Вы находитесь на странице: 1из 16

A Priori Knowledge,

E xperience and D efeasibility1


H am id Vahid

A bstract
Some recent discussions of a priori knowle dge, taking their departure from
Kant’s characterization of such knowledge as being absolutely independent
of experience, have concluded that while one might delineate a concept of
a priori knowledge, it fails to have any application as any purported case
of such knowledge can be unde rmined by suitably recalcitrant experiences.
In response, certain defenders of apriority have claimed that a priori justi-
Ž cation only requires that a belief be positively dependent on no experience.
In this paper, I begin by showing how the exchange of arguments between
the disputants comes down, in the end, to no more than a display of
con icting intuitions. I shall then provide a diagnosis explaining how our
explication of a priori justiŽ cation depends on our standards for applying
the term ‘a priori’ with this, in turn, re ecting our prior intentions as to
whether we are willing to allow the existence of such warrants. I shall further
argue that the claim that such knowledge can be affected by subversive
experience is not entirely compatible with the spirit of apriority. Finally I
conclude by making some methodological remarks about the prospects of
a positive characterization of a priori knowledge by comparing it to the
concept of knowledge.
Keywords: a priori knowledge; certainty; experience; defeasibility

In recent philosophical writings the notion of a priori justiŽ cation (knowl-


edge) has often been viewed with suspicion. Issues stemming from the
works of Q uine and other proponents of naturalized epistemology have
generally been regarded as having undermined this species of justiŽ ca-
tion.2 U nder the in uence of Kripke and others, on the other hand, there
has been a resurgence in the attempts to revive this notion, aiming at a
minimal conception of a priori that would gu arantee its existence.3 These
attempts start by delineating the concept of a priori from the cluster of
such metaphysical and semantical concepts as necessity and analyticity,

International Journal of Philosophical Studies Vol. 7 (2), 173– 188


© R outledge 1999 0967–2559
I N T E R N A T I O N A L J O U R N A L O F P H I L O S O P H I C A L ST U D I E S

and then seek to lay down some (minimal) conditions whose satisfaction
would ensure the existence of a priori knowledge (justiŽ cation).
D espite the increasing sophistication of the arguments on both sides,
one is left with the impression that the real issue is being shroud ed in a
mist of confusion and misunderstanding, with philosophers often appearing
to be talking at cross purposes. While some categorically deny the exis-
tence of a priori warrants, others begin by proposing an analysis of the
concept that seems to guarantee that its extension will not be empty, thus
bringing about a deadlock. My aim in this paper is to uncover the grounds
for this stalemate and Ž nd out why the debate has tended to slip, in the
end, to no more than a display of con icting intuitions on both sides. I
shall end the discussion by attempting to evaluate the prospects for a char-
acterization of the a priori that is in harmony with our most basic intuitions
about the concept. Let us, however, begin by quickly going through the
(more recent) history of the controversy.

I D enying the A Priori


Contemporary discussions of a priori knowledge usually take their depar-
ture from Kant’s famous characterization of such knowledge. A ccording
to Kant a priori knowledge is that which is absolutely independent of
experience, that is to say S’s knowledge that p is a priori just in case S’s
justiŽ cation for his belief that p is independent of experience. This deŽ -
nition, however, leaves a lot to be desired, especially in view of the fact
that Kant ties the (epistemic) concept of a priori to such metaphysical
concepts as necessity. In the light of Kripke’s in uential distinction
between the three realms of epistemology, metaphysics and semantics,
recent theorists have tended to characterize a priori justiŽ cation with a
view to avoiding the Kantian confusion s. The most painstaking attempt
at a characterization of the notion of a priori is the one put forward by
Philip Kitcher within a reliabilist framework.4
Kitcher begins by introducing the term ‘warrant’ to refer to those
processes which produce belief ‘in the right way’, and then deŽ nes ‘knowl-
edge’ as belief that is produced by a process which is a warrant for it.
A lthough a priori knowledge is usually characterized as knowledge which
is independent of all experience, Kitcher nonetheless allows a minimal
role to experience where it is needed at the level of concept formation.
A nother notion that is introdu ced is S’s ‘life’ at t, and it is deŽ ned as the
total sequence of experiences S has had up to time t. If S could have had
that life and could have gained sufŽ cient understanding to believe that p,
then it is said to be sufŽ cient for S for p. Kitcher then construes Kant’s
account of a priori knowledge as requiring that in order to produ ce knowl-
edge which is independent of experience, a process must satisfy the
following conditions: (i) it must be available independently of experience;
174
A PR I O R I K N O W L E D G E , E X P E R I E N C E A N D D E F E A SI B I L I T Y

(ii) it must produce warranted belief independently of experience; and


(iii) it must prod uce true belief independently of experience. The intuition
behind the latter is that a priori warrants must be ultra-reliable. The idea
is that only when a person possesses a method of arriving at belief which
guarantees true belief is she entitled to ignore the experience she is
exposed to. Conditions (i), (ii) and (iii) are then incorporated into the
following account of a priori knowledge (as clauses (3a), (3b) and (3c)
respectively).

