Journal of Philosophy, Inc.

Knowledge, Belief, and Rationality

Author(s): David B. Annis
Source: The Journal of Philosophy, Vol. 74, No. 4 (Apr., 1977), pp. 217-225
Published by: Journal of Philosophy, Inc.
Stable URL: http://www.jstor.org/stable/2025439
Accessed: 20-01-2016 13:10 UTC

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

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.

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

http://www.jstor.org

This content downloaded from 204.235.148.92 on Wed, 20 Jan 2016 13:10:05 UTC
All use subject to JSTOR Terms and Conditions
 KNOWLEDGE, BELIEF, AND RATIONALITY 217

fail to qualify as natural objects. Yet 'natural object' in the fa-

miliar philosophical use of this expression is true of these things.
As far as I can see, nothing stands in the way of describing the
myth of Pygmalion as a case in which 'natural object' and 'ma-
terial object' apply to an individual temporarily; hence we need
not recognize these two expressions as E terms. Since any general
term true of a persisting thing that is more general than 'material
object' would not be a descriptive term, we can justifiably conclude
that descriptive terms do not satisfy the doctrine N.
MARJORIE S. PRICE
New York City

KNOWLEDGE, BELIEF, AND RATIONALITY

M OST contemporaryphilosophers offering an analysis of
propositional knowledge accept the following require-
ment:
Entailment Thesis: Necessarily,for any person S, time t, and prop-
osition h, if S knows that h at t, then S believes that h at t.
Although there have been a number of attacks on this thesis, they
have not been persuasive.' The basic reaction to such attacks tends
to be: If S does not believe then he must not really know; other-
wise why would S not believe?
The problem is that philosophers who have attacked the entail-
ment thesis have not offered an account of the relation of knowl-
edge and belief which would explain our basic reaction. Well-
entrenched tenets, be they scientific or not, are rarely rejected, even
if they involve persistent anomalies, unless there is a competing
alternative to fill the void. What I propose to do then is to give
an alternative account of the relation of knowledge and belief
which will explain our basic reaction.
1 See for example Colin Radford, "Knowledge-By Examples," Analysis, xxvii,
I (October 1966): 1-11; Zeno Vendler, Res Cogitans (Ithaca, N.Y.: Cornell, 1972),
ch. v; D. W. Hamlyn, The Theory of Knowledge (Garden City, N.Y.: Double-
day, 1970), ch. 4; Joseph Margolis, Knowledge and Existence (New York: Oxford,
1973), ch. I. For more classical attacks, see Plato, Republic, Francis Cornford,
trans. (New York: Oxford, 1966), bk. v, 474b-480; J. Cook Wilson, Statement
and Inference (New York: Oxford, 1926), vol. I, ch. ii of part I and chs. i-iii
of part II; H. A. Prichard, Knowledge and Perception (New York: Oxford,
1950), pp. 80-94, 96-98; H. H. Price, "Some Considerations about Belief," Pro-
ceedings of the Aristotelian Society, xxxv (1935): 229-240.

This content downloaded from 204.235.148.92 on Wed, 20 Jan 2016 13:10:05 UTC
All use subject to JSTOR Terms and Conditions
2i8 THE JOURNAL OF PHILOSOPHY

