This article was downloaded by: [Canadian Research Knowledge Network]

On: 29 March 2009

Access details: Access Details: [subscription number 782980718]
Publisher Routledge
Informa Ltd Registered in England and Wales Registered Number: 1072954 Registered office: Mortimer House,
37-41 Mortimer Street, London W1T 3JH, UK

Australasian Journal of Philosophy

Publication details, including instructions for authors and subscription information:
http://www.informaworld.com/smpp/title~content=t713659165

The psychological turn

Hilary Kornblith a
a
University of Vermont,

Online Publication Date: 01 September 1982

To cite this Article Kornblith, Hilary(1982)'The psychological turn',Australasian Journal of Philosophy,60:3,238 — 253
To link to this Article: DOI: 10.1080/00048408212340661
URL: http://dx.doi.org/10.1080/00048408212340661

PLEASE SCROLL DOWN FOR ARTICLE

Full terms and conditions of use: http://www.informaworld.com/terms-and-conditions-of-access.pdf

This article may be used for research, teaching and private study purposes. Any substantial or
systematic reproduction, re-distribution, re-selling, loan or sub-licensing, systematic supply or
distribution in any form to anyone is expressly forbidden.

The publisher does not give any warranty express or implied or make any representation that the contents
will be complete or accurate or up to date. The accuracy of any instructions, formulae and drug doses
should be independently verified with primary sources. The publisher shall not be liable for any loss,
actions, claims, proceedings, demand or costs or damages whatsoever or howsoever caused arising directly
or indirectly in connection with or arising out of the use of this material.
 Australasian Journal of Philosophy
Vol. 60, No. 3; September 1982

THE PSYCHOLOGICAL TURN

Hilary Kornblith

Traditional epistemological theorists tried to develop a set of rules which an

agent might use to determine what he ought to believe; this was conceived to be
Downloaded By: [Canadian Research Knowledge Network] At: 20:41 29 March 2009

the central project of any epistemological theory. These traditional theories

were apsychologistic; questions about justified belief were thought to have
nothing to do with questions about the psychological processes which in fact
occur when beliefs are acquired.
Recently, epistemology has taken a psychological turn. It is now widely
believed that questions about the justification of belief cannot be answered
independently of questions about a belief's causal ancestry. What effect, if any,
does the psychological turn have on the central project of traditional
epistemology?
In sections I and II, I outline the differences between psychologistic and
apsychologistic epistemology. In section III, I discuss some reasons for thinking
that the proper form of an epistemological theory is psychologistic, and in the
remaining sections, I explore the implications of psychologising epistemology
for traditional epistemological projects.

Let me begin by giving a precise characterisation of the two different kinds of

theory of knowledge at issue.
An apsychologistic theory of knowledge is characterised by a rule for
determining a privileged class of beliefs, and a set of epistemic rules. Given the
set of an agent's beliefs at a particular time, we first apply the rule for
determining the privileged class of beliefs, the result of which, of course, is a set
of beliefs. To this set of beliefs we apply the various epistemic rules. The set of
beliefs thus determined are those which the agent is justified in believing at the
time in question. An agent is justified in believing a particular proposition at a
particular time just in case that proposition is a member of the set determined
by applying the epistemic rules to the privileged class of beliefs for that agent at
that time.
Foundationalism is one obvious variety of apsychologistic theory.
Foundationalism in general may be seen as an attempt to specify a rule for
determining a privileged class of beliefs as well as a set of epistemic rules for
deriving justified beliefs from that privileged class. Foundationalists have
sometimes claimed, for example, that the set of beliefs an agent has about his

238
 Hilary Kornblith 239
present state of mind form a privileged class and that the beliefs in this class are
justified, as are those which can be derived from them, say, by deduction and
induction.
The coherence theory of knowledge is also a kind of apsychologistic theory.
One form of the coherence theory has it that all of the beliefs an agent has at a
particular time are members of the privileged class. It is then claimed that there
is a single epistemic rule: the coherence rule. The coherence rule tells us that
those members of the privileged class which 'fit together' in the appropriate way
are justified for the agent in question at the time in question. Coherence
theorists may thus be seen as attempting to explain precisely what their one
epistemic rule amounts to, that is, as attempting to say precisely what 'fitting
Downloaded By: [Canadian Research Knowledge Network] At: 20:41 29 March 2009

together' is.
Psychologistic theorists, however, eschew epistemic rules. Instead, a
psychologistic theory of knowledge is characterised by a description of a set of
justification-conferring processes. It is claimed that if the presence of a belief in
a particular agent at a particular time is to be accounted for by one of these
processes, then the agent is justified in holding that belief at that time. Thus far,
the details of the psychologistic theory have yet to be worked out. No one has
yet offered a complete characterisation of the set of justification-conferring
processes, nor, indeed, has a complete account been offered of what makes a
process justification conferring. Nevertheless, the outlines of such a theory have
already clearly emerged, especially in the work of Alvin Goldman. 1
It is not clear from the manner in which I have characterised these two kinds
of theory thus far that there is any genuine difference between the two. Indeed,
it might seem that the psychologistic theory and the apsychologistic theory
come down to very much the same thing. Doesn't each epistemic rule specify a
justification-conferring process for acquiring beliefs? And doesn't each
justification-conferring processs specified by the psychologistic theory allow us
to formulate an epistemic rule for determining justified beliefs? If these two
kinds of theory are to be seen as differing in anything more than style of
formulation, more needs to be said about the difference between epistemic rules
and justification-conferring processes.

