Bonjour The Myth of Knowledge

PHILOSOPHICAL PERSPECTIVES
Philosophical Perspectives, 24, Epistemology, 2010
THE MYTH OF KNOWLEDGE
Laurence BonJour
University of Washington
It is my conviction that epistemology has, in recent times, gone seriously

astray. The main symptom of this has been ever more complicated and convoluted
accounts of the supposedly central concept of knowledge, combined with less and
less attention to the question of why knowledge is supposed to be intellectually
important and valuable. In effect, epistemologists have been more concerned with
preserving the supposed common-sense applications of the term knowledge
than with making sense of why there should be such a concept and of why it or
its application should be topics of central concern to philosophy.
The starting point for these developments was the widespread abandonment
of the strong Cartesian conception of knowledge in favor of what is sometimes
referred to as the weak conception1 : the fallibilist conception, according to
which one can have knowledge that a particular claim is true even though
ones justification (evidence, warrant, or supporting grounds) for that claim is
less than conclusive even though the having of that level of justification is
compatible with the claim in question actually being false. It is widely believed
that the correct specific concept of knowledge, the concept that is assumed to be
embodied in common sense and ordinary thought and to be of central concern
to epistemology, is one that satisfies this latter conception. My main thesis in this
paper is that this whole fallibilist view of knowledge is mistaken, that there
simply is no well-defined, intellectually significant concept of knowledge fitting
this general conception: none that can be genuinely found in common sense or
indeed can be constructed or stipulated in a satisfactory way. The supposed weak
concept of knowledge in question is, I am suggesting, a philosophical myth.
The great majority of philosophical discussions of knowledge in recent times
have assumed such a concept, and in this way are predicated upon this myth.
When the myth is abandoned, some of the most widely discussed problems that
focus explicitly on the concept of knowledge, I will suggest, simply disappear,
with the Gettier problem and the lottery paradox being the most obvious
examples.
58 / Laurence BonJour
In Part I of this paper, I will attempt to develop and support these claims.
Then, in Part II, I will suggest, rather more tentatively, that the correct concept
of knowledge is in fact some version of the strong Cartesian conception, under
which knowledge requires a conclusive level of justification, one that guarantees
the truth of the relevant claim. This conception is usually rejected, often with
very little discussion, for reasons that seem to me to have in the end much less
substance than is usually thought, however initially persuasive they may be. These
reasons and the underlying issues they involve will be discussed below. But the
main concern of the paper is to criticize the view that advocates the concept of
knowledge that I regard as mythical rather than to defend this, to my mind, more
plausible alternative; and it still seems to me possible that in the end the right
thing to say is that there simply is no coherent concept of knowledge to be found
in common senseor at least none that is worthy of any serious philosophical
concern.
Part I: Against the Weak Conception

1. Preliminaries
I begin with an account of the conception of knowledge that I am opposing.
Here the obvious starting point is what is often referred to as the traditional
account or definition of knowledge as justified true belief , an account that
still seems to me broadly correct, albeit obviously in need of much additional
clarification and specification. There are many different issues that could be
raised in relation to this conception, including issues pertaining to the belief and
truth conditions, but for present purposes, four central points will suffice.
First, knowledge in the sense that is allegedly delineated by this conception
is supposed to be a supremely valuable and desirable cognitive state, one whose
possession marks the difference between full cognitive success and at least some
degree of cognitive failure: knowledge is the epistemic summum bonum. And
it is because knowledge has this status that skepticism about knowledge seems
intellectually threatening in a way that raises serious philosophical concerns.
Thus, I suggest, any satisfactory concept of knowledge must be such as to
make sense of this supposed supreme value; and any features or requirements
that pertain to it must be ones that contribute in some intelligible way to this
evaluative status. I will assume throughout that this is an absolutely essential
condition for any adequate account of knowledge.
Second, whether a person has knowledge of something in this sense is
supposed to be an all or nothing matter, not a matter of degree. While beliefs or
opinions can be justified or rational to various degrees, intuitively one either has
knowledge or one does not, with no room for gradations in between.2
The Myth of Knowledge / 59
Third, as just suggested, the relevant notion of justification, in contrast, is

emphatically a matter of degree, allowing for a wide range of variations from
very weak to very strong.
Fourth, the prevailing view, which I will adopt with no further discussion
here, is that justification in the relevant sense (epistemic justification) has to
do with likelihood or probability of truth, with stronger degrees of justification
making the truth of the relevant belief correspondingly more likely or probable.
Here I will think of justification in internalist terms: as a reason or reasons that
a person has for thinking that the belief in question is true.3
Given this initial picture, the obvious question that needs to be asked is:
what specific degree of justification is required to achieve this exalted cognitive
state of knowledge? And the most initially obvious answer to this question,
one taken for granted by Descartes (and also by Locke and a host of other
historical philosophers), is that knowledge requires the highest possible degree
of justification: justification that is conclusive, that guarantees the truth of the
claim that is believed, that makes it impossible in relation to that justification that
the belief is false. This is the strong conception of knowledge referred to above.4
In recent times, this strong conception of knowledge has, for reasons to be
considered momentarily, almost always been dismissed as untenable. But it is
worthwhile to pause for just a moment to reflect on its internal structure and,
as it might be termed, its conceptual cohesiveness on the way in which the
three ingredients of the conception relate to each other and fit together to make
an initially plausible, well-motivated combination, and one that is obviously of
great intellectual and philosophical significance. We have one ingredient, belief,
that brings a specific claim or proposition into the purview of the cognitive
agent in question. We have an appealing cognitive goal, namely bringing it about
that one believe such a claim only if it is true. And, since we are in general
unable to directly determine the truth of a belief, we have a less direct means
for achieving this goal, namely seeking justification of the epistemic sort for
the claim in question and relying on the degree of justification that results as
a guide for which beliefs to accept. Given such a picture, it clear that complete
cognitive success has been achieved only when the justification attained for the
claim is sufficient to establish completely that it is true. And it is thus natural
and eminently reasonable to identify the situation of such complete success as
the most valuable cognitive state, namely knowledge. No one could reasonably
deny the epistemic value of such a state, and both the intuitive rationale for its
constituents and the way that they fit together seems entirely clear.
But of course there is an obvious and familiar problem: familiar sorts of
considerations show that the strong conception is extremely difficult to satisfy,
indeed in many or even most cases apparently impossible to satisfy. Perhaps
one can have conclusive justification for simple a priori claims and for claims
about some aspects of immediate experience, but, most recent epistemologists
would agree, for little if anything else. And that means that under the strong
conception, there will be at best very little knowledge and that only in quite
restricted areas. How serious a problem this is and indeed even what exactly the
problem is are issues that seem to me to be much less straightforward than is
usually thought; I will return to them in Part II of this paper. But the way in
which the discussion typically goes is something like this: the standard definition
of knowledge is invoked, the strong conception is mentioned almost in passing,
and then is dismissed at once as leading to an unacceptable skepticism, as being
unreasonably stringent or demanding, as contrary to common sense, and the like.
Usually all this is done in a quick sentence or two, with the central idea being that
very many of the ascriptions of knowledge that ordinary people seem to make
would be false if the strong conception were correct, and that it is implausible
(in some way that is usually not very fully specified) that this is so.
The conclusion that is then most commonly drawn is that the correct concept
of knowledge is still a version of the traditional definition, but one according to
which the degree of justification required for knowledge is one that falls short
of conclusive justification but is still fairly strong, substantially stronger than
one that merely makes the claim barely more likely to be true than false. This
general sort of view is often referred to as the weak conception of knowledge,
and I will adopt this label for it as well. (In addition to the three conditions
specified in the traditional definition, virtually all recent versions of the weak
conception also incorporate a further condition aimed at warding off Gettierstyle counterexamples. The specification just given should be interpreted as
allowing for this sort of addition, though I will suggest below that the need
for such a condition is merely one of the unfortunate consequences of adopting
the weak conception.)
2. The central problem for the weak conception