(2) S knows a priori that p if and only if S believes that p and S’s
belief that p was produced by a process which is an a priori
warrant for it.
(3) a is an a priori warrant for S’s belief that p if and only if a is
a process such that, given any life e, sufŽ cient for S for p,
(a) Some process of the same type could produce in S a belief
that p,
(b) If a process of the same type were to produce in S a belief
that p, then it would warrant S in believing that p,
(c) If a process of the same type were to produce in S a belief
that p, then p.

O n this account, a priori warrants are expected to prod uce true beliefs
in all counterfactual situations. They are never to lead us astray. The above
deŽ nition can be seen as a challenge to the a priorist. H e must show
that any item of knowledge he claims to be a priori is produced by an
ultra-reliable process which could guarantee the same (true) belief in all
counterfactual situations, and which could continue to warrant the belief
in the face of all conceivable experience which might be thought to under-
mine it. In other words, the apriorist must show that no alternative
experience could undermine our conŽ dence in the belief produced by
the process responsible. Conditions (3a), (3b) and (3c) (together with the
cognizer’s awareness of their satisfaction) in effect rule out the coexist-
ence of a priori knowledge with rational uncertainty and fallibility. They
are meant to capture the sense in which a priori knowledge is said to be
independent of exp erience. A ny warrant that deserves to be called ‘inde-
pendent of experience’ must, on this account, be indefeasible. A n a priori
warrant is a process whose ability to warrant the belief that p for S is
never undermined in any world in which S gains sufŽ cient understanding
to believe that p.
A s can be seen, central to Kitcher’s assumptions is the thesis – call it
the ‘unrevisability thesis’ (U T) – that rational revisability is incompatible
with a priori justiŽ cation. H e illustrates this point by arguing against a
remark made by Kripke to the effect that a priori knowledge is compat-
ible with uncertainty.
175
I N T E R N A T I O N A L J O U R N A L O F P H I L O S O P H I C A L ST U D I E S

Something can be known, or at least rationally believed, a priori,


without being quite certain. You read a proof in a math book; and,
though you think it is correct, maybe you have made a mistake.
. . . You have made a computation, perhaps with an error. 5

Kitcher’s point, pace Kripke, is that while the absence of certainty is


compatible with knowledge, it rules out a priori knowledge. In the above
case, for example, the process of following a proof fails to satisfy condi-
tion (3b) . E xperiences under which the process in question would fail to
warrant the belief may include such circumstances as receiving criticism
rather than applause from the learned world, Ž nding error in many other
theorems, eminent mathematicians denying the validity of the proof and
so on.

II Modest A Priori Knowledge


In response to Kitcher’s criticisms, the prop onents of a priori knowledge
have staged a comeback by trying to cast doubt on Kitcher’s unrevisability
thesis according to which a priori justiŽ cation entails rational unrevis-
ability. Before considering their reasons, it is necessary Ž rst to take note
of two variants of U T, strong and weak versions. 6

(SU T) If S is justiŽ ed in believing that p a priori, then p is ration-


ally unrevisable in the light of any future evidence.
(WU T) If S is justiŽ ed in believing that p a priori, then p is ration-
ally unrevisable in the light of any future em pirical
evidence.

It seems uncontroversial that WU T is more plausible than SU T. For all


that WU T requires is unrevisability in the light of experiential evidence,
and that Ž ts very well with our intuitive conception of a priori justiŽ ca-
tion as something which, on the face of it, is sensitive to the distinction
between experiential and non-experiential evidence. A belief is said to
be justiŽ ed a priori if experiential evidence plays no role either in the
original justiŽ cation or in its subsequent revision. What a priori justiŽ ca-
tion, thus, seems to rule out is revision on the basis of exp eriential
considerations rather than non-experiential ones. SU T is implausible
precisely because it is insensitive to this fact. What, then, of WU T?
WU T is not susceptible to this problem as it distinguishes between
revisions based on empirical evidence and revisions resulting from non-
experiential considerations. Kitcher’s characterization of a priori knowl-
edge seems to involve only a commitment to WU T. WU T is, neverthe-
less, rejected by the proponents of a priori knowledge, who call their
position ‘modest apriorism’.
176
A PR I O R I K N O W L E D G E , E X P E R I E N C E A N D D E F E A SI B I L I T Y

A ccording to modest apriorists, a priori knowledge (justiŽ cation) is


compatible with fallibility and revisions based on empirical evidence.
Crucial to their position is a distinction between positive and negative
dependence.7 To maintain that the cognizer’s belief is justiŽ ed a priori is,
according to the proponents of modest apriorism, only to make a claim
about the type of evidence that justiŽ es the belief, whereas to say of the
cognizer’s belief that it is rationally revisable in the light of experiential
evidence is to make a claim about the type of evidence that can defeat
the justiŽ cation. O ne is, thus, naturally led to distinguish between posi-
tive and negative dependence on experience. S’s belief that p is said to
be positively dependent for its justiŽ cation on the type of process which
produces and sustains it, but it is only negatively dependent on processes
that would undermine its epistemic status under counterfactual circum-
stances.
A ccordingly, modest apriorists would maintain that a priori justiŽ cation
only requires that a belief be positively dependent on no exp erience and
empirical beliefs. It may, nonetheless, be negatively dependent on expe-
riences and empirical beliefs in the sense that their occurrence could
undermine its justiŽ cation in counterfactual circumstances. G iven the fact
that neither kind of dependence entails the other, modest apriorists are
able to claim that the vulnerability of a priori beliefs to empirical defeaters
(i.e., their negative dependence on experience) need not upset their a
priori status. This is how modest apriorists put this point:

