DEFINITIONS AND METHODOLOGICAL PRINCIPLES IN THEORY OF KNOWLEDGE

Author(s): Bertrand Russell

Source: The Monist , OCTOBER, 1914, Vol. 24, No. 4 (OCTOBER, 1914), pp. 582-593
Published by: Oxford University Press

Stable URL: https://www.jstor.org/stable/27900507

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
https://about.jstor.org/terms

Oxford University Press is collaborating with JSTOR to digitize, preserve and extend access to
The Monist

This content downloaded from

154.161.142.91 on Mon, 22 Jul 2024 11:43:26 UTC
All use subject to https://about.jstor.org/terms
 DEFINITIONS AND METHODOLOGICAL PRIN
CIPLES IN THEORY OF KNOWLEDGE.

IN the
nitionsfollowing article,
resulting from the analysis ofI experience
proposesugto suggest some defi
gested in previous articles, and some methodological pre
cepts which I believe to be useful in the later portions of
theory of knowledge. It was scarcely possible to state
either the definitions or the precepts until we had decided
that the relation of acquaintance is fundamental to expe
rience, for almost all definitions and principles that we can
adopt will presuppose this theory.
Subjects have been already defined in a previous article
as entities which are acquainted with something, and ob
jects as entities with which something is acquainted. In
other words, subjects are the domain, and objects are the
converse domain, of the relation acquaintance. It is not
necessary to assume that acquaintance is unanalyzable, or
that subjects must be simple ; it may be found that a further
analysis of both is possible. But I have no analysis to
suggest, and therefore formally both will appear as if they
were simple, though nothing will be falsified if they are
found to be not simple.
Cognitive facts will be defined as facts involving ac
quaintance or some relation which presupposes acquaint
ance.

It is to be observed that we can define cognitive facts,

not mental particulars, since we have no reason to assume

This content downloaded from

154.161.142.91 on Mon, 22 Jul 2024 11:43:26 UTC
All use subject to https://about.jstor.org/terms
 THEORY OF KNOWLEDGE. 583
that subjects are in fact always or ever of a different kind
from other particulars. The distinguishing mark of what
is mental, or at any rate of what is cognitive, is not to be
found in the particulars involved, but only in the nature
of the relations between them. Of these relations acquaint
ance appears to be the most comprehensive and therefore
the most suitable for the definition of cognitive facts.
There is, however, a certain difficulty in defining the
word "presuppose" which occurs in the above definition.
What is intended is to include such relations as (e. g.)
attention and judgment, which seem plainly to involve ac
quaintance with the object attended to or the objects enter
ing into the judgment. But it is not intended to include
relations which merely happen, as an empirical fact, to
occur only where there is acquaintance. In the cases which
it is intended to include, the fact that acquaintance is in
volved seems a priori evident, and not merely a statistical
occurrence. I do not know what, precisely, is the logical
difference between these two kinds of connection, but it
seems plain that there is a difference. The kind of con
nection intended in the definition is what we may call the
a priori kind; and since the difficulties of the distinction
belong solely to logic, we need not investigate them further
at present.
Theory of knowledge, or epistemology, is more difficult
to define. To begin with, no definition can be satisfactory
which introduces the word "knowledge," both because this
word is highly ambiguous, and because every one of its
possible meanings can only be made clear after much
epistemological discussion. Then again, difficulties arise
as regards the relations of epistemology to psychology and
to logic respectively. It is obvious that much of episte
mology is included in psychology. The analysis of experi
ence, the distinctions between sensation, imagination, mem
ory, attention, etc., the nature of belief or judgment, in

This content downloaded from

154.161.142.91 on Mon, 22 Jul 2024 11:43:26 UTC
All use subject to https://about.jstor.org/terms
584 THE MONIST.

