Journal of Philosophy, Inc.

Two Conceptions of Knowledge

Author(s): Ernest Sosa
Source: The Journal of Philosophy, Vol. 67, No. 3 (Feb. 12, 1970), pp. 59-66
Published by: Journal of Philosophy, Inc.
Stable URL: http://www.jstor.org/stable/2023791
Accessed: 05-02-2017 23:02 UTC

JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of content in a trusted
digital archive. We use information technology and tools to increase productivity and facilitate new forms of scholarship. For more information about
JSTOR, please contact support@jstor.org.

Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at
http://about.jstor.org/terms

Journal of Philosophy, Inc. is collaborating with JSTOR to digitize, preserve and extend access to The
Journal of Philosophy

This content downloaded from 132.248.9.8 on Sun, 05 Feb 2017 23:02:47 UTC
All use subject to http://about.jstor.org/terms
 THE JOURNAL OF PHILOSOPHY
VOLUME LXVII, NO. 3, FEBRUARY I2, I970

COMMENTS AND CRITICISM

TWO CONCEPTIONS OF KNOWLEDGE

K NOWLEDGE of the nature of knowledge is deplorably

scarce. Fortunately, the reason is not lack of interest. On
the contrary, the bewildering variety of competing theories
is part of the problem. It is to, be hoped, however, that intensive dis-
cussion of such theories will help reduce the scarcity. In what fol-
lows I want to contribute to this end by briefly discussing two of
the theories.
Extending certain ideas of Chisholm's concerning the justification
of belief and its defeat, Keith Lehrer and Thomas Paxson, Jr. de-
velop an ingenious definition of propositional knowledge, well worth
discussing.* They first distinguish between "basic" and "nonbasic"
knowledge as follows. A belief is a piece of nonbasic knowledge if
and only if it is a case of knowledge because some other statement
justifies it. A belief is a piece of basic knowledge if and only if it is
knowledge and is not nonbasic. Basic knowledge is completely justi-
fied true belief, but nonbasic knowledge goes beyond that. For it re-
quires not only that the belief in question be true and completely
justified, but also that the justification be undefeated. Nonbasic
knowledge is undefeated justified true belief. Knowledge, finally, is
basic or nonbasic knowledge.
Obviously, a most important notion for this analysis is that of the
defeat of a justification. Following a detailed discussion of several
alternatives, the authors propose this definition: the complete justi-
fication that p provides S to believe h is defeated by q "if and only
if (i) q is true, (ii) the conjunction of p and q does not completely
justify S in believing h, (iii) S is completely justified in believing q
to be false, and (iv) if c is a logical consequence of q such that the

* Keith Lehrer and Thomas Paxson, Jr., "Knowledge: Undefeated Justified

True Belief," this JOURNAL, LXVI, 8 (April 24, 1969): 225-237.

59

This content downloaded from 132.248.9.8 on Sun, 05 Feb 2017 23:02:47 UTC
All use subject to http://about.jstor.org/terms
6o THE JOURNAL OF PHILOSOPHY

