Goldman Intern Exp

Journal of Philosophy, Inc.
Internalism Exposed
Author(s): Alvin I. Goldman
Reviewed work(s):
Source: The Journal of Philosophy, Vol. 96, No. 6 (Jun., 1999), pp. 271-293
Published by: Journal of Philosophy, Inc.
Stable URL: http://www.jstor.org/stable/2564679 .
Accessed: 03/12/2012 10:46
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 support@jstor.org.
Journal of Philosophy, Inc. is collaborating with JSTOR to digitize, preserve and extend access to The Journal
of Philosophy.
http://www.jstor.org
This content downloaded by the authorized user from 192.168.52.75 on Mon, 3 Dec 2012 10:46:50 AM
All use subject to JSTOR Terms and Conditions
THE JOURNAL OF PHILO SOPHY

VOLUMEXCVI, NO. 6, JUNE 1 999
INTERNALISM
EXPOSED*
tn recent decades, epistemology has witnessed the development
| and growth of externalist theories of knowledge and
Sjustification.1 Criticsof externalism have focused a bright Spc)tlight on this approachandjudged it unsuitablefor realizingthe true
and original goals of epistemology.Their own favoredapproach,internalism, is defended as a preferable approach to the traditional
concept of epistemicjustification.2I shall turn the spotlight toward
internalismand its most prominent rationale,revealingfundamental
problems at the core of internalismand challenging the viabilityof
its most popular rationale. Although particularinternalist theories
such as (internalist)foundationalismand coherentismwill occasionallybe discussed,those specific theories are not my primaryconcern.
B An earlier version of this paper was presented in Pittsburgh,at the CentralDivision meeting of the AmericanPhilosophicalAssociation,April 2S, 1997. My commentator on that occasion was MatthiasSteup, and I am much indebted to him for
valuable correspondence on this topic. I am also grateful to Tim Bayne and Holly
Smith for veryuseful suggestions.
l Prominent statements of externalism include D. M. Armstrong, Belief,Tr2ttXl
clad Knowledge(New York:Cambridge,1973); Fred Dretske, Kno7lJledge
clnd tXleFI071J
of Informl.tion (Cambridge:MIT, 1981); Robert Nozick, PhilosopXlicalExplclecltions
(Cambridge:Halward,1981); my Elvist.emology
clnd Cognition (Cambridge:Halward,
1986); and AlvinPlantinga, Waw7-rlnt
and Prope>^Function
(NewYork:Oxford, 1993).
Majorstatemenl:sof internalisln include Roderick Chisholm, Theosyof Kno71X1eclge (Ellglewood Cliffs, NJ: Prentice-Hall, 1966, 1st edition; 1977, 2nd edition;
1989, 3rd edition); Laurence BonJour, The Sttucture of Emlvi7ical1
Kno7U1ede(Cambridge: Halward,1985);John Pollock, Conte@orcl7yTZ2.eories
of Kno7lJledge(Totowa,
NJ:Rowmanand Littlefield, 1986); RichardFoley, T/le T/leoryof EpistemicRcltionallit.ffe
(Cambridge:HarsZard,
1987); and IseithLehrer, Theowyof Knowleclge(Botllder:Westview, 1990). In addition to relativelypure versions of externalismand internalism,
there are also mixturesof the two approaches,as found in WilliamAlston, Elvistemic
Justificaltion(Ithaca:Cornell, 1989); Ernest Sosa, Kno7uJleclge
in Perslvective(New York:
Cambridge, 1991); and Robert Audi, T/le Strztctuseofhtstificrltion
(New
York:Cambridge, 1993).
9
0022-362X/99/9606/271-93
(C)1999 The Journal of Philosophy,Inc.

271
272
THEJOURNALOF PHILOSOPHY
The principalconcern is rather the general architectureof internalism, and the attempttojustify this architectureby appeal to a certain
conception of whatjustificationconsistsin.
I. DEONTOLOGY,ACCESS,AND INTERNALISM
I begin with a certain rationalefor internalismthat has widespread

support.It can be reconstructedin three steps:
(1) The guidance-deontological
(GD)conceptionofjustificationis posited.
(2) A certain constrainton the detel-minersof justification is derived
from the GDconception, that is, the constraintthat alljustification
determinersmust be accessibleto, or knowableby,the epistemicagent.
(3) The accessibilityor knowabilityconstraintis taken to implythat
only internalconditionsqualifyas legitimatedeterminersof justification.Sojustiticationmustbe a purelyinternalaffair.3
Whatmotivatesor underlies this rationalefor internalism?Historically,one central aim of epistemologyis to guide or direct our intellectual conduct, an aim expressed ill Rene Descartes'stitle, "Rules
for the Direction of the Mind."4Among contemporarywriters,John
Pollockexpressesthe idea this way:
I have taken the fundamentalproblemof epistemologyto be that of deciding what to believe. Epistemicjustiflcation,as I use the term, is concerned with this problem. Collsiderations of epistemic justification
guide us in determiningwhat to believe. We might call this the "beliefguiding"or"reason-guiding"
sense of'justiElcation'(op. cit., p. 10).
The guidance collception of justification is commonly paired with

the deontological conception of justification.John Locke5wrote of a
person's "dutyas a rationalcreature"(ibid., p. 413), and the theme
of epistemicdutyor responsibilityhas been echoed by manycontem3 Plantinga also traces internalism to the deontological conception: "If we go
back to the source of the internalist tradition,...wecan see that ilaternalismarises
out of deontologw7;
a deontological conception of warrant...leadsdirectly to internalism"(lv cil., pp. 2625). Alston proposes a slightlydifferent rationale for internalism, although his rationale also proceeds via the knowabilityconstraint (olv. cit.,
p. 236). He suggeststhat the concept of justification derivesfrom the interpersonal
practice of criticizingone another's beliefs and askingfor their credentials.A person can appropriatelyrespond to other people's demandsfor credentialsonly if he
knows what those credentials are. So it is quite undel standable,says Alston, that
justifiers must meet the requirement of being accessible to the agent. Clearly,this
is one wayto derive the accessibilityconstraintwithout appeal to the deontological
conception. But Alston is the only one I know of who advancesthis grotlnd for the
accessibilit constraint.In any case, most of the problems I shall identif,vpertain to
the accessibilityconstraint itself, which Alston's rationale shares with the deontological rationale.
4 P/lilosolvXlical
Worksof Descartes, lAolumeI, ElizabethHaldane and G. R. T. Ross,
trans. (NewYork:Dover, 1955).
An Esscl.yConacernisag
Human Unclerstanaslisa$
Volu?leII, A. C. Fraser, ed. (New
York:Dover, 1955).
r)
INTERNALISMEXPOSED
273
porary epistemologists, including Laurence BonJour (op.cit.),Roderick Cllisholm (op. cit.), Carl Ginet, Paul Moser, Matthias Steup,
Richard Feldman, and Hilaly Kornblith.6 Chisholm defines cousins
of the concept of justification in terms of the relation 'more reasonable than', and lle re-expresses tlle relation 'p is more reasonable
than q for S at t' by saying: "S is so situated at t that his intellectual requirement,
his responsibility
as all intellectual being, is better fulfilled
by p tllan by q."7Similarly, Feldman says that one's epistemic duty is
to "believe what is supported or justified by one's evidence and to
avoid believing what is not supported by one's evidence" (op. cit.,p.
254).
The guidance and deontological conceptions ofjusuXicationare intimately related, because the deontological conception, at least when
paired with tlle guidance conception, considers it a person's epistemic
duty to guide his doxasuc atatudes by his evidence, or by wllatever factors determine the justficational status of a propositon at a given time.
Epistemic deontologists commonly maintain that being justified in believing a proposition p consists in being (intellectually) required or permitted to believe p; and being unjusuEiedin believing p consists in not
being permitted, or being forbidden, to believe p. When a person is unjustified in believing a proposition, it is his duty not to believe it.
It is possible to separate the deontological conception from tlle
guidance idea. In etllical theolW,a distinction has been drawn between
accounts of moral duty that aim to specify wllat makes actions rigllt
and accounts of moral duty tllat aim to provide practical decision pr}
cedures for wllat to do.8 If an account simply aims at the first desideratum, it need not aspire to be usable as a decision guide. Similarly,
accounts of epistemic duty need not necessarily be intended as decision guides. When the deontological concepton is used as a rationale
for epistemic internalism of tlle sort I am sketching, llowever, it does
incorporate the guidance conception. Only if tlle guidance concez
G Ginet, KZ.ozlJleclge,
Peacelvtiosl,
.nal MeRzmy
(Dordrecht: Reidel, 1975); Moser,
EmlviricalJ>stification
(Dordrecht:Reidel, 1985); Steup, "The Deontic Conception
of EpistemicJustification,"P/lilosolvXlicrl.l
Stz.dies,LIII (1988): 65-84; Feldmall, "Epistemic Obligations,"in J. Tomberlin, ed., Philosolvhical
Perslvectives,
Volume II (Atascadero, CA: R;dgeview, 1988), pp. 235-56; and Itornblith, "JustifiedBelief and
EpistemicallyResponsibleAction,"P/liloso/vXlical
Reviezu,
XCII (1983): 33-48.
7 T/le0'sy
of Knowleclge,
2nd edition, p. 14 (emphasisadded).
8 For example, R. Eugene Bales distinguishesbetween two possible aims of actutilitarianism:as a specifier of a light-makingcharacteristicor as a decision-making
procedure. See "Act-utilitarianism:
Account of Right-makingCharacteristicsor Decision-makingProcedure,"Aslericczn.
PhilosophisaI
bl.claterly,
VIII (1971): 257-65. He
defends utilitarianismagainst certain critics by saying that it does not hlve to pelform the latterfunction.
274
tion is incorporated can the argument proceed along the intended