Given the account so far of the weak conception of knowledge and how it
is dialectally arrived at, the immediate and urgent question should be: what then
is this specific level of justification, the attaining of which transforms what is up
till then at best merely increasingly probable or likely true belief into the exalted
state of knowledge?
Here it is perhaps unreasonable to demand a precise numerical specification
of a level of probability and perhaps not even entirely clear that numerical
probabilities are the right way to think of degrees of justification. But if we are
to suppose that there is a definite concept of knowledge which when satisfied
yields the exalted cognitive state in question, it is surely not good enough to say
merely, as is commonly said, that the level of justification in question is strong
or high or adequate or enough to make it highly likely that the belief
in question is true, for nothing this vague is enough to specify a definite level
of justification and a corresponding definite concept of knowledge.5 And yet
the striking fact is that philosophical discussions that either explicitly invoke or
tacitly presuppose the weak conception of knowledge almost never have anything
much more helpful than this to say about what this magic level of justification,
as I will somewhat tendentiously refer to it, might be or, even more important,
about why it has this very special status. Indeed, it is fair to say that nothing like
a precise specification of the magic level has ever been seriously suggested, let
alone more widely accepted.6
This failure on the part of those who either espouse or presuppose the
weak conception of knowledge to offer anything like a clear specification of the
magic level of justification or even to give any reasons for thinking that such
a specification might in some way be possible seems to me to constitute in itself
a very serious objection to that conception.7 But an even more serious objection
is that it is very difficult or, I believe, impossible to see what could give any level
of justification that is short of being conclusive the kind of special significance
that the weak conception requires it to have.
Perhaps the clearest way to think about this problem is to suppose that
there is some issue about which it is important to you to find the truth, and that
there is some specific proposition that constitutes the best candidate you can find
for the truth about this issue. It is easy to understand how finding higher and
higher levels of justification for the claim in question improves your cognitive
situation, making it more likely that the corresponding belief is true, and also
how finding genuinely conclusive justification, if that were possible, would be
the best situation of all. But the claim of the weak conception is that there is
some specific level of justification that is less than conclusive but that nonetheless
transforms your cognitive situation in a much more radical way than did increases
in justification up to that point (or further increases above it). Before this level is
attained, you merely have a belief that is more and more likely or probable, but
at that point you suddenly have knowledge. But why does achieving this specific
level of justification make such a difference and what exactly is this difference
supposed to amount to?
One thing that it clearly does not amount to is that further increases in
justification cease to matter. If the issue in question is indeed an important one,
then as long as the justification you have achieved remains less than conclusive,
further increases in justification will still be cognitively valuable, will still further
improve your cognitive situation by making the truth of your belief even more
likely. Indeed, it is hard to see why such further increases are not valuable in
exactly the same way, to precisely the same extent, as those that came earlier,
before the supposed magic level was reached. And this means that the sorts
of inquiry that lead to such increases in justification are also no less valuable,
no less urgent than they were before. But why then is it supposed to be so
supremely important whether or not the magic level has been reached? Why
are we supposed to care so much about this or indeed care about it at all?
(And of course, if we really do care so much, why is so little attention devoted
to determining what that level actually is?)
It is entirely possible, of course, that in relation to a particular issue, you
might decide that the level of justification and corresponding likelihood of truth
attained so far is good enough and might be willing to rest content with that
result. In some cases, when such a satisfactory level has been reached, you might
even be moved to say that you know the claim in question where this means
roughly, I would suggest, that you are sure enough of it, that your justification
approximates closely enough to being conclusive, for the issue at stake. (See the
further discussion below, in Part II.) But clearly what that good enough level
might be will vary widely from case to case, depending on the specific issue
and the surrounding context, and thus your willingness to say that you know in
such a case or for others to ascribe knowledge to you on the same sort of
basis provides no real reason for thinking that there is one magic level of
justification of the sort that the weak conception requires.8
But what then could give the supposed magic level of justification the
kind of special significance that it is supposed by the weak conception to have?
What could explain why a gradual increase of justification, rather than merely
yielding gradually increasing levels of probability or likelihood or rational belief,
at some point results in a qualitatively different state that is cognitively exalted
in the way that knowledge is supposed to be? I believe that there is simply no
answer available to this question, no account of what would give any level of
justification that is short of conclusive this very special status. And this is the
most fundamental reason for thinking that the weak conception of knowledge
is mistaken and that the supposed concept that would satisfy it is a myth.
Paradoxically enough, the weak conception, though it seeks to preserve the
truth of common-sense attributions of knowledge, makes it impossible to give
an account of why those attributions have any real epistemic significance or why
they should matter very much for epistemology.
Before turning to other issues, there is one more fairly obvious reason for
thinking that there is no non-conclusive level of justification that can play the role
that the weak conception of knowledge requires. Suppose that I have attained
just the supposed magic level of justification, but no more, for two independent
claims, P and Q, which I believe on that basis and which are in fact true. And
suppose that any needed further condition, such as an anti-Gettier condition
see the next section is also satisfied, so that according to the supposed weak
concept of knowledge, I know that P and know that Q. And suppose that I now
infer the conjunction of P and Q, and then validly infer some further claim R
from that conjunction (understanding why and how the inference is valid), that
I proceed to believe R on that basis, and that R is also true (and that any further
required condition is again satisfied). Do I thereby achieve the exalted state of
knowledge with respect to R?
The answer, of course, is that if the levels of justification in question can
be thought of as levels of probability or as behaving anything like levels of
probability, then the level of justification that I thereby acquire for R will in
general fall short of the magic level, because the product of two probabilities
less than 1 is always smaller than either of them. Thus under the weak conception,
there is no guarantee that when I know two things and infer even the simplest,
most obvious conclusion from them, I will then have knowledge of the result.
(Of course I may have such knowledge, if the levels of justification are, contrary
to what I have been assuming, sufficiently greater than the magic level.) And
this seems to be an intuitively unacceptable result: what is the supposed state
of knowledge really worth, if even the simplest inference from two pieces of
knowledge does not lead to further knowledge?
Here we are, of course, in the vicinity of the much-discussed issue of cognitive
or epistemic closure. Reasons have been given for denying that closure always
holds. But, though I have no space for a detailed discussion here, those reasons
usually pertain to relatively restricted cases: usually to the issue of whether antiskeptical claims that follow from more ordinary claims must be known if the
more ordinary claims are known. The denial of closure that results from the
weak conception is, however, far more radical more radical, I believe, than
any that has ever been seriously advocated. And here we have an additional
reason for thinking that there is no non-conclusive level of justification that can
have the kind of cognitive significance that the weak conception of knowledge
requires it to have, and that the supposed specific concept of knowledge that
would satisfy that conception is a myth.
In addition to the problem of specifying the magic level of justification
and of explaining what could possibly give it such a status, the assumption
that the correct concept of knowledge satisfies the weak conception also gives
rise to a number of collateral problems. Here I will focus on the two best
known and most wisely discussed of these: in the next section, the Gettier
problem; and, in the following section, the lottery paradox. My suggestion
is basically that these by now familiar problems are at bottom not really
genuine epistemological problems at all. Unlike the more traditional problems of
epistemology, these supposed problems do not reflect genuine intellectual issues
that require deep philosophical insight and creativity for their solution. Instead,
they are entirely artificial difficulties, created by the mistaken assumption that
the weak conception of knowledge is correct, and can be easily dissolved by
simply rejecting that assumption.9 The ease with which this can be done seems
to me to constitute a further argument, albeit a less direct one, against the weak
conception.
3. The Gettier problem

