Justification and Truth

Author(s): Stewart Cohen

Source: Philosophical Studies: An International Journal for Philosophy in the Analytic
Tradition, Vol. 46, No. 3 (Nov., 1984), pp. 279-295
Published by: Springer
Stable URL: http://www.jstor.org/stable/4319711
Accessed: 08/02/2009 19:06
Your use of the JSTOR archive indicates your acceptance of JSTOR's Terms and Conditions of Use, available at
http://www.jstor.org/page/info/about/policies/terms.jsp. JSTOR's Terms and Conditions of Use provides, in part, that unless
you have obtained prior permission, you may not download an entire issue of a journal or multiple copies of articles, and you
may use content in the JSTOR archive only for your personal, non-commercial use.
Please contact the publisher regarding any further use of this work. Publisher contact information may be obtained at
http://www.jstor.org/action/showPublisher?publisherCode=springer.
Each copy of any part of a JSTOR transmission must contain the same copyright notice that appears on the screen or printed
page of such transmission.
JSTOR is a not-for-profit organization founded in 1995 to build trusted digital archives for scholarship. We work with the
scholarly community to preserve their work and the materials they rely upon, and to build a common research platform that
promotes the discovery and use of these resources. 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

STEWART COHEN

JUSTIFICATION AND TRUTH*

(Received16 December,1983)

Epistemologistsgenerally agree that two components of knowledge are justification and truth. If S believes that P, then S knows that P only if S is justiefied in believingthat P and it is true that P. A centralissue in epistemology
concems the connection between justification and truth. While one could
view them as conceptually distinct components of knowledge, a variety of
philosophers from Descartes to the present have presupposedthe view that
justificationand truth are conceptuallyrelated - that there is an intemal connection between a belief being justified and being true. The appeal of this
view is no doubt rooted in the conviction that knowledgedoes not arisewhen
a belief merely happens to be true. The motivation for requiringthat a true
belief be justified in order for it to count as an instanceof knowledgejust is,
in some sense, to providea connection to truth.
A furtherconsiderationthat supportsthis view is the fact that justification
is a genericnotion. In additionto being epistemicallyjustified, a belief can be
pragmaticallyor morally justified. The belief of a defense attomey who, in
order to providea better defense, convinceshimself that his client is innocent
may be said to be justified in one of these latter senses. One might contend
that the connection to truth is what distinguishesepistemicjustification from
these other senses of justification.
The burden of this view is to characterizethe nature of the connection
between epistemicjustificationand truth. One connection that is immediately
apparentis that S is epistemicallyjustified in believing that P only if S is
epistemicallyjustified in believing that P is true. But this trivialconnection
does not distinguishepistemicjustification from moralor pragmaticjustification. For one is morally or pragmaticallyjustified in believingthat P just in
case one is morally or pragmaticallyjustified in believing that P is true. This
is not surprisingsince this trivialconnection seems to be more a point about
belief than justification. To believe that P, just is to believe that P is true. The
Philosophical Studies 46 (1984) 279-295. 0031-81 16/84/0463-0279$01.70
? 1984 by D. Reidel Publishing Company

280

STEWART

COHEN

connection between epistemicjustification and truth must amount to something beyond this trivialconnection.
One might classify theories of epistemic justification according to how
they construe the truth connection. In what follows, I propose to do this,
and to consider some of the problemsthat might arisefor the variousclassifications. I will argue that the issues concerningthe truth connection present
epistemologywith a dilemma.

II

The strongest view one could take regardingthe truth connection is that
taken by Descartes. The Cartesianview is that justification logically entails
truth. To put it schematically:It is a conceptual truth that, if conditions C
justify belief B for subjectS, then C logically entailsthat B is true.'
The legacy of the Cartesianview is skepticism.Descartesdemonstratedin
the first meditation that no such connection is forthcoming(the arguments
of the later meditations notwithstanding).Given any plausiblespecification
of C for any S, it will always be logically consistent to suppose that not B.
This is what the evil demon argumentshows. Where,e.g., C comprisesfacts
about sensory data, and whereB is a belief about the truth of some empirical
proposition, it is always logically possible that the evil demon has arranged
for C to obtain whereB is false. Not wishingto be saddledwith this skeptical
result, most contemporaryphilosophershave rejected the Cartesianview and
have opted instead for a fallibilist theory of justification. A fallibilisttheory
allows that whereC makesB justified for S, it is still possible that B is false.
While fallibilismdoes seem to avoid skepticism, one might still inquireas
to what the truth connection comes to on a fallibilist view. A natural
proposal would be to construe the connection as probabilistic.This tack has
been taken by severalphilosophers(Goldman,Swain, et al.) whose theories
can be grouped under the general heading of Reliabilism.For purposes of
exposition, I focus on Goldman'sview.2
Goldman's brand of Reliabilismexplicates the connection between justification and truth in terms of the truth-frequencyof the justified belief.
There is no logicalguaranteethat a justified belief is true on this view. Whatis
guaranteedis that a justified belief is produced by a cognitiveprocessthat is
reliable, i.e., by a cognitive process that tends to produce true beliefs. Gold-