I
The question "How do you know that h?" is a request for a state-
ment of the person's evidential position, that is, what puts the
person in a position to know. I have argued elsewhere that S's
nonbasic belief that h is epistemically justified if and only if there
is a set of propositions e1, e2,. . ., en such that
(1) For each e, (i = 1, 2,..., n) S believes that e,.
(2) For each e,, S is justified in believing ei.
(3) S believes that h on the basis of e1, e2, . . ., en.
(4) The set e1, e2,..., en provides adequate support for h.
(5) There is no other set of propositions el, e2',. . ., e,' such that S
believes the elements of the set and the conjunction of these ele-
ments with e1, e2, .. ., en does not provide adequatesupportfor h.2
I reject the further condition that, for S to be justified in believ-
ing h, S also must believe that the set e1, e2, .. ., en provides ade-
quate support for h. First, the set el-en may be extremely large,
and, hence, S may not be able to specify all the reasons for which he
believes. In this case, although S may believe each ej, he may not
explicitly believe the conjunction of them. This is not to say that
S disbelieves the conjunction, but only that he lacks the belief.
Second, whether or not certain beliefs can be imputed to a person
depends upon whether the individual has the necessary concepts.
Suppose that a four-year-old boy is given a toy for Christmas. We
ask him whose toy it is and he answers that it is his. When we
inquire where he got it, he responds that his parents gave it to
him for Christmas. If we consider most four-year-old boys and their
knowledge of their toys, it seems clear that the boy knows that the
toy is his. But he might very well not have the conceptual sophis-
tication necessary to have the belief that the evidence he has pro-
vides adequate support for the proposition that it is his toy. Thus,
although the boy has beliefs about the world, he may not yet have
beliefs about epistemic concepts. The empirical evidence on the
conceptual development of children supports this.3
Condition 3 above is a complex condition for which it is difficult
to provide an analysis. It is intended to ensure that there is a cer-
tain relationship between S's belief that h and his evidence. Hence
it excludes the following situation: There is a set of propositions
2 "Epistemic Justification," paper presented at the Eastern Division APA meet-
ings, December 1974, and published in Philosophia, VI, 2 (June 1976): 259-266.
3 For the best review of the literature see John H. Flavell, "Concept Develop-
ment," in Paul H. Mussen, ed., Carmichael's Manual of Child Psychology (New
York: Wiley, 1970), vol. I, pp. 983-1059.

This content downloaded from 204.235.148.92 on Wed, 20 Jan 2016 13:10:05 UTC
All use subject to JSTOR Terms and Conditions
 KNOWLEDGE, BELIEF, AND RATIONALITY 2I9

satisfying conditions 1, 2, 4, and 5, but these are not the reasons

for which S believes h. The set fi, f2' ... I, f, which are poor reasons,
are the reasons for which S believes. Here S is epistemically cul-
pable in that he has failed properly to appraise his evidence. If
el-en constitute good evidence for h, properly appraising the evi-
dence does not require S to believe that el-en provide adequate
support for h. A proper appraisal of the evidence can be mani-
fested by S reasoning correctly and not fortuitously from the evi-
dence to h.
The above analysis of justification has doxastic import; that is,
it entails that S believes that h. What is needed is a doxastically
neutral account of "S is justified in believing that h." To achieve
this we may replace condition 3 by its counterfactual analogue:
(3')If S were to believe that h, then his belief would be based on el,
e2 .C . . Ien.

(Alternatively this might be expressed as: If S were to believe that

h, then the reasons for which S would believe would be el-en.) The
counterfactual analogue is meant to capture the same epistemic
conditions relevant to h as the original analysis of justification ex-
cept for the condition of S's believing h. Hence in the analogue we
are not allowed to assume, for example, that at t (the time at which
the various conditions are supposed to be true) S fails properly to
appraise el-en even though he might come to believe that h. At t S
must properly appraise the evidence, as would be manifested in his
correctly and not fortuitously reasoning from his evidence to h and
being prepared to believe h on that basis.
A person's evidential position at a time with respect to the non-
basic statement h (a statement for which to be justified the person
must have evidence) is specified when the person is described in
terms of conditions 1, 2, 3', 4, and 5, that is, whether or not there
is a set of propositions el-en, such that the person believes them,
whether or not the person is justified in believing them, and so on.
A knowledge-giving evidential position for a person S at t with
respect to the nonbasic statement h is a position that justifies S in
believing h at t (that is, a position that satisfies the above five con-
ditions) so that, if the other conditions of knowledge obtained, S
would know that h. (I assume that these other conditions would at
least include the truth of h and S's evidential position being in-
defeasible.4) Hence, to answer the question "How do you know
4 For an excellent discussion of defeasibility see Marshall Swain, "Epistemic
Defeasibility," American Philosophical Quarterly, xi, 1 (January 1974): 15-26.

This content downloaded from 204.235.148.92 on Wed, 20 Jan 2016 13:10:05 UTC
All use subject to JSTOR Terms and Conditions
220 THE JOURNAL OF PHILOSOPHY