O rdinarily, we could say that the striking of the match causes it to


light, even though (a) there are certain background conditions that
have to be met (e.g., oxyge n in the room) and (b) there is a ceteris
paribus clause (e.g., not-x, y, etc.). Further we may say that the
match’s lighting causally depends both on the striking of the match
and on the presence of oxyge n in the room in a way that it does
not depend on the absence of conditions that would, if they obtained,
prevent the match from lighting. 8

In the same fashion, modest apriorists claim that a belief is independent


of empirical evidence if the evidence plays no part similar to the condi-
tions of kind (a), even though it may play an analogous role to that of
the conditions excluded by the ceteris paribus clause.
A t this point the following question arises for the modest apriorist. If
the cognizer can now (at t) lose her justiŽ cation for p because of some
subversive empirical beliefs and experiences that tend to undermine it,
then we cannot say that her justiŽ cation at any time (before t) is posi-
tively independent of those experiences since it is only a contingent fact
that those experiences fail to occur before t and, thus, a contingent
fact that the justiŽ cation in question is not affected by them. To put it
177
I N T E R N A T I O N A L J O U R N A L O F P H I L O S O P H I C A L ST U D I E S

more bluntly, if the cognizer’s a priori justiŽ cation is eliminated under


those conditions, how could it have really been independent of experience
after all? H ow could the epistemic status of the belief undergo such a
transformation at time t? Let us call this ‘the transform ation problem’. In
response, some modest apriorists distinguish between the existence of a
priori warrants and their strength.9 While the strength of a process may
be dependent on experience, this is not so with respect to its existence.
So the fact that the cognizer’s justiŽ cation may undergo an epistemic trans-
formation at time t does nothing to undermine the existence of the process
in question as an a priori warrant.
H aving made these distinctions the modest apriorist would, then, be in
a position to go on to show that some processes satisfy his strictures on
a priori warrants and that there is a priori justiŽ cation after all. This is
not, however, a satisfactory situation in regard to the question of the
viability of the concept of a priori as each side seems to be in a position
to make a legitimate claim about whether the existence of a priori justi-
Ž cation does or does not follow from his or her proposed analysis. H ow
are we to break out of this deadlock? A re we supposed simply to stipu-
late what constitutes apriority, and then draw the corresponding existential
claims, or are there rational arguments to which one can appeal to settle
the issue? In what follows I shall try to show that the controversy has
actually lost its focus, resulting in a situation where the theorists tend to
presuppose what they intend to prove. Before I present my diagnosis,
however, let us Ž rst see if the exchange of (seemingly) rational arguments
by the parties involved amounts to anything more than a display of
con icting intuitions.

III Modest A priorism Criticized: A D iagnosis


A s Kitcher’s strictures on a priori warrants Ž nd one of their most exten-
sive critical evaluations in Casullo’s writings, let us proceed by considering
how he judges their plausibility.10 The idea of having condition (3a) is, as
we saw, to ensure that the a priori warrants are available independently
of experience. Casullo modiŽ es (3a) in order to allow for the possibility
that some non-experiential conditions may be necessary for the produc-
tion of the belief in question. Let us go along with this modiŽ cation and
concentrate on those conditions ((3b) and (3c)) that feature prominently
in Kitcher’s analysis of a priori justiŽ cation. By imposing a very strong
defeasibility condition on a priori warrants, (3b) claims that such processes
must produce warranted beliefs independently of experience. A ccording
to Casullo (3b) entails a defeasibility condition (D C) which is actually a
close relative of WU T.

178
A PR I O R I K N O W L E D G E , E X P E R I E N C E A N D D E F E A SI B I L I T Y

(D C) If a is an a priori warrant for S’s belief that p, then the


warrant that a confers on p cannot be defeated by any experiences
compatible with S’s acquiring the concepts involved in p.