JUSTIFICATION

AND TRUTH

281

man leavesit open just how reliablea cognitiveprocessmust be in orderfor it

to confer justification on the belief it produces. However,he is explicitly a
fallibilist.3 Since reliability is a probabilisticnotion, for our purposesGoldman's position amounts to: It is a conceptual truth that, if C justifies B for
S, then C makes it probable that B is true.4 So, for example, if the belief
"There is somethingred before me" is justified in virtue of its being produced
by a perceptualprocess, then the fact that the belief was so producedmakes
it probablethat it is true.
It is clear that Reliabilismavoids the skepticismof the Cartesianconception of the truth connection. For the evil demon hypothesis only demonstrates the possibility that there are belief forming processeswhich are not
reliable. The fact that this is consistent with their actual reliabilityobviates
any skeptical conclusion. But skepticism aside, I think the evil demon
hypothesis (or its contemporaryneurophysiologistversion)uncoversa defect
in the Reliabilistposition. We can see this by supposingthe hypothesisto be
true. Imaginethat unbeknownto us, our cognitiveprocesses(e.g., perception
memory, inference) are not reliableowing to the machinationsof the malevolent demon. It follows on a Reliabilist view that the beliefs generated by
those processesare neverjustified.
Is this a tenable result?I maintainthat it is not. Of course, we are not here
supposing that we know that the demon hypothesis is true. Certainlyif we
were to know that our cognitiveprocessesare unreliablethen the beliefs they
generatewould not be justified. Whatwe want to supposeis the mere truth of
the deinon hypothesis. Now part of what the hypothesis entails is that our
experience is just as it would be if our cognitiveprocesseswere reliable.Thus,
on the demon hypothesis,we would haveevery reasonfor holding our beliefs,
that we have in the actual world. Moreoversince we actually have reasonto
believe that our cognitive processesare reliable,it follows that in the demon
world we would have every reason to believe that our cognitive processes
were in fact reliable.We might even imaginethat a brilliantphilosopherhad
seemingly demonstrated(a la Descartesof the latermeditations)the falsity of
the demon hypotheses, to the extent that anyone who could follow the
reasoningwas (intuitively)justified in acceptingthe conclusion.
It strikes me as clearly false to deny that under these circumstancesour
beliefs could be justified. If we have every reason to believe e.g., perception,
is a reliable process, the mere fact that unbeknown to us it is not reliable

282

STEWART

COHEN

should not affect it's justification - conferringstatus (a fortiori if we have

good reason to belief that the conditions which in fact make perception
unreliabledo not obtain.5'6
My argument hinges on viewing justification as a normative notion.
Intuitively,if S's belief is appropriateto the availableevidence,he is not to be
held responsiblefor circumstancesbeyond his ken. Goldmancan be viewed as
having illustrated this when he discusses a possible counter example to his
analysis. He considers whether wishful-thinkingwould be a justification conferringprocess in world W,if it were a reliableprocessin W.In a footnote
he points out that "if people in world W learn inductively that wishfulthinkingis reliable,and regularlybase their beliefs on this inductiveinference,
it is quite unproblematicand straightforwardthat their beliefs arejustified".7
Goldman's point is that in this case, the inductively inferred reliability of
wishful thinking can be sufficient for beliefs producedby wishful-thinkingto
be justified in W. But if this is true, the pictureshould not changeif wishfulthinkingturns out, contraryto the inductiveevidence, to be unreliablein W.
Now Goldman might contend that inductive inference is itself a reliable
belief-formingprocess and that this fact explains the justifiedness of the
beliefs in question. But again,I fail to see how the picturechangesepistemically
if as a result of infelicitious circumstances,beliefs produced in accordance
with the canons of inductive inference turn out to be inscrutablyfalse most
of the time.8
An entrenchedReliabilistsmay be unmovedby this appealto the normative
characterof epistemic justification. One might insist that a world where the
evil demon hypothesis is true is a world where there are no justified beliefs.
I will arguethat a theory with this consequencefails to capturea central,perhaps the central distinctionin epistemology.Moreover,I think it will be clear
that the distinction is most plausibly construed as markingthe difference
betweenjustified and unjustifiedbelief.
When Goldman is marshallingdata for his theory, he lists certain beliefforming processes whose outputs we would consider to be unjustified. He
gives these examples: confused reasoning,relianceon emotional attachment,
mere hunch or guessworkand hasty generalization.Goldmannotes that these
processes all share the feature of unreliability.He contrasts these with processes that intuitively issueinjustifiedbelief, e.g., standardperceptualprocesses,
remembering,good reasoning, and introspection. Noting that these latter
process all have reliabilityas a common feature,Goldmangoes on to suggest,