II

Let us say that an epistemic rule is a rule which, when applied to a set of beliefs,
yields a set of beliefs as output. The central project of traditional epistemology
has been to discover those rules which yield justified beliefs as output when
applied to the class of privileged beliefs. Thus we will have a rule which states
that beliefs which have properties PI . . . . . P, are members of the privileged
class of beliefs ~r, and we will have a rule which states that if or is the set of one's

Alvin Goldman, 'Discrimination and Perceptual Knowledge,' Journal of Philosophy, LXXIII

(1976), pp. 771-791; 'What is Justified Belief?,' in G. Pappas, ed., Justification and Knowledge,
(Dordrecht: Reidel, 1979), pp. 1-23.
 240 The Psychological Turn

privileged beliefs, one is justified in believing the members of zr', where ~" is
the closure of ~r under operations O1. . . . . Om.
(1) The Applicability Constraint: The rule which specifies the class of
privileged beliefs as well as the epistemic rules specifying acceptable operations
on beliefs must be rules which the agent is in a position to apply. Thus,/'1 . . . . .
Pn must be properties which an agent will recognise his beliefs to have, if they
have them, and O I. . . . . O,, must be operations which an agent is in a position
to perform on the members of set 7r of his privileged beliefs.
Why should one believe that epistemic rules must obey the Applicability
Constraint? I believe the motivation for this constraint to be roughly as follows.
A responsible epistemic agent wants to have only those beliefs he is justified in
Downloaded By: [Canadian Research Knowledge Network] At: 20:41 29 March 2009

having; he must therefore be able to sort out the propositions he ought to

believe from those he ought not. The agent thus needs a set of rules which he is
in a position to apply which will tell him what transitions among beliefs transmit
justification, as well as a rule which, again, he is in a position to apply, and which
will tell him to which of his beliefs these rules of transition may justifiably be
applied. Rules which the agent is not in a position to apply will do the agent no
good.
(2) The A Priority Constraint: It is a priori knowable that if operations 01,
. . . . 0 m are performed on the members of a set of privileged beliefs 7r, the
resulting set of beliefs ~-' will be beliefs which the agent in question is justified in
believing.
Let me draw a distinction between basic and derived epistemic rules. I will
say that an epistemic rule is basic if an agent would be justified in applying it no
matter what experience he might have or have had; an epistemic rule is derived
just in case it is not basic. Now it seems that in order to justifiably adopt a
derived epistemic rule, one must be justified in believing that one has had the
experiences which make its adoption justifiable; this, of course, means that
one's adoption of the belief about one's experience must be licensed by the
appropriate epistemic rule. If all episternic rules were derived rules, this would
lead to an infinite regress; this would, of course, violate the Applicability
Constraint. Therefore, there must be some basic epistemic rules.

2 Some examples of recent defenders of each of these constraints are as follows.

Applicability Constraint: Roderick Chisholm, Theory of Knowledge, second edition
(Englewood Cliffs: Prentice-Hall, 1977); p. 62; James Cornman, 'On Justifying Non-Basic
Statements', in G. Pappas, ed., op. cit., pp. 142-143; Roderick Firth, 'Ultimate Evidence', in R.
Swartz, ed., Perceiving, Sensing and Knowing (Berkeley: University of California Press, 1965),
p. 495; Carl Ginet, Knowledge, Perception andMemory (Dordrecht: Reidel, 1975), p. 34; Keith
Lehrer, Knowledge (Oxford: Clarendon Press, 1974), p. 198; C. I. Lewis, An Analysis of
Knowledge and Valuation (La Salle: Open Court, 1946), p. 30; John Pollock, Knowledge and
Justification (Princeton: Princeton University Press, 1974), pp. 33-49; Bertrand Russell,
Problems of Philosophy (Oxford: Oxford University Press, 1912), pp. 130-140.
A Priority Constraint: Roderick Chisholm, Perceiving (Ithaca: Cornell University Press,
1957), p. 100; R. Firth, Ioc. cit.; Mark Pastin, 'Meaning and Perception', Journal of Philosophy,
LXXIII (1976), p. 574; J. Pollock, loc. cit.; B. Russell, op. cir., p. 134.
Necessity Constraint: R. Chisholm, Theory of Knowledge, p. ~/1;R. Firth, Ioc. cit.; M. Pastin,
Ioc. cit.; J. Pollock, Ioc. cir.; B. Russell, op. cit., p. 134.
 Hilary Kornblith 241
The set of basic epistemic rules are thus ones which any person is justified in
applying; which derived rules a person is justified in applying will vary from
person to person, depending on one's experience. When epistemologists speak
of discovering the set of epistemic rules, they are of course speaking of the set of
basic epistemic rules, the epistemic rules which do not vary from person to
person; I will follow this usage. Since (basic) epistemic rules are those which an
agent is justified in applying no matter what experience he may have or have
had, it follows immediately that it must be knowable a priori that these rules
may be justifiably applied.
(3) The Necessity Constraint: It is a necessary truth that if operations 01,
. . . . Or. are performed on the members of a set of privileged beliefs ~-, the
Downloaded By: [Canadian Research Knowledge Network] At: 20:41 29 March 2009

resulting set of beliefs ~r' will be beliefs which the agent in question is justified in
believing.
Since the justifiable application of these rules is knowable a priori, our
knowledge that these rules are so applicable cannot depend on any feature
peculiar to the actual world. But if the applicability of the rules does not depend
on features peculiar to the actual world, that they may be justifiably applied
must be necessarily true.
While I do not endorse the arguments of this section, I believe that they are
quite important. Even a brief look at an incomplete list of those who have
endorsed arguments like those above should convince the reader that the
importance of these arguments is not negligible, and that their influence on the
development of recent epistemology is immense. 2