conjunction of c and p does not completely ju

then S is completely justified in believing c to be false" (231).
Some examples may help clarify this definition. Consider first the
case (228/9) of Tom Grabit who, S is sure, is the man who stole a
library book. S is not only sure of Tom's having taken the book,
moreover, but is also justifiably sure. However, unbeknownst to S,
Tom's mother has said that Tom was not in the library that day,
and that John Grabit, Tom's identical twin, was. (In fact Mrs.
Grabit is insane, and there is no John.) Here it is believed by
the authors that S's justification does not "depend" on his being
completely justified in believing Mrs. Grabit not to have said the
things in question, and it is also believed that, in consequence, the
fact of Mrs. Grabit's testimony does not defeat S's justification.
Justification is presumed defeated, however, when S believes that
Nogot, who is present, has a Ford, and infers that someone present
has a Ford, where although no one else present is thought by S to
own a Ford, Havit, who is also present, does in fact own one, and
Nogot does not. Here S's justification for believing that someone
present owns a Ford is supposed defeated by the fact that Nogot has
no Ford (229).
The relevant difference between the Grabit case and the Nogot
case is thought to lie in the fact that S's justification for believing
that someone present owns a Ford depends on his justification for
believing it false that Nogot has no Ford, whereas his justification
for believing that Tom Grabit took the book does not depend on
his justification for believing it false that Mrs. Grabit said the things
in question (229). And this difference accounts for clause (iii). Since
S is not completely justified in believing it false that Mrs. Grabit
said what she did, her having done so fails to defeat his justification
(for anything).
That concludes my exposition and clarification of the analysis of
knowledge under discussion. I hope that the foregoing examples
clarify it sufficiently to make the following assessment a fair one.
My first objection can be explained by reference to the Grabit
case. It turns out that, according to the present analysis, nothing
(q) can defeat S's complete justification (p) for believing (h) that
Tom took the book. For even if (i) q is true, even if (ii) the conjunc-
tion of p and q does not completely justify S in believing h, even if
(iii) S is completely justified in believing q to be false, yet it must
always be false that (iv) for every c, if c is a logical consequence of q
such that the conjunction of c and p does not completely justify S in
believing h, then S is completely justified in believing c to be false.

This content downloaded from 132.248.9.8 on Sun, 05 Feb 2017 23:02:47 UTC
All use subject to http://about.jstor.org/terms
 TWO CONCEPTIONS OF KNOWLEDGE 6i

For consider the proposition (c) that either q is true or Mrs. Grabit
said Tom was not in the library, etc. This c is a logical consequence
of q such that (in accordance with the author's hypotheses) the con-
junction of c and p does not completely justify S in believing h, but
S is not completely justified in believing c to be false, since by hy-
pothesis S is not completely justified in believing it false that Mrs.
Grabit said Tom was not in the library, etc.
The foregoing objection does depend on the assumption that if a
man is completely justified in believing a disjunction false, he is
completely justified in believing each disjunct false. But even if we
don't make this assumption there will be innumerable cases where
the present account of knowledge gives the wrong result. Suppose,
for example, that S's justification (p) for believing Tom took the
book depends on his justified belief that Tom now has a library
copy of the book (with the appropriate library markings, cardholder,
and so on). Yet, suppose further (q) that the copy Tom now has
happens to be a fake and is not the library copy. Even so, S's justifi-
cation need not be defeated. For consider the disjunction: either the
copy is not the library copy, etc., or Mrs. Grabit said that Tom was
not in the library, etc. This is a proposition c logically implied by q,
such that the conjunction of c with p does not justify S in believing
that Tom took the book, but we may surely suppose that S is not
completely justified in believing c to be false, since by hypothesis S
is not completely justified in believing it false that Mrs. Grabit said
Tom was not in the library, etc. Therefore, in the present case S has
undefeated justified true belief that Tom took the book. And yet
surely S does not know that Tom took the book, as the analysis
under review would have it. For S's complete justification for believ-
ing that Tom took the book depends essentially on his justified belief
that Tom now has a library copy of the book. But the copy Tom
has is by hypothesis a fake.
Let me emphasize that the foregoing type of objection does not
depend essentially on the particular example. Specifically, it does
not depend on the assumption (shared by the authors) that the con-
junction of S's justification with the fact of Mrs. Grabit's testimony
does not justify S in believing that Tom Grabit took the book. For
suppose any proposition r that someone S is not completely justified
in believing false, whose conjunction with S's justification for believ-
ing h does not justify S in believing h. The same problem will then
remain, that S's justification for believing h cannot be defeated or
at least may consistently be supposed undefeated in cases where this
should not be open to us. To surmount the present difficulty, there-