JUSTIFICATION

AND TRUTH

283

"the justificational status of a belief is a function of the reliability of the

processor processesthat cause it" (p. 10).
I have arguedthat the evil demon case shows that reliabilityis not necessary for justification. This claim can be bolsteredby consideringthe contrast
between the two categoriesof belief-formingprocessesGoldmanlists.
Since Goldman cites reliability as the relevant feature that distinguishes
the class of justification-conferringprocessesfrom the class of non-justification conferringprocesses,he is committed to the view that in the evil demon
world, the cognitive processeshe lists are indistinguishablefrom the perspective of epistemic justification. Thus we can imaginetwo inhabitantsof this
world, A, who is a good reasoner,i.e., reasonsin accordancewith the canons
of inductive inference, and B, who engages in confused reasoning,wishful
thinking, reliance on emotional attachments,guesswork,etc. Since the beliefs
of A and B are both producedby unreliableprocesses(the evil demon sees to
this), a reliabilist theory of justification must render identical epistemic
appraisalsof both sets of beliefs.
Plainly, this cannot be correct.A's beliefs are conditioned by the evidence
whereasB's beliefs are not. A is a good reasonerwhereasB is not. A's beliefs
are reasonablewhereasB's beliefs are not. There is a fundamentalepistemic
difference between the beliefs of A and the beliefs of B. But the Reliabilist
does not have the theoreticalmeansto displaythis difference.
I would claim that the distinction between the beliefs of A and B is
marked precisely by the concept of justified belief. Beliefs producedby good
reasoningare paradigmcases of justified belief and beliefs arrivedat through
fallacious or arbitraryreasoning are paradigm cases of unjustified belief.
Whether or not reasoningresults in false belief, even if this happensmore
often than not, is irrelevantto the question of whetherthe reasoningis good.
To maintainotherwisewould be on a parwith confusingtruth and validity.9
The Reliabilist might acknowledgethat there is a clear epistemic distinction between the beliefs of A and B, but that the distinction has nothingto
do with their justificational status. Rather the differenceis that A's beliefs
arereasonableor rationalwhereasB's are unreasonableor irrational.
This maneuverwould be of no help. First of all, 'reasonable'and 'rational'
are virtual synonyms for 'justified'.But we need not quibbleover semantics.
If the Reliabilistwants to distinguish'justified'from 'reasonable'or 'rational'
he may do so. But clearly the important epistemic concept, the one epistemologists have been concerned with, is what the Reliabilist would call

284

STEWART

COHEN

'reasonability'or 'rationality.'The difference between the beliefs of A in the

evil demon world and the beliefs of B in the evil demon world seems to
capture the concept that epistemologisthave been worriedabout ratherthan
the difference between the beliefs of A in the actual world and the beliefs of
A in the evil demon world. In effect, the Reliabilistwould be changingthe
subject.
Another move open to the Reliabilist is to limit his theory to non-discursivebelief. The above objection hinges on the role of reasoningin justification. We noted that one can be justified in holding a belief in virtue of
having arrivedat the belief through good reasoningeven if the reasoningis
unreliable. However this standard of assessmentwould not apply to nondiscursiveprocesseslike perception. (We can assume for purposesof this discussion that certain processeslike perceptiondo not involvereasoning).Here
one might claim that the only critierionavailablefor assessingsuch processes
is reliability. Thus the only way a perceptual belief can be appraisedwith
respect to its justificationalstatus is in termsof its reliability.
Suppose perceptionis a non-discursiveprocess.It does not follow that perceptual beliefs can only be assessed in terms of their reliability. Again the
evil demon hypothesis will help to clarify this. Consideragaintwo inhabitants
of the evil demon world, A and B. Both have unreliableperceptualprocesses.
Suppose both A and B believe their is something0 before them on the basis
of being appeared to 0-ly. WhileA has no evidence to the contrary, B is
presented with strong evidence that owing to a clever deception there is
nothing 0 before him.
I think it's clear that there is a fundamentalepistemic differencebetween
A's perceptual belief and B's perceptual belief - a difference which again
underscoresthe normativecharacterof epistemicjustification.Notice that we
need not assume that B disregardsthe evidence as a result of any discursive
process. He may just arbitrarilyignore it. But from an epistemic point of
view, B ought not to have proceeded in the way he did. We might say that
contraryto A, B has been epistemicallyirresponsiblein acceptingthat there is
something k before him. As a result, while A is justified in his perceptual
belief, B is not. Thus we see that even for beliefs generatedby non-discursive
processes, a reliabilitytheory of justification cannot account for differences
in theirjustificationalstatus.

