the mind; for com-pelling psychological and epistemological reasons we should deny that there areobjectsof the mind.

The source of the trouble is the dogma that to have a thought is to have anobject before the mind. Putnamand Fodor (and many others) have distinguishedtwo sorts of objects, those that are trulyinnerand thus 'beforethe mind' or 'grasped'by it, and those that identify the thought in the usualway. I agreethat no objectscan serve these two purposes. Putnam (and some of the other philosophersI havementioned) think the difficulty springsfrom the fact that an object partly identifiedin termsof externalrelationscannot be counted on to coincidewith an objectbeforethe mind because the mind may be ignorant of the external relation. Perhaps hisis so. But it does not follow that we can find other objects which will insurethedesired coincidence. For if the object isn't connected with the world, we can neverlearn about the world by having that object before the mind; and for reciprocalreasons, it would be impossibleto detect such a thought in another. So it seemsthat what is before the mind cannot include its outside connections--its semantics.On the other hand, if the object is connected with the world, then it cannotbe fully'before the mind' in the relevantsense. Yet unless a semanticobject can be beforethe mind in its semanticaspect, thought, conceivedin terms of such objects, cannotescapethe fate of sense data.The basic difficulty is simple: if to have a thought is to have an object 'beforethe mind', and the identity of the object determineswhat the thought is, then itmust always be possibleto be mistakenabout what one is thinking. For unless oneknows everything about the object, there will always be senses in which one doesnot know what object it is. Manyattemptshave been madeto find a relationbetweena person and an object which will in all contexts hold if and only if the personcanintuitively be said to know what object it is. But none of these attemptshassucceed-ed, and I think the reason is clear. The only object that would satisfy the twin re-quirementsof being 'before the mind' and also such that it determines what thecontent of a thought must, like Hume'sideas and impressions, be what it seems andseem what is is'. There are no such objects, public or private,abstractor concrete.The argumentsof Burge, Putnam, Dennett, Fodor, Stich, Kaplan, Evans andmany others to show that propositions can't both determine the contents of ourthoughts and be subjectively assuredare, in my opinion, so many variants on thesimple and generalargumentI have just sketched. It is not just propositions thatcan't do thejob; no objectscould.When we have freed ourselves from the assumption that thoughts must havemysterious objects, we can see how the fact that mental states as we commonly

455


17/19


18/19

PRESIDENTIALADDRESS- DAVIDSON(11) DanielDennett, "BeyondBelief', in Thoughtand Object,p. 76.(12) Tyler Burge, "Other Bodies", in Thought and Object; "Individualismandthe Mental", nMidwestStudies inPhilosophy, Volume4, PeterFrench,Theodore

Uehling,HowardWettstein,eds., Universityof MinnesotaPress,1979; "TwoThoughtExperimentsReviewed",Notre Dame Journalof FormalLogic, 23 (1982), pp. 284-93; "Individualism ndPsychology".(13) JerryFodor, "MethodologicalSolipsismConsideredas a ResearchStrategyin CognitivePsychology".(14) Steven Stich, From Folk Psychology to Cognitive Science, M.I.T. Press,1983.(15) See, for example, GarethEvans, The Varietiesof Reference, Oxford Uni-versityPress,1982, pp. 45, 199, 201.(16) TylerBurge,"Individualism nd the Mental",p. 83.(17) Ibid.,pp. 90, 102, 104.(18) Burge suggeststhat the reason we normallytake a personto mean by hiswords what others in his linguistic community mean, whether or not the speakerknows what others mean, is that 'People are frequently held, and hold themselves,to the standardsof the community when misuse or misunderstanding re at issue.'He also says such cases '...dependon a certainresponsibility o communalpractice'.("Individualismand the Mental", p. 90) I don't doubt the phenomenon, but itsbearingon what it is supposed to show. (a) It is often reasonable o hold peopleresponsiblefor knowingwhat their words mean;in such cases we may treat them ascommitted to positions they did not know or believe they were committed to. Thishas nothing (directly) to do with what they meant by their words,norwhat they be-lieved. (b) As good citizens andparentswe want to encouragepractices hat enhancethe chances for communication;using words as we think others do may enhancecommunication. This thought (whether or not justified) may help explain whysome people tend to attribute meaningsand beliefs in a legalistic way; they hopeto encourageconformity. (c) A speakerwho wishes to be understood must intendhis words to be interpreted(and hence interpretable)along certainlines; this inten-tion may be served by using words as others do (though often this is not the case).Similarly, a hearer who wishes to understanda speaker must intend to interpretthe speaker'swords as the speaker intended (whether or not the interpretation s'standard'). These reciprocalintentions become morally importantin endless situa-tions which have no necessaryconnection with the determinationof what someonehadin mind.(19) See, for example,"TwoThoughtExperimentsReviewed",p. 289.(20) Burgehas described 'thought experiments'which do not involve languageat all; one of these experimentsprompts him to claim that someone broughtup inan environmentwithout aluminumcould not have 'aluminumthoughts'. ("Indivi-dualismand Psychology", p. 5.) Burgedoes not say why he thinks this, but it isby no meansobvious that counterfactualassumptionsareneeded to make the point.In any case, the new thought experimentsseem to rest on intuitions quite differentfrom the intuitions invoked in "Individualism nd the Mental"; t is not clearhowsocial norms feature in the new experiments,and the linguistic habits of the com-munity are apparentlyirrelevant. At this point it may be that Burge'sposition isclose to mine.(21) "TwoThoughtExperimentsReviewed",p. 288.

457


19/19

458 APAPROCEEDINGS(22) "TheMeaningof 'Meaning"', . 227.(23) AndrewWoodfield, n Thoughtand Object,p. viii.(24) "Individualismndthe Mental",p. 111.(25) "Individualismand Psychology", p. 15, note 7. Cf. "Individualismandthe Mental",p. 111.(26) "Mental Events", in Donald Davidson, Essays on Actions and Events,OxfordUniversityPress,1982.

Note: I am greatly indebted to Akeel Bilgramiand Ernie LePorefor criticism andadvice. Tyler Burgegenerously ried to correctmy understanding f his work.

davidson knowing one s own mind

Documents