Consider first the Gettier problem. Gettier famously showed that there are
easily constructed examples that satisfy the traditional definition of knowledge,
construed along the lines of the weak conception, but that nonetheless do
not seem intuitively to be cases of knowledge.10 These cases may be roughly
characterized as ones in which a belief is strongly justified (exactly how strongly
matters little, as long as the justification is less than conclusive) and also true,
but true in a way that intuitively is merely accidental or fortuitous in relation
to the justification in question true in some unexpected way that is other

than the way that the justification seems to point to. Thus in one of the most
widely discussed examples, you have strong evidence that one of your colleagues,
Professor Nogot, owns a Ford automobile and on this basis come to accept the
general claim that at least one of your colleagues owns a Ford; the evidence
is in some way misleading or deceptive, and Nogot does not in fact own a
Ford; and yet a different colleague, Professor Havit, about whom you have no
relevant evidence at all, does in fact own a Ford. Your belief that one among
your colleagues owns a Ford is both true (because of Havits ownership) and
strongly justified (because of your evidence pertaining to Nogot), but does not
seem intuitively to be a case of knowledge, with the intuitive reason being that
Havits possession of a Ford is not in any way suggested or indicated by your
evidence and so is a mere accident in relation to it. As Gettier makes clear in his
own original discussion, the possibility of this sort of case rests squarely on the
assumption that the level of justification required for knowledge is some level
that is less than conclusive and thus presupposes that the correct concept of
knowledge is a version of the weak conception.
There is no room in this paper to even begin to recapitulate the extended
and extremely convoluted discussion that was provoked by Gettiers argument.
Fortunately, what matters for present purposes is not the details but rather
the general character of the discussion: Many, many different revisions of the
conception of knowledge were proposed, usually but not always taking the form
of adding a fourth condition to the standard three. Many of these proposed
revisions then fell victim to further counterexamples of different and often more
complicated sorts, leading to still further proposals, and so on, and so on. By
the time this amazingly voluminous discussion had at least largely run its course,
two things were clear. First, though there were several fairly prominent general
lines of solution, there was no single solution to the problem that came very close
to being generally accepted. This fact by itself would perhaps not differentiate
the Gettier problem from most of the standard problems of philosophy. But
second, and much more important though less noticed, the discussion yielded
almost nothing by way of theoretical insight into the nature of the problem and
its implications for the concept of knowledge and for epistemology generally: no
insight into how and why such a problem arises in the first place, no account of
why the concept of knowledge should involve further conditions of the various
kinds proposed, no understanding of how these further conditions might fit
together with the remaining conditions to make a cohesive and intellectually
significant concept, and thus no real understanding of why any of the modified
conceptions of knowledge that emerged from the discussion should be regarded
as intellectually important or significant as worthy of the very extensive
philosophical concern that has been lavished on them. Instead, the discussion
of the Gettier problem was focused almost entirely, in a shallow, essentially
ad hoc way, on piecemeal revisions and specific examples. And thus all of this
discussion, I would suggest, yielded very little in the way of genuine philosophical
understanding of knowledge or of anything else.
The concern I am raising here may be illustrated by focusing on one of the
most widely proposed fourth, anti-Gettier conditions, the requirement that the
persons justification not be defeasible, where the justification for proposition P
is defeasible if there is some true proposition Q such that if the person were to
become justified in believing Q, he or she would no longer be justified (to the
requisite degree) in believing P.11 Thus in the example above, there are various
choices for Q: perhaps some description of the ways in which the evidence about
Nogot is misleading, perhaps some stronger evidence showing that Nogot does
not own a Ford, or perhaps just the denial of the original claim about Nogots
ownership. And it seems quite plausible that in all of the familiar sorts of Gettier
examples, such a true proposition, or usually more than one, can be found.
Thus the addition of such a condition would yield the intuitively delivered
result that the various Gettier examples are not in fact cases of knowledge and in
this way would solve the problem. But does the addition of such a condition
really yield any insight into what is going on? Do we understand why knowledge,
understood as a supremely and uniquely valuable cognitive state, should require
or involve such a condition? In relation to the strong conception of knowledge,
the fact that justification is in this way defeasible would of course show at
once that it is less than conclusive and would thus show that the concept of
knowledge is not in fact satisfied. But if justification need not be conclusive, why
exactly must it not be in this way defeasible? What, other than merely avoiding
the counterexamples, could be the rationale for such a requirement? And my
suggestion is that without such a rationale, the significance of the supposed
concept of knowledge that results from adding the indefeasibility requirement
is obscure. (I would suggest that here, and in other places as well, the relevant
intuitions in fact derive tacitly from the strong conception, which is, like many
intuitively compelling ideas, much easier to explicitly repudiate than to fully
escape from.)
There is no space here to consider further anti-Gettier conditions, but my
suggestion is that the same general sort of objection applies to virtually all of
them. In addition to various other problems, there is simply no clear intuitive
rationale for them and so also no clear intuitive rationale for the supposed
weak concept of knowledge that results from incorporating them: no reason
why the resulting concepts should be thought to be intellectually important and
worthy of serious philosophical attention.12
Here is my diagnosis of the situation. Part of what gives rise to the supposed
problem is obvious enough. If a claim is justified to some less than conclusive
degree, then it will obviously be possible for it to still turn out to be false. And
if this is possible, then it will also be possible for it to be true in some way
other than the way that the justifying evidence or reason suggests and in this
way to be accidentally or fortuitously true. In such a case, a person who believes
that claim on the basis of the evidence in question will have a true belief, but
in an accidental or fortuitous way. But what is needed to generate a seemingly

serious, indeed extremely intractable problem is the further assumption, made by
proponents of the weak conception, that there is a definite concept of knowledge,
a supremely valuable, all-or-nothing cognitive state, that requires only some less
than conclusive degree of justification. For if this assumption is made and a case
of the sort just described occurs in which the level of justification is taken (on
what basis?) to be at the magic level or greater, the intuitive unsatisfactoriness
of the situation from an epistemic standpoint will require saying that the exalted
state of knowledge has not in fact been achieved. And this in turn will require
assuming that some further requirement for that state that has not been satisfied,
and we will be off to the unrewarding Gettier races.
Without that assumption, in contrast, there is, I suggest, simply no discernible philosophical problem to be solved. Cases of the sort in question, cases
where truth is accidental or fortuitous in relation to the justification available
for a claim, may at times be surprising, and it is useful to see how and why they
may occur. But without the assumption of the weak conception, there is simply
no need to seek a definite, precisely defined condition that will exclude them.13
Whereas if the weak conception of knowledge were correct, there would have to
be such a condition, no matter how difficult it is to clearly define it or, more
importantly, to assign it a plausible rationale. Thus my suggestion is in effect
that in light of the very unsatisfactory results of the Gettier-inspired inquiry,
we should apply modus tollens rather than modus ponens to the just italicized
conditional.
4. The lottery paradox