JUSTIFICATION

AND TRUTH

285

III

If my argumentsare correct, then epistemicjustification is not conceptually

connected with objective truth-frequency.We need not conclude from this
that justification and truth are in no way conceptually related. In fact, the
foregoing considerationssuggestan alternativeaccount. The salient featureof
those considerationsseemed to be this: From the perspective of epistemic
responsibility, a person S, can have an impeccablebelief that certain conditions make the truth of a proposition,B likely. Andunderthese circumstances
S can be justified in believingB, even if those conditions do not in fact make
the truth of B likely. As such, we might considerthe possibilitythat the connection between justification and truth is to be found at the subjectiveor
doxastic level. To put it schematically,one might proposethat it is a conceptual truth that if C justifies B for S, then C entails that S believe that certain
conditions obtain which make it probablethat B is true.
This sort of account is congenialto coherence theories of justification. ?
Although coherence theories vary considerably,it is not unreasonableto cite
as a characteristicfeature, the requirementthat the justificationof any belief
be a function of its relation to other beliefs.-1 So if the justification of a
belief B for S entails that S have some other belief that certain conditions
obtain which make the truth of B probable,then any justified belief will owe
its justificationto a relationit bearsto anotherbelief.12
As a representativeof this generalapproach,we can examine a theory of
justification advancedby Keith Lehrer.13 For Lehrer,justification residesin
the coherence of propositions accepted for the purposes of attainingtruth
and avoiding error. He analyzes the coherence relation in terms of probability.14 Ignoringcertain sophistications,a propositionB is justified for S iff
S accepts that given the system of propositions which he accepts in the
interest of obtaining truth and avoidingerror,B has a higherprobabilityof
being true than any statement with which it competes.1s Thus S is justified
in believingB only if S accepts that certain conditions R (viz, those conditions described by the beliefs in S's acceptance system) make it more
probable that B is true than any proposition that competes with B. This
entails that S accepts that R makes it more probable than not that B is true
since one of the competitorsof B is not B. Here then is a representativeof the
view that a connection between justification and truth lies at the subjective
or doxastic level.

286

STEWART

COHEN

While this view would seem to give us some insight into the nature of
epistemic justification, a major difficulty lies in its reliance on an intellectualist model of justification.16 Recall that on this view, in orderfor S to be
justified in believing B, S must believe that certain conditions obtain which
make the truth of B probable.One might object that these supportingbeliefs
are not always necessaryfor justification to arise. Casesinvolvingperceptual
knowledge are often invoked in behalf of this weaker claim. Suppose that S
perceivessomething which is red and so comes to believe "Thereis something
red." Some philosophershave maintainedthat it is sufficient for S to be (at
least prima facic) justified in this perceptualbelief that S merely be appeared
to redly17 or that S merely believe that he perceivessomethingto be red.18
The proponent of the intellectualist model would arguethat the mere fact
that S is appearedto redly cannotjustify a belief for S that somethingis red.
S must believe he is appeared to redly and moreoverbelieve that his being
appearedto redly makes it likely that there is something red. Similarly,the
mere fact that S believes he perceivessomething to be red cannotjustify the
belief in question for S, unlessS believesthat his believinghe perceivessomething to be red makes it likely that he does perceive something to be red.
(i.e., S must believe that his perceptualbeliefs are reliable.)
Although the claims of the intellectualisthave some intuitive appeal, one feels that something like this must be involved in justification - the
strongestargumentagainstthem seems to be that they run the risk of skepticism. For in many cases of what we generally take to be e.g., perceptual
knowledge, the psychologicalreality of the supportingbeliefs requiredby the
intellectualistmodel is questionable.Thereis no obvioussense in which most
adult persons, not to mention children and animalshave beliefs about how
they are appeared to as well as beliefs about their reliabilityas perceivers,
when they have perceptualbeliefs.
To avoid skepticism,the defenderof the intellectualisthas two alternatives.
He can arguethat the relevantsupportingbeliefs are presentunconsciouslyor
he can arguethat the beliefs are presentdispositionally.
The first alternativeis ambiguousbetween two days in which beliefs can
be said to be unconscious. According to psychoanalytic theory, beliefs are
unconscious insofar as a certain psychological mechanism, e.g., a defense,
blocks conscious accessto them. If the mechanismceases to operate,then the
belief surfacesin consciousness.It would be peculiarto claim that the beliefs
requiredby the intellectualistmodel are unconsciousfor this reason.

JUSTIFICATION

AND TRUTH

287

It may also be claimedthat certainbeliefs are unconsciousby appealingto

models of human cognition that posit unconciousinference.The premisesof
these inferencesmay be construedas unconsciousbeliefs.19
Could the beliefs requiredby the intellectualistmodel be unconsciousin
this way? Plainly an affirmative answer to this question cannot be based
solely on the requirementsof the intellectualistmodel. Cognitivepsychologists
currently view human cognition on the model of an information-processing
mechanism.If the intellectualistseeks to appropriatea computationalmodel
of human cognition, he must providesome independentreasonfor supposing
that such a model will be consonant with his own. At present,the details of
these models are both sketchy and controversial.Thus, whether the intellectualist can forge a doxastic connection out of the states posited by these
psychologicalmodels is not a matter that can be glibly assumed.Ratherit is
a matter for furtherempiricalresearch.
The second alternativefor the intellectualistis to arguethat the requisite
beliefs are psychologicallyreal in some dispositionalsense. The problemwith
this approachis the difficulty in detenniningjust when a dispositionalbelief
is present. The clearest cases of dispositionalbelief are those in which the
belief was previously occurrent and is presently stored in memory. But,
obviously this won't do in perceptualcases. This cannot be the sense in which
beliefs like 'S is appearedto redly' are dispositional.20It seems that the intellectualist must claim that the supportingbeliefs are dispositionalin the sense
that if the subjectwere to considerthe relevantpropositions,he would assent
to them. So if S is justified in believingthat there is something red before
him, S would assent to the proposition that he is appearedto redly (or the
proposition that he believes he sees something red), and the propositionthat
this fact makes it likely that that there is something red before him, if he
were to considerthese propositions.
[s this a psychologically more plausiblepicture?Again, it is not obvious
that subjectshave these requiredbeliefs even in this dispositionalsense. It is
not always an easy matter to characterizethe way in which you are appeared
to. This is a skill that artists have to work at acquiring.But even if the
requiredbeliefs were present in this dispositionalsense, it is easy to see that
this sense of belief is too weak to subserveany epistemologicalend. Suppose
it were true of S at (some time) t that S would assent to P if S were to
considerP. This fact would not suffice forP to function as one of S's reasons
for holding some belief at T. For it may be that S's assent to P would be