D C, Casullo claims, should have little appeal for anyone who subscribes
to a reliabilist account of warrant, for it ignores the possibility that our
experiences, which may include m isleading evidence, might defeat the
warrant conferred on a belief by a reliable process. D C, in effect, rules out
the existence of possible worlds in which S acquires the concepts involved
in his (a priori) belief p but also has evidence, perhaps misleading, that the
process producing the belief is not reliable. But Casullo goes on to claim
that ‘there appears to be no inconsistency in the supposition that such worlds
exist’.11 Now since (3b) entails D C and D C is false, (3b) should be rejected.
Casullo’s rejection of (3b) sounds, however, more like the reiteration
of his intuitions than a real argument. H e sufŽ ces to say that there is no
inconsistency in supposing that there are worlds in which the cognizer has
evidence, perhaps misleading, that the processes prod ucing a priori beliefs
are not reliable. But that is exactly what Kitcher is denying through his
condition (3b) . H e wants to say that for any warrant to be a priori, it
should be strong enough to discharge its warranting function in the face
of recalcitrant (‘unkind’) experiences, thus denying that there may be a
world in which a priori warrants coexist with unkind experiences. Casullo’s
‘argument’ against condition (3c) fares no better. (3c), we may recall,
requires a priori warrants to be ultra-reliable, meaning that they must not
produce any false beliefs. Why does Kitcher require this of a priori
warrants? The most likely possibility, according to Casullo, is that it is the
very notion of an a priori warrant which necessitates such a strong connec-
tion with truth. The reason why a person can ignore empirical information
about the world is that she possesses a method of arriving at belief which
guarantees true belief. This intuition, claims Casullo, ‘rests on very shaky
grounds’. For in order to be able to ignore empirical information about
the world, one would have to commit oneself not only to the existence
of a priori warrants but also to the following thesis (S) concerning the
strength of such warrants:

(S) The warrant conferred on a belief by an a priori process cannot


be affected by experience.

But S, he says, is no more than a variant of D C,which, ‘as argued earlier


. . . should be rejected’. Yet, as we saw above, Casullo’s rejection of D C
did not seem to be any more than a restatement of his fallibilist intuitions
in regard to a priori knowledge. This means that in rejecting (3b) and (3c)
all that Casullo is doing is to contrast his own intuitions against those of
Kitcher’s. We thus seem to have reached an impasse with each side simply
179
I N T E R N A T I O N A L J O U R N A L O F P H I L O S O P H I C A L ST U D I E S

insisting on its own intuitions. H ow are we to adjudicate between the


opposing parties? To provide an answer to this question, I shall now
proceed to present my diagnosis of why this situation arises. To do this
we need to return to the modest apriorist’s distinction between positive
and negative dependence on exp erience.
Modest apriorists, we may recall, typically insist on making a distinc-
tion between positive and negative dependence, arguing that a priori
warrants need only be positively independent of experience. The situa-
tion, they argue, is analogous to the structure of our ordinary causal talk.
Just as, ordinarily, we say that a match’s lighting causally depends on the
striking of the match or the presence of oxyge n in a way that it does not
depend on the conditions excluded by the ceteris paribus clause, we can
equally say of an a priori justiŽ cation that it depends only negatively on
the empirical conditions that would, if they obtained, upset the justiŽ ca-
tion in question. To motivate the distinction SummerŽ eld quotes R obert
Audi, who gives the following illustration:

when one takes a walk in Washington Square, one’s safety depends


on one’s not being harmed by certain rufŽ ans who are several
miles away stalking Central Park but could have been stalking
Washington Square. This is not obviously false and if one has very
high standards of ‘safety’, one may accept it. But it is perfectly
reasonable to say that one’s safety on a walk in Washington Square
depends simply on what is happening there – or relevantly near there
– at the time.12

A nalogously, according to the modest apriorist, one can say that one’s
justiŽ cation for believing p depends on the processes that produce and
sustain the belief in such a way that it does not depend on the processes
that may undermine its justiŽ cation in counterfactual circumstances. A
priori warrants, thus, need only be positively independent of experience.
They may, however, be negatively dependent on experience and empir-
ical beliefs.
D espite the initial plausibility of the above example, Audi’s remarks, I
believe, highlight a point that undermines the fallibilist’s progra mme. The
point appears where Audi talks about the standards for applying the term
‘safety’. If we take very high standards about safety, e.g., being a sceptic
about safety, we are, says Audi, ‘likely to take the stronger view’, asserting
that when we take a walk in Washington Square our safety depends on
our not being harmed by rufŽ ans who are stalking Central Park. H e illus-
trates his point in the context of the fou ndationalism/coherentism
controversy. Foundationalism has often been charged with the claim that
no belief is really independent of other beliefs, and thus properly basic,
because the justiŽ catory status of any belief may vary with changes in
180
A PR I O R I K N O W L E D G E , E X P E R I E N C E A N D D E F E A SI B I L I T Y

