Springer
http://www.jstor.org/stable/4321775 .
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].
Springer is collaborating with JSTOR to digitize, preserve and extend access to Philosophical Studies: An
International Journal for Philosophy in the Analytic Tradition.
http://www.jstor.org
PhilosophicalStudies(2006) 129:575-603
DOI 10.1007/si1098-004-3616-8
? Springer2006
JOSEMEDINA
WHAT'S SO SPECIALABOUT SELF-KNOWLEDGE?
1. INTRODUCTION
What is so special about self-knowledge? In a nutshell, my an-
swer to this questionwill be "Nothing."I will try to show that
knowledgeof one's own mind is a particularlyinterestingtype
of knowledge for various reasons, but that there is nothing
specialabout it: that is, thereis no epistemicprivilegeor special
normative force that attaches to self-knowledgealone. I will
argue that knowledge of one's own mind is as fallible and
corrigible as any other area of human knowledge. As my
argument will show, the persistent insistence that this is a
special area of knowledge that enjoys special epistemicprivileges is an unwarrantedresidue of the Cartesian picture of
human knowledgeas groundedin subjectivecertainties.I will
analyzeour self-ascriptionsof mental states as speechacts, and
I will draw on Wittgenstein'selucidationsof these speech acts
to account for their performativeand normativedimensions.
When these ascriptions are properly situated in their home
languageand consideredas moves within particularlanguagegames, they cease to be mysteriousand we are no longer inclined to attributeto them a specialnormativeforce or special
epistemicpropertiessuch as incorrigibility.Ultimately,I will try
to show, these self-ascriptionsare no differentin kind from the
ascriptionsof intentionalstates to others.
However,manypeopleremainconvincedstill today that selfascriptionsor avowals of mental states enjoy a specialkind of
authority;and this is taken to be a datum that any epistemic
account must accommodate and explain. As Bilgrami - for one -
puts it (1992a, p. 247), our self-knowledgeof intentionalstates
seems to have a special characterwhich "cries out" for an
576
JOSE MEDINA
explanation. For Descartes, the explanation came from the
model of inner observation.On the Cartesianpicture,we have
directaccess to the objectsbefore the mind, and it is the immediacy of inner perceptionthat gives self-knowledgeits special
characterand its special certainty:while our knowledgeabout
anythingexternalis inferentialandfallible,ourknowledgeabout
the objectsbeforethe mind is noninferentialand infallible.This
Cartesianpicturewas heavilycriticizedby Ryle (1949;esp. pp.
167-181). He famously arguedthat self-knowledgeis an inferentialmatterand that thereis nothingspecialaboutit. On Ryle's
view, self-knowledgeis in principleas fallibleas any other kind
of knowledge.It is rarelychallengedjust becausewe are usually
the ones who have more (and better)evidenceabout ourselves.
We are usuallybetterplacedto judge (on the basis of evidence)
what we think, believe, desire,etc. But our (usually)privileged
knowledgeabout ourselvesdoes not amount to more than our
(usually)privilegedpositionto observeourselves.And, for Ryle,
self-observationis not a matterof immediatephenomenological
access to our intentionalstates.We ascribeintentionalstates to
ourselveson the basis of what we say and do, just as we ascribe
intentionalstatesto otherson the basisof what they say and do.
Accordingto Ryle, the immediacyof self-knowledgeis an illusion, the naturalproduct of the expertisewe have acquiredin
ascribingbeliefs, desires,etc. to ourselves.Self-ascriptionsbecome so automaticthat theirinferentialcharacterusuallypasses
unnoticed.
If the Cartesianparadigmunderscoresthe subjectivefeatures
of intentionalstatesand our knowledgeabout them, the Rylean
paradigmstressesthe dispositionalfeaturesof intentionality.In
the contemporary literature on self-knowledge, both the
introspectiveparadigmof Descartesand the inferentialistparadigm of Ryle have come under attack. The introspective
paradigmis said to have gone too far becauseit rules out the
defeasibilityof self-judgments.It is commonlyrecognizedthat
there is usually a presumption of first person authority, that it is
generallyreasonableto assumethat the subjectknows what she
thinks, believes,desires,etc. unless thereis positive evidenceto
the contrary. But this first person authority falls short of
WHAT'SSO SPECIALABOUTSELF-KNOWLEDGE?
577
incorrigibility. As Davidson - for one - puts it, "we do not
always have indubitable or certain knowledge of our own
attitudes.Nor are our claimsabout our attitudesincorrigible.It
is possible for the evidence available to others to overthrow
self-judgments"(Davidson, 1984, p. 103). On the other hand,
the inferentialistparadigmis said to have gone too far because
it does away with the undeniable spontaneous character of
self-judgments.Generally,self-judgmentsare more a matterof
"avowability" than a matter of inference. So neither the
introspective nor the inferentialist paradigm seem to fit
self-knowledge,and yet they seem to be the only epistemic
paradigmsavailablefor the elucidationof self-knowledge,the
only games in town. Drawing on Wittgenstein'slater philosophy, I will propose another game. In order to develop this
alternative,I will proceedindirectly,going from the negativeto
the positive,from the criticalto the constructive.I will focus my
critical discussion on recent attempts to vindicate the special
characterof self-knowledgeeither on epistemicor on practical
grounds. From this critique there will emerge an alternative
picture of self-knowledgeand its relation to other types of
knowledge.
In the last two decades there have been some attempts to
account for the special character of self-knowledge. Among the
most notorious are those of Burge (1988), Davidson (1984,
1986) and Bilgrami (1992a, b, 1998). Of these three, only
Burge's account of self-knowledgeis introspective and still
Cartesianin spirit.Burgearguesthat his social externalismstill
leaves some (though limited)room for infallibleCartesianselfjudgments.The other two accounts of self-knowledge,Davidson's and Bilgrami's,give up the whole idea of "objectsbefore
the mind" and discard infallibility.They set out to provide a
nonepistemic rationale for the presumption of first person
authority.Both Davidson and Bilgramitry to capturewhat is
special about self-knowledgein a "constitutivethesis" (i.e. the
thesis that "it is constitutive of the very notion of intentionality
that we by and large have self-knowledge";Bilgrami, 1992a,
p. 248; 1998,pp. 212ff).For Davidson, firstperson authorityis
a necessary presuppositionof our interpretativepractices:"4an
578
JOSE MEDINA
unavoidable presumptionbuilt into the nature of interpretation" (Davidson, 1984, p. 111). For Bilgrami, first person
authority is an unavoidablenormativerequirementbuilt into
the notion of responsibleagency, for "it is a presuppositionof
the generalidea of being responsiblefor somethingwe do, that
we know what we are doing" (Bilgrami,1992b,p. 250). In what
follows I examine:first,Burge'sepistemicaccountof the special
characterof self-knowledge;and, second, Bilgrami'sattemptto
flesh out Davidson's constitutivethesis by groundingfirst person authorityin the normativepresuppositionsunderlyingthe
practicesin which we exerciseresponsibleagency.