288

STEWART

COHEN

prompted by his consideration of P. By reflecting, people can discover

reasons for beliefs they hold. Such reasonscannot be said to have justified
those beliefs prior to their discovery. For example, S may believe that it is
going to rain on the basis of the pronouncementsof a ouija board. Now it
may be that S is a skilled meteorologist who has neglected to consider the
proposition that the present weather conditions portend the approachof a
rainstorm.Even if it were true that S would assent to this proposition if he
were to consider it, this does not constitute grounds for saying that S is
justified in believingthat it is going to rain.
At this point, the intellectualist may simply want to insist that in those
cases where the relevant beliefs are lacking, the subject is not justified. If
and to what extent this position yields skepticismdepends on the empirical
issue of the extent to which those beliefs are lacking in actual cases. As a
last resort, the intellectualist model can be viewed as an ideal to which as
a matter of empiricalfact, we only approximate.To take this tack, the intellectualist would, to a certain extent have to impugnsome of our common
sense judgments about when our beliefs are justified. This position would
involve a kind of "soft" skepticism.Strictly speaking,a lot of the beliefs we
think are justified are not. But in some sense, the beliefs arejustified insofar
as they approximatethe ideal model.21
Admittedly this is a vague position which places the burden squarelyon
the shouldersof the intellectualist.Whatis needed is an accountof the sensein
which our belief systems do approximatethe ideal model and an explanation
of how this can in some sense confer justification on those beliefs. I think
this would prove to be a difficult task.

IV

To reject the intellectualist model is to reject the thesis that justification is

connected to truth at the doxastic level. Since we have previously seen that
there is reason to deny that any objectiveconnection exists, we may wonder
what remainsof the thesis that justification and truth are connected. It can
still be maintainedthat S is justified in believingthat B only if S is justified in
believing that B is true. But as we noted earlier,this trivial connection does
not seem to capture the basic intuition that justification must be connected
to truth. Can one build a theory of justification that employs only this trivial
connection? In fact, this is how one might plausiblyconstruecertainFounda-

JUSTIFICATION

AND TRUTH

289

tions theories. These theories typically start by sanctioning common sense

judgments concerning when our belief are justified. They then proceed to
posit a series of epistemic principleswhich validatethese judgmentsby specifying just how those beliefs arejustified. It is importantto see that epistemic
principles do not specify any connection between justification and truth
beyond the trivialone.
Consideran epistemicprincipleproposedby RoderickChisholm: 22
(C)

For any subjectS, if S believeswithout groundfor doubt, that he

is perceivingsomething to be F, then it is evident for S that he
perceivessomethingto be F.

John Pollock has proposed:23

If S is appearedto redly, then S is primafacie justifiedin believing
that there is somethingred before him.
These principles simply state that under certain conditions (stated in the
antecedent) a certain belief (stated in the consequent) is justified. But the
conditions describedin the antecedentdo not involveeither an objectiveor a
doxastic connection to the truth of the belief describedin the consequent.
Pollock is quite explicit about this. For him, epistemic principlesare true in
virtue of a meaning connection. But this connection is not between a
statement and its truth conditions, ratherthe connection is between a statement and its justification conditions.24 Of course in each case the epistemic
principlesdescribeconditions under which S is justified in believingthat P is
true, but this is just the trivialconnection.
Is a theory that employs only this trivial connection adequate?I believe
we can see that it is not. This can be most clearlybroughtout by considering
the details of Pollock's theory.25 For Pollock, epistemic principlesdescribe
conditions under which beliefs are primafacie justified. To say that a justification is prima facie is to say that it can be defeated under certain conditions.26 Pollock characterizestwo types of defeaters (construed propositionally) for P being a prima facie reason for S to believe that Q. Type I
defeaters are reasons for S to believe that Q is false. Type II defeatersare
reasons for believingthat the truth of P is not an indicationof the truth of Q,
independentlyof being a reasonfor believingthat Q is false. So take Pollock's
principle:

290

STEWART

COHEN

If S is appearedto redly, then S isprimafaciejustifiedin believing

that there is something red (before him) (i.e., 'beingappearedto
redly' is a primafacie reasonto believe "thereis sometingred.")
The proposition'S is in a room with no red objects'is a type I defeaterof this
prima facie justification, since it is a reason for S to believe it is false that
'There is something red.' The proposition'S is in a room with a red light' is a
type II defeater. Although it is not a reason to believe 'There is something
red' is false, it is a reasonto believe that the truth of 'S is appearedto redly' is
not an indicationof the truth of 'Thereis somethingred (before S)'.
The issue I want to raiseis whethera theory like Pollock'scan account for
defeaters of prima facie justification. Type I defeatersare explainableby the
trivialjustification - truth connection. If S has a reason to believe that Q is
false, then its clear why S is not justified in believing that Q is true. But
matters are not so clear for type II defeaters. Thereis no doubt that Pollock
has correctly identified a source of defeat for primafacie justificationin type
II defeaters. If S has reason to believe there is a red light shiningon X, then
the fact that X appearsredly to S does not justify S in believingthat X is red.
The difficulty is that there seems to be no way for a theory like Pollock'sto
account for the fact that type II defeatersdefeat. WhereP is a primafacie
reason for S to believe that Q, essentiallytype II defeatersattack the connection between P and the truth of Q. But we have just seen that a theory like
Pollock's does not requireany such connection between P (the justification
conditions for Q) and the truth of Q.27 If a theory requiresthat the truth of
P logically entail the truth of Q (like Descartes);or if a theory requiresthat
P make the truth of Q probable(like Goldmans);or if a theory requiresthat
the subject believe that P makes the truth of Q probable(like Lehrer's);then
such a theory can account for why the subjecthavinga reasonto believethat
P is not connected to the truth of Q, defeats his justification for believing
that Q is true on the basis of P. But in the absence of any such requirement
the existence of type II defeatersis utterly mysterious.
It looks as if the existence of type II defeaters gives us good reason to
think that any complete theory of epistemicjustificationmust requirea nontrivial truth connection. Now Pollock and Chisholmmay want to claim that
the existence of type II defeaters is a primitivefact about epistemicjustification.28 This raises the issue of what explanatory requirementswe should
place on a theory of epistemicjustification. For the failureto explain type II

JUSTIFICATION

AND TRUTH

291

defeatersis an instance of a more generalexplanatorydeficiency that certain

philosophers have attributed to Foundations theories like Chisholm's and
Pollock's. The objection is that those theories are not framedat a sufficiently
general or abstract level to provide any insight into the nature of epistemic
justification. Accordingto Keith Lehrer:29
The most important function [of a theory of justification], in my opinion, is to explain
why certain beliefs are justified and others are not. Hence a theory of justification must
be judged in terms of how well it explains this. A system of principles may be presented
as a theory of justification when the principles are presented as ultimate ... If a belief is
justified according to the system, the only explanation for why it is justified is - that is
what the principles tells us. But the principles are unexplained.

Chisholm and Pollock do not really provide analyses of justification in the

sense that Goldman and Lehrer do. Their theories just amount to sets of
epistemic principles,one concerningperception,one conceming memory etc.
These principlestell us that beliefs of certaintypes arejustified undercertain
conditions. But the principles are nowhere united by a generaltheory that
explains why those beliefs are justified under those conditions. In a sense,
Foundation theories do not tell us what justification is. This objection to
Foundationalism is reminiscent of the complaints of Socrates who in the
Platonic dialogues, repeatedly chides his dialectical opponents for merely
citing instances of a particularconcept ratherthan providingan account of
the concept itself.
In all fairnessto Chisholmand Pollock, it does not seem to be their intent
to provide analysesof justification in the sense that Goldmanand Lehrerdo.
Their project is to account for how knowledge is possible. Their epistemic
principlesare intended to servethispurpose.Butcertainlyit wouldbe desirable
to have a generaltheory that would motivate epistemicprinciples- a theory
that would explain why, e.g., certain perceptual conditions yield justified
beliefs.
My present purpose is to show that the truth-connectionprovidesa basis
for such an explanatorytheory. This is illustratedby the theories of Goldman
and Lehrer, both of which make essential use of a non-trivialtruth connection. Those theories can explain the truth of epistemicprinciplesby showing
how the justifying conditions described in the antecedent are connected to
the truth (construed objectively or doxastically) of the beliefs describedin
the consequents. It will be helpful to consideran example.ConsiderPollock's
principle:

292

STEWART

COHEN

If S is appearedto 4-ly,then S is primafacie justified in believing

there is something0 before him.
Presumingthat Goldmanwould concede the truth of a principlevery close to
this, how would he explain its truth?Goldmanholds that justified beliefs are
produced by reliable cognitive processes. Since perception is one example
Goldman gives of a reliableprocess, he would explain the truth of the principle in those terms. If S's belief that there is something pbefore him results
from S's being appearedto 4-ly,then S's belief is produced by the reliable
cognitiveprocess - perception.As suchS's belief is justified.
Lehreras well would concede the approximatetruth of Pollock'sprinciple.
How would he explain its truth? Lehrerholds that justification arisesout of
probabilisticrelations within an acceptance system. Generally when one is
appearedto 0-ly, ones acceptancesystem is such that the proposition 'thereis
something4' is assigneda higherprobabilitythancompetingproposition.So on
Lehrer'stheory, it will generallybe true that if S is appearedto 0-ly, S is justified in believingthat there is something / before him.
The argument would proceed in the same way for epistemic principles
involving memory, introspection, et al. Thus Goldman and Lehrer claim a
theoretical advantageover Foundationalistssince they can in this way explain
the principlesthat the Foundationalistmust take as basic. And they achieve
this advantage by constructing their theories on the basis of a non-trivial
truth connection.