background beliefs (e.g., it may be undermined by beliefs involving scep-


tical possibilities like being deceived by a demon, etc.). U nless we are
able to rule out these sceptical alternatives, continues the argument, we
cannot lay any claim to having beliefs independent of other beliefs. But
this is the case only if we have, with the sceptic, a very high minimal stan-
dard of justiŽ cation. If, on the other hand, we lower our standards, e.g.,
by being a modest founda tionalist, we would not need to identify inde-
pendence from other beliefs with indefeasibility. We can agree with the
sceptic that we cannot determine the falsity of certain sceptical hypotheses,
but deny that our justiŽ cation for believing propositions whose falsity is
entailed by these hypotheses requires knowing them to be false. Audi goes
on to remark that neither founda tionalism nor coherentism need be taken
to entail that there is any knowledge or justiŽ ed belief, for they are only
meant to reveal the structure of our justiŽ ed beliefs (if any). We can
choose our standard of justiŽ cation in such a way that it makes it plau-
sible to say that there are justiŽ ed beliefs.
Putting Audi’s remarks in the context of our current discussion, we can
conclude that the view we adopt with regard to the structure of our beliefs
will be in uenced by, among other things, our prior intentions as to
whether we are willing to admit or deny the existence of justiŽ ed beliefs.
If we are exclusively concerned with sceptical possibilities, then no belief
will be independent of other beliefs, just as by being a sceptic about safety,
we will say that one’s safety, when walking in Washington Square, depends
on one’s not being harmed by certain rufŽ ans who are stalking Central
Park.13 What goes for knowledge and safety, goes, I believe, for a priori
knowledge and justiŽ cation. Just as the description of our walk in
Washington Square was very much dependent on our standards in applying
the term ‘safety’, our explication of a priori justiŽ cation will very much
depend on our standards for applying the term ‘a priori’. If we are being
sceptical about a priori knowledge (justiŽ cation), we will (with Kitcher)
adopt sufŽ ciently high standards of apriority to ensure that no process
will satisfy them. If we are, on the contrary, willing to allow the existence
of a priori knowledge (justiŽ cation), we want to choose our standards
(with the modest apriorist) in such a way that this species of knowledge
can be produced by our current cognitive processes much in the same
way that ‘founda tionalists might, when they turn to the question whether
there are justiŽ ed beliefs, want to use a standard of justiŽ cation which
makes it plausible to say there are’.14 O ur analysis of apriority thus re ects
our priori intentions as to whether we wish to allow the existence of
a priori justiŽ cation. This explains why the (purported) exchange of argu -
ments between the disputants comes down, in the end, to a mere display
of con icting intuitions.
The above diagnosis also highlights an asymmetry in the analogy of
modest apriorism with the structure of our ordinary causal talk. It seemed
181
I N T E R N A T I O N A L J O U R N A L O F P H I L O S O P H I C A L ST U D I E S

all too easy for the modest apriorist to argue for the negative dependence
of a priori warrants on experience from the analogy with, say, the lighting
of matches, which, in a perfectly coherent sense, can be said to depend
only negatively on the conditions excluded by the ceteris paribus clause.
The asymmetry consists in the fact that in the case of the lighting of
matches, our observing it in daily life and noting the conditions in which
it occurs (or fails to occur) provide us with enough grou nds to embark
on an initial but non-arbitrary analysis of its causal structure. But we lack
such a grip on the issue of the structure of our (purported) a priori beliefs
because not only do we fail to have any (settled) intuitions about the
intension of the concept of a priori, but we are also uncertain whether
there are any processes that confer a priori justiŽ cation on the beliefs they
produce. We can plausibly claim, in the case of the matches, that their
lighting depends only negatively on the conditions being excluded by
the ceteris paribus clause because our description of the factors involved
corresponds to an already existing pattern. Not so in the case of our
(purported) a priori beliefs, whose very existence is in dispute here.15 It
is, therefore, not open to the modest apriorist to appeal Ž rst to such analo-
gies to argue for a similar treatment of a priori beliefs, and then to proceed
to show that there are processes that satisfy his strictures. It is only when
we are equally certain of the existence of a priori warrants that we can
appeal to such analogies to delineate their structure. To do otherwise is
to put the cart before the horse. A part from highlighting the fact that
one’s views on apriority are more or less dictated by one’s intuitions
about whether there is any a priori knowledge, the points made in this
section also seem to suggest that modest apriorism is not an entirely stable
position. This conclusion can be further supported by a re-examination of
the transform ation prob lem, to which I shall now turn. Finally, I conclude
by making some methodo logical remarks about the prospects for settling
the debate.

IV The Transformation Problem


The transform ation problem, we may recall, resulted from the fallibilist’s
distinguishing between positive and negative dependence on exp erience:
if the cognizer’s a priori justiŽ cation of his beliefs is to be affected by
subversive experience (after time t), then it cannot really be claimed that
those beliefs were ever independent of experience at any time before t.
A s we saw, the modest apriorist (Casullo) then resorted to a distinction
between the existence of a process and its strength in order to solve the
problem.16 But, like similar attempts on his part, this one also begs the
question against the sceptic (Kitcher) who would deny that any distinc-
tion can be made between the existence and strength of a priori
belief-forming processes. In view of the preceding remarks, however, it
182
A PR I O R I K N O W L E D G E , E X P E R I E N C E A N D D E F E A SI B I L I T Y