III
In light of recent work, it can no longer be denied that a certain psychological
element must enter into epistemological theorising. As G. Harman and A.
Goldman have shown, 3 questions about the justification of a belief cannot be
answered independently of questions about its causal ancestry. For any
favoured logical connection among beliefs, one can describe cases where the
logical connection holds but is not mirrored in psychological connections; where
putative epistemic rules tell us that p provides good reason for q, we may
describe cases where an agent believes both p and q, but does not believe p
because he believes q. Familiar examples involve mad neuro-physiologists,
people who believe propositions because they like the sound of particular
sentences expressing them, and so on. An adequate theory of justification must
thus take account of the psychological connections among beliefs. The question
thus facing contemporary epistemologists is no longer, 'Is the proper theory of
knowledge psychologistic or apsychologistic?', but rather, 'How much
psychology must we allow into our epistemology?'.
In order to further explore the real difference between psychologistic and

3See especially Gilbert Harman, Thought (Princeton: Princeton University Press, 1973),
pp. 24-33; Alvin Goldman, 'Discriminationand Perceptual Knowledge'; Alvin Goldman, 'What
is Justified Belief?' op. cit.
 242 The Psychological Turn
apsychologistic theories, we would do well to consider a question which arises
for psychologistic theories of knowledge. The remainder of this section will be
devoted to motivating that question.
Roughly, psychologistic theories are those which hold that a belief is justified
just in case its presence is due to the workings of the appropriate sort of belief
forming process. Psychologistic theorists are thus committed to a certain
interest in the processes by which beliefs are formed. Now it should be clear
that these processes are ones the details of which are known to few, if any,
persons; certainly the details of these processes are not accessible to
introspection. For example, in the case of depth perception, it has been
determined that changes in relative shadow length influence a subject's beliefs
Downloaded By: [Canadian Research Knowledge Network] At: 20:41 29 March 2009

about the relative distance of objects; nevertheless, subjects are unaware that
relative shadow length has this effect. This phenomenon can only be accounted
for if we suppose that there is some state of a perceiver, inaccessible to
introspection, which 'registers' relative shadow length and plays a crucial role in
determining the agent's beliefs about the relative distance of objects. What is
important about this case, of course, is that there is nothing peculiar to the role
shadows play in depth perception; there are numerous mechanisms which play
a crucial role in the production and retention of beliefs which are such that an
agent has no direct access to them.
We are now in a position to formulate the question with which psychologistic
theorists must deal. Psychologistic theorists claim that a person has a justified
belief just in case the process responsible for the presence of that belief is of the
appropriate sort; if we trace back along the causal chain which leads to the
production of a belief, we will find not only transitions among accessible states,
but also transitions among mental states which are not directly accessible to the
agent. Question: To the extent that psychologistic theorists are correct in
claiming that a belief producing process must be of a certain sort if it is to issue
in justified belief, must we be concerned with only that part of the causal chain
which involves transitions among accessible states, or must we look at the
causal chain from start to finish, i.e., in its production of and transitions among
mental states of whatever sort?

IV
If the arguments in favour of the three constraints on epistemic rules given in
section II are sound, then there is good reason to take the psychological turn in a
most conservative manner. If the motivation for the three constraints on
epistemic rules remains untouched when we take the psychological turn, then
the justification conferring processes of psychologised epistemology are nothing
more than those processes which accord with the epistemic rules of
apsychologistic theories. In short, while the notion of performing an operation
on a set of beliefs is clearly psychological - - the fact that an agent has a belief
which could have been arrived at,by performing a certain operation on his other
beliefs does not guarantee that it was arrived at by performing that operation, --
determination of the set of operation O 1 , . . . , Omis a task which has nothing to
 Hilary Kornb#th 243
do with the contingencies of human psychology. Someone who adopts a
position like this grants the psychological point of Section III, but denies that it
has any import for epistemological theorising. There are, after all, two distinct
notions of justification, corresponding to the notions of believing for a reason
and having a reason for believing. 4 Being justified, in the sense of believing for a
reason, is clearly a psychological notion. Someone who opposes psychologism,
however, may well grant this point and yet deny that being justified, in the
sense of having a reason for believing, is to be analysed in psychological terms.
In order to determine what the relations among our beliefs ought to be, we need
only logic; psychology comes in only in determining whether these logical
relations are, in fact, instantiated. There is a deceptively attractive argument
Downloaded By: [Canadian Research Knowledge Network] At: 20:41 29 March 2009

which seems to demand this conservative move. Let us examine it. 5

I would like to bring to the fore a principle which was implicit in the
motivation for the Applicability Constraint on epistemic rules. People are
responsible for the beliefs they have; if a person has an unjustified belief, that
person is epistemically culpable. Let me call this the Principle of Epistemic
Responsibility. 6 This Principle plays a crucial role in the argument I am about to
give. I believe this Principle to be true, but I will not argue for it here; I will
simply adopt it as a working hypothesis.
We want to know, in examining a belief forming process to determine if it is
justification conferring, whether we should examine the process only insofar as
it deals with transitions among accessible states, or whether our examination
must extend to the production of and transitions among all mental states within
the realm of the complete psychological description of the belief forming
process. If the latter is true, the Principle of Epistemic Responsibility tells us
that we are responsible not only for transitions among states accessible to us, but
also for transitions among states which are inaccessible to us (e.g., the state
which registers relative shadow length and its kin). This does not seem fair.
How can one reasonably hold a person responsible for the transitions that
person makes among mental states which he doesn't know about? There is
such a thing as culpable ignorance, of course, but it doesn't seem fair to require
of a responsible epistemic agent that he know a complete psychological theory;
this, and more, would be required if we were to insist that justification has to do
with transitions among all mental states involved in the complete psychological
description of the belief forming process. Hence, the reliability required for
justification must deal only with transitions among beliefs.
Indeed, the proponent of this argument might argue for a much stronger