There are a number of different versions of the lottery paradox, not all of
them directly concerned with knowledge. But the version I am concerned with
here can be set out very simply, with the weak conception of knowledge as an
essential ingredient.
Assume, following the weak conception, that there is a less than conclusive
magic level of justification that is, together with belief and truth (and the
satisfaction of an anti-Gettier condition hereafter I will take this for granted),
sufficient for knowledge. And now suppose that a lottery is to be held, in which
there will be one and only one winning ticket, drawn in a fair and random way.
Whatever that magic level of justification may in fact be, it seems clear, so
long as what is essential to epistemic justification is something in the vicinity of
likelihood or probability of truth, that if the number of tickets is large enough,
the probability of, and so the justification for, the claim that any particular ticket
will lose will reach or exceed the magic level. Suppose then that the number of
tickets is indeed at least that large (even though we have no clear idea how large
that would have to be). It seems to follow that if a person were to believe, on
the basis of this high probability and resulting justification, that some particular
ticket will lose, and it will in fact lose, then that person knows, in advance
of the actual drawing, that the ticket in question will lose. This already seems
counterintuitive: given the weak conception, we may suppose that the person
knows that one ticket will win, and the chosen one has as good a chance as any
of the others. But, still worse, the same reasoning would obviously apply to all
the other losing tickets, so that a person who somehow happened to form the
corresponding set of beliefs on the same sort of basis would thereby know of
each losing ticket that it would lose. Indeed, such a person would apparently be
in a position to infer from this other knowledge, together with the knowledge
that there will be a winning ticket, that a certain specific ticket will win, and have
knowledge of this result!
Because it is abundantly clear on an intuitive basis that such a person does
not genuinely know any of these things, the lottery example may well seem to
constitute a simple and decisive reductio of the weak conception of knowledge
with the supposed paradox deriving only from the untenable assumption
that the weak conception is correct. In the end, I believe that this is the correct
assessment. But before accepting it as such, we must first ask whether there
is any tenable response to the problem that preserves the weak conception: an
intuitively plausible way of adding some further condition or requirement (in a
way parallel to that which is attempted in relation to the Gettier problem) that
will rule out the various claims in the lottery case as satisfying the requirements
for knowledge, while still leaving the commonsensical knowledge claims that the
weak conception is intended to preserve untouched.
One response is to accept that the person indeed knows the claims about
the individual tickets, but to block any further inferences by rejecting the view
that one always knows the validly inferred consequences of ones knowledge.
As already argued above, such a view seems to me to drastically undercut the
supposed value of knowledge.
Another sort of response is to impose some further condition on justification
that will rule out lottery cases, while leaving other knowledge claims that involve
the weak conception untouched. In fairly striking contrast to the situation with
the Gettier problem, however, suggestions for how to accomplish this are pretty
thin on the ground. I will look at two possibilities, starting with an earlier attempt
of my own, one whose deficiencies seem to me to reveal well enough some of
the pitfalls in this area. In The Structure of Empirical Knowledge (and an earlier
paper),14 I suggested that what goes wrong in the case of the initial claim that
some specific ticket in the lottery will lose is that although this claim can be
assumed to be strongly enough justified to satisfy the justification requirement
for knowledge (which I was assuming to fit the weak conception), it is also one
of a class of relevantly similar, equally justified alternatives, all individually very
unlikely, but such that the person in question knows (according to the correct
version of the weak conception) that one of these unlikely alternatives will in
fact obtain and has no relevant way of distinguishing among them. And thus, I
suggested further, we should add a further requirement for knowledge that rules
out knowledge in this specific sort of situation, with the background assumption
being that such a requirement will cause no trouble in most other cases where
the justification meets the standard of the assumed to be correct version of the
weak conception.
There are many problems that can be raised for this proposal. One objection
is that the problem posed by the lottery does not really require knowledge that
one ticket will win: thus even if there are a few unsold tickets and no one will
win if one of these is drawn, the intuitions behind the problem seem unaltered.15
Another is that the problem seems to persist even if the chances of winning and
so the resulting justifications are not precisely equal. I think that the proposed
additional requirement could be modified to allow for these more complicated
sorts of cases, but will not attempt to do so here, since there are more serious
problems ahead.
A further, much more serious objection is that ruling out knowledge in
lottery cases turns out to impugn a much wider range of ordinary, common-sense
knowledge than might at first have been thought.16 This is because many cases
of common-sense knowledge entail that various lottery-type propositions will
not hold and so, assuming closure, these common-sense claims cannot be known
unless one can also know the denials of those lottery-type propositions with
the latter knowledge being ruled out by the proposed additional requirement.
Thus, to take a rather salient example, I think I know that (P) I am going to
continue working on this paper tomorrow. But I also may well know (by the
standard of the correct version of the weak conception) that I am one of a large
group of people, relevantly similar on my information, at least one of which
will suffer a fatal heart attack today and I cannot know, given the proposed
requirement, that (Q) I will not be one of the victims. But P obviously entails
Q, and so, given closure, I cannot know P without also knowing Q.17 How far
this sort of problem extends is uncertain, but a little exercise of imagination
will suggest that it extends very far indeed, threatening at least a large swath of
supposed common-sense knowledge.
But the deepest and most serious objection to the proposed additional
requirement, in my view, is that it and so the proposed version of the weak
conception that incorporates it again lacks any plausible intuitive rationale.
One way of putting this is to say that it seems rather ad hoc: ruling out the
objectionable cases merely by giving a slightly more abstract version of the
situation in which they arise.18 But while this way of putting the point has some
intuitive force, it does not seem to me to go to the heart of the matter. What
is really objectionable about the suggested additional requirement, as was the
case with the first of the anti-Gettier requirements discussed above, is that we
have no clear understanding of why the supremely valuable cognitive state of
knowledge (understood as fitting the weak conception) should involve such a
requirement or better, turning the same point around, that we have no idea
of why a state that satisfies this requirement (along with the others involved in
the weak conception) should be regarded as supremely valuable from a cognitive
or epistemic standpoint, why it should be thought to have the exalted evaluative

status that knowledge is intuitively supposed to have.
As with the first anti-Gettier condition, this would not be so if it was
the strong conception of knowledge that was being assumed. In relation to
that conception, a failure to satisfy the proposed requirement would show
immediately that the justification was less than conclusive, thus failing to satisfy
the justification condition, but this would also mean of course that no further,
independent requirement would be required. But if it is the weak conception that
is supposed to be correct, why exactly should the failure to satisfy the suggested
requirement rule out knowledge? If the justification for knowledge need not be
conclusive (so that knowledge is fallible), why should the knowledge that this
possibility of mistake will in fact realized in one of a set of indistinguishable
cases mean that the rest of them cannot still be genuine knowledge? How is this
significantly different from the more general fact that some proportion of our
claims of knowledge will turn out to be mistaken because the justification was
less than conclusive? (Here again, it seems to me that intuitions whose real source
is the strong conception are creeping in, making this attempt to accommodate
them seem even more obviously ad hoc.)
These last two problems also afflict a different and rather simpler additional
requirement proposed by Dana Nelkin (with something like this also being
suggested by a number of others).19 Nelkins idea is essentially that mere
statistical probability, of the sort involved in a lottery or similar example, never
satisfies the justification condition for knowledge.20 This proposal will rule out
many cases of seemingly ordinary common-sense knowledge in the same way
as did the previous one, because they entail lottery-type propositions that could
only be justified and known on a statistical basis. Thus, in the specific example
offered above, I will be unable to know on the basis of high probability that I will
not be the one in the relevant group who suffers a fatal heart attack and hence
will also be unable to know that I will continue to work on this paper tomorrow.
And again, the intuitive rationale for imposing this additional condition
(apart from merely its success in ruling out all and only the cases we want
to rule out) is far from obvious. Why, if the point of epistemic justification
is finding truth and if (as the weak conception insists) conclusive justification
is not required for knowledge, should high enough statistical probability not
suffice? Nelkin suggests that the problem is the lack of a causal or explanatory
connection between the probabilistic justification and the fact that makes the
claim true (assuming that it satisfies the other requirements for knowledge and
so is in fact true) and suggests further (though pretty vaguely) that this points in
the direction of some sort of externalist account of justification and knowledge.
Externalist views are beyond the intended purview of this paper, but of
course many internalist views (including my own) have appealed to explanatory
connections as at least one basis for justification. But the problem with using
the idea of explanation as a rationale for the proposed requirement is that the
justificatory relevance of explanatory connections seems itself to be that they
make truth more likely or probable. To be sure, the resulting probability or

likelihood is not in any obvious way statistical in character and certainly does
not in general come with a specific numerical value attached. But I can find
no clear reason for thinking that it is fundamentally different in kind from that
which results from statistical considerations and thus no clear way in which
the justificatory relevance of explanation provides a good rationale for ruling out
statistical probability as also yielding a basis for justification.
All of this seems to me to suggest strongly that the lottery paradox is indeed
best viewed as a simple reductio of the weak conception of knowledge. If less
than conclusive justification were enough for knowledge, then we would have
knowledge in the lottery case, and it is intuitively very clear that we do not. My
further suggestion is that the underlying reason for this is that it is utterly clear
in a lottery case that the justification is less than conclusive; whereas in other
cases that allegedly constitute knowledge under the weak conception, this lack
of conclusiveness is less conspicuous and so more easily ignored.21
5. Summary of Part I
I have argued that the view that common sense embodies a specific concept of
knowledge that both satisfies the weak conception and is of central philosophical
importance faces a number of serious objections. There is no apparent way to
specify or determine, even to a reasonable measure of approximation, what the
required non-conclusive level of epistemic justification might be and, more
importantly, no way to say why any such level should have the sort of significance
that the weak conception attributes to it. There is also no plausible intuitive
rationale for the further requirements that are needed by the weak conception to
solve the Gettier problem and the lottery paradox, and so no plausible account
of why the supposed concept that would include these would be intellectually or
philosophically important. All this suggests that a reconsideration of the strong
conception is in order, which is what the second part of the paper will attempt
to do.
Part II: A Tentative Defense of the Strong Conception