lines to the accessibility constraint, and from there to internalism. This
is why I shall henceforth speak of the GD conception ofjustification.
I turn now to the second step of the argument for internalism. Following William Alston,9 I shall use the term justifersfor facts or states
of affairs that determine the justificational status of a belief, or the
epistemic status a proposition has for an epistemic agent. In other
words, justifiers determine whether or not a proposition is justified
for an epistemic agent at a given time. It seems to follow naturally
from the GD conception of justification that a certain constraint must
be placed on the sorts of facts or states of affairs that qualify as justifiers. If a person is going to avoid violating his epistemic duty, he
must know, or be able to find out, what his duty requires. By know,in
this context, I mean only: have an accurate,or true,belief. I do not
mean: have a justifiedtrue belief (or whatever else is entailed by the
richer concept of knowledge). Admittedly, it might be possible to
avoid violating one's duties by chance, without knowing (having true
beliefs about) what one's duties are. As a practical matter, however, it
is not feasible to conform to duty on a regular and consistent basis
without knowing what items of conduct constitute those duties. Thus,
if you are going to choose your beliefs and abstentions from belief in
accordance with your justificational requirements, the facts that make
you justified or unjustified in believing a certain proposition at a
given time must be facts that you are capable of knowing, at that
time, to hold or not to hold. There is an intimate connection, then,
between the GD conception of justification and the requirement that
justifiers must be accessible to, or knowable by, the agent at the time
of belief. If you cannot accurately ascertain your epistemic duty at a
given time, how can you be expected to execute tllat duty, and how
can you reasonably be held responsible for executing that duty?l
The knowabilityconstrainton justifierswhich flows from the GD
conception may be formulated as follows:
KJ:The only facts that qualifyasjustifiersof an agent's believing ,@at
time t are facts that the agent can readilyknow, at t, to obtain or
not to obtain.
" "Internalismand Externalismin Epistemology,"reprinted in his of. cit., pp.
185-226,here p. 189.
"' Some internalistsexplicitly reject externalism on the grounds that it cannot
be used as a decision guide. For example, Pollock says:"[I]t is in principle impossible for us to actuallyemploy externalistnorms. I take this to be a conclusive refutation of belief externalism" (of. cit., p. 134). He would not subscribe to the full
argument for internalismI am discussing,however,because it is committed to the
"intellectualistmodel"of epistemology,which he disparages.
INTERNMISMEXPOSED
275
How can an agent readilyknow whether candidatejustifiers obtain

or do not obtain? Presumably,the agent tnust have a way of determining, for any candidate class of justifiers,whether or not they obtain. Such a way of knowing must be reliable, that is, it must
generate beliefs about the presence or absence of justifiers that are
usually (invariably?)correct. Otherwise,the agent will often be mistaken about what his epistemic duty requires. The way of knowing
must also be "powerful,"in the sense that whenjustifiers obtailait is
likely (certain?) that the agent will believe that they obtain; at least
he will believe this if he reflects on the matteror othexzviseinquires
into it.ll As we shall soon see, internaliststypicallyimpose additional
restrictions on how justifiers may be known. But the minimal,
generic version of KJ simply requires justifiers to be the sorts of
facts that agents have some wayof knowing.In other words,justification-conferringfacts must be the sorts of facts whose presence or
absence is "accessible"to agents.19
Given the KJconstrainton justifiers,it becomes fairlyobviouswhy
internalismaboutjustificationis so attractive.Whereasexternalfacts
are facts that a cognitive agent might not be in a position to know
about, internal facts are presumablythe sorts of conditions that a
cognitive agent can readilydetermine. So internal facts seem to be
the right sorts of candidatesforjustifiers.This considerationleads to
the third step of our rationale for internalism. Only internal facts
qualifyasjustifiersbecause they are the only ones that satisfythe KJ
constraint;at least so internalistssuppose.
One possible way to criticize this rationale for internalism is to
challenge the GD conception directly. This could be done, for example, by arguing that the GD conception of justification presupposes the dubious thesis of doxastic voluntarism, the thesis that
doxastic attitudes can be "guided"by deliberate choices or acts of
will. This criticism is developed by Alston,l3 and I have sympathy
with many of his points. But the voluntarism argument against
the GD conception is disputed by Feldman (op. cit.) and John
" For the distinction between reliability and power (phrased slightly differently), see my op. cit., chapter 6.
') Jack Lyons points out that to comply with one's epistemic duty it suffices to
know that one has (undefeated) justifiers for proposition p; one does not have to
know zvAich
justifiers these are. So the argument is not entitled to conclude that
knowledge of particularjustifiers is required by epistemic duty. Practicallyspeaking, however, it is difficult to see how a cognitive agent could know that relevant
justifiers exist without knowingwhich particularones exist. So I shall pass over this
objection to the internalistline of argument.
'3 "The Deontological Conception of Justification,"reprinted in his op. cit., pp.
115-52.
276
OF PHILOSOPHY
THEJOURNAL
example, argues that epistemic

among others. Feldman, for
Heil,l4
of doxastic volunis not wedded to the assumption
deontologism
force, he points out, even
Many obligations remain in
still
tarism.
discharge them. A person is
an agent lacks the ability to
when
situation
financial
when his
obligated to repay a debt even epistemic obligations have
legally
Perhaps
it.
him unable to repay
makes
topic of doxastic volunproperties.l5Since the complex in its own right, I set
analogous
treatment
would require article-length
tarism
to other issues. Although
attention
my
issue aside and confine
tlais
of justification, I take it as
it
do accept the GD conception
I not
discussion and explore where
present
the
of
purposes
for
given
crucial for internalism is the
In any case, what is ultimately conception hopes to rationleads.
requirement that the GD
accessibility
fails to provide a good rationale,
Even if the GD conception
alize.
rationale could be prowould be viable if some other
internalism
for a suitable accessibilityrequiretnent.
vided
STRONGINTEMISM
II. DIRECTKNOWABILIIYAND
in termsof knowabilityplain
initialKJconstraintwasformulated often add the furtherqualThe
simple,but proponentsof internalism must be directly
knowable
and
that determinantsof justificationwritesas follows:
ificatiola
example,
the cognitiveagent. Ginet,for
by
that minimallysuffices
about S's pOSitiOll
Everyone of everyset of facts
that p must be diconfident
in being
to make S, at a given time,justified cit., p. 34).
to S at that time (o,@.
recognizable
rectly
Chisholmwrites:
Similarly,
immediate in
justificatiollis...internaland
in believ[T]he concept of epistemic
by reflection,whatone isjustified
that one can find out dit^ectly,
ing at any time.lG
KJcondo not endorsejust the minimal which