4 My way of formulating this distinction is due to Robert Audi, 'Foundationalism and Epistemic
Dependence', Journal of Philosophy, LXXVII (1980), pp. 612-613. The same distinction is
drawn, using different terminology, by Roderick Firth and Alvin Goldman. See R. Firth, 'Are
Epistemic Concepts Reducible to Ethical Concepts ?', in Alvin Goldman and Jaegwon Kim, eds.,
Vah~,es and Morals: Essays in Honor of William Frankena, Charles Stevenson and Richard
Brandt (Dordrecht: Reidel, 1978); Alvin Goldman, 'What is Justified Belief?', op. cit.
I will examine only one type of argument given for these constraints; this is not to deny that
arguments of a radically different sort might be given. Nevertheless, as the references cited in
footnote 2 indicate, this kind of argument is quite influential.
6 This Principle was suggested to me, in another context, by Patricia Kitcher.
 244 The Psychological Turn
conclusion. What kinds of transition among beliefs make for justification?
Suppose that someone believes that p and on that basis comes to believe that q.
Let us further suppose that p provides good reason to believe that q, but the
agent in question does not know this. Is the transition from p to q adequate for
justification?
It might seem that an argument can be given based on the Principle of
Epistemic Responsibility for the conclusion that the inference in question is not
adequate for justification. Suppose that the agent who makes this inference is
examining his beliefs. He notices that he infers q from p, and he doesn't know
whether p proves good reason for q. We have supposed, of course, that p does
provide good reason for q, but the agent doesn't know this. Now if the agent is
Downloaded By: [Canadian Research Knowledge Network] At: 20:41 29 March 2009

in fact justified in believing q on the basis of p, the fact that he doesn't know that
p provides good reason to believe q is irrelevant. But now consider a situation in
which the same agent infers s from r, where r does not provide good reason for
s. Here, again, we may imagine the agent examining his beliefs. He notices that
he has inferred s from r, but he doesn't know whether r provides good reason
for s. For all he can tell, he is in exactly the same situation with respect to r and s
that he is in with respect to p and q. The move from r to s is clearly unjustified
and thus, by the Principle of Epistemic Responsibility, we must hold the agent
culpable for his epistemic error. But if the agent is justified in making the
inference from p to q, then the agent has done something epistemically
praiseworthy in one case and epistemically blameworthy in another, where the '
agent is not in a position to tell the difference between the two! Again, the
notion of culpable ignorance does not seem to apply, and this application of the
Principle of Epistemic Responsibility seems unfair. The only way out of this
problem, it seems, is to deny that the inference from p to q is adequate for
justification when the agent does not know that it is adequate for justification.
If this argument is valid, then all those features of a person which make him
justified in believing a proposition must be directly accessible to him; in short,
this is an argument for the Applicability Constraint. Taking the psychological
turn does not seem to block the arguments of section II which led us from the
Applicability Constraint to the A Priority and Necessity Constraints, and thus
the search for epistemic rules once again takes its place at the forefront of
epistemological theorising. Naturalising epistemology, it seems, can be
accomplished merely by appending a footnote to apsychologistic theories.

V
Certain very general considerations about responsibility play a crucial role in the
argument just given, and it will be important to bring them out into the open.
Let it be granted that the Principle of Epistemic Responsibility is true; we are
responsible for the beliefs we have. The beliefs we have are generated by certain
processes which take place in us. In order to fulfill our epistemic responsibility,
must we be able to monitor these processes every step of the way? The
argument just given in defence of epistemic rules presupposes that we must. It
should be clear, however, that we need not.
 Hilary Kornblith 245
Consider a person who works in a factory and is responsible for the quality of
products produced by a certain machine. One way in which she might fulfil her
responsibility is by mastering all the details of the mechanism by which the
product is produced, and periodically checking each part of the machine to
insure that the mechanism is operating properly. This is, however, a costly and
inefficientway for her to fulfill her responsibility. A far more sensible approach
would be for her to periodically check the products for defects. If defects are
found, the machine may need overhauling; if not, the machine may reasonably
be assumed to be working properly. No knowledge of the inner workings of the
machine is necessary in order to perform this task.
The same is true with beliefs. Each of us is responsible for the beliefs he has.
Downloaded By: [Canadian Research Knowledge Network] At: 20:41 29 March 2009