1. Introduction
If the weak conception is indeed untenable for the reasons indicated, then
there are, as far as I can see, only two possibilities left within the limitations
assumed in this paper. One is that the common-sense concept of knowledge is
indeed one that fits the strong conception (with the further questions about that
concept that this leaves open being of little moment here). The other is that there
is after all no coherent, intellectually satisfactory concept of knowledge to be
found in common sense: none that can both satisfy the essential constraints on
such a concept and make good sense of ordinary usage and practice. I will return
to this last alternative at the very end of the paper. But the main task of this
second part will be to see what can be said in favor of the view that the strong
conception is, after all, the correct one, with the main task being to reconcile
that view with the ordinary usage of the terms know and knowledge.
The problem, of course, is that if the strong conception is indeed the correct
account of knowledge, then most of our ordinary claims to know and attributions
of knowledge will turn out to be mistaken. The precise extent to which this is so
can be debated in relation to many different epistemological issues, but the details
of this do not matter very much for our present purposes. It is enough to say
that on the prevailing views of these various issues, the proportion of knowledge
attributions that can possibly be correct according to the strong conception will
be quite small at best.
As already briefly noticed above, there are many who would regard this
result by itself as a conclusive refutation of the strong conception, with the idea
being apparently that it is either impossible or at least very, very unlikely that
common-sense knowledge attributions could be so badly mistaken. But such
a conclusion seems to me to be much too quick especially in the absence
of any very appealing alternatives. It is no doubt surprising that most of the
employments of a common-sense concept should be mistaken, but it does not
seem to me self-evidently impossible that there could be an adequate explanation
of why this is so. And this is especially the case where the concept in question
is a broadly evaluative one that characterizes a kind of ideal state or situation,
one that might often be approached but rarely fully realized. I believe that there
are in fact resources available to a proponent of the strong conception that can
go quite a long way toward an explanation of this situation though whether
they go far enough remains to be seen.
2. Explaining false knowledge attributions: some alternatives

How then might we try to explain, in a way compatible with the strong
conception, why ordinary people make so many false attributions of knowledge,
both to themselves and to others? I will suggest that there are a number of
plausible explanations for individual cases of this sort. No one of these can by
itself adequately handle enough of these cases, but when they are taken together, it
is at least much less obvious that they do not succeed. I will begin by considering
these individually.
(i) Simple epistemological error. Perhaps the most obvious explanation, one
that plausibly extends very widely and that seems largely unnoticed by many
philosophers, is that an ordinary person may reasonably regard the justification
for a belief as conclusive even where deeper philosophical insight shows, or at
least seems to show, that it is not. Consider, for example, a case where an ordinary
person seems to himself or herself to be perceiving a standard sort of mediumsized physical object at close range and under good conditions, and believes
on this basis that such an object is there. Even if the persons justification in
such a case is not in fact conclusive (because of subtle philosophical objections
having to do with, for example, Cartesian demons or the possibility of being a
brain-in-a-vat), it is easy to see how it might nonetheless seem to him or her to
be conclusive,22 leading to a self-attribution of knowledge (and to attributions
of knowledge to others whose situation is similar).
This sort of explanation extends plausibly to a great many of the commonsense knowledge attributions that turn out, according to the strong conception,
to be false, including many of those that seem intuitively most compelling. Very
often the reasons why the justification in such cases is less than conclusive
involve philosophical considerations, especially skeptical possibilities of various
sorts, which ordinary people cannot be assumed to be familiar with (and whose
obviousness is easily exaggerated by philosophers). Cataloging the full range
of such cases is too large a task to be attempted here, but it would plausibly
include, in addition to beliefs about the existence of physical objects on the basis
of direct perception: many simple, perceptually based causal beliefs (e.g., that
the stroke of an axe caused a cut in a tree), many beliefs about the mental states
of other people (e.g., that the accident victim who is screaming and bleeding
profusely is in great pain), many beliefs about the contents and structure of the
physical world (e.g., that Chicago is north of St. Louis, or that chickens exist in
the world), and so on. The general formula here is that the more complicated,
subtle, and unobvious are the philosophical reasons for thinking that a given
sort of justification is inconclusive, the more plausible it is that ordinary people
may be led to make mistaken attributions of knowledge while still adhering in
their thinking to a concept of knowledge that fits the strong conception.23
(ii) Exaggeration. As Butchvarov points out in a very suggestive discussion
of this issue,24 certain concepts are often used in a way that involves exaggeration,
and it is clear that knowledge is one of these. Butchvarovs characterization is that
the concepts in question are ones that pertain to evaluative standards or ideals
and that have serious practical significance. His leading example is the concept of
romantic love. It is obvious that romantic love represents a kind of ideally perfect
situation with respect to a certain sort of interpersonal relation and also that the
applicability of this concept has a range of practical implications that are usually
viewed as highly desirable. This creates a strong pressure to attribute the state
of being in love to oneself and to others even when the strong (though rather
elusive) requirements for being in such a state are not clearly met and even
when they are clearly not met. Moreover, unlike some other kinds of exaggeration
that are, one might say, less serious, many such attributions of love are not very
readily withdrawn (even though there is very often at least a background worry
about whether they are correct).
The analogy with knowledge is not perfect, but it does seem to me to be
quite helpful. As viewed by the strong conception, knowledge is also an ideal
state of a certain sort and one with even more pervasive important practical
significance. (See the discussion, in the next section, of assertion and practical
reasoning.) Thus it is easy to see how exaggerated attributions of knowledge
might be made in cases where the desirability of achieving this ideal is particularly
high, especially when the gap between the actual case and the ideal is, or is viewed
as being, relatively small, but sometimes even when this is not the case. Many
such attributions will be withdrawn if strongly challenged or if the issue comes
to seem especially urgent, but this will again not be done as readily as with other,
more casual sorts of exaggeration.25 (Knowledge may also be attributed in ways
that involve more casual sorts of exaggeration, including the case, mentioned
earlier, where the degree of justification is viewed as being close enough for the
purposes or issues at hand. These attributions will be more readily withdrawn if
seriously challenged or if the stakes turn out to be more serious.)
(iii) Close approximation. The third sort of explanation is in effect a special
case of exaggeration, but one that is different enough to warrant separate
discussion. Sometimes a case may seem so close to an ideal standard as to make
it seem unreasonable or needlessly fussy to insist on the difference. It seems
plausible to suppose that there are many cases where the strength of a persons
justification, though not genuinely conclusive, is so close to being conclusive (or
is mistakenly assessed by common sense as so close to being conclusive) that
there seems to be no serious intellectual point in refraining from an attribution
of knowledge. This is not a matter of being close enough for some specific,
contextually relevant purpose; nor is it a matter of meeting some independently
specifiable standard that is just short of conclusiveness. Instead it simply reflects
a general common-sense impatience with fine hair-splitting. And this, I suggest,
accounts for a significant class of mistaken knowledge attributions in cases where
the justification really is overwhelmingly strong even though not quite conclusive
and perhaps especially for cases, such as very strongly confirmed inductive
generalizations, where it is clearly realized that complete conclusiveness is not
possible, even in principle.
(iv) Knowledge claims as conditional. A final sort of explanation, suggested
by Richard Fumerton,26 is that knowledge claims are sometimes, perhaps even
often, elliptical. Someone claims to know that P, but the proposition he or
she really thinks is conclusively justified (or, perhaps, justified in a way that
closely approximates conclusiveness or is being claimed in an exaggerated way
to be conclusive) is not P simpliciter, but rather that if various possibilities that
are being assumed to be in some way not worth taking seriously are indeed
false, then P. One common version of this is one where several possibilities
are taken seriously and all but one of those is thought to be excluded by the
evidence (conclusively or in one of the other ways just noted), but there are
other remote possibilities that are dismissed without any real consideration. An
example: Given that the last piece of cake is gone and that Ann, David, and
Jennifer were the only ones besides myself in the house, and that I am sure (or
close to sure or am claiming in an exaggerated way to be sure) that neither I nor
Ann nor David ate it, I claim to know that Jennifer did. I am assuming, without
any real consideration, that various other remote possibilities do not hold: that
the piece of cake did not evaporate in some mysterious way; that it was not
entirely devoured by a horde of ants that then departed, leaving no trace behind;
that a very clever but strangely motivated burglar did not sneak in and eat it;
and so on. Fumertons suggestion is that the knowledge claim in such a case is
better represented as a conditional, with the denials of the possibilities that are
not being seriously considered conjoined in the antecedent and the proposition
contained in the original knowledge claim as the consequent.
3. Two further factors: assertion and practical reasoning