2. FIRST PERSON AUTHORITYAND EPISTEMICPRIVILEGE
Burge has argued that the content of our intentional states is
crucially dependent on (socially mediated) external factors (see
Burge, 1979). The content of our thoughts about water or
arthritis cannot be determined independently of the stuff out
there called "water" or "arthritis," and more specifically, not
independently of the criteria of one's linguistic community for
what counts as water or arthritis. But Burge (1988) has argued
that his social externalism does not pose a threat to selfknowledge, that it is in fact compatible with a restricted version
of the thesis of privileged access of the Cartesian picture of the
mind. This is his argument in "Individualism and SelfKnowledge," which will be the focus of my critical examination
of his view. There Burge emphasizes that it would be a mistake
to think that "a person could not know what thoughts he had
unless he undertook an empirical investigation of the environment" (p. 69). For Burge, there is a sense in which the subject
has special knowledge of her intentional states: she has privi-
leged accessto what intentionalstates she has, even though she
has no special authorityon the necessaryand sufficientconditions for the individuation of their content. Burge emphasizes
that there are "basic cases" of self-knowledge which call for a
Cartesian treatment. These cases are incorrigible "cogito
thoughts": cases where our intentional states become "objects
WHAT'S SO SPECIAL ABOUT SELF-KNOWLEDGE?
579
before the mind." They are expressed in judgments about one's
thoughts (beliefs, desires, etc.) such as "I think that I am
thinking that p" (p. 70). In these basic cases, in thinking that
one is thinking that p, one is also thinking that p. And since
whatever goes into the individuation of the thought that p
carries over to the iterated thought, the correctness of the second-order thought does not depend on the subject's knowledge
about the correct individuation of the intentional content that
she expresses with "p."
Burge (1988) argues that the self-referential character of
second-order self-ascriptions makes them incorrigible:"No errors at all are possible in strict cogito judgments; they are selfverifying" (p. 74). When one thinks (and hence knows) that one
is thinking that p, one can be wrong about the enabling conditions of the thought that p, but there is one thing one cannot
be wrong about, namely, that it is the thought that p that one is
thinking: "One is thinking that p in the very event of thinking
knowledgeably that one is thinking it. It is thought and thought
about in the same mental act." (p. 70) It is this immediacy that
constitutes, for Burge, the epistemic ground for the special, selfverifying character of basic self-knowledge: "our knowledge of
our own thoughts is immediate, not discursive. Our epistemic
right rests on this immediacy, as does our epistemic right to
perceptual beliefs" (p. 72; emphasis added). According to
Burge, the justification of basic self-knowledge is in no way
discursive.
But there is something extremely peculiar about the second-order thoughts that Burge appeals to in his account of
self-knowledge. "In basic self-knowledge," he says, "one
simultaneously thinks through a first-order thought (that water is a liquid) and thinks about it as one's own" (Burge,
1988, p. 75; emphasis added). This seems to require the
capacity to entertain two thoughts at the same time: the
thought that p, and the thought that one thinks that p. Even
if these complex mental acts that involve both a first-order
and a second-order thought were possible, it seems that they
would occur under very exceptional circumstances. I do not
usually go through an explicit mental act to identify an
580
JOSE MEDINA
occurrentintentional state as an intentional state of mine (as
opposed to what?), while at the same time entertainingthat
very intentional state (e.g. a belief or a desire).At any rate, it
is not clear why it should matter whether these complex
mental acts do in fact occur, for it is not clear what their
occurrencewould show. After all, what is self-verifyingand
incorrigible,according to Burge's argument, is the "cogitolike judgment" as a whole (e.g. "I think that I believe that
p"), not the embeddedjudgment(e.g. "I believethat p"). And
it is not clear that the alleged epistemic privilege of these
second-orderjudgments can explain the general presumption
of first-personauthority, which concerns our first-orderselfjudgments (judgmentsthat express our beliefs, desires, etc.,
not judgmentsabout our thinkingthat we believe, desire,etc.,
something or other). The infallible characterof the secondorder self-judgmentcannot be transmittedwilly-nilly to the
first-orderself-judgment.
Burge(1988)emphasizesthat we have special"second-order,
self-ascriptivepowers"(p. 72). But what do these powersshow
about self-knowledge?According to Burge, my incorrigible
second-orderthought that I think that p automaticallycertifies
two differentthings: (a) that it is p that I am thinking,that I
cannot be wrong about the content of my thought (p. 70); and
(b), that it is me who is thinking the thought that p, that I
cannot be wrong about who owns that first-orderthought (p.
75). These are two (related but) analytically distinct issues
concerningintentionalstates:the issue of content and the issue
of ownership.The thesis that self-ascriptionsare immunefrom
error of misidentificationof its possessor is trivial. For it is a
trivial phenomenon that the occurrentintentional states expressedin my avowals are intentionalstates of mine. What is
not trivialis how to explain this phenomenon.Burge seems to
suggestthat this immunityfrom errorspringsfrom our special
"second-order,self-ascriptivepowers," from the incorrigible
capacity to identify one's intentionalstates "as one's own" (p.
75). But it is not clear that we can analyze avowals into the
identificationof a content and the identificationof its possessor. In self-ascriptions,the use of "I" introducesa performative
WHAT'S SO SPECIAL ABOUT SELF-KNOWLEDGE?
581
element that is not easily assimilableto a mental act of identification. As Brandom (1994) has pointed out, "I" is an
expressive locution used to undertakecommitments, not to
identify commitmentsas one's own in a special, self-verifying
sort of way. As Brandomputs it, the immunityfrom error of
misidentificationof avowals amountsto nothing more than the
fact that "the undertakingby acknowledgingof a commitment is
somethingthat no one but I can do" (p. 553). So the immunity
from error of misidentificationderives from the performative
aspects of self-ascriptions,not from a special mental act of
identificationof one's intentionalstates as one's own.
As Wittgensteinsuggested,avowals are immunefrom errors
of misidentificationpreciselybecausethey do not involve an act
of identificationat all: "I don't choose the mouth which says
'I ..."'; "It has no sense to ask 'How do you know it's you who
sees. . .'." (Wittgenstein, 1968, p. 311) Wittgenstein's analysis of
avowals puts the emphasis on the performativeaspect of the
indexical "I": for example, the expression "I promise..." does
not say whopromises;it is used by a speakerto makea promise.
On this performative analysis, "I" is not a referring expression
that results from an act of identificationand thereforetacitly
makes a claim about the correctidentificationof a subject:I do
not say "I" to pick myself out from among others but to draw
attentionto myself, as when I raisemy hand or clearmy throat
loudly before I speak (Wittgenstein,1958,p. 67). Accordingto
my Wittgensteinian,performativeview, self-attributionsare not
decomposableinto a content attributedand a person identified
as its possessor;they are avowals.So when I say "I am in pain,"
I do not identifymyselfas the possessorof a sensation;I simply
draw attention to myself and the audienceidentifieswho is in
pain. As Wittgensteinputs it:
Thereis no questionof recognizinga person when I say I have toothache.
To ask 'are you sure that it's you who have pains?'would be nonsensical.