One way for us to fulftll this responsibility is to develop a complete theory of

psychological processes and a complete theory of what makes such processes
justification conferring. Needless to say, this cannot be done. Fortunately, it
need not be done in order to fulfill our epistemic responsibility. All we need do
is monitor the output of our psychological processes. This amounts to nothing
more than being on the lookout for false beliefs. Nothing need be known of the
manner by which beliefs are produced.
Consider, for example, the case of the chicken sexer. 7 Apparently it is
possible to train people to distinguish between male and female two day old
chicksby having them look at labelled photographs of such chicks. After a bit of
practice, chicken sexers achieve a very high degree of accuracy in classifying
young chicks by sex. Interestingly enough, not only are the chicken sexers
themselves incapable of explaining what visual cues they rely on in performing
this feat, but no one has yet determined what visual oaes are relied upon.
Nevertheless, the task is performed.
Now chicken sexers, like everyone else, are responsible for the beliefs they
have. They do not fulfill this responsibility, however, by knowing anything
about the processes by which their beliefs about the sex of chicks are formed.
Rather, in the process of training, they periodically test themselves to make
sure that their judgments about the sex of photographed chicks are correct.
Once established in business, a responsible chicken sexer will keep in touch
with his customers to assure that few mistakes have been made. A chicken
sexer who finds that he has been making a fair number of errors will enroll in a
refresher course.
What is true of the chicken sexer is true of epistemic agents generally. Most
of us know little if anything of the processes by which our beliefs are formed.
This does not, however, impede us in fulfilling our epistemic responsibility. All
we need do is keep on the lookout for false beliefs. To the extent that we find
them, we must revise our ways of acquiring beliefs; this does not, however,
require any knowledge of the details of the processes by which we acquire them.
For lack of a better term, we might say that chicken sexers acquire a certain
epistemic habit, namely, the habit of acquiring beliefs about the sex of young

7 This example was brought to the attention of philosophers by Douglas Gasking's 'Avowals', in
R. Butler, ed., Analytical Philosophy (New York: Barnes & Noble, 1962).
 246 The Psychological Turn
chicks on the basis of those chicks' visual appearances. In well trained chicken
sexers, this epistemic habit is a good habit and might thus better be described as
an ability - - the ability to discriminate male from female chicks. Poorly trained
chicken sexers do not acquire good epistemic habits; they do not acquire
abilities.
The Principle of Epistemic Responsibility tells us that chicken sexers are
responsible for their beliefs. In light of the above discussion, we might put the
Principle in the following way: People are responsible for their epistemic habits.
Chicken sexers are thus responsible for the manner in which they come to form
beliefs. The complete psychological description of the process via which chicken
sexers acquire their beliefs about chicks will undoubtedly include transitions
Downloaded By: [Canadian Research Knowledge Network] At: 20:41 29 March 2009

among states the very existence of which the chicken sexer is unaware. By
holding the chicken sexer responsible for the reliability of his belief producing
processes, we hold him responsible for these transitions. This amounts to
nothing more, however, than holding him responsible for his epistemic habits,
and this is no more unreasonable than holding him responsible for his other,
non-epistemic, habits.
People are thus responsible for acquiring good epistemic habits just as much
as they are responsible for acquiring other good habits. Of course, in acquiring a
habit, one acquires a tendency to move from mental state to mental state, and
these transitions are not always accessible, much less known, to the person who
goes through them. This does not, however, make the burden of responsibility
for one's habits unduly heavy. If it did, we could never be responsible for any of
our acts, epistemic or otherwise.
The epistemic ability possessed by the chicken sexer is an acquired ability,
and the remarks I have made thus far have an obvious application to all
acquired epistemic habits. Does the Principle of Epistemic Responsibility apply
to innate epistemic habits as well? I believe it does. What the Principle of
Epistemic Responsibility requires, in effect, is that we constantly be on the look-
out for false beliefs. If we have some habit which tends to produce false beliefs,
we are responsible for giving up that habit. But this has nothing to do with
whether our habits are acquired or innate; that a habit is innate does not mean
that it cannot be given up. We are thus responsible for all our epistemic habits.
The truth of the Principle of Epistemic Responsibility thus in no way requires
that we apply the Applicability Constraint to our account of justification
conferring processes.
There is a deeper reason, however, for rejecting the Applicability Constraint,
for, on one straightforward reading of what it requires, it cannot be satisfied. If it
is held that no belief is justified unless it was arrived at under the guidance of
epistemic rules, and in order for a belief to be arrived at under the guidance of
an epistemic rule, the epistemic rule itself must be justifiably believed, then we
are confronted with a very familiar regress. In order for any belief to be
justifiably held one must justifiably believe some epistemic rule, but in order to
believe that one must have arrived at it under the guidance of some further
epistemic ruie, which in turn requires justified belief in thatepistemic rule, and
SO o n .
 Hilary Kornb# th 247
A defender of the Applicability Constraint may insist that my reading of the
phrase 'under the guidance of epistemic rules' is too strong. In order to arrive at
a belief under the guidance of epistemic rules, it will be insisted, one need not
actually believe the rules. This reading need not be so weak as to allow that
whenever p, if p, then q, and q are believed that q is arrived at under the
guidance of the epistemic rule which licenses the operation of modus ponens.
Rather, in determining whether a particular belief is justified, we look at the
particular inferences which were actually made m this is where the psychology
comes in m and we then determine whether these inferences are licensed by
the epistemic rules. If they are, then the inferences take place under the
guidance of the epistemic rules.
Downloaded By: [Canadian Research Knowledge Network] At: 20:41 29 March 2009