Thus,Ginet aladChisholm
but a more restrictiveversion,
straintas earlier formulated,
mightbe writtenas follows:
Studies,XL ( I983):355-64. GD conception of
"DoxasticAgency,"Philosophical
undercuts the step from the not require that
however,
response,
does
]5 Feldman's
constraint. If epistemic duty
justificationto the knowability this duty, there is no longer a rationale for the
worry is
the agent be ableto discharge
line of response to the voluntarism
different
A
on benot
constraint.
focus
should
knowability
that epistemologicalanalysis that is subject to the
action
taken by Lehrer,who suggests
of
sort
where acceptance is some
Reidel, 1981),pp.
lief but on acceptance,
Bogdan, ed., KeithLehrer(Dordrecht:
R.
in
Self-Profile,"
"A
will
3-104.
added and original emphasis
3rd edition, p. 7; emphasis
") Theoryof Knozvledge,
deleted.
14
INTERNitISM EXPOSED
277
KJdil:
The only facts that qualifyasjustifiersof an agent's believing17
at time t are facts that the agent can readilyknow directly,at t, to
obtain or not to obtain.
An initialproblemarisingfromKJCIil.
is this:Whatwarrantsthe imposition of KJclir
as opposed to the looser constraint,KJ?KJwas derived
frorllthe GD conception on the grounds that one cannot reasonably
be expected to complywith epistemic duties unless one knowswhat
those dutiesare. How does such an argumentwarrantthe furtherconclusion that directknowledgeof justificationmust be available?Even
indirectknowledge (whateverthat is) would enable an agent to coznply with his epistemicduties. So the second step of the argumentfor
internalismcannot properlybe revisedto featureKJdilin place of KJ.
Proponents of KJclir
might reply that direct forms of knowledge are
more powerfulthan indirect knowledge,but this reply is unconvincing. The powerrequirementwasalreadybuilt into the originalversion
of KJ,and it is unclearlaowdirectnessaddsanythingof significanceon
thatscore.WhetlaerKJCIil.
can be derivedfroznGD is a seriousproblem,
because tlaeargumentfor internalismrests on something like the directnessqualification.I shall say more about this later;for now I set
this point asidein orderto explorewhereI(Jdil.
leads.
Whattnodes of knowledgecount as direct?At least one form of direct knowledge is introspection. A reason for thinking that introspection is what Chisholm means by direct knowledge is that he
restrictsall determinersof justificationto consciousstates:
A consequence of our "internalistic"
theoryof knowledgeis that, if one
is subject to an epistemic requirementat any time, then this requirement is imposed by the consciousstatein which one happens to find oneself at that time (ibid., pp. 59-60).
Since he restrictsjustifiers to conscious states, it is plausible to assume tlaat direct knowledge, for Chisholtn, means introspective
knowledge,and knowledgeby "reflection"coincides with knowledge
by introspection.l7At least in the case of Chisholm,then, KJclil.
might
be replacedby:
KJint:
The only facts tllat qualifyasjustifiersof an agent's believing1wat
time t are facts that the agent can readilyknow byintrospection,at t,
to obtainor not to obtain.
17 Other epistemologists who restrict justifiers to conscious states or discuss
access in terms of introspection include Moser, p. 174; Feldman, "Having
Evidence,"in D. Austin, ed., Philosophical
Analysis(Dordrecht:IRluwer,1988), pp.
83-104; and Audi, "CausalistInternalism,"AmeaiGan Philosophical
Quarterly,
XXVI,4
(1989): 309-20.
278
OF PHILOSOPHY
THEJOURNAL
by introspectionare facts
the only facts that an agent can know
Now,
currentlyin, so
what conscious states he is (or is not)
concerning
under KJint.
justifiers
as
are the only sorts of facts that qualify
these
internalism:
form of internalismmaybe called strong
This
SI:Only facts concerningwhat conscious
arejustifiersof the agent'sbeliefs at t.
states an agent is in at time t
approach to justifiinternalism,however,is an unacceptable

Strong
This is
consequences.
for it has serious, skepticism-breeding,
cation,
given time, the
by the problemof storedbeliefs.At any
demonstrated
memoryratherthan occurmajorityof one's beliefs are stored in
vast
(for example, one's social
data
or active.Beliefs about personal
rent
geography, or about
about
number), about world history,
security
professionalcolleagues, are altheinstitutionalaffiliationsof one's
at a given moment. Furthermostall stored rather than occurrent
one's conscious state at the
more,for almost any of these beliefs,
No perceptualexperience, no
timeincludes nothing thatjustifies it.
entertainedat
memoryevent, and no premisesconsciously such a beconscious
sufficientfor
theselected moment will be justificationallynone of these beliefs is
then,
lief.According to strong internalism,
threatens a drastic
justifiedat that moment. Strong internalism
justified, and
deemed
diminutionin the stock of beliefs ordinarily
thatjustificationis neceshencein the stock of knowledge,assutning
againstthis type of theory.
saryfor knowledge.This is a tnajorcount
his own account of havbecause
Feldmananticipatesthis problem
consciously entertained factors
ing evidence also implies that only
Feldman tries to meet the
haveevidentialforce (op. cit.,pp. 98-99). and dispositionalsenses
threatby distinguishingbetween occurrent
knowledgeratherthanjusof epistemicterms. (He actuallydiscusses terms of justification bein
tification,but I shall address the issue
Feldman is not simply
cause that is the target of our investigation.)occurrent and disposihas
restating the familiarpoint that 'belief'
'justified'is ambiguous
term
tional senses. He is proposing that the
Feldmanapparently
sense.
between an occurrentand a dispositional
most have disposiat
people
claimsthat in the case of stored beliefs,
tlonalJustltlcatlon,not occurrentJustltlcatlon.
First,if havinga dispoThere are two problemswith this proposal.
statesqualifiesas a justifier of
sition to generate conscious evidential
memorialto perceptualdisa belief, whywould this not extend from
awakesfrom a nap but has not
positions?Suppose a train passenger
believing propositions about
yet opened his eyes. Is he justified in
Surely not. Yet he is disthe details of the neighboring landscape?
INTERNALISMEXPOSED
279
posed, merely by opening his eyes, to generate conscious evidential
statesthat would occurrentlyjustifysuch beliefs. So the dispositional

approachis far too permissiveto yield an acceptablesense of 'justified'.l8Second, can an internalist,especiallya strong internalist,live
with the idea that certain dispositionscount asjustifiers?Having or
not having a disposition (of the requisitetype) is not the sort of fact
or condition that can be knownby introspection.Thus, the proposal
to supplement the occurrentsense of 'justified'with a dispositional
sense of 'justified'is simplythe abandonmentof stronginternalism.
III. INDIRECTKNOWABILITY
AND WEAKINTERNALISM
The obvious solution to the problem of stored beliefs is to relax the