[...] it is as impossiblethat in making the statement'I have toothache' I
should have mistakenanotherperson for myself,as it is to moanwith pain
by mistake,havingmistakensomeoneelse for me. To say, 'I havepain'is no
more a statementabouta particularperson than moaningis (Wittgenstein,
1958,p. 67; emphasispreservedand added)
582
JOSE MEDINA
But, at any rate, the issue of ownershipcannot be Burge's
primaryconcern. For his externalismabout content is trivially
compatiblewith the immunityfrom error of this kind. What
Burgewants from his accountof self-knowledgeis to neutralize
the challengethat his externalismhas the unwelcomeimplication that people do not usually know what they are thinking
(not that they usually don't know whether what they are
thinking is their own or someone else's thought). In order to
answer this challenge, it is the issue of content that Burge's
account of self-knowledgemust address:the issue of whether
we have any specialauthorityon whatwe think, believe,desire,
etc. To determinewhetherBurge has answeredthis challenge,
let's examinemore closely the peculiartwist that he gives to the
introspectiveparadigmin his account of self-knowledge.
On the Cartesianpicture,"the objectsbeforethe mind'seye"
were contentful objects of awarenesstowards which we can
have differentpropositionalattitudes, that is, the contents of
our thinking,believing,doubting,etc. By contrast, the objects
that Burge claims to be immediatelypresent to the mind in a
self-verifyingway are the intentionalstates themselves(thinking that p, believingthat q, etc.) ratherthan theircontents(p, q,
etc.). Burge'sclaim is that in a "basiccogito-likejudgment"we
cannot be wrong about the intentionalstate we are in. But if I
am incorrigibleabout what intentionalstate I am in, I must be
able to distinguish(infallibly)that intentionalstate from other
intentionalstates. We can distinguisha belief from a desire(or
from any other intentional state) without appealing to their
content, since they can in fact have the same content. But we
can only distinguishone belief from another(e.g. the beliefthat
p from the belief that q) by its content. So sincethe subjectwho
entertainsa basic cogito-like thought is incorrigibleabout the
first-orderthought containedin it, it seems that she must also
be incorrigibleabout the content of that first-orderthought.
But Burge's externalismmaintains that the contents of our
intentional states "ain't in the head." So how can the
immediatepresenceof an intentional state in the mind of its
possessorenableher to identifyit, infallibly,in a second-order,
self-ascriptiveact?
WHAT'S SO SPECIAL ABOUT SELF-KNOWLEDGE?
583
Burge emphasizes that we must distinguish between
authoritativeself-knowledgeand the knowledge requiredfor
the proper specificationof one's thought: "One clearly does
not have first-personauthority about whether one of one's
thought is to be explicatedor individuatedin such and such a
way" (p. 78). Burgearguesthat the knowledgeinvolved in the
"explication"of one's thoughts is not required "in order to
know that one is thinking them" (p. 78). When a thought
about p comes before my mind in a second-order, selfascriptive act, what I know infallibly is nothing about "p";
what I know infallibly is simply that it is "p" that I am
thinking. But how are we to understand"p" in the claim "I
know at least one thing: that it is p that I am thinking"?
Elsewhere Burge has drawn a sharp distinction between "a
lexical item and the explication of its meaning," "the word"
and "the entry for the word," arguingthat the "explicational
meaningis the semanticalanalogue"of the lexical item (1989,
p. 181). Now, if we apply this distinction to Burge's basic
cases of self-knowledge,it seems that my incorrigibleknowledge about my thoughtsconcernsthe lexical items that are the
objects of my thinking, but not their explicationalmeaning.
On this syntacticinterpretationof Burge'sview, what I know
incorrigibly is only the string of signs that occupies my
thinking,not its content. Another way to put it is to say that
in the self-verifyingjudgment "I know I think "p","p" is
mentioned but not used.
This readingof Burge'sview suggeststhat he is proposinga
syntacticcriterionfor the individuationof thoughts on which
the subjecthas privilegedauthority(as opposedto the semantic
criterionon which the subjecthas no authority).But if this is
Burge'sproposal, he needs to provide an account of this syntactic criterionand to show its significancefor self-knowledge.
This proposal would requirea reformulationof the concept of
"propositionalattitudes"as attitudestowardssyntacticstrings,
ratherthan towardscontentfulpropositions.But is thereroom
in Burge'sview for the introductionof a syntacticcriterionfor
the individuationof content along with the semanticcriteriaof
his externalism? The use of double standards for the
584
JOSE MEDINA
individuationof content would commit him to the bifurcation
of intentional content. For the distinction between syntactic
and semanticcriteriaof individuationis just one way of construingthe distinctionbetweennarrowand broad content (e.g.
Fodor's 1981 way). But, as is well known, Burge rejects this
distinction;he explicitlyopposes the bifurcationof content (see
Burge, 1986). So it seems unlikely that in his account of selfknowledge he would appeal to the distinction between the
syntactic and semantic "contents" of our intentional states.
But, even more importantly,the retreatto the syntacticlevel to
defend first-personauthoritywon't do for Burge,for it doesn't
answerthe challengeposed to his externalism.The challengeis
to explainhow, on Burge'sexternalistview of content, subjects
can have self-knowledgeof the contentsof their thoughts.The
(indeedpeculiar)claim that subjectsat least know the syntactic
strings of signs that occupy their minds doesn't address this
challenge. It is in fact besides the point, for self-knowledge
about the words that go through one's mind was never in
question. So the authoritativeself-knowledgeof Burge'sview
must concernmore than the syntacticelementsthat constitute
the content of one's thought.
In his account of self-knowledgeBurge says: "Authority
concerns those aspects of the thought which have intentional
(aboutness)properties.For me, those are the only aspectsof the
content of a thought"(Burge, 1988,p. 74). So Burge'sstrategy
is not bifurcation;for him, thereis only one notion of content,a
semantic one. His aim is not to distinguishbetween two different kinds of content, and to identify one which falls under
the privilegedepistemicauthorityof the subject.Rather,Burge
seeks to identify, among the various ways in which contentful
intentionalstates can be specifiedfrom a first-personpoint of
view, one kindof specificationwhich is not open to challenge,a
kind of self-ascriptionwhich is incorrigible.He claims to have
found this in self-ascriptionsof a second order. Let's take an
examplethat Burgeexaminesin a footnote (p. 74). Considerthe
"cogito-likejudgment" "I judge that I am thinking that my
aunt is charming."Burgeadmits that I can be mistakenabout
the properspecificationof the contentof my thoughtin all sorts
WHAT'S SO SPECIAL ABOUT SELF-KNOWLEDGE?
585
of ways: for instance, maybe the person that I am actually
judging to be charming is not in fact my aunt ("I have no
particularauthority about that"). But even in that case, he
adds, "I am not making a mistake about what I am thinking
about that person;thereis no mistakeabout the intentionalact
and intentionalcontent of the act" (p. 74). But in what sense is
there no mistakeabout the "intentionalcontent of the act?"If
the subject does not have a belief about his aunt at all, but
about someone else who he mistakenlybelievesto be his aunt,
say, his neighbor,doesn't he make a mistakeabout the content
of his beliefwhen he attributesto himselfa beliefabout his aunt
in the second-orderself-attribution?What kind of attribution
of contentis this second-orderself-ascriptionwhose correctness
does not dependon what, from an objective,third-personpoint
of view, would be the proper specificationof content?It is, of
course, a de dicto attributionof content.
There are two differentstyles of specifyingcontent:de dicto
and de re (see, e.g., Brandom 1994, pp. 520ff). In the de dicto
style, the ascriberattributesa commitmentto a contentwhichis
specifiedin the subject'sown terms,not in those of the ascriber.