We seem to have come full circle. We began by noting that there is good
reason to suppose that there is an internalconnection between justification
and truth. We then examined several proposals as to how to construe that
connection and found them to be problematic. Finally we considered a
theory that does not employ any such connection and found it wanting for
preciselythat reason.
What can we conclude? Intuitively, it's hard to see how some truth connection could not exist. Being epistemicallyjustified in believing a proposition is quite a different matter from being morallyor pragmaticallyjustified.
And there is a strongtemptation to say that the differencelies in the fact that
unlike moral or pragmaticjustification, epistemicjustificationis connected in

JUSTIFICATION

AND TRUTH

293

some important way to the truth of the justified belief. Nonetheless it has
turned out to be a difficult matter to say precisely what that connection
comes to. While theories that ignore the connection incur an explanatory
deficit, theories that are based on some initially plausible construal of the
connection are problematic. This leaves open severalpossibilities.Perhapsa
theory based on some other formulationof the truth-connectionwill avoid
the shortcomingsof those I have considered.On the other hand, perhapsan
explanatory theory can be achieved without the use of a truth-connection,
although it is hard to see how such a theory could explain type II defeaters.
Finally we might suppose that there is no greater level of generalityto be
attained by a theory of epistemicjustification then that providedby Foundtions theories. There may be no more general truths about the nature of
epistemicjustification beyond those stated in the Foundationalist'sepistemic
principles.

NOTES
* I have benefitted greatly from commentsmade in discussionby Keith Lehrer,John
Pollock,and KeithQuillen.
' Probably, Descartes is most reasonably interpreted as holding this view vis-a-vis
some notion of philosophicalcertainty ratherthan ordinaryjustification. I will sidestep the scholarlyissues by simply referringto this view as the Cartesianview whether
or not Descartesactuallyheld it.
2 Alvin I. Goldman, 'Whatis justified belief, in George S. PappasJustificationand
Knowledge(Dordrecht:1979); MarshallSwain,Reasonsand Knowledge(Ithaca:1981).
Two philosophershave proposedreliabilityanalysesof knowledgeratherthan justiflcation, viz, D. M. Armstrong,Belief, Truth,and Knowledge(London: 1973) and Fred I.
Dretske,Knowledgeand the Flow of Information(Oxford:1981). Becausetheseanalyses
are aimed at knowledge rather than justification, it is not clear whether what I say
appliesto them.
3

lbid,p.

1.-

Unlike Goldman'sfrequencyprobabilityanalysis,Swainanalyzesreliabilityin terms

of inductiveprobability.On this interpretation,C makesB more probablethan not B
only if C constitutes better evidence for B than not B, op. cit. p. 100. This epistemic
interpretationwould seem to renderthe analysiscircular.
5 Goldmanmight reply that in the evil demon case, the cognitiveprocessesthemselves
are reliable.It is only becauseof the interventionof the evil demonthat the beliefs they
produce turn out false. Equivalently,Goldman might hold that reliabilityin a nonmanipulatedenvironmentis a necessarycondition of justification.I don't believe anythingcrucialhangson this. Certainlywe canimaginea worldwherethe cognitiveprocesses
are unreliablein a non-manipulatedenvironment.Wemightimaginea worldwhere,e.g.,
the perceptualprocessesnaturallyproduceconsistentsets of hallucinations.
6 In responseto this objection,Goldmanhas suggested(in conversation)
that his theory
be interpretedin the following way. S's belif at t is justifiedonly if S's belief at t results
from a cognitivebelief-formingprocessthat is reliablein the actualworld.(He makesthe
4

294

STEWART

COHEN

same point in a differentcontext in 'Whatis justified belief?', op. cit., p. 17). On this
interpretation,the e.g., perceptualbeliefs of the inhabitantsof the Demonworldwould
be justified,sinceperceptionis a reliablebelief-formingprocessin the actualworld.
However,the theory so construedwould be too strong.Wecan imaginea worldwith
being's whose perceptualfacultiesoperateaccordingto differentnaturallaws than the
ones that hold in the actualworld.Sincethe perceptualfacultiesof suchbeingswouldbe
unreliablein the actual world, they wouldbe precludedfromhavingjustifiedperceptual
beliefs.
Op. cit., p.23.