KJconstraint:allowjustifiers to be merely indirectlyknowable.This
yields:
KJind:
The only facts that qualit asjustifiersof an agent's believingp at
time t are factsthat the agent can readilyknowat t, either directly
or indirectly,to obtainor not to obtain.
The danger here is that indirect knowledge might let in too much
from an internalistperspective.How are externalistforms of knowledge for example, perceptualknowledge to be excluded?Clearly,
internalismmust propose specific forms of knowledgethat conform
with its spirit. It is fairlyclear how internalismshould deal with the
problem of stored beliefs:simplyallow knowledgeof justifiersto include memoryretrieval.Storedevidencebeliefscan qualifyasjustifiers
becausethe agentcan knowthattheyobtainby the compoundrouteof
first retrievingthem from memoryand then introspectingtheir consciouscontents.Thisyieldsthe followingvariantof the KJconstraint:
KJint+let:
The only facts that qualit asjustifiersof an agent's believingp
at time t are facts that the agent can readilyknow,at t, to obtain
or not to obtain, byinta^ospection
and/ormemory
retrieval.
This KJconstraintallowsfor a more viable form of internalismthan

strong internalism.We may call it weakinternalism,and initiallyarticulate it through the followingprinciple:
WI:Only facts concerning what conscious and/or stored mental states
an agent is in at time t arejustifiersof the agent'sbeliefsat t.
WIwill certifythe justificationof manystored beliefs, because agents

often have other stored beliefs that evidentiallysupportthem. A per18 Feldman might reply that there is an importantdistinction between memorial
and perceptual dispositions;but it is not clear on what basis a principled distinction can be drawn.
280
OF PHILOSOPHY
THEJOURNAL
the capital of the United

who believes that Washington,D.C. is
son
that a map of the U.S. he
mayhave a stored belief to the effect
States
capital. The latter
consulted slaowedWashington as the
recently
internalismis
weak
So
one.
belief is whatjustifies the former
stored
beliefs. Weakinterplagued with the problem of storedjustified
not
because even
seems to be a legitimate form of internalism
nalism
states.
beliefs qualify,intuitively,as internal
stored
strong internalism,it too
than
Althoughweakinternalismis better
Many
offorgottenevidence.l9
severe problems.Firstis the problem
faces
once had adequate evibeliefs are ones for which an agent
justified
At the time of epistemic apdencethat she subsequently forgot.
that is retrievable
she no longer possessesadequateevidence health benefits
praisal,
the
about
story
a
memory.Lastyear, Sallyread
from
of the New YorkTimes.She then
ofbroccoli in the "Science"section
effects. She still reformed a belief in broccoli'sbeneficial
justifiably
her original evidentialsource
tainsthis belief but no longer recalls
or undermining
corroborating
(andhas never encountered either
and, lf
justified,
still
is
belief
sources).Nonetheless, her broccoli
Presumably,this is because her
true,qualifiesas a case of knowledge.
proper. But past acpastacquisition of the belief was epistemically (includingweakininternalism
quisitionis irrelevantby the lights of
mental states are justifiers
ternalism),because only her current events are "external"and
past
relevantto her current belief. All
thereforeirrelevantaccordingto internalism. possess evidence in
currently
It might be replied that Sally does
her backgroundbeliefs, we
of
supportof her broccoli belief. One
rememberswas learned in an
maysuppose, is that most of what she
she not, after all?now have
epistemicallyproper manner. So does she has someevidence, but
groundsfor the targetbelieR Admittedly? Surely not. In a variant
is this evidence sufElcientfor justification? background belief
same
case, suppose that Sally still has the waslearned in an epistemnamely,that most of whatshe remembers
acquired her broccoli belief
ically proper manner but she in fact
the New YoakTimes.So her
from the NationalInquirerrather than
corroborated,in an epistemibroccoli belief was never acquired, or
the indicated current backcally sound manlaer.Then even with
withjustifiablybelieving that
ground belief, Sallycannot be credited is still relevant,and decibroccoli is healthful. Her past acquisition
Gilbert Harman, C/lan.gein Viezv(CamThis sort of problem is discussed by
Justifica"InternalistFoundationalismand the
bridge:MIT, 1986); Thomas Senor, XCIV (1993): 453-76; and Audi, "Memorial
Synthese,
tion of Memory Belief,"
Topics,XXIII ( 1995 ): 31-45.
Justification, Philosophical
19
EXPOSED
INTERNALISM
281
we are consideringthe "epistemAt least it is relevantso long as

sive.
justificationcarriesa true belief
sense of justification,in which
izing"
belief in the healthfulness
distancetowardknowledge.Sally's surelyshe does not know
agood
sense, for
broccoliis not justified in that
sole
of
the NationalInquirerwasher
that
broccoli is healthful given
that
ot lntormatlon.
source
for weakinternalof forgotten evidence is a problem
category
The
strong interfacing
stored beliefs
because, like the problem of
ism
of what is
sector
A large
it threatens skeptical outcomes.
have
nalism,
people
which
for
counted as knowledge are beliefs
ordinarily
their originalevidence.
forgotten
evidence, Steup20offers the
In reply to the problem of forgotten
for memorialstatesto
solution.An additionalrequirement
following
the agent have adequate evia belief that p, saysSteup, is that
justify
"Ifshe had encounfor believingthe followingcounterfactual:
the
dence
she would not have formed
p in a questionable source,
varitered
Inquirer
National
that in the
that p." Steup's suggestion is
belief
for this counterfactual,and
evidence
Sallyfails to have adequate
this
ant,
not justified. My response to
is why her broccoli belief is
to
that
strong
too
is
requireinent
is twofold.First,the proposed quite difficultto get adeproposal
It is
on memoriallyjustified belief.
impose
Second, the procounterfactual.
quateevidence for the indicated weak as well. Sally might have
too
posedrequirement seems
but still be unjustifiedin
counterfactual
the
for
evidence
for
adequate
She Inight have adequate evidence the
holdingher broccoli belief.
and
true
not
is
it
if
being true;but
So
thecounterfactualwithout its
her broccoli belief is not justified.
it,
told
I
as
is
story
the
of
rest
does not work.
Steup'sinternalist-stylesolution weak internalismis what I call the
A second problem confronting WIsaysthat onlyconsciousand
Principle
entretrieval.
of concur7
or
problem
but it does not say that all sets
justifiers,
are
states
mental
stored
as justifiers.21Presumably,which
conjunctionsof such states qualify to be decided by reference to
a matter
resets of such states qualifyis
beliefs can all be concurrently
stored
of
set
certain
a
If
pass
KJillt+l.et.
introspected,then they would
trievedat time t and concurrently asjustifiersunder the principle
and could qualify
rethe test of KJillt+ret,
if they cannot all be concurrently
But
knowability.
indirect
of
it is clear
they would fail the test. Now
trievedand introspectedat t,
cannot
time
a
at
stored credal corpus
that the totalityof an agent's
(see acknowledgmentnote above).
of fact
His proposalwaspart of his commentary
principle that the knowabilityA & B.
the
one would need to reject
fact,
conjunctive
21 Obviously,
the
of
B entail the knowability
A and the knowabilityof fact
2n
282
be concurrentlyretrievedfrom memory.So that set of stored beliefs