As Brandomputs it, "de dicto ascriptionsspecify the contents
of attributed commitments in terms that, according to the
ascriber, the one to whom they are ascribed would acknowledge
as specificationsof the contents of commitmentsundertaken"
(p. 523). In de dicto ascriptions, the ascriber undertakesno
commitmentas to what the properspecificationof the content
attributedwould be, all things considered;the only commitment that the ascriberundertakesis that the content so specified is somethingthat the subjectto whom it is ascribedwould
say, a (possible,if not actual)dictumof that subject.In Burge's
example, a de dicto ascriptionwould be "He thinks that his
aunt is charming."In this ascriptionthereis no commitmenton
the part of the ascriberas to whetherthe personwho the subject
considerscharmingis actuallyhis aunt, or whethershe is better
described as his neighbor. A de re ascription would be "he
thinks of his neighborthat she is charming."In the de re style
of specifyingcontent, the ascriberattributesa commitmentand
at the same time undertakesone herself:in our example, the
586
JOSE MEDINA
commitment about the content attributed that the ascriber
undertakes is that the subject's thought is a thought about his
neighbor. Burge's treatment of second-order self-ascriptions
clearly emphasizes their de dicto character. This is what he
seems to underline when he says that these self-ascriptions do
not carry with them "an ability to explicate correctly one's
thoughts or concepts via other thoughts and concepts" (Burge,
1988, p. 78; emphasis added). Indeed, if - as Brandom (1994)
puts it - de dicto ascriptions use "only what the agent is aware
of or acknowledges" (p. 525), then all self-ascriptions are de
dicto. For when it comes to ascribe contents to oneself, there is
nothing one can go by other than what one is aware of or
is willing to acknowledge. In self-attributions the de dicto style
is not an option, but a necessity, since here the perspective of
the ascriber and that of the person to whom a content is
ascribed coincide.
So, given that Burge's "cogito-like judgments" are de dicto
attributions of content, we can now understand why their
correctness does not depend on what the right specification of
that content would be, all things considered. We can also
understand now why this analysis does not require distinguishing between two different notions of content (broad and
narrow), but only distinguishing between two different styles of
specifying content (de re and de dicto). But the de dicto character of second-order self-ascriptions falls short of showing that
there is something special about self-knowledge. Burge
emphasizes that the correctness of second-order self-ascriptions
in no way depends on our ability to individuate intentional
states from a third-person point of view (see p. 78). This is true,
but it is true of all de dicto ascriptions, whether they are of firstorder or of second-order, and whether they are self-ascriptions
of content or ascriptions of content to others. And the reason is
trivial: they do not involve a commitment on the part of the
ascriber as to what the correct specification of content would be
from an objective point of view. So what is it that Burge takes
to be so special about self-ascriptions of a second order?
The obvious difference between a first-order self-ascription
such as "I believe that p" and a second-order self-ascription
WHAT'S SO SPECIAL ABOUT SELF-KNOWLEDGE?
587
such as "I think that I believe that p" is that in the formerthe
content that occurs de dicto is "p," while in the latter the content that occurs de dicto is "I believe that p." That is, in the
second-orderself-ascriptionthe propositional-attitudeverb itself ("believe")occurs de dicto. It is not just the content of an
intentionalstate, but the intentionalstate itself, that is ascribed
de dicto, that is, as something that the subject to whom it is
ascribedis awareof or is willingto acknowledge.But what is so
special about this?After all, there can also be second-orderde
dicto attributionsfrom a third-personpoint of view: e.g., "He
thinks that he believes that p." In these second-orderattributions, whetherto oneself or to another, the only commitment
that the ascriberundertakesis that, under the appropriatecircumstances, the subject would avow "I believe that p"; no
commitmentis undertakenas to whether"p" correctlyspecifies
the intentionalcontent ascribed,or as to whetherthe propositional attitude of the subjecttowards that content is correctly
characterizedby "believes"(ratherthan by "doubts,""desires
to believe but doesn't," etc.). The correctnessof these secondorderascriptionsrests only on the subject'savowal of the firstorderintentionalstate ascribed.When the subjectto whom the
first-order intentional state is ascribed de dicto is not the
ascriber,there is always the possibilitythat the subjectwill not
avow the dictumascribedto him. The ascriptionis therefore
open to challenge;and when challenged,it requiresa discursive
justification.By contrast,second-orderde dicto self-ascriptions
do not seem to be open to challenge in this way: here mismatches between the perspectiveof the ascriberand the perspectiveof the subjectto whom a first-orderthoughtis ascribed
cannot occur, simply because they are one and the same perspective.In self-ascriptionswhat, accordingto the ascriber,the
subject will acknowledge and what the subject actually
acknowledgescoincide.
So there seems to be something special about the secondorder de dicto self-ascriptionsthat Burge calls "cogito-like
judgments."What is special is that in a self-judgmentof this
kind, such as "I think that I believethatp," the dictumascribed
is already acknowledgedin the ascriptionof the dictum.In a
588
JOSE MEDINA
self-ascription,the de dicto attributionis at the same time an
acknowledgement of the dictum. So what distinguishes selfascriptionsof intentionalcontents or of contentfulintentional
states from their ascriptionsto othersis that the formerare not
mere attributionsof commitments:they are also (and simultaneously) acknowledgements of those commitments. Selfascriptionshave a hybridcharacter:they have both a reporting
aspect and a performativeaspect. This performativeaspect,
introducedby the use of "I," is what has led manyphilosophers
(Burge included) to view self-ascriptionsas a special kind of
attribution.But, in their reportingaspect, as de dicto attributions, self-ascriptionsare no differentfrom de dictoattributions
from a third-personpoint of view. The surplusof self-ascriptions comes from their performativedimension:a self-ascription is not only the attributionof a commitmentbut also "the
undertaking by acknowledging of a commitment" (Brandom
1994, p. 553). The asymmetrybetween first-personand thirdperson ascriptions reflects what Brandom calls "the irreducibilityof the deonticattitudeof acknowledgingcommitmentsto
that of attributingthem": "undertakingcommitmentscannot
be reduced to attributing them, even to oneself. [...] these are
two differentkinds of practicaldoing" (p. 554).
However, according to Burge, the impossibilityof a mismatch between the commitments that a subject ascribes to
herself and the ones that she actually has is guaranteedby a
self-verifyingmental act in which one recognizes"a first-order
thought [...] as one's own" (Burge, 1988, p. 75). But no such
cogito-like mental act seems to be needed for first person
authority.A mismatchbetweenthe commitmentsthat a subject
ascribesto herselfand those that she undertakesis ruledout by
the logic of self-ascriptions.There can be no mismatchsimply
becausein a self-ascriptionthe subjectdoes not merelyattribute
a commitmentto herselfbut also undertakesit. And notice that
this is true of all self-ascriptions,whether of first-orderor of
second-order.In normal circumstances,the use of the firstperson pronoun "I" has the expressiverole of acknowledging
commitments.In the self-ascriptionof first-order"I believethat
p," the subject acknowledges and hence undertakes a
WHAT'S SO SPECIAL ABOUT SELF-KNOWLEDGE?