In addition to the alternatives just discussed, there are two further factors
that, I suggest, help to explain the prevalence of knowledge attributions that
do not accord with the strong conception of knowledge, even though (as I am
provisionally assuming in this part of the paper) that conception is correct as
an account of the common-sense concept of knowledge. Both of these have
the effect of increasing the pressure to make knowledge claims even where the
standard for knowledge is not fully met, especially if the gap is or is perceived
to be relatively small. (The two factors in question have in fact often been taken,
sometimes only implicitly, to support the weak conception, but I will argue that
this way of construing them is less plausible, given the other problems with the
weak conception, than the one I will suggest.)
(i) Much has been written recently about the idea that knowledge is the
norm of assertion: that is, roughly, that assertions are properly made only
where the assertor has knowledge of the asserted proposition (so that an assertion
is in effect an implicit claim to have knowledge).27 There is something that seems
intuitively right about this view: it is, after all, odd to say something of the form
P but I do not know that P. But if I am right that the strong conception is correct
and is indeed the only really intelligible conception of knowledge that we have,
and assuming that justification is indeed at best rarely conclusive, we would be
led to the result that assertions (without some probability qualification or the
like) can almost never be properly made.
But such a result is obviously intolerable from a practical standpoint. We
have urgent needs for communicating information even uncertain, perhaps
only apparent information to each other, and it would be extremely burdensome to have to couch virtually all assertions in probabilistic terms (in addition
to being almost pointless where, as is usually the case, we have no way of
determining the specific degree of probability with any real precision). Thus
there is a strong practical pressure to make unqualified assertions even where
this is not really (fully) warranted by the epistemic situation (and to back them
up, when challenged, with attributions of knowledge). Sometimes this leads to
loose assertions that are unacceptable in a really patent way, but it also helps
to explain why knowledge attributions falling under categories (ii), (iii), and
(iv) of the previous section are so prevalent. (Such a further explanation is, of
course, not needed for category (i), where the person believes, mistakenly, that
the standard for knowledge, and so for correct assertion, is fully met.)
(ii) A further connected but still distinguishable pressure in the direction
of knowledge attributions that do not fully meet the standard of the strong
conception is provided by the role of knowledge in practical reasoning. Much
of the reason that we seek true beliefs about the world is so that we can use
them in deciding what to do in various practical situations. Here it is obvious
that what we really need and want is knowledge that fits the strong conception:
knowledge that involves conclusive justification and so an assurance of truth.
For only knowledge of this sort can provide a secure basis for determining the
course of action that is really best in relation to our desires and purposes. And
this makes it at least plausible that, as some have suggested, knowledge is also
the norm for what can be properly used in practical reasoning.
But given our assumption that the strong conception is correct, the resulting
problem is very obvious. Lacking very much genuine knowledge and still faced
with the unavoidable need to deliberate and act, what are we to do? It would
perhaps be possible in principle to proceed for a time in probabilistic terms, and
Bayesians, among others, have tried to say how such reasoning might work.
But the problems with this are obvious and overwhelming: we rarely know
probabilities with enough precision to arrive at anything very definite on this
basis; and in any case, we ultimately have to actually do some particular thing
and not just probably do it.
Thus there is again a practical pressure in the direction of reasoning from
and acting upon claims whose justification does not fully meet the standard
for knowledge and then in effect buttressing such reliance by attributing the
corresponding knowledge to oneself and to others. And this again can further
explain the attributions of knowledge that fall under categories (ii), (iii), and
(iv) above though once more no such further explanation is needed for
category (i).
Thus, I am suggesting, the intuitive connection between knowledge, on
the one hand, and assertion and practical reasoning, on the other, rather than
supporting the weak conception, can instead contribute to an explanation of why
knowledge attributions that are incorrect according to the strong conception are
so commonly made (even though the strong conception is, as I am presently
assuming, correct). The fundamental point with regard to both assertion and
practical reasoning is that while the intuitive connection of each with genuine
knowledge is strong, this by itself in no way guarantees or even makes especially
likely that there will in fact be enough genuine knowledge to meet the practical
pressure to make assertions and the even stronger practical pressure to engage
in practical reasoning. Rather than inferring from the existence of obvious cases
of assertion and practical reasoning to a correspondingly weak standard for
knowledge that is satisfied for the relevant claims, it is, I submit, at least as
plausible in itself and more plausible given other considerations to suppose

that the practical needs involved lead to many instances of (less than fully proper)
assertion and (less than optimum) practical reasoning where the standard for
genuine knowledge is not satisfied for the claims involved and hence, given
the intuitive connection, to corresponding inaccurate, exaggerated attributions
of knowledge.
4. Summary and tentative assessment

My suggestion is that the elements in the previous two sections, taken
together, go a long way toward explaining how the view that the strong
conception of knowledge is correct might be reconciled with the range of
knowledge attributions that ordinary people actually make. There are strong
pressures, both intellectual and practical, that favor relatively loose and generous
attributions of knowledge to others and, especially, to ourselves; and these lead
to many exaggerated knowledge attributions, especially but not only where our
epistemic situation closely approximates, or seems to us to closely approximate,
the correct standard. They also may lead to elliptical claims of knowledge, where
the claim should really be conditioned on the falsehood of a set of possibilities
that are being assumed not to hold. And, it must be emphasized again, there
is also a wide range of cases where though the justification is, on fairly subtle
philosophical grounds, inconclusive, this cannot be assumed to be apparent to
ordinary people, who make knowledge attributions accordingly.
It will be obvious that all of this is still highly schematic and impressionistic
at best, needing to be filled out and amplified by a detailed consideration of
many more examples than can be offered here. But the question that still needs
to be asked, albeit in a tentative and provisional way, is whether this sort of
explanation, thus amplified, would be good enough. Even without the needed
amplification, such an explanation is surely good enough to call into question the
quick, almost casual dismissals of the strong conception that are so prevalent,
but is it (or can it be made) good enough to strongly support the view that
the strong conception of knowledge is indeed the correct one, the one that is
operative in ordinary thought?
I have to admit to a good deal of uncertainty about the correct answer to
this question, and thus the result of this paper will itself be, alas, inconclusive.
What I want to insist on, however, is that this uncertainty gives no real support
to the weak conception, which still remains impossible to flesh out in a tenable
way (or, I think, to views like contextualism). The alternative is rather that there
simply is no coherent conception of knowledge to be found in common sense
and ordinary thought or at least none that has the kind of intellectual and
philosophical significance that such a concept is supposed to have.
To many, this last suggestion will surely seem intolerably puzzling or
even paradoxical: how could a widely used common-sense concept be simply
incoherent? But in light of the discussion here, I think we can see at least the
outlines of a kind of quasi-historical explanation of how this might be so. Putting
the point very roughly and impressionistically, common sense thought in this
area began with the strong conception of knowledge, which is both intuitively
compelling and admirably fits our practical needs. But the increasing realization
that this conception is very difficult to satisfy led, not to the substitution of
an ersatz weak conception that would be impossible to specify in a meaningful,
intuitively satisfying way, but rather to an essentially irresolvable discord between
the basic concept of knowledge and the attributions of knowledge that we want
and need to make. The result is a situation in which there is simply no conception
of knowledge that can make coherent sense of all the things we think and want
to say about it.28
It will be obvious that the difference between the situation just described
and the one suggested earlier (in which the strong conception is correct but
widely misapplied) is anything but sharp and that the two possible situations
may well shade into each other in ways that make it hard to draw clear lines.
This is perhaps the main reason for my own uncertainty about which conclusion
is correct.
5. The implausibility of skepticism