does not qualifyas a justifier for purposes of weak internalism.Unfortunately,this sort of belief set is preciselywhat certain typesof internalist theories require by way of a justifier. Consider holistic
coherentism,which saysthat a propositionp is justified for person S
at time t if and only if p coheres with S's entire corpus of beliefs at t
(including, of course, the stored beliefs). A cognitiveagent could ascertain, at t, whether p coheres with her entire corpus only by concurrently retrieving all of her stored beliefs. But such concurrent
retrievalis psychologicallyimpossible.22Thus, the criticallyrelevant
justificational fact under holistic coherentism does not meet even
the indirect knowabilityconstraint,much less the direct knowability
constraint.Here is a clash, then, between a standardinternalisttheory of justificationand the knowabilityrationale under scrutiny.Either that rationale is indefensible, or a familiar type of internalism
must be abandonedat the outset. Nor is the problem confined to coherentism. Internalist foundationalism might also require concurrent retrieval of more basic (or low-level) beliefs than it is
psychologicallyfeasibleto retrieve.
IV. LOGICALAND PROBABILISTIC
RELATIONS
As these last exainples remind us, eveiy traditionalform of internalism involvessome appeal to logical relations,probabilisticrelations,
or their ilk. Foundationalismrequires that nonbasicallyjustified beliefs stand in suitablelogical or probabilisticrelationsto basicbeliefs;
coherentism requiresthat one's systemof beliefs be logicallyconsistent, probabilisticallycoherent, or the like. None of these logical or
probabilisticrelationsis itself a mental state, either a conscious state
or a stored state. So these relationsdo not qualifyasjustifiersaccording to either SI or WI. The point may be illustratedmore concretely
within a foundationalistperspective.Suppose thatJones possessesa
set of basic beliefs at t whose contents logically or probabilistically
support proposition p. This property of Jones's basic beliefs- the
propertyof supportingpropositionp-is not ajustifierunder WI,for
the propertyitself is neither a conscious nor a stored mental state.
Nor is the possessionof this propertyby these Inentalstatesanother
mental state. So WI has no wayof authorizingor permittingJonesto
believe p. Unless Wl is liberalized, no nonbasic belief will be justified, whichwould again threatena seriousform of skepticism.
22 The "doxasticpresumption"irlvokedby BorlJour(0f7.cit., pp. 101-06)
seems
to assume that this is possible, but this is simply arl urldeferldedassumptiorl.Pollock (op. cit., p. 136) also raises the problem iderltifiedhere, though irl slightlydiffererltterms.
INTERNALISMEXPOSED
283
Can this problem be remedied by simply adding the proviso that

all propertiesof conscious or stored mental statesalso qualifyasjustifiers?23
This provisois unacceptablypermissivefor internalism.One
propertyof many conscious and stored mental states is the property
of beingcausedbya reliable
process,
yet surelyinternalismcannot admit
this archetypicallyexternalisttype of propertyinto the class of justifiers. How should the class of properties be restricted?An obvious
suggestionis to include only formal propertiesof mental states,that
is, logical and mathematicalpropertiesof their contents. But should
all forinal properties be admitted?This approach would fly in the
face of the knowabilityor accessibilityconstraint,which is the guiding theme of internalism.Only formal propertiesthat are knowable
by the agent at the time of doxastic decision should be countenanced as legitimatejustifiers under internalism. Such properties,
however, cannot be detected by introspection and/or memory retrieval.So some knowing operations suitable for forinal properties
must be added, yieldinga liberalizedversionof the KJconstraint.
How should a liberalizedKJconstraintbe designed?The natural
move is to add some selected computational operations or algorithms,proceduresthat would enable an agent to ascertainwhether
a targetedpropositionp has appropriatelogical or probabilisticrelations to the contents of other belief states he is in. Preciselywhich
computational operations are admissible?Again, problems arise.
The firstis the problem
of thedoxasticdecisioninterval.
The traditionalidea behind internalismis that an agent isjustified
in believing p at timet if the evidential beliefs (and perhaps other,
nondoxastic states) possessed at t have an appropriate logical or
probabilisticrelation to p. In short,justificationis conferredsimultaneously with evidence possession.Feldman makes this explicit: "For
any person S and propositionp and time t, S epistemicallyought to
believe p at t if and only if p is supportedby the evidence S has at t"
(op. cit.,p. 254). Once the knowabilityconstraintis introduced,however, simultaneousjustification looks problematic.If justification is
contingent on the agent's abilityto know whatjustifiers obtain, the
agent should not be pertnittedto believe a propositionp at t unless
she can know byt whether the relevantjustifiersobtain. Since it necessarilytakessome time to compute logical or probabilisticrelations,
the simultaneitymodel of justificationneeds to be revisedso that an
agent's mental states at tjustify her in believing only p at t + , for
23 More precisely, the corltemplated proviso should say that the possessiorl of
arlypropertyby a merltalstate (or set of merltalstates) qualifies as a justifier. This
readirlgwill be urlderstoodwhereverthe text talksloosely of "properties."
OF PHILOSOPHY
THEJOURNAL
large, of course, lest

e cannot be too
of
stavalue
e. The
suitable
some
to affect the justificational
as
so
change
deterto
mental states
agent time
agent's
the
enough to allow the
large
be
must
e
But
tus of p.
(2) allow
relevantformalrelations.
the
mine
mental change, but unsatisavoid
(1)
two conditions
maywell bejointly
These
formalrelations
Mental
compute
to
time
enough
problem for internalism. change
severe
a
pose
new evidence,
which would
fiable,
statesthat generate required computaperceptual
including
states,
change before
and tlley could easilyother hand, although mental states
rapidly
very
On the
not be epistemically
couldbe executed.
tions
belief systemmight
agent's
the
interval e has
rapidly,
change
do
to each change ulltil feasible.
respond
or
to reflect
mightbe
required
decisioninterval,then, during which justificadoxastic
Some
elapsed.
decision interval
Coherentism
therea short enough
Is
can be computed?
properties
formal
if p coheres with
pertinent
tionally
propositionp only
believing
in
that coherence
thatS isjustified
says
at the time. Assume that tlle logiheld
system
requires
restof S's belief
the
Then coherentism
with S's belief
consistency.
logical
implies
any propositionp
of
inconsistency
consistencyor
consistencyor
cal
But how quicklycan Christopher
justifier.
a
as
As
must qualify
system
by mental computation? consistency
ascertained
be
tautological
inconsistency
of even
points out, determinationin the general case. Using the
Cherniak24
task
complex
a belief system
a
iscomputationally
the consistency of
for
check
to
method
an ideal computer
truth-table
propositions,even
atomic
table in the
138 independent
with
each row of a truth
(checking
would
speed"
proton)
at "top
working
the diameterof a
traverse
to
ray
"big-bang"
light
the
it takes a
time from
time
years, the estimated
billion years
billion
twenty
take
Presumably,twenty
present.
the
to
of the universe
dawn
doxasticdecision interval!
for garden-variety
not
is an acceptable
then, is too constraining
one of
interval,
again clashes with
Anyreasonable
constraint
The knowability
internalistsmight
coherentism.
Dyed-in-the-wool
for tradiinternalism.25
of
thestock brands with this result. "So much the worse
its
live
to
live with demise."
beprepared
might say, "we can
There
they
tionalcoherentism," internalism entirely off the hook. qualify
get
not
do
not
factsthat
Butthis does
logical and probabilistic
threatento be many
284
of Logic," this
Universal Acceptance
Complexity and the
"The
24 'iComputational
1984): 739-58.
12 (DeCelllher
is identified by Kornblith,
Coherence
coherentisrn
JOURNAL,LXXXI,
for
the
of
State
computationaldifficulty
Bender, ed., T/zeCurrent
25 This
Coherence,"in J.
Unattainabilityof Kluwer,1989), pp. 207-14.
Theoey(Dordrecht:
INTERNALISMEXPOSED
asjustifiersbecause they require too long a doxasticintervalto compute. Furthermore,it is unclear what is a principled basis for deciding what is too long. This quandary confronting internalism has
apparentlyescaped its proponents'attention.
A second problem for logical and probabilisticjustifiers is the
availability
problem.
Suppose that a particularset of computational
operations call it COMP is provisionallyselected for inclusion along
side introspection and meluory retrieval.COMPmight include, for
example, a restricted(and hence noneffective) use of the truth-table
method, restrictedso as to keep its use within the chosen doxastic
decision intezxal.26
This yields a new versionof the KJconstraint:
KJillt+let+coAlp:
The only facts that qualifyas justifJlers
of an agent's believing p at time t are facts tllat the agent can reaclily
know within a suitablecloxasticdecision intelwalviczintrospection,memotyretrieval,and/or COMP.
Now, the KJconstraintis presumablyintended to applynot only to the