This weak reading of 'under the guidance of epistemic rules' solves one
problem for the defender of the Applicability Constraint, but, at the same time,
destroys its motivation. The motivation behind insisting that epistemic rules be
applicable is simply that unless an agent can consider these rules, recognise their
truth and then apply them, the agent will not be able to decide what he ought to
believe. On the weak reading of proceeding 'under the guidance of epistemic
rules', however, the agent need neither consider the rules nor recognise their
truth. What then is the point in insisting that the rules be applicable? It must, of
course, be possible for our reasoning to be governed by the epistemic rules, but
once the strong reading of 'under the guidance of' is dropped, as it must be, that
we act under the guidance of epistemic rules requires no more in the way of our
applying them than that we digest under the guidance of the laws of chemistry.
There is no need, however, to revert to the weak reading if the scope of the
strong reading is restricted. When we insist that all beliefs must be arrived at
under the guidance of epistemic rules in order to be justified, the strong reading
leads to an infinite regress. If, however, we exempt beliefs in the epistemic rules
from this requirement, the regress is halted. It will not do to claim that beliefs in
the epistemic rules are justified no matter how they are arrived at, for this is not
true. Any belief can be had for bad reasons. Nevertheless, while other beliefs
require some special action on the part of the agent to be justified m namely,
reasoning under the guidance of epistemic rules ~ it might be claimed that
beliefs in the epistemic rules, while not justified whenever held, need not be
subject to the same strict requirements as other beliefs. If applicability is to be
preserved and the regress avoided, it must be possible to chart this middle
course.
It is here that the A Priority Constraint comes in. If the epistemic rules are a
priori knowable, one may be justified in believing them without actually
applying epistemic rules. There are, to be sure, ways in which one might come
to believe the rules which would fail to justify one's belief ~ for example, if one
were to believe them on the basis of bad authority - - but as long as these ways
can be ruled out, the belief in the epistemic rule is justified. The requirement for
justified belief in epistemic rules is purely negative, for they are justified so long
as they are not the product of illusion, bad authority, etc.; other beliefs,
however, are only justified if they are arrived at under the guidance of epistemic
rules. This is, I believe, the position one must adopt if the Applicability
 248 The Psychological Turn
Constraint is to be honoured.
In order to fill out this position, one must give some account of the property
which illusion, bad authority, and so on, have in common in virtue of which
beliefs in the epistemic rules arrived at by these means are not justified; let us
call this common property P. It must be sufficient for a belief in an epistemic
rule to be justified that the means by which it was arrived at lacks property P,
we cannot require that the agent believes that it lacks property P without a
vicious regress. The result is thus a psychologistic theory of justification for
beliefs in the epistemic rules together with an apsychologistic theory of
justification for all other beliefs.
It is not possible to adopt an apsychologistic theory through and through-- to
Downloaded By: [Canadian Research Knowledge Network] At: 20:41 29 March 2009

require even of beliefs in the epistemic rules that they be arrived at by

application of epistemic rules - - for this results in a regress. This forces us to
grant, howevei, that despite the attractiveness of the argument for the
Applicability Constraint, it is impossible to apply it to all beliefs. If it is possible
to be justified in some beliefs without having applied epistemic rules in arriving
at them, why need we apply epistemic rules in order to be justified in any
beliefs? Isn't the motivation for the Applicability Constraint completely lost
once it is discovered that it is impossible to arrive at all beliefs by applying
epistemic rules?
Here it will be pointed out that the epistemic rules enjoy a special epistemic
status insofar as they are a priori knowable. It is in virtue of this special status
that the standards for justified belief may be relaxed in the case of beliefs in the
rules themselves. Alternatively viewed, it is in virtue of the fact that most other
beliefs lack this special status that the standards for justified belief must be
raised in the case of beliefs other than those in the epistemic rules. The two-
tiered approach of apsychologistic theories is thus not an ad hoc move to avoid
the regress, but just what one would expect in virtue of the special epistemic
status of the rules.
Even if one were to grant that there are a priori knowable epistemic rules, it is
not at all obvious why beliefs in these rules should be judged by different
standards than other beliefs in determining their justificatory status. A priori
knowable or not, different processes of acquisition may result in a different
justificatory status. The claim that the rules are a priori knowable, or enjoy some
special epistemic status, does no work here.
Thus far, I have limited my discussion to the Applicability and A Priority
Constraints. Insofar as the argument for the Necessity Constraint depends on
both of these, it is doubly damned. Moreover, the move from 'p is knowable a
priori' to 'p is necessarily true', which lies at the heart of the motivation for this
constraint, is now widely recognised to be fallacious, in light of the pioneer~
work of Saul Kripke. 8

See especially, 'Naming and Necessity', in G. Harman and D. Davidson, eds., Semantics of
Natural Language (Dordrecht: Reidel, 1972), pp. 253-355. For a particularly illuminating
account of the defect in this argument, see Philip Kitcher, 'A Priori Knowledge', Philosophical
Review, LXXXIX (1980), pp. 17-18, and 'A Priority and Necessity', Australasian Journal of
Philosophy, 58 (1980), pp. 89-101.
 Hilary Kornblith 249
If the arguments I have offered are correct, the importance of taking the
psychological turn in epistemology has been vastly underrated. The central task
of apsychologistic epistemology was to discover the correct set of epistemic rules
- those rules in accordance with which an epistemic agent would have to
-

reason if he were to be justified. If we adopt a psychologised epistemology, as I

believe we must, the motivation for this project disappears.
What I have argued is not that there can be no epistemic rules which are
applicable, a priori knowable or necessary; rather, I have argued that these
constraints are unmotivated. Until we see special reason for applying these
constraints, our epistemological theorising should not be limited by them. I
would not be surprised if it should turn out that, in the sense in which there are
Downloaded By: [Canadian Research Knowledge Network] At: 20:41 29 March 2009

epistemic rules, they are necessary truths, though my reasons for thinking this
are based on a Kripkean notion of necessity and not on the notion of necessity
which, I believe, motivates the apsychological account described. In any case,
such a claim requires an argument, and until arguments are given, I see no
reason for placing any constraints like those discussed on epistemological
theorising.
My suggestion that epistemologists ought to give up the project of looking for
applicable, a priori knowable and necessary epistemic rules is not meant to
amount to the view that epistemology is dead. A naturalistic epistemology
brings with it a new set of problems and a new set of projects. If I am right, the
old project of finding epistemic rules is dead, but the field of inquiry of which
that project was a part is very much alive. Work has only begun on an adequate
account of what it is that makes a belief forming process justification conferring.