589
commitment(the belief that p). If that self-ascriptionis consideredas a de dictoascription,its correctnessis guaranteedjust
because the dictumascribed("p") has already been acknowledged by the subjectin the very act of self-ascription.And the
same goes for second-orderascriptions.The only differenceis
that in a second-orderself-ascription(e.g. "I believe that I
believe that p"), a further commitmentis acknowledgedand
hence undertaken(the belief that I believe that p), a further
dictum("I believethat p") is avowed. But nothing is gainedby
going to a metalevel;in particular,one does not acquiremore
authority on one's thoughts with this move. For first-person
authoritydoes not dependon a complexact of innerperception
in which a first-orderthought becomes the object of introspection. Burge himself seems to suggest that it is the logic of
self-judgments (their self-referentiality)that supports their
"authoritative"character(e.g. Burge, 1988,p. 75). But if that is
true, it is not clear why we need to postulate a self-verifying
mental act underlyingeach authoritativeself-judgment.Such
an act seems superfluous (an "idle wheel"). First person
authoritydoes not rest on a specialepistemiccapacity,on "our
second-order,self-ascriptivepowers" (Burge 1988, p. 72), but
rather, on our practical capacity to undertake commitments by
acknowledgingthem. There is nothing special about selfascriptionsbeyond the hybrid characterthat they have in our
discursivepractices.
It is true that self-ascriptionsdo not usuallystand in need of
discursivejustification. But they are not self-verifying.They
have only a defaultstatus. Because of its performativeaspect,
because it is an acknowledgement,a self-ascriptionis usually
regardedas primafacie evidencefor the subject'scommitment
and entitlementto the intentionalstate she claims to have. But
of course acknowledgementdoes not guarantee entitlement.
The correctnessof the reportingaspect of a self-ascriptionsis
not guaranteed by its performative aspect. Although
self-ascriptionsare simultaneouslyattributionsand acknowledgementsof commitments,what a subjectacknowledgesin a
self-ascription may be at odds with everything else she
acknowledges;and when this is pointed out to the subject,she
590
JOSE MEDINA
must vindicate discursivelyher entitlementto the intentional
state she ascribesto herself. If, for instance, evidence is cited
that the subjectdesiresto believe that p but doesn't really believe that p, then the subjectmust defendher claimdiscursively,
that is, she must give evidencein supportof her claim that she
actually believes that p. When an appropriatechallenge has
been raised, the subject does not earn entitlementsimply by
saying "I think that I believe that p." A second-orderinsight
into one's own intentional states does not help here. For, by
definition,self-deceptioncannot be ruled out by introspection.
So self-ascriptionsinvolve a discursivecommitmentto defend
the content of the ascriptionwhen appropriatelychallenged.
To say that self-ascriptionshave a defaultstatus is simplyto
say that they are usually taken at face value: unless there is
positive evidence to the contrary, they are not challenged.
Whenan agentacknowledgesa commitment,entitlementto it is
grantedby default(this is what Brandomcalls the "defaultand
challengestructureof entitlement";see Brandom,1994,esp. pp.
177ff).But this presumptionof first-personauthorityfalls short
of the infallibilitythat Burge attributesto "basic self-knowledge." Therecan be cases of self-deception.If the interlocutors
do have evidenceat odds with the subject'sself-ascription,this
can be challenged;and when it is, the subjecthas to earn discursivelyhis entitlementto the commitmentundertakenin his
self-ascription.But Burge and other critics of Ryle's inferentialist paradigm are right that one's authority on what one
thinks (believes, desires, etc.) is a noninferentialauthorityin
some ways similar to the authority on what one perceives.
Thoughinferentiallyredeemable,self-judgmentsare not usually
inferentiallyderivedfrom the evidenceone has about oneself.
They usually are direct and spontaneous.But the authorityof
self-judgmentsdoes not derivefrom infallibleintrospectiveacts.
As Brandom (1994) has suggested (see esp. p. 558), the presumptionof first-personauthorityis contingenton the logic of
self-ascriptionsinstituted by our discursivepractices.To have
first person authority,to be a reliablereporterof one's intentional states, is to be a responsibleagent in those practices.To
WHAT'S SO SPECIAL ABOUT SELF-KNOWLEDGE?
591
the link betweenself-knowledgeand responsibleagency I now
turn. Can this link give a special status to self-knowledge?
3. THE NORMATIVE SIGNIFICANCE OF SELF-ASCRIPTIONS:
SELF-KNOWLEDGE AND RESPONSIBILITY
Unlike Burge,Davidson(1984, 1986)has triedto accountfor the
special status of self-knowledgeoutside the Cartesianframework. Davidson's claims about first-personauthorityare antiCartesianin at least two respects.In the first place, Davidson
does not ground the subject's authority over her intentional
attitudes in self-verifyingintrospective acts or the intuitive
powers of the mind. Instead, he insists that "first-person
authorityis dependenton, and explainedby, social andpublic
factors" (Davidson, 1986, p. 453). And in the second place,
Davidson holds a much weaker view of first-personauthority
than Descartes and his followers (including Burge and his
"limitedCartesianism").For Davidson, first-personauthority
does not amount to incorrigibility:"we do not always have
indubitableor certainknowledgeof our own attitudes.Nor are
our claimsabout our attitudesincorrigible.It is possiblefor the
evidence available to others to overthrow self-judgments"
(Davidson, 1984,p. 103).But nonetheless,Davidson treatsselfascriptionsof intentionalattitudesas a specialkindof attribution
(e.g., Davidson, 1984, p. 108). For Davidson, first-person
authority is a necessary presupposition of interpretation: "an
unavoidablepresumptionbuiltinto the natureof interpretation"
(Davidson, 1984,p. 111). This is what he termshis "constitutive
thesis."It is withinthis Davidsonianframeworkthat Bilgrami's
account of self-knowledgeis developed.
Bilgrami(1992a,b and 1998) has emphasizedthat there is a
"constitutive"relationbetweenself-knowledgeand responsible
agency. He has argued that first-personauthority has to be
groundedin "considerationshaving to do with moral responsibility rather than epistemology" (Bilgrami, 1992a, p. 250).
Once we abandonthe idea that "beliefs,wishes and doubts are
entities in or before the mind," as Davidson recommends
592
JOSE MEDINA
(Davidson, 1986, p. 454), Burge's claim that the ultimate
ground of our "epistemic right" to self-knowledge is the
immediacy of our intentional states in introspectiveacts becomes suspicious.Bilgramiarguesthat the specialcharacterof
self-knowledgehas to be grounded in the presuppositionsof
our normative practices. His account of self-knowledge
emphasizesthe normativedimension of first-personauthority
and its link to the notion of responsibility. According to
Bilgrami,our authorityon our intentionalstates is inextricably
interwovenwith our status as responsibleagents.His argument
goes as follows: (1) "we reactto each other in moral terms";(2)
"these reactionspresupposethat we hold each other responsible"; and (3) in order to be responsible,"we must, in general,
know our beliefs and desires and our intentions because it is
these states which bring about and explain our doing"
(Bilgrami,1992a,p. 250).