cleverestor best-trainedepistemicagentsbut to a11epistemicagents,includingthe most naiveand uneducatedpersonson the street.Afterall,
the point of the knowabiliW
constraintis thatjustiElersshould be facts
withinthe purviewof everyepisteinicagent.Under the GDconception,
compliancewith epistemicdutyor responsibilityis not intended to be
the privatepreserveof the logicalor mathematicalelite. It is something
that ought to be attained-and should thereforebe attainable by any
human agent. The truth-tablemethod, however,does laotseem to be
in the intellectualrepertoireof naive agents,so it is illegitimateto include COMPoperationswithin a KJ constraint.Unlike introspection
and memoryretrievaln
it is not availableto all cognitiveagelats.
It may be replied that computational operations of the contemplated sort would be within the capacityof normal human agents. No
super-humancomputationalpowersare required.Computingpower,
however,is not the issue.A relevantsequence of operationsmust also
be availablein the agent's intellectual repertoire; that is, she must
know which operations are appropriateto obtain an answer to tlae
relevant (formal) question.27Since truth-tablemethods and other
96 Because of the contemplated restriction, there will be many questions about
formal facts to which COMPcannot deliver answers.Thus, formal facts that might
othel wise qualify as justifiers will not so qualify under the version of the KJ constraintthat incorporatesCOMP.
"7 Propositional (or "declarative")
knowleclgeof the appropriatesequence of operations is, perhaps, an unduly restrictive requirement. It would suffice for the
agent
to .have "procedural"
skills of the right sort. But even such skillswill be lack.
.
. .
lllg ln nalve cognltlve
agents.
285
286
OF PHILOSOPHY
THEJOURNAL
of ordinarycognialgorithmsare probablynot in the repertoire

such
constraint.
KJ
a
in
agents,they cannot properlybe included
tive
that should be
methoc;ology
Athird problem concerns the proper
computationalopin selecting a KJconstraintthat incorporates
used
a KJconstraintthat
As we see from the first two problems,
erations.
the basic cog
reflect
must
to the spiritof the GD rationale
conforms
these basic
What
beings.
skills or repertoiresof actual human
nitive
priori. It
a
determined
be
consist in, however, cannot
repertoires
empiricalscience. This fact
only be determinedwith the help of
can
of internalundermines the methodological posture
fundamentally
VII.
section
a subjectto which I shall returnin
ism,
accessibilityconstraint,one
Until now, I have assumed a universal
potentialjustifiersfor
perhaps
holds for all cognitiveagents. But
that
for another.Justifiersmight
agent need not be potentialjustifiers
one
on what is knowdepending
allowedto varyfrom agent to agent,
be
logical or
different
have
by the particularagent. If two agents
able
not qualifyas justiskills, then some propertiesthat do
probabilistic
other. Indeed, the
the
for
for one might yet qualifyasjustifiers
fiers
agent-relativized
this
precisely
mightbe readin
KJillt+let+coMp
constraint
routes
knowledge
indicating
The subscriptsmaybe interpretedas
way.
agents.
all
to
necessarily
not
are availabletotheagentin question,
that
as a function of differencesin
relativized
agent
are
If I(Jconstraints
the same eviskills,this means that two people in precisely beliefs,and
knowledge
background
dentialstate (in termsof perceptualsituation,
But if the two
entitlements.
soon) might have different epistemic
each must
duties,
epistemic
agentsare to complywith their respective
second
the
parallels
simply
which knowledgeskillsshe has. This
know
episone's
If
argument.
stepof the internalist'soriginal three-step knowledgeskills(for exone's
on
temicdutiesor entitlementsdepend
then compliance with one's
skills),
computational
one's
ample,on
one possesses.There are two
dutiesrequiresknowledgeof whichskills
unlikelythat manypeople-esproblemswith this approach.First,it is
have this sort of knowledge,
peciallyordinarypeople on the street
Second,whatis now reandthis againthreatenslarge-scaleskepticism. about the truth-getting
quired to be known by the agent is something of her skillsin detectthe power
powerof her cognitiveskills that is,
the sort of externalproperty
ing justifiers.This seems to be precisely
How can theyacceptthis solution
thatinternalistsregardas anathema.
internalism?28
while remainingfaithfulto the spiritof
spirit leads to a similar reqelirement
It might be argued that internalism's not just for agent-relativizedversions.
constraint,
KJ
a
even for universalversionsof
the general form of internalism.
Perhapsso; but so much the worsefor
28
EXPOSED
INTERNALISM
V. EPISTEMICPENCIPLES
287
not clear exactly

of justifiers, it is
speaks
epistemic
constraint
the KJ
When
do justifiers include included,
Specifically,
be
comprehend.
these
what
that principles should
themselves?I believe among the items that determine
principles
are
which
epistemic principles
because
in believinga proposition, prinjustified
is
agent
epistemic
an
or not
whether
Furthermore,true
'justifiers'was defined. if she is going to determine her
just
ishow
mind
Inustknow
areitems an agent Knowledgeof her current states of
ciples
duties
her epistemic
duties correctly.
epistemic
not instructher about
will
properties
epistemicprinciples.
their
and
she also knowstrue according to internalunless
entitlements
and
to be known,
are epistemic principles epistemic principlesare normative
How
saysthat central
necessarily true.
Chisholm29
ism?
(when true) are
which
in parprinciples,
supervenience
can be known a priori
they
truths,
they are necessary "byreflection."
Since
they can be known
ticular,
vIponhis own con-
that, merely by reflecting

that will eninternalistassumes
The
epistemicprinciples whether
of
set
a
state, he can formulate to any possiblebelief he has,
scious
respect
him to find out, with belief.30
able
that
isjustifiedin llaving
he
epistemic princi-
as to whether (correct)
or accessible
passageis ambiguous
This
just to epistemologists, must be rereflection
on
however,
are accessible
ples
as well. The latter,
to be deteF
supposed
are
naiveepistemic agents
to
justifiers
because
by internalism,
quired
by all epistemicagents.reallycapableof formulatingand recminable
agents
dubious.
Areordinaryor naive
This seems highly
epistemic principles?have failed to articulateand
correct
ognizing
many career-longepistemologists
Since different epistemoloEven
for episepistemic principles.
correct
appreciate
conflicting candidates can be
mutually
and
epistemologists
offer disparate
gists
a fraction of these
tendered by epistemolprinciples, at most
far
temic
thus
principles
the
of
record by
none
Perhaps
and possiblydismal people,
right.
shaky
this
of
light
ordinary
ogistsis correct!In
how can we expect
pitfalls,to
epistemologists,
multiple
and its
professional
at
epistemology
of
that they should succeed
whoare entirelyignorant
plausible
it
is
Nor
among
states,since
succeedat this task?31
on their conscious
reflection"
computational
"by
what
purely
is
thistask
principlesmust resolve
epistemic
the matters
209-15.
xxrv ( 1990):
No^X.s,
EpistemicPrinciples,"
Episemphasisomitted.
"' "TheStatusof
3rd edition, p. 76; in ''Internalismand Externalismin
Knowledge,
of
Theorw
Alston
by
expressed
is
worly
31 A similar
pp.221-22.
temology,'
288
OF PHILOSOPHY
THEJOURNAL
cognizers.I do not see