If the strong conception of knowledge is indeed the correct one, then the
result is clearly a fairly strong form of skepticism not, I think, total skepticism,
but still a skepticism according to which a very high proportion of ordinary
knowledge claims and attributions are in fact false.29 As noted at the beginning
of this paper, such a result is often dismissed as too implausible or even absurd
to be accepted. One objection appeals directly to the ordinary uses of the terms
know and knowledge, and it is this objection that I have been mainly dealing
with in this part of the paper.
But I want in this section to consider what seems to be a somewhat different
objection, one that appeals directly to the alleged intuition that we do have
knowledge of many things that the strong conception would seem to exclude. I
have in mind here what is sometimes referred to as the Moorean objection,
named after its perhaps most famous proponent, G. E. Moore. Moore simply
asserted that we do have knowledge of such things his famous example is
his knowledge, as he holds them out in front of himself, that he has hands
and that any view that denies this must be rejected. (Part of the idea is that any
premises upon which the denial of such knowledge is grounded will inevitably
be less certain than the knowledge attributions themselves.)30
The intuition to which Moore is pointing is indeed a powerful one, not
easily dismissed. But before concluding on this basis that the strong conception
of knowledge must, after all, be mistaken, I think it is important to ask, more
carefully than is usually done, just what the content of Moores intuition really
is. Is Moore claiming merely that it is correct to apply the words know and
knowledge to such cases, even if this is done on the basis of low epistemic
standards of the sort that the contextualist claims to often apply? Is he perhaps
claiming only that the justification for the claims in question, while less than
conclusive, exceeds some magic level that can be neither specified nor explained
as to its significance? Or is the intuition not rather that the justification in
such cases really is conclusive, even though various philosophical arguments and
possibilities might seem to show that it is not?
Those who have such an intuition will have to ask themselves what its
content really is, but my suggestion is that it is the last of these three alternatives
that is correct. And in this case, the Moorean intuition does not really count
against the strong conception of knowledge and provides no genuine motivation
for alternatives like the weak conception and contextualism. Instead it is a direct
challenge to the philosophical grounds for thinking that justification is only very
rarely conclusive. Though intuitively very powerful, it seems to me that this
challenge can be met by appeal to the various arguments in question. But the
important point here is that the correctness of the strong conception is not really
at issue in the resulting discussion.
6. Conclusion
I have tried to show in this paper that the weak conception of knowledge is
untenable and that the specific common-sense concept that is often thought to
realize it is a myth.31 My tentative conclusion is that the correct common-sense
concept of knowledge is indeed a version of the strong conception, though a
complete case for this result would require a much more extensive survey of
examples than has been possible here. A further implication, I believe, is that
focusing on knowledge and especially on preserving common-sense attributions
of knowledge yields little that is of deep epistemological and philosophical
interest. What this further suggests is that the really important philosophical
issues pertain to justification and not to knowledge.
Notes
1. Conception rather than concept, because a specific concept would have to
say more precisely just what this lesser level of justification is something that
I will claim cannot satisfactorily be done. A conception, as I will use the term
here, is thus a general outline or characterization which one or more specific
concepts might fit. The weak conception of knowledge is, of course, perfectly
real: what I am claiming to be a myth is the idea that that there is a specific
common-sense concept that satisfies it. (The earliest use of the terms weak and
strong to distinguish two conceptions of knowledge or senses of know
and knowledge seems to be in Norman Malcolms paper Knowledge
2.
3.
4.
5.
6.
and Belief, reprinted in Malcolm, Knowledge and Certainty (Englewood Cliffs,

N.J.: Prentice-Hall, 1963). Malcolms actual account of the distinction is more
complicated and a bit idiosyncratic, having to do with something like the
intentions of the person using the term.)
I will assume for the most part that the degree of justification required for
knowledge does not vary from context to context (as claimed by contextualists)
or from subject to subject (as claimed, in effect, by John Hawthorne, Knowledge
and Lotteries, Oxford: Oxford University Press, 2004). I have no space here to
explore these recently popular views (apart from a very brief discussion in a later
footnote). But I also think it obvious that an account that avoids complexities
of this sort is preferable if it can be made to work.
Thus the paper is mainly directed at those who share the internalist point of view
or at least agree that knowledge involves a substantial internalist element. But
I believe that the main line of argument would apply with little if any modification
to an externalist view like Goldmans reliabilism, where degrees of justification
correspond to degrees of reliability. I also think that it would apply to many if not
all alternative externalist and quasi-externalist accounts of the third condition
for knowledge (the one that differentiates knowledge from mere true belief); but
there is no room here for a discussion of these further alternatives.
Sometimes it is said that the justification must entail the truth of the claim in
question, but I do not want to assume that anything that can justify belief must
be capable of standing in an entailment relation (and so would presumably have
to be propositional in nature). Little needs to be done to turn this conception into
a specific concept: clarification of the relevant concept of truth and perhaps a
specification of a requisite degree or strength of belief. But it is worth noting that
this conception of knowledge assumes that the concept of justification, and the
further concepts needed to explicate justification, are conceptually prior to the
concept of knowledge and can be understood independently. This assumption
seems to me to be correct, but I have no space for an independent defense of
it here. For a contrary view, see Timothy Williamson, Knowledge and its Limits
(Oxford: Oxford University Press, 2000), chapter 9; and for an objection to the
strong conception based on this point, see Hawthorne op. cit., pp. 13740.
Discussions that say something close to this are legion. Included among them
is my book The Structure of Empirical Knowledge (Cambridge, Mass.: Harvard
University Press, 1985).
Perhaps the only philosopher whom might be viewed as having seriously tried
to define the degree of justification required by the correct realization of the
weak conception is Roderick Chisholm, in the three editions of his Theory of
Knowledge (Englewood Cliffs, N.J.: Prentice-Hall, 1966, 1977, 1989), together
with other works written in the same period. Chisholms final account, in the 3rd
Edition, is that a proposition p is evident (his term for the level of justification
required for knowledge) for a subject S just in case For every proposition q,
believing p is at least as justified for S as is withholding q [that is, suspending
judgment with respect to q]. But after all of Chisholms rather painstaking
discussion in these works, neither what the resulting level of justification really
amounts to nor, more importantly, the rationale for that level being the right one
is at all clear.
7. It could, I suppose, still be insisted that common sense somehow tacitly involves
or incorporates a definite concept of knowledge that involves such a specification,
even though there is no apparent way to determine what it is or how it is thus
embodied. But it is hard to see why such a claim should be taken seriously.
8. Here, of course, we see one of the motives for contextualism. I will have a little
to say about contextualism in a later footnote, though, as already noted, a full
discussion is impossible here.
9. I would also say something similar about, for example, the Preface paradox.
10. Edmund Gettier, Is Justified True Belief Knowledge? Analysis, vol., 23 (1963),
pp. 12123.
11. This is a very simple version of such a condition; there are also more complicated
versions.
12. A rather different sort of problem pertains to a much simpler anti-Gettier
condition that has sometimes been offered: the requirement that it not be an
accident (in relation to the justification) that the original belief is true. As
discussed above, the idea that the truth of the relevant belief is merely an accident
in relation to the justification is a good intuitive characterization of the various
Gettier-type examples, so this condition also might seem to work in the sense
of ruling out counterexamples though it may seem rather more obviously ad
hoc, amounting as it does to little more than the requirement that the case in
question not be a Gettier case. But a deeper difficulty for this condition is that
as long as the justification is less than conclusive, it isnt clear why it isnt always
to some degree an accident that the belief is true. To say that the justification
is less than conclusive is to say that there are some possible ways that the claim
in question could have been false even though that justification obtained. But
then why isnt it to at least some small degree accidental, in relation to that
justification, that one of these alternatives didnt actually occur? And to say
merely that the truth of the belief must not be too accidental, that there must
be a high though not complete degree of non-accidentality pertaining to it, is
to introduce a problem parallel to that which we saw concerning the degree
of justification required for the weak conception, and one that seems no more
tractable. (In relation to the anti-defeasibility requirement, we might also ask,
somewhat more speculatively: if justification is less than conclusive, wont there
always be truths which if carefully selected could defeat the persons justification
especially if it is just barely at the magic level? I do not know how to show
that this is so, but the suggestion strikes me as quite plausible. And if it is so, then
the solution in question, for which we could find no clear rationale anyway,
doesnt seem after all to work.)
13. Without the assumption of the weak conception, there is also no need to ignore
the pretty obvious possibility that the way in which truth can be accidental or
fortuitous may very well be a matter of degree, rather than the all or nothing
situation that an anti-Gettier condition has to assume.
14. Op. cit., pp. 23536, n. 21.
15. See Keith DeRose, Knowledge, Assertion, and Lotteries, Australasian Journal
of Philosophy, vol. 74 (1996), p. 570.
16. This was originally pointed out by Jonathan Vogel, in his paper Are There
Counterexamples to the Closure Principle? in M. Roth. and G. Ross (eds.),
17.
18.
19.
20.
21.
Doubting: Contemporary Perspectives on Skepticism (Dordrecht: Kluwer, 1990).