Bilgrami's account underscores crucial aspects of selfknowledge that have been largely ignored. But it is not clear
that the normative dimension of self-knowledgeand its link
with responsibleagency are sufficientgrounds to establishthe
special characterof self-knowledge.Bilgramiwants to exploit
the "constitutive" relation between self-knowledge and
responsibilityto give a "substantialgrounding"to Davidson's
claim that self-knowledgeis a necessarypresuppositionof our
interpretativepractices(see Bilgrami,1992a,p. 250). Puttinghis
normativeaccount at the serviceof this task, Bilgramimisunderstands the nature and scope of the link between selfknowledge and responsibility:on the one hand, by claiming
that "self-knowledgeis a necessary condition for the implementation of practicessurroundingassignmentof responsibility" (Bilgrami, 1998, p. 215), he seems to intellectualize
responsibleagency;on the other hand, by claimingthat its link
with responsibleagency is something exclusive of self-knowledge, he seems to miss the general normative aspects of
knowledge, groundingthe special characterof self-knowledge
in somethingwhich (on closer inspection)may turn out to be
common to all knowledge.
WHAT'S SO SPECIAL ABOUT SELF-KNOWLEDGE?
593
Bilgrami(1998) argues that the special significanceof selfknowledgeis practicaland ethical and can only be justifiedon
normative grounds. He summarizeshis position as follows:
"Self-knowledgeis necessary for responsibilityfor no other
reason [...] than that our evaluative justifications of the
practices of assigning punishmentand blame seem to be apt
only when self-knowledgeis present" (p. 215). Bilgramicontends that the following conditional (which is supposed to
capturethe core idea behindfirstpersonauthority)is true:"If I
believe that I believe that p, then I believe that p." He argues,
however, pace Burge, that this conditional stands in need of
qualification.The "qualifyingcondition" that Bilgrami proposes is that the conditions for responsibleagency be fulfilled.
According to Bilgrami, whenever a subject is a responsible
agent, we must assume that she has the intentional states she
claims to have; for to assume otherwisewould be to question
the subject's status as a responsible agent. For Bilgrami, to
regardsomeone as a responsibleagent is to attributeto her a
commitmentto knowing her beliefs, desires,or intentions;and
to be responsiblefor an action is to be accountablefor that
action, that is, to be able to explainit (if need be) in terms of
one's beliefs, desires, or intentions. Therefore, Bilgramiconcludes, self-knowledge is a necessary conditionfor responsibility.
But does this claim really amount to a special and unique
normativerelation betweenself-knowledgeand responsibility?
Being responsiblefor one's words and actions is certainly
requiredfor self-knowledge(in fact, for knowledgein general).
In order to have self-knowledge,one must be a responsible
agent. However, the conversedoes not seem equally plausible.
In order to be a responsibleagent, must one have self-knowledge? Perhaps, but what kind of self-knowledge?,and under
what conditions?Unless properlyqualified,the claim that selfknowledgeis a necessarycondition for responsibilityseemstoo
strong. Doesn't it sufficeto have some self-knowledgesome of
the time?Or do we alwayshave to befully knowledgeableabout
our beliefs, desires,intentions, etc. in order to be responsible?
The latter seems to be Bilgrami'sclaim. Accordingto Bilgrami,
in orderto be responsibleat all, one must be able to accountfor
594
JOSE MEDINA
one's actions (when called upon them) in terms of her beliefs,
desires or intentions. He is certainlyright that in order to be
responsible,one must in some sense know what one is doing.
But knowing what one is doing in the minimal sense required
for being responsibleof one's actions should not be equated
(and usually isn't) with knowing one's beliefs, desires, or
intentions and being able to explain one's actions through
them.We do not exemptagentsfrom responsibilityjust because
they do not have the proper knowledge about the beliefs,
desiresor intentionsthat led to theiractions.Beingaccountable
for one's actions is not the same as being able to give a correct
account of one's actions, as Bilgramiclaims (see esp. 1992, p.
250). It is not clear that one ceases to be a responsibleagent
when one lacks self-knowledge.Cases of insanityaside, we still
hold someone responsiblefor her actions even if she has mistaken beliefs about why she did what she did, and even if she
simplydid it blindlyand has no clue of why she did it. Bilgrami,
however, defines responsible agency as agency explicable in
terms of the intentional states that the agent claims to have.
When agency is so defined, it is of course true (but just by
definition) that if the conditions for responsible agency are
fulfilled,the agent actuallydoes have the intentionalstates she
claims to have. For these conditionsare fulfilledonly when the
agent actuallyhas the intentionalstatesthat she claimsto have!
If self-knowledge is incorporated into the very notion of
responsibleagency,then (but only then) it is trivialthat "under
the conditions of responsibleagency," self-knowledgeis guaranteed.
There is indeed an important connection between selfknowledgeand responsibleagency. But this connectionis less
rigid and more generalthan Bilgramiseems to think. And the
importantpoint is that he could not ground the special character of self-knowledgein its connection with responsibilityif
this were a connectionthat holds wheneverwe have knowledge
about something. Unless he shows that responsibilityis involved in self-knowledgein a way in which it is not involvedin
any other kind of knowledge,he hasn't shown anythingspecial
about self-knowledge.That "epistemic responsibility"is in-
WHAT'S SO SPECIAL ABOUT SELF-KNOWLEDGE?
595
volved in one way or another in any instance of knowledge
seems to be suggested by the justification condition usually
imposed on knowledge.A belief amountsto knowledgeonly if
it is justified;and that means that its possessor must be "epistemically responsible"and have the adequate evidence. As
Brandom (1994, pp. 202ff) puts it, knowledge is a hybrid
deontic status that involves both commitmentand entitlement.
When a knowledge claim is appropriatelychallenged,the asserterhas a justificatoryresponsibility:she must vindicateher
commitmentby demonstratingentitlementto it. This kind of
responsibilitydoes not distinguish self-knowledgefrom any
other kind of knowledge.In this sense, all knowledgeis equally
normative.
But Bilgrami'sidea seems to be that in the case of selfknowledge it is not just a specific epistemic or justificatory
responsibility,but the very status of being a responsibleagent
that is at stake. Accordingto Bilgrami,it is the very status of
being a responsibleagent that groundsthe commitmentto firstperson authority. Being a responsible agent is a normative
status that involves normative expectations. Bilgrami (1998)
emphasizesthat we blame agents for failing to live up to their
commitmentto self-knowledge,but we do not blameagentsfor
their lack of knowledgeof the environmentor from their lack
of knowledgeof other people's intentionalstates (we can only
"blame" causal mechanisms in those cases). According to
Bilgrami (1998), there is a "philosophicalground" (i.e. the
"constitutive"relation between self-knowledgeand responsibility) upon which we can alwaysdemandself-knowledgefrom
a responsible agent, and this is what makes self-knowledge
special. For, he contends, there is no corresponding"philosophical ground" that supports a similar demand for other
types of knowledge.According to Bilgrami,there are a priori
reasonsto demandthat a responsibleagent have knowledgeof
her intentional states; but we can only demand that she have
knowledgeof the environmentor knowledgeof the intentional
states of others under special circumstances.Epistemic demands in these cases can only be supported by contextual
596
JOSE MEDINA
considerationsregardingthe agent's needs, desires,intentions,
etc.