This content downloaded from 132.248.9.8 on Sun, 05 Feb 2017 23:02:47 UTC
All use subject to http://about.jstor.org/terms
62 THE JOURNAL OF PHILOSOPHY

fore, Lehrer and Paxson must reconsider their account of when a

justification is defeated. If so, however, it is to be hoped that they
will do so on the basis of a filled-out concept of complete justifica-
tion. In fact, the problem of explaining complete justification and,
particularly, the relation of justification-for-believing that applies
to ordered pairs of statements, in my view, takes precedence over
the more specific problems here detailed. It is true that the authors
do not wish to commit themselves to any substantive theory of
justification. And yet one wishes they had said more about their
relation of justification.
A second problem for the present account of knowledge is sug-
gested by the explanation of the difference between the Grabit and
Nogot cases given above. It is presumed by the authors that the im-
portant difference lies in the fact that S is completely justified in
believing it false that Nogot has no Ford but not completely justified
in believing it false that Mrs. Grabit said what she did. It is not dear
to me, however, that this a difference that is essential in explaining
why it is that S knows Tom to be the culprit but does not know that
someone present owns a Ford. For suppose that Mrs. Grabit is not
insane after all, that Tom does have an identical twin John, that
Mrs. Grabit honestly believes that John and not Tom was in the
library at the time and testifies to that effect, and that everyone else
concerned has heard of Mrs. Grabit's testimony, and has conse-
quently changed his mind about Tom. Surely it would not do to
grant S entitlement for knowledge that Tom took the book while
denying it to everyone else concerned, when they are supposed to
have all the data that S has plus the additional data concerning Mrs.
Grabit's testimony.' And yet this would be a consequence of the
account of knowledge under discussion. For, by the authors' hy-
pothesis, S is not completely justified in believing it false that Mrs.
Grabit said the things in question (and I presume we may add "to
the people in question," etc.). And this rules out her testimony to
the others as a defeater of S's complete justification for blaming
Tom.
A third difficulty emerges if we modify Skyrms' pyromaniac exa
1 See my "Analysis of 'Knowledge that p'," Analysis, xxv.l, 103 (October
1-8, esp. page 5, where it is argued that any adequate analysis of knowledg
account for this its "social aspect," as Harman has called it in agreement. S
"Knowledge, Inference, and Explanation," American Philosophical Quar
v, 3 (July 1968): 164-173; see esp. pp. 172-173. Actually, both Harman and
further in arguing for the "relevance of evidence one does not possess." S
paper by Harman forthcoming in Marshall Swain, ed., Induction, Accepta
and Rational Belief (Dordrecht, Holland: D. Reidel, Synthese Library), an
3-6 of my paper cited above.)

This content downloaded from 132.248.9.8 on Sun, 05 Feb 2017 23:02:47 UTC
All use subject to http://about.jstor.org/terms
 TWO CONCEPTIONS OF KNOWLEDGE 63

ple, endorsed by Lehrer and Paxson. "The pyromaniac has found

that Sure-Fire matches have always ignited when struck. On the basis
of this evidence, the pyromaniac is completely justified in believing
that the match he now holds will ignite upon his striking it." 8 Let
us suppose further that on the basis of the evidence derived from
his past experience he is also completely justified in believing that
upon the match's ignition he will smell powder when appropriately
situated and will feel pain on contact with the match, etc. But he
will not smell powder or feel pain, etc., since he will be struck with
temporary sensory paralysis. Presumably, if we take the proposition
(q) that he will not smell powder or feel pain, etc., even when ap-
propriately situated, and conjoin it with the pyromaniac's evidence
(p), the resulting conjunction (q 8c p) will not completely justify him
in believing (h) that the match will ignite. But should this preclude
his knowledge that the match will ignite, as it does according to the
present definition? 4
Because of the three problems indicated, I must conclude that the
account of knowledge under review is inadequate as it stands. Never-
theless, I am in agreement with its broad outline. In giving an ac-
count of knowledge it is surely correct to make allowance for a meta-
physical element (truth), a psychological element (something like
belief), and a sui generis epistemological element (warrant, for ex-
ample, or complete justification).
An alternative analysis that agrees in this general outline with that
of Lehrer and Paxson is the following:
S knows p if
(i) p is true;
(ii) S believes p;
(iii) p is evident to S; and
(iv) there is a set of propositions that
(a) fully renders p evident to S. and
(b) includes no subset that is epistemically defective with
respect to S and p.