that h?" when h is nonbasic is simply to explain how one satisfies

our five conditions.
In the next section I shall use the notion of a person's evidential
position to explain the relation of knowledge and belief.
II
Consider two people, S1 and S2, who are in the same knowledge-
giving evidential position with respect to h, where S1 believes that
h and S2 lacks the belief. For example, suppose that Inspector
Smith is asked to investigate a murder. He discovers certain facts
(fingerprints, motive, opportunity, and so on) and reasons properly
from them to the conclusion that Jones killed the victim. As a
result of this Smith comes to believe that Jones is guilty (h). Con-
ditions 1, 2, 3', 4, and 5 are thus satisfied. If we assume that the
other conditions for knowledge obtain, then Smith knows that h.
Now suppose Jones Jr. is Smith's assistant. He too is aware of all
the evidence and that it indicates his father is guilty; that is, he
has correctly appraised the evidence. Nor does he have any counter-
evidence. Furthermore, if he were to believe h, then the reasons
for which he would believe it would be the evidence that both he
and Smith possess. Since they are in the same knowledge-giving
evidential position with respect to h and Smith knows, it seems
that Jones Jr. knows too. But suppose that he does not come to
have the belief that h. Here our reaction is: If he knows that his
father is guilty, he must believe it. Perhaps he is not aware of all
the facts; perhaps he has failed to see the force of the evidence;
maybe he has what he at least thinks is counterevidence. If he does
not believe, he must not really know; otherwise why would he not
believe?
How are we to explain our basic reaction? Is it that belief is a
necessary condition of knowledge? There is another explanation.
To be in Jones's evidential position and yet not to believe would
be supremely irrational from an epistemic point of view. (From a
moral point of view perhaps he ought to believe in his father's
innocence.) We are being asked to suppose that Jones is in the
same evidential position as someone who knows that h and yet
that Jones does not believe h. But to suppose that Jones does
not believe is to call into question his rationality.
A basic principle of rational belief is:

(PI) If S is in a knowledge-givingevidential position with respect to h,

then from an epistemic point of view S ought to believe that h;
that is, to withhold or disbelieve is irrational.

This content downloaded from 204.235.148.92 on Wed, 20 Jan 2016 13:10:05 UTC
All use subject to JSTOR Terms and Conditions
 KNOWLEDGE, BELIEF, AND RATIONALITY 221

Since we assume that an agent is rational, our only explanation of

why Jones does not believe is that he must not really know.
Another principle of rational belief is:
(P2)If S has the nonbasic belief that h, then from an epistemic point
of view S ought to have reasonsfor which he believes.
Since we assume that an agent is rational, we ask the question
"Why do you believe?" If the person answers that he has no
reasons, we are disturbed. Our response is: What do you mean you
have no reasons? You must have some reasons for believing that.
If we find that the person has no reasons, we are inclined to say
that he does not really believe.
Given the assumption that a person is rational and principle P2,
in a situation where a person has no reasons for his nonbasic be-
lief, we should reason as follows: If he is rational, he will have
reasons for his belief. He has no reasons; so he is not being rational
in this case. Instead we tend to reason in the following way: If he
is rational, he will have reasons for this belief. We assume that he
is rational; hence he must have reasons. But he has no reasons;
so he must not really believe. In a parallel fashion, given principle
PI and a situation where the person is in a knowledge-giving evi-
dential position with respect to h but where he fails to believe, we
ought to reason: If he is rational and in a knowledge-giving evi-
dential position, he will believe that h. Since he does not believe,
he is not being rational in this case. Instead we reason: If he is
rational, he will believe. We assume that he is rational; hence he
must believe. But he lacks the belief; so he must not really know
(that is, be in a knowledge-giving evidential position).
The assumption that an agent is rational is also made when we
assess people in other ways. Consider our concept of a rational
action. An action is rational relative to a person's goals if it con-
stitutes a reasonable choice of means for attaining his goals given
his beliefs. Suppose then that S has goal x and knows that y is the
way to obtain x. y is in S's power, no ill effects will result from
doing y, and so on. But suppose S does not do y. Again we find
this disturbing. We are inclined to say that x must not really be
his goal. If it were he would do y. Here we assume that S is ra-
tional, and, hence, our only explanation of his behavior is that x
must not really be his goal.
In the above cases, from the fact that S did not have reasons for
a nonbasic belief h or did not adopt those means optimal for
achieving goal x given his beliefs, we inferred that the person did