are within the competence of ordinary
skills
this can be answereda priori, "byreflection."If epistemic princihow
A crippling problem emerges for internalism.
no such principles can
are not knowable by all naive agents,
ples
If no epistemic princias justifiers under the KJ constraint.
qualify
believed by any
so qualify, no proposition can be justifiably
ples
Wholesaleskepticismfollows. THREE-STEPARGUMENT
agent.
VI. THE COREDILEMMAFORTHE
there is any cogent inferential
Iraisedoubts here about whether
via an acceptable KJ
from the GD conception to internalism
route
unvarnishedverHere is the core dilemma.The minimal,
constraint.
That siminternalism.
sionof the KJconstraintdoes not rationalize
must be readily knowable,
pleconstraint merely says that justifiers
external ratherthan interbe
andsome readilyknowablefacts might
are allowed, then knowlnal.If all routes to knowledge of justifiers
by perception is
knowledge
If
edgeby perception must be allowed.
could qualifyfor the statusof
allowed,then facts of an external sort
claims that purely external
justifiers.Of course, no epistemologist external facts are nomipartly
factsshould serve as justifiers. But
Considerpropertiesof
justifiers.
of
rank
the
for
natedby externalists
indicator of a certain environtheform: being a reliable perceptual
least partlyexternal because
at
mentalfact. This sort of propertyis
the usual assumption)is exreliabilityinvolvestruth, and truth (on
perceptualstate has the
ternal.Now suppose that a certain auditorypresence of a mourning
of the
propertyof being a reliableindicator
possessionof this reliableindidovein one's environment.Mightthe
on the grounds that it is incatorshippropertyqualifyas a justifier
to knowledge is legitimate, I
deed readilyknowable?If every route
excluded. After all, one could
do not see how this possibilitycan be
doves and their songs to deteruse past perceptions of mourning
is a reliable indicator of a
mine that the designated auditorystate
knowledge is allowed,
mourning dove's presence. So if unrestrictedbe perfectly knowable.
might
the (partly) external fact in question
KJconstraintdoes not exclude
Thus, the unvarnishedversion of the
externalfactsfrom the ranksof thejustifiers.
then, does not supportinThe simple version of the KJconstraint,
what undoubtedlyleads interternalism.Tacit recognition of this is
constraint.Unfortunately,this
nalists to favor a "direct"knowability
GD conception. The GD conextra rider is not rationalizedby the
agents must know whatjustiception at best implies that cognitive
typesof knowledge,or paths
fiers are present or absent.No particular
GD conception cannot rationalto knowledge,are intimated.So the
ISMEXPOSED
INTERNAL
289
constraint that unambiguously
KJ
the
of
version
restrictive
a
ize
implies
internalism.
yields
way. The GD conception
another
always
point
facts
meput the
Let
but are internal
knowable,
probareadily
be
must
justifiers
As discussedearlier, seem
that
than externalfacts?
do not
knowable
readily
more
as internal,but they
qualify
traditionof psyrelationspresumably
bilistic
beings. An entire
human
naive
by
readily
be knowable "biasesand heuristics"suggests that "conto
on
the
as
research
fallacies, such
chological
commit probabilistic
commonly
judgmental heuristics,
agents
formallyincorrect
use
and
anchoring-and-adfallacy,"
junction
heuristic and the
at detecting
representativeness
abilities
the
as
such
is right, people's
this
If
People's percepheuristic.39
justment
actuallyratherweak.
are
far superior.
relationships
probabilistic
factsseem, by contrast, holds little
external
detect
to
therefore,
capacities
tual
of the KJ constraint, conditions in preferversion
unqualified
The
to internal
for restrictingall justiElers
requires.33 PRIORI?
promise
internalism
as
OR A
conditions,
EMPIRICAL
external
to
ence
OF EPISTEMOLOGY:
epistemology
METHODOLOGY
that
VII.THE
doctrine
incorporatesthe
that needs
one
standardly
than
rather
Internalism
enterprise
armchair
way:
point this
a a priorior
ispurely
Chisholmputs the
science.
empirical
from
llelp
formulatesare prin-
that [the epistemologist]

in one's
Theepistemicprinciples upon and apply melnelyby sitting
come
assistance.In
ciplesthat one may and witlloutcallingfor any outside
so to speak,
alnmchair,
own state of mind.34
only considerone's
aword,one need
of apriorism
the merits
raiseddoubts about
already
sections
the theoreticalarchitecture
Previous
in
in the context of
even
epistemology,
the viabilityof apriorism
in
challenge
to
want
here. I now
presented
restricdepth.
greater
reservations,an internalist
earlier
rlly
allows only
Assumetllat, despite
derived, one that
been
to serve
somehow
has
tionon justifiers certain of their nonexternalproperties
consciousstatesand
under Uncertainty:
under
Ikahneman, "Judgment eds., Juclg*slent
Daniel
and
Tversky,
Amos Tversky
P. Slovic, and
32 See
IZhneman,
and
Ikahneman,
Tversky
in
and
Heuristicsand Biases,"Cambridge,1982), pp. 3-20;
Fallacy in Probability
(New York:
The Conjunction
Uncertainty
Reasoning:
Intuitive
293-315.
"Extensionalversus
Review,XCI (1983): logical or probabilisticfacts intuitively
which
why
Judgment,"Psychological
moreover,
the same sense in
not really clear,
are not internal inabout the contours of in33 It is
certainly
They
facts.
problem
count as "internal"
This is an additional
mental states are internal.
ternalism.
3rd edition, p. 76.
ofKnowledge,
34 Theoey
29o
as justifiers. How should the epistemologistidentify particularconscious states and propertiesasjustifiersfor specific propositions (or
typesof propositions)? In other words,how should specific epistemic
principlesbe crafted?Should the taskbe executed purelya priori,or
can scientificpsychologyhelp?
For concreteness, considerjustifiers for memorybeliefs. Suppose
an adult consciously remembers seeing, as a teenager, a certain
matinee idol. This ostensible memory could have arisen from imagination, since he frequently fantasizedabout this matinee idol and
imagined seeing her in person. What clues are present in the current memory impression by which he can tell whether or not the
recollection is veridical?This is preciselythe kind of issue which internalistepistemic principles should address.If there are no differences in features of memory states that stem from perceptions of
real occurrences versus features of states that stem from mere
imagination, does this not raise a specter of skepticismover the domain of memory? If there are no indications by which to distinguish veridical from nonveridical memory impressions, can we be
justified in trusting our memory impressions? Skepticism aside,
epistemologists should surely be interested in identifying the features of conscious memory impressions by which people are made
more or less justified (or prima facie justified) in believing things
about the past.
Epistemologistshave said very little on this subject. Their discussions tend to be exhausted by characterizationsof memory impressions as "vivid"or "nonvivid."There is, I suspect, a straightforward
reason for the paucityof detail. It is extremelydifficult,using purely
armchairmethods, to dissect the microfeaturesof memory experiences so as to identify telltale differences between trustworthyand
questionablememories.On the other hand, empiricalmethods have
produced some interesting findings, which might properly be infused into epistemic principles in a way entirely congenial to internalism. Important research in this area has been done by Marcia
Johnson and her colleagues.35I shall illustratemy points by brief reference to their research.
Johnson calls the subjectof some of her researchreality monitoring.
She tries to characterizethe detectable differences between (conscious) memorytracesderivedfrom veridicalperceptionof eventsver35 See Johnson and Carol Raye, "Reality Monitoring," PsychologicalReview,
LXXXVIII (1981): 67-85;andJohnson, MaryFoley,AuroraSuengas, and Raye, "Phe-
nomenal Characteristicsof Memories for Perceived and Imagined Autobiographical Events,"JournalofExperirzzental