VI
Thus far I have argued that the psychological turn should effect a substantial
change in epistemological theorising. I do not mean to suggest, however, that
the concerns which motivated traditional theories can be ignored. There is, I
believe, a real worry for psychologistic theories of knowledge which derives
from the arguments presented in section IV, and I do not believe that this worry
has been adequately addressed yet.
There seem to be at least two kinds of case in which a process which seems to
be justification conferring does not yield a justified belief:
(a) While the agent's belief is caused by what seems to be a justification
conferring process, the agent believes that the belief was caused by a
process which is not justification conferring.

(b) While the agent's belief is caused by what seems to be a justification

conferring process, it is almost universally held among the members of
the agent's community that this process is not justification conferring,
and the agent is unable to answer the challenge which members of the
community would pose him.
One example of type (a) finds an agent with perfect pitch who believes himself
to be tone deaf unable to prevent himself from coming to have beliefs about the
 250 The Psychological Turn
pitch of various tones. An example of type (b) finds a genuine telepath who has
never checked out any of his telepathic deliverances relying on his 'sixth sense'
in spite of being a member of a community which has perfectly sensible
objections to the possibility of telepathy. In both cases, processes which seem to
be justification conferring issue in unjustified belief.
The problem here, I believe, stems from supposing that psychologistic
theories of knowledge and justification are committed to the view that the
justificatory status of belief is a function of nothing more than the type of
process by which the belief was produced. It should be clear, however, that
psychologistic theories are not so committed. What makes an epistemological
theory psychologistic is that it is committed to the view that the justificatory
Downloaded By: [Canadian Research Knowledge Network] At: 20:41 29 March 2009

status of a belief is, at least in part, a function of the process by which the belief
was produced; this is not to say that it is not a function of other things as well.
Indeed, it should be clear that if we divide psychological processes into types
in any reasonable sort of way, the justificatory status of a belief must depend on
something more than the type of process by which it was produced. If you and I
both come to believe that a certain object in front of us is an orange, and we
both come to believe this on the basis of the object's visual appearance, then on
any intuitively reasonable~account of how to divide psychological processes into
types, you and I have just undergone the same type of psychological process.
This does not, however, guarantee that our beliefs have the same justificatory
status. If I have independent reason to believe that the object which appears to
be an orange is in fact a piece of artificial fruit, and yet, for whatever reason, this
bit of information does not in any way affect my belief about the object in front
of me, then I will not be justified in my belief while you are justified in yours.
Similarly, if I have special reason to doubt my ability to recognise oranges and
yet still come to have the belief that the object in front of me is an orange, I will
again be unjustified in my belief while you are justified in yours.
An adequate account of justification will thus have to allow that the
justificatory status of a belief is at least a function of two variables: the process
which is responsible for the presence of the belief and the agent's background
beliefs. Two processes of the same type may yield beliefs of different
justificatory status if they operate in environments of suitably different
background beliefs.
Cases Of type (a) are merely one kind of example which show the importance
of background beliefs in helping to determine the justificatory status of a belief.
Examples of this sort do not exhibit a defect in psychologism; they exhibit a
defect only in a crude psychologistic theory which insists that the justificatory
status of a belief depends on the type of process responsible for its presence, and
nothing else.
Cases of type (b) have a different sort of solution. What is at issue here has
come to be known as the social character of knowledge. Harman, for example,
has urged that an otherwise unexceptional means of arriving at beliefs will not
be justification conferring in a suitably hostile environment; in particular, a
belief cannot be justified, whatever the process which produced it, if there is a
widely known challenge to the belief in the agent's community which the agent
 Hilary Kornblith 251

is not in a position to answer. 9

The Principle of Epistemic Responsibility has a great deal to do with cases of
this sort. Although the focus of epistemological discussions is often on
perceptual knowledge, much, if not most, of the knowledge we acquire is due to
our interaction with other members of our community. To the extent that we
recognise the information gathering ability of other epistemic agents, we must
recognise that when we find others disagreeing with us, we must re-think our
own position. We ignore this at our peril; to do so would be epistemically
irresponsible.
This kind of example thus introduces a third factor of which the justificatory
status of a belief is a function: the agent's environment. Once again, this does
Downloaded By: [Canadian Research Knowledge Network] At: 20:41 29 March 2009

not show a defect in psychologism; it merely shows that a proper psychologistic

theory must be more complicated than it might at first have seemed.
The version of psychologism proposed by Alvin Goldman ~° provides a nice
illustration of how these three factors combine to determine the justificatory
status of a belief. Goldman claims that a belief is justified just in case its
presence is due to the workings of a reliable belief forming process. A process is
reliable to the extent that it tends to produce true beliefs in actual as well as
relevant counterfactual situations.
How does Goldman's account square with our intuitions about examples of
type (a) ? Consider the case of the person with perfect pitch who takes himself
to be tone deaf. Now if a person has perfect pitch, he will, on hearing a tone,
undergo a process of belief formation which is reliable; that is, the process will
tend to form true beliefs. If that person believes himself to be tone deaf,
however, a situation which was not otherwise relevant becomes relevant:
namely, the situation in which the person is tone deaf. Since in such a situation,
true beliefs would not be formed, the process in question is not reliable in the
circumstances defined. The resulting belief is thus unjustified.
Cases of type (b) are handled similarly. Insofar as we recognise that taking
no account of the opinions of others is epistemically irresponsible, we must
allow that when a community differs with an agent over the truth of a particular
proposition and that agent is not in a position to answer the community's
challenge, situations which would not otherwise be relevant alternatives for the
agent become so. Since the process by which the agent arrived at his belief
would not tend to produce true beliefs in these situations, the process which
produced the belief is not reliable, and so the belief is not justified.
Perhaps part of the reason some philosophers have been hesitant to take the
psychological turn is this; 11since we hold an agent responsible for the beliefs he
has, whether an agent is justified in believing a proposition cannot turn on
something which merely happens to an agent, it must turn on something which
the agent either does or fails to do; if questions of justification come down to
questions about the kinds of belief producing processes which go on in us, then
justification is something which merely happens to an agent.