So Bilgrami (1998) argues that while a lack of knowledge
about, for example,Indianhistoryin no way underminesone's
status as a responsible agent, a lack of self-knowledgedoes
disqualifythe subject as a responsibleagent. But it may be a
mistaketo separateradicallyone's statusas a responsibleagent
in discursivepracticesfrom the specificresponsibilitiesone has
with respect to specific commitments.A lack of knowledge
about Indian history does disqualify the subject in some respect:it does curtailher responsibilitiesin discursivepractices,
namely,as an epistemicallyresponsibleand reliableinformerof
Indian history. On the other hand, a lack of self-knowledge
does not (or at least not always) disqualify the subject as a
responsibleagent in all respects,but only as an epistemically
responsibleand reliablereporterof her own intentionalstates.
Lack of self-knowledge(of the explanatorykind) shows nothing more than that the agent is just not good at explainingher
own mind, but not that she is not good as a responsibleagent
tout court. It is true, however, that a total lack of self-knowledge is sufficientto disqualifysomeone as a competentparticipant in discursivepractices.But so is a total lack of knowledge
of the environmentor a total lack of knowledgeof the intentional statesof others.When an agent is systematicallydeceived
about her own intentionalstates, we do not considerher as a
responsibleagent. But, similarly,when an agent is systematically wrong about the environmentor about other people's
beliefs, desires and intentions, we do not considerher a competent agent in discursive practices. This point has been
emphasizedby Wittgenstein'sholistic view of knowledgein On
Certainty.
In On CertaintyWittgensteinemphasizesthat there are certain things I cannot doubt about myself. Some of these things
have to do with my mental states:I cannot doubt when I have
certainfeelingsor sensations(e.g., when I am in pain), or when
I have certainbeliefs, desires,and intentions.But other things
about myself that seem to be indubitableare factual matters:
what my name is (?328),or that I have two hands(?250),or that
WHAT'S SO SPECIAL ABOUT SELF-KNOWLEDGE?
597
I have never been in the stratosphere (?222), or in China (?339)
or Bulgaria(?269)for that matter.Doubting these thingswould
result in "madness" (?281). But the same is true if I doubt
certain basic facts about the world and about the people
around me. There are certain doubts and certain beliefs that
seem irreconcilable with rationality - for example, the belief
that cars grow out of the earth (?279) or that people have
sawdustin their heads (?281).As Wittgensteinputs it:
In certaincircumstancesa man cannot make a mistake.[.. .] If Moore were
to pronouncethe opposite of those propositionswhich he declarescertain,
we shouldnot just not sharehis opinion:we shouldregardhim as demented.
(Wittgenstein,1969,?155;emphasispreservedand added)
It is indeed remarkablethat our status as rational agents and
thinkers seems to depend on not doubting certain things, on
being able to hold on to some certainties.But as Wittgenstein
insistentlyargues,we should not concludefrom this that these
certaintiesare a special kind of infallible knowledge. In fact,
they are not knowledgeat all. They are, rather,the presuppositions of our knowledge;what needs to be taken for granted,
to stay put or "held fast" (?144),for our epistemicpracticesto
get off the ground. These certaintiesconstitutethe background
against which our investigativepractices are carried out and
our knowledgeclaims are raised, challenged,and defendedor
defeated. These backgroundcertaintiesdo not have a special
epistemic status, but they do have a special relation to rationality: "The reasonable man does not have certain doubts."
(?220) On Wittgenstein'sview, reasonable thinking and reasonable actingpresupposethe holistic aspect of knowledgeand
thought. We do not simply believe one thing at a time; our
thoughtsand actionsrequirea systemor framework:"a totality
of judgments"(?140),"a whole systemof propositions"(?141).
This totality or systemcontains all kinds of certaintieswe have
accumulatedthrough our own experienceand through that of
others, certaintiesabout the world, about ourselves,and about
the people around us. This is how Wittgensteindescribesthis
holistic frameworkof empiricalcertainties:
598
JOSE MEDINA
Thereare countlessempiricalpropositionsthat count as certainfor us. One
such is that if someone'sarm is cut off it will not grow again. Another,if
someone'sheadis cut off he is dead and will neverlive again.Experiencecan
be said to teachus thesepropositions.However,it does not teachus themin
isolation: rather, it teaches us a host of interdependentpropositions.
(Wittgenstein,1969;??273-274)
The holistic nature of our system of beliefs or frameworkof
judgments is of utmost importance,both for what is in the
foregroundof this system or frameworkand for what is in the
background.My beliefs about the world can only be formed
and made sense of in connectionwith many other beliefsabout
the world as well as backgroundcertaintiesthat "stand fast"
(see ?116). The latter may include propositions such as "The
earth has existed long before my birth" (?84);the former include all kinds of beliefsI have about how things behavein the
surroundingsI am familiarwith. Similarly,my beliefsabout my
fellowhumanbeingsalso constitutea system.Eachbeliefin this
system presupposesgeneral certaintiesthat go unquestioned:
for example,that people have heads, or that they have parents,
or that they have feelings, sensations,emotions, etc. But each
belief in this systemalso presupposesmany other beliefs about
people's physical and mental states, their production,interaction, consequences,etc. Subjectscan be wrong about any of
their individualbeliefs about the world and about the people
around them. Here they can make mistakes, individualmistakes, although they rarely do and by and large we can trust
them on these mattersand they can trustthemselves(see ?150
and ?159). But for them to be wrong about all or most of their
beliefs about the world and about their fellow human beings
would be for them not to have a meaningfultotality or belief
systemabout thesemattersat all. This would disqualifythem as
competentthinkersand agents,for it would leave them without
a system or frameworkwith which to think and act, just as
doubting the backgroundcertaintiesof their beliefs would do.
Discarding most of our individual beliefs or rejecting the
backgroundcertaintieson which we rely would have the same
paralyzingeffect and for the same reasons:it would disablethe
entire belief system and renderit ineffective.The same can be
WHAT'S SO SPECIAL ABOUT SELF-KNOWLEDGE?
599
said about the beliefs concerningour own selves and minds.
They also form a system: a system that is sustained by tacit
backgroundcertainties(e.g. that I have a body, that I have a
past, that I have sensations,emotions,beliefs,desires,etc.); and
a system that consists in a complex network of interrelated
beliefsabout oneself. (And it is worth noting that, accordingto
Wittgenstein, the belief system that makes self-knowledge
possibleincludesnot only beliefsabout one's mentalstates, but
also beliefs about factual matters such as one's name, where
one has been in one's lifetime,etc.) By and largewe are correct
about ourselves and about the contents of our minds. And
indeed,as Bilgramihas argued,this presumptionof authorityis
guaranteedby our status as rational and responsibleagents.
Some self-knowledge is indeed required by rationality and
responsible agency. But the same is true of other types of
knowledge,such as knowledgeof the world and knowledgeof
our fellows and their minds: they are also requiredby rationality and responsibleagency;and no one can be considereda
rational and responsibleagent unless she has minimalknowledge about the empiricaland the social world, that is, some
correct beliefs about her surroundings.Indeed, it is a trivial
truth that, in the light of a massivecognitivefailure(no matter
in what area), people cease to be considered rational and
responsibleagents.