This content downloaded from 204.235.148.92 on Wed, 20 Jan 2016 13:10:05 UTC
All use subject to JSTOR Terms and Conditions
222 THE JOURNAL OF PHILOSOPHY

not really believe h or did not really have x as a goal. But in

neither case is having reasons for h or adopting those means to x
a necessary condition of believing h or having x as a goal. Instead,
they are conditions that a person must satisfy if the person is to
be rational. Similarly, if S is in a knowledge-giving evidential posi-
tion with respect to h and yet S fails to believe, we infer that S
must not really know that h. But belief is not a necessary condition
of knowledge; again it is a condition that S must satisfy to be
rational.
The strength of the assumption that an agent is rational is evi-
dent from the fact that rarely do we ever call into question a
person's rationality. In our examples above, what we instead called
into question was whether S really believed h, knew h, or had x as
a goal. We seek an explanation of a person's behavior consistent
with the assumption of rationality. So strong is this assumption
that it is frequently taken as a defining property of a human be-
ing. (Compare Boethius' "Persona est naturae rationalis individua
substantia.")
At this point one might wish to raise the objection that, in our
imagined cases, S, and S2 are not really in the same evidential
position. When we examine these cases closely we find that S2 fails
to satisfy all the conditions for being in a knowledge-giving evi-
dential position. For example, S2 will fail to believe all the ele-
ments of the set el-en, where the remaining evidence does not pro-
vide adequate support for h, or S2 will believe some proposition i
where the conjunction of i and el-en does not provide adequate
support for h.
The problem with this objection is that it commits us to saying
that there is only externally irrational belief, but no such thing as
internally irrational belief. That is, if S2 fails to believe h, this is
always due to his not believing part of the evidence or his having
counterevidence, so that from his evidential position it is rational
not to believe h (internally rational) whereas from our evidential
position it is irrational for S2 to fail to believe h (externally irra-
tional). But there is such a thing as internally irrational belief
(and action). For example, a person may believe that he is Napo-
leon even though his other beliefs do not support this. In this case
his belief is both externally and internally irrational. (The same
can be true of a person's actions.)
Another objection that might be made is that philosophers who
have accepted the entailment thesis have assumed that the agent

This content downloaded from 204.235.148.92 on Wed, 20 Jan 2016 13:10:05 UTC
All use subject to JSTOR Terms and Conditions
 KNOWLEDGE, BELIEF, AND RATIONALITY 223

is rational. For example, in the Theory of Knowledge 5 Roderick

Chisholm defines knowledge as evident true belief. He then de-
fines the evident in terms of one attitude being more reasonable
than another. And when he explains what he means by one atti-
tude being more reasonable than another, he includes the condi-
tion that the agent is rational. A weaker approach would be to
conditionalize the entailment thesis on the assumption of rational-
ity. But neither approach will succeed.
There are examples where a person's state of belief makes little
or no difference to our knowledge attributions. For example, some
"idiots savants" are able to answer complex mathematical questions,
and yet when it comes to explaining the meaning of a simple word
they fail. They may produce the answer to a mathematical ques-
tion by running around the room in circles while they beat their
heads against the wall. Suddenly they blurt out the answer. They
cannot explain how they do it; generally they cannot even explain
the simplest of mathematical operations. Ayer, Malcolm, Hamlyn,
Unger, and others all have said that in such cases the person has
knowledge.6 The principle for attributing knowledge in such cases
appears to be (roughly): If within a specific question-type domain
(certain types of mathematical questions, questions predicting the
winners of horse races, etc.) an individual is consistently correct,
then, after a sufficient consistency, the person is said to have knowl-
edge of the answers to questions within that specific domain. At
least part of the underlying reason for attributing knowledge in
such cases is that there is no other reasonable explanation for such
consistent success. Consistent correctness to questions of high spec-
ificity rules out chance as an explanation.
In such cases it seems to make no difference to our knowledge
attributions whether the person believes or not. If we are told that
the idiot savant does not believe, our reaction is not: If he does
not believe, he must not really know; otherwise why would he not
believe? In this case we expect irrationality, so we are not sur-
prised. If it is correct that these people know, then we cannot
define knowledge by including rationality. Nor can we condition-
alize the entailment thesis on the assumption of rationality. As-
sume that Jones is able to say what cards another person is turning
5 Englewood Cliffs, N.J.: Prentice-Hall, 1966; pp. 21-23.
6 A. J. Ayer, The Problem of Knowledge (London: Penguin Books, 1956),
pp. 32-35. Norman Malcolm, Knowledge and Certainty (Englewood Cliffs, N.J.:
Prentice-Hall, 1963), pp. 226-228. Hamlyn, Theory of Knowledge, pp. 81-82.
Peter Unger, "Experience and Factual Knowledge," this JOURNAL, LXIV, 5 (March
16, 1967: 152-173.