9 Gilbert Harman, Thought, pp. 142-154.

~0 See the works cited in footnote 1.
 252 The Psychological Turn
I hope the discussion of these two types of example has done something to
dispel the view that a psychologistic theory of knowledge is committed to the
view that justification just happens to an agent. As these two examples
illustrate, what an agent does or fails to do bears on questions of justification. In
the first example, it was the agent's failure to take account of his belief that he
was tone deaf that blocked his justification; in the second example, it was the
agent's failure to consult with and take seriously the opinions of others which
prevented him from being justified. As we saw in the chicken sexer example of
section V, part of what makes a chicken sexer justified in his beliefs has to do
with the steps he takes to assure that misidentifications will come to his
attention. All of these cases illustrate the active role an agent plays in making
Downloaded By: [Canadian Research Knowledge Network] At: 20:41 29 March 2009

his beliefs justified.

We may thus better understand the motivation behind the Applicability
Constraint. An agent must be in a position to do something about the
justificatory status of his beliefs. It is, of course, a mistake to think that this
requires that we be able to monitor the processes by which our beliefs are
formed at every step of the way; nevertheless, the emphasis on the role the
agent plays in making beliefs justified is not misplaced. I will explore this point
further in the final section of the paper.

VII
I would like to turn to one final point, and this has to do with the motivation for
the project of discovering the proper set ofepistemic rules. My point here is
substantially similar to a point made by Alvin Goldman. n Suppose that, like
Descartes, one believes that human beings are in a position to choose their
beliefs; in short, that there is free choice of belief. If this were so, then a
responsible epistemic agent would have to develop a set of rules for deciding
what beliefs he ought to adopt. Thus, we see the need for epistemic rules.
The view that there is free choice of belief, however, is no longer widely held,
and I will take for granted here that it is false. If it is, then once again we see that
the motivation for the project of discovering the proper set of epistemic rules
disappears. Epistemic beliefs are no longer seen as a necessary prerequisite for
justified non-epistemic beliefs; they are more plausibly viewed as products of
reflection o n the non-epistemic beliefs one already has and on the manner in
which one has acquired them. 13
If we allow that beliefs are not subject to direct voluntary control, must we
also give up the Principle of Epistemic Responsibility? Isn't it contradictory, or
at best unfair, to insist that we are responsible for the beliefs we have and yet
deny that the processes by which we form our beliefs are subject to our direct
voluntary control? I believe that the answer to both of these questions is, 'no'.
The possibility of reconciling our lack of free choice of belief with the Principle
of Epistemic Responsibility rests on the fact that while our beliefs are not

1| This was suggested to me by Philip Kitcher.

12 ' T h e Internalist Conception of Justification', section IX.
13 Ibid.
 Hilary Kornblith 253
subject to direct voluntary control, they are not entirely out of our control
either. While it is not possible simply to will to believe, say, that snow is black
and so come to believe it, our epistemic habits are not uninfluencable. This
should not be terribly surprising, for in this respect our epistemic habits are very
much like our other habits. If I should develop some bad driving habits, for
example letting my mind wander while driving on the highway and
consequently weave slightly from lane to lane, it simply will not do for me to
will that I should give up this habit. This is not to say, however, that the habit is
not subject to my voluntary control and that once I develop it I can only hope
that I will somehow lose it. Rather, I might do any number of things which
might bring about a change in my habits. I might ask those with whom I drive to
Downloaded By: [Canadian Research Knowledge Network] At: 20:41 29 March 2009

point out to m e w h e n e v e r I start weaving in and out; I might make a point of

not thinking about philosophical problems when I drive; I might go for short
drives on the highway when I know I can keep my attention focussed, and
slowly increase the length of my driving time. Similarly, if I develop bad
epistemic habits, though I will not be able to simply will them away, they are not
completely out of my control. Those, for example, who find that they tend to
think of women they meet as less intelligent than men and yet realise that this is
grossly irrational cannot simply will this bad epistemic habit away; nevertheless,
something can be done about it. Bad epistemic habits can be given up by
contrivances similar to those used for giving up bad driving habits. The
Principle of Epistemic Responsibility is entirely compatible with the view that
beliefs are not subject to direct voluntary control.
I thus conclude that the psychologistic theory of knowledge does not allow
for minor modifications in traditional, apsychologistic epistemological theories.
With the naturalisation of epistemology, we must completely rethink the
motivation for our epistemological projects) 4
University of Vermont Received June 1981

~4I am verymuch indebted to the work of Alvin Goldman. I receivedhelpfulcommentson distant

ancestorsof this paper from SydneyShoemaker,Richard Boydand Carl Ginet. More recently, I
have received valuable suggestions from Patricia Kitcher and Arthur Kuflik. An anonymous
referee for the Australasian Journal of Philosophy pointed out a number of errors and made
usefulclarificatorysuggestions. I am especiallygrateful to Philip Kitcher. Research on this paper
was in part supported by a Summer Research Grant from the University of Vermont.