In short, our general competenceas rational thinkers and
agents involves minimal knowledgeabout ourselvesas well as
about the world around us and the people around us. For
normalfunctioningagentswho are consideredto be competent
in our practices,there is the presumptionof minimalepistemic
authority in all the differentareas of ordinaryknowledge requiredfor everydayactivities- an automaticpresumptionthat
of course does not extend to expertknowledge.Insofaras they
are consideredto be rational thinkersand responsibleagents,
subjects are presumed to have some authority about those
things that are within their power to know. In everydayaffairs
epistemicauthorityis the defaultstatusof competentsubjects,a
status that is not lost until there is evidence to call it into
question. Enjoying the status of a rational and responsible
600
JOSE MEDINA
agent presupposesthat we minimallyknow what goes on with
us, with our fellows, and in the world. One cannot lose all
reliabilityand epistemic authority and retain this status. Rational and responsibleagents must have some correct beliefs
about the subjective, intersubjective,and objective domains.
Rationality and responsibleagency require some first-person
authority,some social authority,and some empiricalauthority
about the world around us. The presumption of epistemic
authority in these areas reflects simply the expectation of
minimal competence that derives from the presumption of
rationalitythat applies to all those who are treatedas responsible agents.Thismakes self-knowledgeno more special than
other types of knowledge.
In conclusion, Bilgrami is certainly right that being a
responsibleagent is a normativestatus that involves normative
expectations concerning self-knowledge(in fact, expectations
concerning knowledge generally). But the only epistemic demand concerning self-knowledge that can be derived from
someone'sstatus as a responsibleagent is that she should have
at least some self-knowledge.And this does not make selfknowledge special, for the status of being responsible also
supportsthe normativeexpectationsthat the agenthave at least
some knowledgeof the environmentand at least some knowledge about other people. Beyond these general normative
expectations,in any particularcase, the agent'scommitmentto
self-knowledgeand her interlocutors'right to demand it are
always contextually bound,just as in the case of any other kind
of knowledge. It is not the case that a responsible agent is
always committedto having self-knowledgeand her interlocutors alwaysjustifiedin demandingit.
Bilgrami'sproposalto considerself-knowledgein connection
with responsibleagency has the attractivefeature of locating
self-knowledgein the practical domain, in the space of our
normativepractices,thus providinga hint as to how to explain
the pragmaticsignificanceof the presumptionof first-person
authority. But the way he unpacks this proposal has the
unwelcome (antipragmatist)consequence of intellectualizing
responsible agency, making it dependent on self-knowledge.
WHAT'S SO SPECIAL ABOUT SELF-KNOWLEDGE?
601
More importantly,Bilgrami'snormativeproposalis not general
enough. For what Bilgrami takes to be the "substantial
grounding"(Bilgrami,1992a,p. 250) of the specialcharacterof
self-knowledge(i.e. its normativedimensionand its connection
with responsibility)is in fact a generalfeatureof knowledge.It
is not just self-knowledge,but all knowledge in general, that
should be consideredin the context of our normativepractices
and viewed as involvingresponsibleagency in discursivepractices. So what's so special about self-knowledge?
CONCLUSION
What indeedis, in the end, so special about self-knowledge?
Nothing so far considered, my discussion shows: neither a
special epistemic status nor a constitutive relation with
responsibility absent in other types of knowledge. Burge's
"limited Cartesianim"fails: first-personauthority cannot be
vindicatedthrough introspectiveacts in which our intentional
statesare recognizedand identifiedas such at a metalevel.What
makes self-knowledgespecial is not the incorrigibilityof the
intuitive powers of our mind when it turns inwards. On the
other hand, Bilgrami'sattempt to find a specialplace for selfknowledgein a non-Cartesianpicturealso fails:the constitutive
relation between this type of knowledge and the notion of
responsibleagencycan be found in other types of knowledgeas
well. But these failures are informative. In particular,there
are two positive conclusions that my critical discussion has
established.
In the first place, as my critique of Burge has shown, the
asymmetrybetween self-ascriptionsof intentional states and
ascriptionsto othersderivesfrom a performativedifference:selfascriptions, unlike ascriptions to others, are simultaneously
attributionsand acknowledgementsof commitments- hence
theirhybridcharacter:they reportthe subject'scommitmentsto
others, but at the same time they performwhat they reportby
undertakingthe commitmentsthroughacknowledgement.The
peculiar performativeforce that self-ascriptionshave derives
602
JOSE MEDINA
from their self-referentiality. This peculiar character is not
restricted to second-order cogito-like judgments or metastatements about our mental states, as Burge has claimed; and
it does not amount to any kind of special epistemic privilege.
In the second place, the normative dimension and the constitutive relation with responsibility which Bilgrami has claimed
to be the special features of self-knowledge, are actually the
crucial constitutive features of all knowledge, for to be a
responsible agent is to be a minimally knowledgeable subject.
Responsibility and epistemic competence are bound up with
each other: there is no responsibility unless there is minimal
knowledge about self, others, and the world. Therefore, corresponding to the defeasible presumption of minimal first-person authority, we find the defeasible presumptions of minimal
authority on the world and on our fellow human beings and
their minds. These presumptions, these different forms of epistemic authority, are part and parcel of the cognitive competence required for responsible agency. Thus I conclude, pace
Bilgrami, that minimal knowledge about ourselves is as necessary for responsibility as minimal knowledge of our surroundings and of the people around us.
REFERENCES
Bilgrami, A. (1992a): Belief and Meaning, Oxford: Blackwell.
Bilgrami, A. (1992b): 'Can Externalism be Reconciled with Self-Knowledge?', Philosophical Topics 20 (1), 233-267.
Bilgrami, A. (1998): 'Self-Knowledge and Resentment', in C. Wright,
B. Smith, and C. Macdonald (eds.), Knowing Our Own Minds (pp.
207-241), Oxford: Oxford University Press.
Brandom, R. (1994): Making it Explicit, Cambridge, MA: Harvard University Press.
Burge, T. (1979): 'Individualism and the Mental', Midwest Studies in Philosophy 6 (pp. 73-122), Minneapolis: University of Minnesota Press.
Burge, T. (1986): 'Individualism and Psychology', Philosophical Review 95,
3-45.
Burge, T. (1988): 'Individualism and Self-knowledge', Journal of Philosophy
85, 649-63. Reprinted in Q. Cassam (ed.), Self-Knowledge pp. 65-79
(1994), Oxford: Oxford University Press.
WHAT'S SO SPECIAL ABOUT SELF-KNOWLEDGE?
603
Davidson, D. (1984): 'First Person Authority', Dialectica 38, 101-111.
Davidson, D. (1986): 'Knowing One's Own Mind', Proceedings and
Addresses of the American Philosophical Association 60, 441-458.
Fodor, J. (1981): 'Methodological Solipsism Considered as a Research
Strategy in Cognitive Psychology', in Representations, pp. 225-253 Cambridge (MA): MIT Press.
Ryle, G. (1949): The Concept of Mind, Chicago: The University of Chicago
Press.
Wittgenstein, L. (1958): The Blue and Brown Books, Oxford: Blackwell.
Wittgenstein, L. (1968): 'Wittgenstein's Lectures on 'Private Experience' and
'Sense Data", edited by R. Rhees, Philosophical Review 77, 275-320.
Wittgenstein, L. (1969): On Certainty, Oxford: Blackwell.
Department of Philosophy
Vanderbilt University
11] Furman Hall
Nashville
TN 37240, USA
E-mail: [email protected]