The point is usefully elaborated by Hawthorne, op. cit., pp. 37.
This is a slight variant of an example offered by Hawthorne, op. cit.
For an elaboration of such a criticism, see Dana Nelkin, The Lottery Paradox,
Knowledge, and Rationality, Philosophical Review, vol. 109 (2000), p. 385.
Ibid., section 5.
Thus this proposal is perhaps best regarded as a modification of the justification
condition rather than as a separate requirement, but this makes no real difference
for the issues that matter here. (Nelkin also sometimes puts the point by saying
that the persons belief must not be based on an inference from high statistical
probability to truth (what she calls a P-inference), but it is at least not obvious
that a person who adopts a claim on the basis of its high statistical probability
must be making such an inference to truth, rather than simply adopting the
claim because it is so very probably true.)
Contextualism is in a way an alternative to the weak and strong conceptions
and also offers a response to the lottery paradox, but I have no space here for
an extended discussion of that, to my mind, rather puzzling view. The central
contextualist claim is that the level of justification required for an attribution
of knowledge to be true varies from context to context (where the relevant
contextual factors pertain to the person attributing knowledge rather than to the
person to whom the knowledge is attributed). Thus in an ordinary context, where
issues of skepticism have not been raised, the required level of justification will
be rather low, low enough that many or most ordinary knowledge claims will be
true. In contrast, the explicit mentioning of skeptical possibilities allegedly has
the effect of dramatically raising the epistemic standard, so that a much higher
level of justification, which many or most ordinary knowledge attributions will
fail to satisfy, comes to be required. And the solution to the lottery paradox is
roughly that the explicit appeal to statistical evidence makes it salient that at least
one ticket that is very unlikely to win will in fact win and thereby has the effect
of again raising the epistemic standard required for knowledge. Contextualism is
not a version of the weak conception, but it does resemble the weak conception
in holding that there are many cases (even though these must be contextually
specified) where some less than conclusive level of justification is adequate to
satisfy the requirement for knowledge (that is, for the correct ascription of
know and knowledge). It thus seems to me to face the same problems,
already discussed in relation to the weak conception, of saying more precisely
just what those standards are and, much more importantly, of explaining why
they are intellectually significant: of explaining just what epistemic difference
reaching or exceeding such a standard (in the relevant context) is supposed to
make and why it is supposed to matter. As far as I can tell, the only answer
that is suggested by the contextualist view is simply that when the contextually
determined level of justification is reached, the terms know and knowledge
come to correctly apply in effect that the knowledge flag can correctly
be waved. And the question is why anyone should care about knowledge (or
knowledge) if no more than that is involved. (It does no good to say that
the difference is that the claim in question can now be properly asserted and
properly used in practical reasoning, since these facts are intuitively supposed
22.
23.
24.
25.
26.
27.
28.
to be consequences of some prior, purely epistemic difference.) I believe that a

similar objection can also be made to the sensitive moderate invariantist view
suggested by Hawthorne, op. cit., chapter 4.
Indeed, on a Moorean direct realist view, it presumably is conclusive, so that the
corresponding knowledge attribution is not in fact mistaken.
It should be added that this general sort of explanation may also combine with
the others to come: epistemological ignorance may make an exaggeration seem
less extreme and so more acceptable; may make the situation seem like a close
approximation to the correct standard when it is not; or may make it seem
some possibilities are unlikely enough to be disregarded or that some remaining
alternative has been conclusively established or ruled out, where this is not the
case. (Some of the other kinds of explanation may also combine with each other
in analogous ways, but I will not bother to specify these explicitly.)
Panayot Butchvarov, The Concept of Knowledge (Evanston, Ill.: Northwestern
University Press, 1970), pp. 5458. I have restated the point in terms of concepts
rather than words.
It is also worth noting that the a view like the one being suggested here can invoke
the salience of various possibilities of error, not to account for shifting standards
for the attribution of knowledge or for knowledge itself (as is done, in different
ways, by contextualism and by Hawthornes sensitive moderate invariantism),
but instead to explain when and why exaggerated knowledge attributions are
likely to be withdrawn. It could also appeal in a similar way to the practical
environment considerations discussed by Hawthorne.
Richard Fumerton, Epistemology (Oxford, Blackwell, 2006), pp. 2324.
The main proponent of this view is, of course, Timothy Williamson. See his op.
cit., chapter 11.
Relevant here is Williamsons intricate, elegant, often brilliant, but to my mind
also rather puzzling view of knowledge in his ibid., though I have no space in
this paper for a detailed discussion. Williamson rejects the idea that knowledge
can be defined or analyzed along the lines of the traditional definition or in
any other way. But his view seems nonetheless to amount to a version of the
weak conception, in that he takes knowledge to be achieved in very many cases
where the persons reasons or evidence yield only justification that is less than
conclusive, including cases of perception, memory, and inference to the best
explanation. What is extremely puzzling is that he then combines that view with
the further views that ones body of evidence is constituted by ones knowledge
and that evidence must be assigned a probability of 1. This seems to amount to
saying that a claim may become part of ones knowledge on less than conclusive
grounds, but then is subsequently to be treated as though those grounds were
conclusive so that its probability or likelihood is thus mysteriously ratcheted
up (and then perhaps used as the basis for further justification and similar
ratcheting up of other claims). Though the problem that Williamson is grappling
with here is perfectly real (how to reconcile inconclusive justification with
the subsequent employment of the claim in question, with no probabilistic
qualification, as evidence for further claims), his solution still strikes me as
incoherent: in effect, a curious and untenable amalgam of the weak and strong
conceptions.

29. The view that the concept of knowledge is simply incoherent would most
plausibly mean that such claims and assertions are neither false nor true, but
instead make no coherent claim.
30. G.E. Moore, Proof of an External World, reprinted in his Philosophical Papers
(London: Allen & Unwin, 1959).
31. I have also suggested, though have not attempted to argue in detail, that
contextualism is not an attractive alternative. See footnote 20.

Bonjour The Myth of Knowledge

Uploaded by

Copyright:

Available Formats

Bonjour The Myth of Knowledge

Uploaded by

Document Information

Original Description:

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Bonjour The Myth of Knowledge

Uploaded by

Copyright:

Available Formats

PHILOSOPHICAL PERSPECTIVES

Philosophical Perspectives, 24, Epistemology, 2010

THE MYTH OF KNOWLEDGE

It is my conviction that epistemology has, in recent times, gone seriously

Part I: Against the Weak Conception

The Myth of Knowledge / 59

Third, as just suggested, the relevant notion of justification, in contrast, is

2. The central problem for the weak conception

The Myth of Knowledge / 61

The Myth of Knowledge / 63

3. The Gettier problem

to the justification in question true in some unexpected way that is other

The Myth of Knowledge / 65

in an accidental or fortuitous way. But what is needed to generate a seemingly

4. The lottery paradox

The Myth of Knowledge / 67

The Myth of Knowledge / 69

or epistemic standpoint, why it should be thought to have the exalted evaluative

make truth more likely or probable. To be sure, the resulting probability or

Part II: A Tentative Defense of the Strong Conception

The Myth of Knowledge / 71

2. Explaining false knowledge attributions: some alternatives

The Myth of Knowledge / 73

3. Two further factors: assertion and practical reasoning

The Myth of Knowledge / 75

plausible in itself and more plausible given other considerations to suppose

4. Summary and tentative assessment

The Myth of Knowledge / 77

5. The implausibility of skepticism

The Myth of Knowledge / 79

and Belief, reprinted in Malcolm, Knowledge and Certainty (Englewood Cliffs,

The Myth of Knowledge / 81

Doubting: Contemporary Perspectives on Skepticism (Dordrecht: Kluwer, 1990).

to be consequences of some prior, purely epistemic difference.) I believe that a

The Myth of Knowledge / 83

You might also like