seems to me that the signiŽ cance of the transformation problem has been
played down by modest apriorists. To explain this, let us Ž rst remind
ourselves of the (rather trivial) point that what is problematic about the
transformation question is not how the change in the epistemic status of
a priori beliefs is brought about, but whether such a change is compatible
with the spirit of apriority. For we know, from the work on the problem
of defeasibility and the nature of defeaters, that one’s epistemic status is
defeasible by all kinds of evidence (including misleading evidence that
one does not possess but is at the disposal of one’s community17). O nce
the epistemic situation of the cognizer becomes the focus of our atten-
tion, the question of how one’s epistemic status can be transform ed reduces
to mere triviality. A lmost all theories of epistemic justiŽ cation include
defeasibility clauses referring to all sorts of defeaters, and it is not clear
why a priori justiŽ cation, being a species of justiŽ cation, shou ld constitute
an exception.18
What is problematic for modest apriorists is to exp lain whether this
phenomenon is compatible with the spirit of apriority as a species of justi-
Ž cation that is supposed to be independent of experience. The prob lem
that still lingers is to explain whether it is possible for a justiŽ ed belief
that is initially independent of experience to be affected by it afterwards
in a way compatible with its apriority. For it is very tempting to conclude
that the transform ation phenomenon is tantamount to admitting that the
state in question was never really (positively) independent of experience
after all. Being eager to combine fallibility with apriority, the modest apri-
orist, of course, denies this. It seems to me, however, that the
transformation phenomenon does not really sit well with apriorism. I think
that the question raised by the phenomenon is a genuine one in that it
alludes to a fundamental weakness in any theory of apriority that wants
to leave some room for revisability (because of empirical evidence).
To see this, consider the question not at the level of belief states but
at the level of their contents (propositions and statements). Posing the
question at the level of statements takes us back to the recent contro-
versy between the so-called ‘veriŽ cationists’ and ‘falsiŽ cationists’, namely,
the problem of demarcation. The issue was specially concerned with the
question whether universal statements should be regarded as having
empirical content. O n the veriŽ ability criterion of demarcation such state-
ments were deemed to lack empirical content because they were not
veriŽ able by any Ž nite number of observation statements. This was,
however, highly unintuitive because some of these statements (such as the
laws of nature) were actually paradigm cases of empirical statements.
FalsiŽ cationists, on the other hand, pointed out that it is not veriŽ ability
but falsiŽ ability by experience that is the hallmark of a scientiŽ c and
empirical statement. To put it differently, it is not dependence on experi-
ence, in a positive sense, that confers empirical status on a statement, but
183
I N T E R N A T I O N A L J O U R N A L O F P H I L O S O P H I C A L ST U D I E S

its negative dependence on experience that makes it scientiŽ c and empiri-


cal. What determines the empirical content of a statement is the class of
its potential falsiŽ ers. This is how Popper makes the point:

But I shall certainly admit a system as empirical or scientiŽ c only if


it is capable of being tested by experience. . . . In other words: I shall
not require of a scientiŽ c system that it shall be capable of being
singled out, once and for all, in a positive sense, but I shall require
that its logical form should be such that it can be singled out, by
means of empirical tests, in a negative sense: it must be possible for
an empirical scientiŽ c system to be refuted by experience [my italic].19

Now what goes for the content of a belief (prop osition) goes for its state.
Just as the falsiŽ ability of statements or propositions (i.e., their negative
dependence on experience) confers an empirical status on them, the
fact that the epistemic status of certain cognitive states is defeasible by
(i.e., negatively dependent on) experience tends to confer an empirical
character on them. It thus seems to me that the claim that a priori justi-
Ž cation is defeasible by experience does, to some extent, conŽ rm the
suspicion of those who do not Ž nd the transformation phenomenon
entirely compatible with the spirit of apriority.

V Conclusion
E arlier I pointed out that the dispute has lost its focus, resulting in a situ-
ation in which the disputants seem to be talking at cross purpo ses. O ur
survey of the Ž eld seems to have provided us with ample support for
this conclusion. A s we have seen, one camp uses the expr ession ‘a priori’
one way – as that which is positively and negatively independent of
experience – while the other camp uses the expression ‘a priori’ in a
different way – as that which is only positively independent. This suggests
that there is no real disagreement at all, and that the dispute is actually
about how one should use the expression ‘a priori’. I shall now end
the discussion by providing some methodological re ections about the
prospects for a positive characterization of a priori knowledge by
comparing it to a parallel attempt in the case of (empirical) knowledge.
In the case of knowledge, attempts at an analysis usually take the form
of a search for necessary and sufŽ cient conditions by providing an explicit
intension to Ž t its intuitive extension.20 We seem to have intuitions about
the intension of the concept of knowledge as well as intuitions about its
extension, that is to say about its ascription to certain cases as opposed
to others. If, after we have propose d an analysis, it turns out that it does
not Ž t our doxastic practices, that will constitute a good reason for thinking
the analysis to be mistaken. The G ettier literature is replete with propo sed
184
A PR I O R I K N O W L E D G E , E X P E R I E N C E A N D D E F E A SI B I L I T Y