Psychology:
General,
CXVII (1988): 371-76.
ISMEXPOSED
INTERNAL
291
Johnof events.36
tracesgeneratedby mere imaginations
memory
sus
memory
which
dimensionsalong
andRaye (op. cit.) proposefour
son
whethertheiroriginwaspercepon
will typicallydifferdepending
from
cues
with memories that originate
or imaginative.As compared
(1)
tual
have
to
tend
memoriesoriginatingfrom perceptionand sound), (2)
imagination,
example, color
perceptual information (for
more
tiine and place, and (3) more
contextual information about
more
is rich along these three didetail.When a memorytrace
meaningful
throughperception.
this is evidenceof its havingoriginated by contrast,tend
mensions,
or thought,
originatingfrom imagination
Memories
more informationabout
contain
they
berich on anotherdimension:
to
in the originalthinkingsor imagincognitiveoperationsinvolved
the
image creation,or search).Per(for example,effortfulattention,
ings
so a memorytrace that
is more automaticthan imagination,
ception
lack attributesconcerningeffrom perceptionwill tend to
originates
thereforesuggestthatdifferences
operations.Johnson and Raye
fortful
can form the basisfor
valuealong these typesof dimensions
inaverage
is perceptualor nonpercepwhetherthe origin of a memory
withdeciding
visualand spatialdetail,and
A memorywith a greatdeal of
tual.
processes,
and organizational
recordsof intentionalconstructive derived.37
out
bejudged to havebeen perceptuallyto borrow these sorts of
should
Epistemologists would be well-advised
epistemicprinciples.A person
and incorporatethem into their the real occurrenceof an osideas
in
is(primafacie) justified in believing
trace is strong on the first
memory
an
tensiblyrecalled event if the
the fourth dimension. Conversely,rethreedimensionsand weakon
the
of
in the real occurrence
agentis unjustified in believing strong on the fourth dimension
is
calledevent if the memory trace
All of these dimensions, of
butweak on the first three dimensions.experience. For this reason,
conscious
these
course,concern features of
should be happy to incorporate
epistemologists
internalist
principles.
kindsof featuresinto their epistemic of epistemologicallysignificant
Let me distinguishtwo categories empirical psychology might
which
factsabout memory experience
with imagof course, to confusions of actual two acblending,
Memoryerrorsare not confined,
or
e1-rorsthat arise from confusing,
ined events. There are also of Johnson's focuses on the actual/nonactual (or
research
this
But
tual events.
perceivedversusimagined) problem.
a target memoly with memories
recognize that people can compare
This kind of "coher37 They also
veridicality.
target's
the
assess
to
of contextuallyrelated events of epistemology,however,and hence not a good exsubject. I
ence"factor is a stock-in-trade
psychology can make to this
ample of the distinctive contributions
thereforepass over it.
36
292
provide. First,as we have seen, it might identify types of representational materialswhich are generallyavailablein people's memoryexperiences. Second, it might indicate which of these representational
materials are either reliable or counter-reliable indicators of the
veridicalityof the ostensibly recalled events. Is the reliability of a
memory cue a legitimate issue from an internalist perspective?It
might be thought not, since reliabilityis usuallyclassed as an external property.But epistemologistsmight use reliabilityconsiderations
to decide which memory characteristicsshould be featured in epistemic principles.They need not insert reliabilityper se into the principles. There is nothing in our present formulationof internalism,at
any rate, which bars the latter approach.Any KJconstraintprovides
only a necessary condition for being a justifier; it leaves open the
possibilitythat additionalnecessaryconditions, such as reliable indication, must also be met. Indeed, many internalistsdo use reliability
as a (partial)basis for their choice of justifiers. BonJour (op. cit., p.
7) says that the basic role of justificationis that of a means to truth,
and he defends coherence as a justifier on the ground that a coherent system of beliefs is likely to correspond to reality. This point
need not be settled definitively, however. There are already adequate grounds for claiming that internalism cannot be optimally
pursuedwithout help from empiricalpsychology,whether or not reliabilityis a relevantconsideration.
VIII. CONCLUSION
Let us reviewthe paradeof problemsinfecting internalismwhich we

have witnessed, though not all in their order of presentation. (1)
The argument from the GD conception of justification to internalism does not work. Internalismcan be derived only from a suitably
qualified version of the KJ constraint because the unqualified version threatens to allow external facts to count as justifiers. No suitably qualifiedversion of the KJconstraintis derivablefrom the GD
conception. (2) A variety of qualified KJ constraints are possible,
each leading to a differentversion of internalism.None of these versions is intuitivelyacceptable.Strong internalism,which restrictsjustifiersto conscious states,is stuckwith the problem of stored beliefs.
Weakinternalism,which allowsstored as well as conscious beliefs to
count asjustifiers,faces the problem of forgotten evidence and the
problem of concurrent retrieval. (3) The question of how logical
and probabilisticfacts are to be included in the class of justifiers is
plagued by puzzles, especiallythe puzzle of the doxasticdecision intervaland the issue of availability.(4) Epistemicprinciples must be
among the class of justifiers, but such principles fail internalism's
INTERNALISMEXPOSED
293
knowabilityrequirement. (5) The favored methodology of internalism-the armchairmethod-cannot be sustained even if we grant
the assumptionthatjustifiersmust be consciousstates.
Internalismis rife with problems.Are they all traceableto the GD
rationale?Could internalismbe salvagedby switchingto a different
rationale?A different rationale might help, but most of the problems raised here arise from the knowabilityconstraint.It is unclear
exactlywhich knowabilityconstraintshould be associatedwith internalism, and all of the availablecandidatesgenerate problematictheories. So I see no hope for internalism;it does not survivethe glare
of the spotlight.
ALVINI. GOLDMAN
Universityof Arizona

Goldman Intern Exp

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Goldman Intern Exp

Uploaded by

Copyright:

Available Formats

Journal of Philosophy, Inc.

THE JOURNAL OF PHILO SOPHY

(C)1999 The Journal of Philosophy,Inc.

I begin with a certain rationalefor internalismthat has widespread

The guidance collception of justification is commonly paired with

tion is incorporated can the argument proceed along the intended

How can an agent readilyknow whether candidatejustifiers obtain

example, argues that epistemic

KJcondo not endorsejust the minimal which

states an agent is in at time t

approach to justifiinternalism,however,is an unacceptable

posed, merely by opening his eyes, to generate conscious evidential

statesthat would occurrentlyjustifysuch beliefs. So the dispositional

The obvious solution to the problem of stored beliefs is to relax the

This KJconstraintallowsfor a more viable form of internalismthan

WIwill certifythe justificationof manystored beliefs, because agents

the capital of the United

we are consideringthe "epistemAt least it is relevantso long as

be concurrentlyretrievedfrom memory.So that set of stored beliefs

Can this problem be remedied by simply adding the proviso that

large, of course, lest

Now, the KJconstraintis presumablyintended to applynot only to the

lllg ln nalve cognltlve

of ordinarycognialgorithmsare probablynot in the repertoire

not clear exactly

that, merely by reflecting

cognizers.I do not see

that [the epistemologist]

nomenal Characteristicsof Memories for Perceived and Imagined Autobiographical Events,"JournalofExperirzzental

Let us reviewthe paradeof problemsinfecting internalismwhich we

You might also like