[A set of propositions A fully renders evident a proposition p to

someone S iff A renders p evident to S, and for every proposition q
such that it is a member of A that q is evident to S, there is a subset
of A that renders q evident to S. (These sets are assumed closed un-
der entailment.) A set of propositions A is epistemically defective
2 Brian Skyrms, "The Explication of 'X knows that p'," this JOURNAL, LXIv, 12
(June 22, 1967): 373-389.
3 Lehrer and Paxson, p. 227.
41 suppose that in this particular case the third and fourth clauses of the
definition of defeat, given above on pp. 59/60, is fulfilled.

This content downloaded from 132.248.9.8 on Sun, 05 Feb 2017 23:02:47 UTC
All use subject to http://about.jstor.org/terms
64 THE JOURNAL OF PHILOSOPHY

with respect to S and p iff it is a (nonempty) set of propositions such

that, for some falsehood f, either A contains the proposition that f
is evident to S or A renders f evident to S but {p is evident to S}
does not. A set renders evident a proposition to a person iff the set
together with epistemic rules but not alone entails that the proposi-
tion is evident to that person. "Epistemic rules" are rules like "If a
proposition is evident and it is evident that its truth entails the
truth of a second proposition, then the second proposition is evi-
dent." Finally, to say that a proposition is evident to someone is to
say roughly this: that if he accepts that proposition he is then quite
justified or reasonable in doing so.] 5
A purported counterexample to this analysis was offered not long
after its appearance (ibid, 37/8). The details are tedious, but essen-
tially it is a case of a Miss Take. She remembers that at an earlier
time something looked black or something looked brown, in evi-
dently normal conditions, where no disconfirming evidence has ap-
peared in the interim. Assuming that in consequence it is now evi-
dent to her that something was then black or brown, and assuming
that something was indeed black or brown, must we not go further,
according to the present account, and say that she now knows that
something was then black or brown? If so, then so much the worse
for the account, inasmuch as we may suppose further that what then
looked black to Miss Take was in fact brown.
It seems to me, however, that this purported counterexample
misses the mark. For a falsehood f can after all be found such that
Miss Take's evidence renders f evident, whereas the set {that it is
evident to her that something was then black or brown} does not.
The falsehood f is the proposition that she then saw that something
was black or saw that something was brown.s
Nevertheless, it stills seems to me that a counterexample lies in the
vicinity. Let us now suppose that Miss Take remembers that at an
earlier time it was evident to her that something was black or
brown. Provided she knows that no disconfirming evidence has ap-
peared in the interim, is it not now evident to her that something
was then black or brown?
An affirmative answer is implied by a principle to the effect that
what is remembered to have been evident must now be evident pro-
vided it is clear that there is no new disconfirming evidence. And
some such principle appears required by much of our knowledge.
We forget the credentials of lecturers and contributors to encyclo-
5I propose this analysis in "Propositional Knowledge," Philosophical Studies,
xx, 3 (April 1969): 33-43; see especially pp. 35-36.)
6 Here I am indebted to Roderick Chisholm and Gilbert Harman.

This content downloaded from 132.248.9.8 on Sun, 05 Feb 2017 23:02:47 UTC
All use subject to http://about.jstor.org/terms
 TWO CONCEPTIONS OF KNOWLEDGE 65