analyses of knowledge (e.g., the indefeasibility and causal accounts) which


are then measured against our reactions to both real and imaginary exam-
ples. A lthough the reaching of a perfect re ective equilibrium in the course
of the process (i.e., identifying an explicit intension that Ž ts the intuitive
extension of the concept) is, as the literature shows, a distant dream, we
have, nevertheless, sufŽ cient grasp of the concept of knowledge to get the
process of evaluation going.
But we do not have a similar grip on the concept of a priori. A pproaches
to the question of a priori knowledge can be divided into two general
groups. The Ž rst group begins by providing a general characterization of
such knowledge as a species that is independent of experience, and then
proceeds to show that there are instances of such knowledge by looking
at the processes that are supposed to satisfy the strictures included in the
characterization. The second group starts by picking up certain (purpo rted)
instances of such knowledge (e.g., mathematical knowledge, modal knowl-
edge, etc.) and then tries to extract their common features as what
constitute apriority. Starting with the intension of the concept of a priori,
our characterization of the concept depends, as we have seen, on our prior
intentions as to whether we wish to allow the existence of such knowl-
edge with this, in turn, hanging on how the notion of ‘independence’ is
to be treated. A strong interpretation (rational unrevisability) would make
it difŽ cult to Ž nd processes which can satisfy it. This leads the propo nents
of a priori knowledge to opt for a weaker interpretation of ‘independence’,
by distinguishing between positive and negative dependence, so as to
ensure its satisfaction by existing cognitive processes. The obvious way
out of this deadlock would be to follow the procedure adopted in the
G ettier literature, and evaluate the prop osals against the intuitive exten-
sion of the concept of a priori. But what sort of beliefs are to be taken
as constituting that extension?
Traditionally intuition has been regarded as the faculty responsible for
producing a priori knowledge, which is usually thought to include math-
ematical and modal beliefs. Since mathematical objects are generally taken
to be abstract, mathematical knowledge, it is argued, cannot be acquired
through sense perception. It is, rather, obtained through the exercise of
intuition. H aving (partly) identiŽ ed the intuitive extension of the concept
of a priori knowledge, we would then be in a position, one might think,
to evaluate the competing proposals for the analysis of that concept by
seeing whether they Ž t that extension. The problem, however, is that we
are provided with very little information, often involving fragmentary
metaphors, about the mechanism through which the (supposedly) cogni-
tive process of intuition produces beliefs.21 There are, on the one hand,
constructivists, who typically construe the mechanism as involving the
apprehension of mental constructions (representations) of mathematical
structures. We are, however, left in the dark about the ways in which we
185
I N T E R N A T I O N A L J O U R N A L O F P H I L O S O P H I C A L ST U D I E S

are suppo sed to apprehend these representations. There are also Platonists,
who, by contrast, tend to liken the process of intuition to sense percep-
tion. It is claimed that, when discovering mathematical truths, we come
into contact with the objects of mathematics, and the axioms (theorems)
‘force themselves upon us as being true’. Nothing, however, is said about
how we ‘perceive’ these abstract objects.
U nless these issues are resolved, we cannot suppose that there is a
faculty of intuition through which we form our mathematical or modal
beliefs. In the absence of some reliable information about the way this
faculty performs its tasks, we shall not be able to know whether we, or
others, have performed intuitions. It cannot, therefore, be claimed that
there is a faculty of intuition which confers a priori justiŽ cation on the
beliefs it produces, and, consequently, its alleged produ cts cannot be taken
to constitute the intuitive extension of the concept of a priori against
which one could evaluate the proposed analyses of that concept.22 It is
because of these complications that the concept of a priori knowledge
seems to resist the kind of treatment that the concept of knowledge has
enjoyed in the G ettier literature.

Institute for Studies in T heoretical Physics


and M athem atics, Tehran

Notes
1 I am indebted to Trenton Merricks, Tim Williamson and the anonymou s IJPS
referees for helpful comments on earlier drafts of this paper.
2 See Q uine, ‘Two D ogmas’ (1980), Kitcher (1983) and G oldman (1986) ,
pp. 299–302.
3 See Kripke (1980), E didin (1984, 1986), H ale (1987, ch. 6), Casullo (1988) ,
SummerŽ eld (1991) and Plantinga (1993, ch. 6).
4 Kitcher (1983) .
5 Kripke (1980), p. 39.
6 Casullo (1988) , p. 190.
7 The terminology is due to SummerŽ eld (1991) ; the distinction is essentially
Edidin’s distinction between forward-looking and backward-looking depen-
dence (see E didin (1984) ). Casullo’s arguments against WU T are discussed
below.
8 SummerŽ eld (1991) , p. 43.
9 This is obviously related to the distinction between positive and negative
dependence. See Casullo (1988) .
10 Ibid.
11 Ibid., p. 205.
12 Audi (1993) , p. 108.
13 The modest apriorist might further add that the factors on which a belief
positively depends, unlike those on which it negatively depends, explain
why the cognizer holds that belief. But this does not help the apriorist for at
least two reasons. First, to characterize positive dependence in terms of what
the belief is based on is to commit oneself to the (controversial) causal/
186
A PR I O R I K N O W L E D G E , E X P E R I E N C E A N D D E F E A SI B I L I T Y