If the evil demon should occasionallyarrangefor a belief to be true, then we would

have a Gettiercase. Thisindicatesthat reliabilityis bettersuited to beinga separatecomponent of knowledgeratherthan a constituentof justification.
processesmust be conditionally
9 Strictly speaking,Goldmanholds only that reasoning
reliablein order for them to yield justified beliefs. "A processis conditionallyreliable
when a sufficientproportionof its output-beliefsaretruegiven that its input-beliefsare
true" ('Whatis justified belief?',p. 13). Howeverwe can easily supposethat the Demon
world is a counter-inductiveworld. Of course this fact is concealedfrom its inhabitants
by the Demon and so does not affect the justificationalstatusof theirbeliefs.In sucha
world,inductionwouldnot even be conditionallyreliable.
10 Explanatorycoherencetheorieswould construethe connection between C andB as
"C explainsB." See GilbertHarman,Thought(Princeton,1973). I believe that what I
say about probabilisticcoherencetheorieswill apply to explanatorycoherencetheories
as well.
" BecausecertainFoundationtheoriesare cast in termsof primafacie justification,this
characterizationis not quite right. For primafacie justified beliefs will haveto bearthe
relationof not beingdefeatedto other beliefsif they areto be completelyjustified.
12 CertainCoherencetheoristsalso requirethat the justificationconditionsfor a belief
include an objective connection to truth. See Keith Lehrer, 'Knowledge,truth and
ontology' (forthcoming) (p. 17) and LawrenceBonjour, 'The coherence theory of
empirical knowledge', Philosophical Studies (1976), pp. 293--294. Insofar as they
requirethis they aresubjectto the sameobjectionthat can be raisedto Reliabilism.
13
In his book Knowledge(Oxford: 1974) Lehrerseems to endorseonly a subjective
connection to truth.In his laterwork, 'KeithLehrer- a self profile',in RaduJ. Bogdan,
Keith Lehrer (Dordrecht: 1981) (pp. 79-85), and 'Knowledge,truth and ontology',
ibid. (pp. 16-18), Lehrerendorsesan objectiveconnectionas well.
14 Lehrerrefers to probabilityas an objectivefeature of the world, e.g., a propensity.
'Knowledge,truth and ontology' (p. 11).
1
Ignoringa technicalemendation,P competeswith Q if prob (P/Q) < prob (P).
16 The term 'Intellectualistmodel of justification'is used by ErnestSosa in 'The raft
and the pyramid',Midwest Studies in Philosophy,ed., French,Vehling,and Wettstein
(Universityof Minnesota:1980) (pp. 3-26).
17 John Pollock,Knowledgeand Justification(Princeton:1974), (pp. 58-64).
18 Roderick Chisholm, Theory of Knowledge,2nd edition (EnglewoodCliffs: 1876)
(p. 78).
19 Cf. GilbertHarman,op. cit. Harmanconstruesthe supportingbeliefsas explanatory
ratherthanprobabilistic.(Chapters7 and 8.)
20 This point is madeby Pollockop. cit. (p. 58).
21 This positionis takenby LawrenceBonjour,'Externalist
theoriesof empiricalknowledge' in MidwestStudies,op. cit., pp. 66-67.
22

Op. cit., p. 78. In 'A version of foundationalism', Midwest Studies, op. cit., Chisholm

seems to acknowledgesome form of nontrivialtruth-connection.After pointingout the

trivialconnectionhe writes:

JUSTIFICATION AND TRUTH

295

There is still anotherpoint about the relationbetween epistemicjustification and truth ... if I want to believewhatis trueand not to believewhatis
false, then the most reasonablething for me to do is to believe what is
justifiedand not to believewhat is not justified(p. 545).
Chisholmmust intend the expression"the most reasonablething for me to do" to be
taken in a pragmaticratherthan an epistemicsense. Otherwisethe statementwouldbe
trivial.It is still unclearwhat Chisholmmeanshere. At any rate,no truthconnectionis
expressedin Chisholm'sprinciplesbeyond the trivialone.
23
Thisprincipleis adaptedfrom remarksPollockmakesin ChapterThreeop. cit.
24
Op. cit. ChapterOne (especiallypp. 11-12). Chisholmis not explicit about this
point.
25

What follows is from Chapter Two, op. cit.

Chisholmdoes not proposeprinciplesof primafaciejustification.Hehandlesdefeaters

by adding clauses directly to the epistemic principle.See Theoryof Knowledge,2nd
edition (pp. 75-76).
27
In Knowledgeand Justification,Pollock does remark"P is a primafacie reasonfor
S to believe-that-Qiff S is prinma
facie justified(in the senseof Section 2.2) in believingthat (P Q)" (p. 42). Pollockuses (P - Q) to sympolizethe subjunctiveconditional"it
wouldnot be true that P unlessit weretrue that Q" (p. 42). Section 2.2 tells us:
"P is prima facie justified for S" means "It is necessarilytrue that if S
(2.2)
believes (or were to believe) that P, and S has no reasonfor thinkingit is
false thatP, then S is (or wouldbe) justifiedin believingthatP" (p. 30).
26

I don't understandthe basis for Pollock's remark,given his account of prima facie
reasonsin Chapters1 and 2. Let P be "S is appearedto redly" and let Q be "Thereis
somethingred beforeS." SupposeS believes(P Q) becausethe OuijaBoardtells him it
is true. This does not giveS a reasonto believeit is false that (P Q). But surelyS is not
justifiedin believing(P Q).
28 Pollockhas takenthis positionin conversation.
29
'The knowledgecycle', Nous, 1977, p. 19. Goldmanmakes the samepoint in 'What
is justified belief?' op. cit., p. 2.

Departmentof Philosophy,
Princeton University,
Princeton,N.J. 08544,
U.S.A.