pedias and the details of proofs and experiments, but we may re-
member that we had the best of reasons to believe something, and
this is often sufficient for present knowledge, even if we forget just
what the reasons were. Assuming I am right so far, isn't there in the
present case a set of true propositions (based on memory) that (a)
fully renders it evident to Miss Take that at t something was black
or brown, and (b) includes no subset that is epistemically defective
with respect to her and the proposition that at t something was
black or brown. But in that case, so long as Miss Take correctly be-
lieves it, she knows at t' that at t there was indeed something black
or brown present, even if at t this became evident to her only be-
cause it was evident to her that there was something black present
where in fact there was only something brown present. Or at least
I fear she knows it according to the present account.
What this shows is that we must make special provision for nested
epistemic modalities. The definition as it stands may be adequate
for cases where there is no nesting. But, as we have seen, it is not
always adequate for cases where there is nesting. (Actually revising
the definition is beyond the scope of this paper.)
A second difficulty with the definition under discussion arises if we
suppose that a proposition to the effect that some other proposition,
x, is evident itself renders evident a proposition y, and is rendered
evident by the proposition that y is evident. (Suppose, for example,
that x is that this triangle is equiangular and y is that this triangle
is equilateral.) Then anything z rendered evident by the proposi-
tion that x is evident is fully rendered evident by the set A {that x
is evident, that y is evident}. Supposing each of x, y, and z to be
evident but not self-evident, however, the existence of A does not
completely explain how it is that z is evident. For the question re-
mains how x and y come to be evident. Indeed, there may be no
nondefective set that fully renders evident either of x and z, in which
case the person in question can hardly be said to know z just because
he correctly believes it and A both is nondefective with respect to
him and z, and fully renders z evident to him.
The third and final difficulty is that a set may be defective because
it renders evident a falsehood in addition to fully rendering it evi-
dent to S that p, where the falsehood rendered evident is, in a sense,
epistemically irrelevant to the proposition p. Thus consider a case
where someone I know to be honest and reliable, etc., tells me that
she is looking in a mirror and sees herself to be incredibly beautiful.
I, a blind man, make two inferences: first, that someone honest and
reliable must indeed be incredibly beautiful, and, secondly, that

This content downloaded from 132.248.9.8 on Sun, 05 Feb 2017 23:02:47 UTC
All use subject to http://about.jstor.org/terms
66 THE JOURNAL OF PHILOSOPHY

someone honest and reliable is being very vain. I am right the sec-
ond time but wrong the first. Would this deny me knowledge in
both cases? Presumably not, despite the present account.
The problem here is that any set that at the moment renders it
evident to me that someone honest and reliable is being very vain
will also render it evident to me that someone honest and reliable
is incredibly beautiful, which is false in our (fortunately hypo-
thetical) example. The present account thus puts both her (pre-
sumed) beauty and her vanity beyond our ken.
These three difficulties make evident the inadequacy of the defi-
nition under review. Improvements are beyond the scope of the
present paper, which has been restricted to critical discussion.7
ERNEST SOSA

Brown University

KNOWLEDGE, CONSISTENT BELIEF, AND SELF-CONSCIOUSNESS

UST knowledge be self-conscious? Hintikka, among others,

has said Yes. In the system (KB) of Knowledge and Be-
lief,* the term 'Paq', which renders the English expression
'q is consistent with all that a knows', is partly defined by a prin-
ciple (APKK*) which requires that, if a set containing Kap [a knows
that p] and Paq is consistent, the set consisting of Kap and q must
be consistent. This principle, which Hintikka refers to as "perhaps
the most important rule" of his essay (17), is adopted in preference
to the weaker rule (APK*) which would require only that under the
above conditions the set consisting of P and q be consistent. A short
proof leads (cf. 105) from APKK* to the principle that a consistent
knower must know that he knows. For brevity I shall refer to this
principle as KK:
(KK) For any person a and proposition p, KaP > KKaP.a
7 It should be possible to develop the definition of "propositional knowledge,"
shown defective on three counts above, into a more adequate, intelligible, and
comprehensive account of propositional knowledge. The three difficulties must be
met, and others besides. I make the attempt in a forthcoming paper on "The
Concept of Knowledge."
For illuminating discussion, I am indebted to Roderick M. Chisholm, Gilbert
Harman. Keith Lehrer. and Robert J. Swartz.
* Ithaca, N.Y.: Cornell, 1962. Parenthetic references are to pages of this book.
1 In Hintikka's system the arrow is read 'virtually implies', which may be in-
formally explained as "implies on the assumption that describer and described
are perfectly consistent." The negation of a virtual implication is said to be
"indefensible," and whatever can consistently be said in a world of perfectly
consistent beings is "defensible."

This content downloaded from 132.248.9.8 on Sun, 05 Feb 2017 23:02:47 UTC
All use subject to http://about.jstor.org/terms