explanatory analysis of the notorious problem of the basing relation according


to which the ground of a belief is that which explains why it is held. It is well
known that this is an inadequate analysis, for various neurophysiological
events and all sorts of deviant causal chains might enter into the formation
of a belief without its being the case that the latter is based on the former.
Second, as we just noted, any partitioning among the factors that are thought
to constitute the justiŽ cation of a belief crucially depends on the sort of stan-
dards of justiŽ cation we initially wish to adopt, just as, to use the analogy
again, what happens in Central Park would enter (positively) into the deter-
mination of our safety in Washington Squa re, if we have very high standards
of safety.
14 Audi (1993) , p. 109.
15 If it can be shown that, for example, intuition is a reliable belief-forming
process conferring a priori warrant on the beliefs it produces, then that will
allow us to address the questions raised here in an, at least, non-arbitrary
manner. But, unfortunately, we do not have enough information about these
faculties and how they (supposedly) produce the relevant beliefs. See section
V of this paper for a discussion of the complications involved in regarding
intuition as a source of a priori beliefs.
16 A somewhat stronger version of the transformation problem (ST*) is discussed
in Casullo (1988) . A ccording to (ST*) , if a defeater d can defeat or override
the warrant conferred on S’s belief that p by e, then e does not justify S’s
belief that p unless S is justiŽ ed in believing –d. To reject (ST*) Casullo
presents a complicated counterexample. But (ST*) can be rejected in a more
principled way on the ground that it imposes unne cessarily strong internalist
constraints on justiŽ cation. For if it were true, we would not even be justi-
Ž ed in believing such ordinary perceptual statements as ‘There is a table in
front of me’ on the ground that we are unable to rule out sceptical possi-
bilities like ‘A n evil demon is now deceiving me’ that would defeat our
justiŽ cation for that belief.
17 See H arman (1980).
18 Towa rds the end of his (1992) paper, Casullo introduces certain constraints
on a priori warrants which he claims the process of intuition fails to satisfy.
In particular he considers the bearing of those types of evidence that the
subject does not possess but are only at the disposal of his community. But
no particular case needs to be made for a priori justiŽ cation once one admits,
as does Casullo, that it is defeasible. For, being a species of justiŽ cation, it
automatically inherits all the features of that concept including its defeasi-
bility by all sorts of evidence (such as misleading evidence, evidence one does
not possess, etc.).
19 Poppe r (1977) , pp. 40–1.
20 For this way of describing the post-G ettier literature on the analysis of knowl-
edge see Craig (1990), p. 1.
21 See Kitcher 1983, ch. 3 for a detailed discussion of intuition as a source of a
priori beliefs.
22 A s I have emphasized before, my main purpose in this paper is to explain
why the exchange of argum ents between the opponents and defenders of a
priori knowle dge has resulted in a stalemate by showing how their accounts
of a priori justiŽ cation depend on their prior intentions as to whether they
wish to allow the existence of a priori warrants. But, as was pointed out in
section IV, the modest apriorist’s claim that a priori justiŽ cation can be
affected by subversive experience is not entirely compatible with the spirit of
apriority. The points raised in this section further support this contention.
187
I N T E R N A T I O N A L J O U R N A L O F P H I L O S O P H I C A L ST U D I E S

The fact that taking the alleged products of intuition as constituting the exten-
sion of the concept of a priori has proved to be quite problematic tips the
balance in favour of the sceptic who denies that there is a priori knowle dge.
It shows, contrary to the modest apriorist’s claim, that it is very difŽ cult to
establish that there are processes which satisfy his constraints on a priori
warrants.

References
Audi, R obert (1983) : ‘Foundat ionalism, E pistemic D ependence, and D efeasibility’,
reprinted in (1993): T he Structure of JustiŽ cation, Cambridge: Cambridge
U niversity Press.
Casullo, A lbert (1988) : ‘R evisability, R eliabilism, and Mathematical Knowledge’,
Philosophy and Phenom enological R esearch, 49: 187–213.
––––– (1992) ‘Causality, R eliabilism, and Mathematical Knowledge’, Philosophy
and Phenom enological Research 52(3): 557–84.
Craig, E dward (1990) K nowledge and the State of N ature, O xford: Clarendon Press.
E didin, A ron (1984) ‘A Priori Knowledge for Fallibilists’, Philosophical Studies
46: 189–97.
––––– (1986) : ‘Language-Learning and A Priori Knowledge’, A m erican Philo-
sophical Q uarterly 23: 383–91.
G oldman, A lvin (1986) : E pistem ology and Cognition, Cambridge, Mass.: H arvard
U niversity Press.
H ale, Bob (1987) A bstract O bjects, O xford: Basil Blackwell.
H arman, G ilbert (1980) ‘R easoning and E vidence O ne D oes Not Possess’, in
P. French, et al. (eds) M idwest Studies in Philosophy, vol. 8, Minneapolis:
U niversity of Minnesota Press.
Kitcher, Philip (1983) : T he N ature of M athem atical K nowledge, O xford: O xford
U niversity Press.
Kripke, Saul (1980) N am ing and N ecessity, O xford: Basil Blackwell.
Plantinga, A lvin (1993) Warrant and Proper Function, O xford: O xford U niversity
Press.
Poppe r, Karl (1977) T he L ogic of ScientiŽ c D iscovery, London : H utchinson.
Q uine, W. V. O. (1980) From a L ogical Point of View, Cambridge, Mass.: H arvard
U niversity Press.
Summ erŽ eld, D onna (1991) ‘Modest A Priori Knowledge’, Philosophy and
Phenom enological R esearch, 51(1) : 39–66.

188

Вам также может понравиться