This content downloaded from 204.235.148.92 on Wed, 20 Jan 2016 13:10:05 UTC
All use subject to JSTOR Terms and Conditions
224 THE JOURNAL OF PHILOSOPHY

over with complete success. Again it makes no difference to our

knowledge attribution whether he believes or not. Furthermore, it
is not even clear that we could hold him as irrational for failing to
believe. The method or process whereby he gains his knowledge is
unknown. Hence he could be rational, know, and yet not believe.
When we evaluate a person's epistemic and doxastic state with
regard to a proposition h, we do so in the context of the follow-
ing assumption:
Presumption of Epistemic Normalcy: (a) The agent S is rational, that
is, satisfiesthe standardsof epistemic rationality (e.g., principles PI,
P2, and others as well), and (b) the agent S has no special cognitive
powers(as do idiots savantsor seers).
Conditions a and b define the normal case. The presumption is
similar to the Anglo-Saxon juridical principle that a person is
assumed innocent until proved guilty. When the question whether
someone knows arises, it is assumed that the person is rational and
has no special powers until otherwise proved. This is why we infer
from the fact that a person does not believe that he does not know.
However, when we discover that the person is not rational or that
he has special powers, we are willing to say that he knows even
though he may not believe.
It is due to this presumption that philosophers have been led to
suppose that knowledge entails belief. The entailment thesis seems
so obvious once we have made this assumption. If Smith knows
that h and Jones is in the same knowledge-giving evidential posi-
tion and yet does not believe, the only explanation of this, once
we have assumed normality, is that Jones must not really know.
Because philosophers have not paid attention to the larger context
in which we evaluate a person's epistemic and doxastic state, they
have been lead to assume the entailment thesis. But believing is
not constituent of knowing. The relation between knowledge and
belief is more complex than simple entailment. Instead it is a
normative relation governed by principles of rational belief and
what constitutes the normal case.

The strength of the account of the relation of knowledge and

belief that I have given is that it explains the same facts that the
entailment thesis does (e.g., our basic reaction to most of the de-
scribed cases of knowing and not believing, why it seems redundant
of say of a person, who knows, that he believes, etc.) and yet it
further explains facts that are inconsistent with the entailment

This content downloaded from 204.235.148.92 on Wed, 20 Jan 2016 13:10:05 UTC
All use subject to JSTOR Terms and Conditions
 BOOK REVIEWS 225

thesis. For example, it explains why a person's state of belief does

not affect our knowledge attributions in special cases and why we
are not surprised by a person's failure to believe something we say
he knows when we are aware of the person's general irrational
behavior.
One last word. We should not mourn the passing of the entail-
ment thesis, for its rejection does not wreak havoc with recent and
important investigations into such notions as justified belief and
defeasible justification. Knowledge still entails truth, the knower's
being in a knowledge-giving evidential position, and this evidential
position's being indefeasible. Furthermore these three conditions
entail knowledge.
DAVID B. ANNIS
Ball State University

BOOK REVIEWS
Truth and Meaning: Essays in Semantics. GARETH EVANS and JOHN
MCDOWELL, editors. Oxford: Clarendon Press. 1976. xxiii, 420 p.
$26.50.
There are thirteen numbered essays, besides a sixteen-page edi-
torial introduction and a twelve-page postscript to Wiggins by
Peacocke. They vary in length from eight or nine pages, at the
hands of Donald Davidson and Peter Strawson, to a characteristic
73 pages by Michael Dummett and 94 pages by Saul Kripke.
The other nine authors are J. A. Foster, Brian Loar, Christopher
Peacocke, Crispin Wright, Michael Woods, Barry Taylor, David
Wiggins, and both editors. Only Strawson's has been previously
printed.
To report on the excellent introduction would mean anticipat-
ing topics of the several essays; so I shall not pause over it except
for a minor criticism. There is disturbing evidence (xi, xiii) of a
budding misconception that the "lively appreciation of the native
speaker's ability to understand new sentences" is modern. It has
long been cited as what distinguishes genuine language from signal
systems, and a lively appreciation of it is reflected in the age-old
pattern of language analysis in terms of grammatical construction
and lexicon.
The most prominent topic in the volume is Davidson's use of
Tarski's truth definition in analyzing meaning. Sentences of the

This content downloaded from 204.235.148.92 on Wed, 20 Jan 2016 13:10:05 UTC
All use subject to JSTOR Terms and Conditions