short all the analytic portion of the subject, in so far as

it does not introduce the distinction between truth and
falsehood, must, I think, be regarded as strictly part of
psychology. On the other hand, the distinction between
truth and falsehood, which is plainly relevant to the theory
of knowledge, would seem to belong to logic, though this
is open to some degree of doubt. And in any case, as soon
as we reach the theory of judgment, even apart from truth
and falsehood, the difficulties encountered are almost en
tirely logical, and logical discoveries are what are most re
quired for the progress of the subject. It can be shown1
that a judgment, and generally all thought whose expres
sion involves propositions, must be a fact of a different
logical form from any of the series : subject-predicate facts,
dual relations, triple relations, etc. In this way, a difficult
and interesting problem of pure logic arises, namely the
problem of enlarging the inventory of logical forms so as
to include forms appropriate to the facts of epistemology.
It would seem, therefore, that it is impossible to assign to
the theory of knowledge a province distinct from that of
logic and psychology. Any attempt to mark out such a
province must, I believe, be artificial and therefore harmful.
The central problem of epistemology is the problem of
distinguishing between true and false beliefs, and of find
ing, in as many regions as possible, criteria of true belief
within those regions. This problem takes us, through the
analysis of belief and its presuppositions, into psychology
and the enumeration of cognitive relations, while it takes
us into logic through the distinction of truth and falsehood,
which is irrelevant in a merely psychological discussion of
belief. We may define epistemology in terms of this prob
lem, as : The analysis of true and false belief and their pre
suppositions, together with the search for criteria of true

*As I have come to know through unpublished work of my friend Mr.

Ludwig Wittgenstein.

This content downloaded from

154.161.142.91 on Mon, 22 Jul 2024 11:43:26 UTC
All use subject to https://about.jstor.org/terms
 THEORY OF KNOWLEDGE. 585
belief. But practically this definition is somewhat wide,
since it will include parts of psychology and logic whose
importance is not mainly epistemological ; and for this rea
son, the definition must not be interpreted quite strictly.
Data are by no means easy to define, since there are
various logically different kinds of data. The simplest
kind will be objects: in this sense, a given person's data are
those particulars with which he is acquainted. Even in
this very narrow sense, there are, as I shall try to show
in subsequent articles, still at least three ways in which
data may be given, namely in sense, in memory, or in imag
ination ; but these different ways of being given do not in
volve corresponding differences in the objects given. On
the other hand, we can understand universals, and any
primitive understanding of a universal (i. e., any under
standing not derived from some other understanding) is
in a sense a case of acquaintance, and is certainly suffi
cient ground for calling the universal in question a datum.
What is the correct analysis of the understanding of uni
versals is a difficult question, but such understanding must
certainly be included among the data of knowledge. Again,
perceptions of facts (as opposed to the corresponding judg
ments) must be included among data; if I see that one
thing is to the left of another, or if I observe that the thun
der is later than the lightning in a particular case, these
perceived facts are certainly among my data. And per
ceived facts are not always thus particular : general logical
facts, for example, are often such as can be perceived. I
do not know whether any judgments (as opposed to per
ceptions) are to be included among the data of knowledge:
arguments may be adduced on either side in this question.
But certainly all epistemologically legitimate inference de
mands that both the premise and the connection of premise
and conclusion should be data, either for perception or for
judgment.

This content downloaded from

154.161.142.91 on Mon, 22 Jul 2024 11:43:26 UTC
All use subject to https://about.jstor.org/terms
586 THE MONIST.

I think we may say broadly that data include all partic

ulars, universals, and facts, which are cognized otherwise
than by inference or by belief not derived from analysis
of a perceived fact. Thus the only cognitions which do
not give data are (a) beliefs obtained by inference, and
(b) beliefs not obtained either by inference or by analysis
of perceived facts. Beliefs of the second kind might be
fitly called prejudices: it will be a mere accident whether
they are true or false. Inferred beliefs, on the other hand,
if derived from data and not from prejudices, will be true.
But in these remarks we are anticipating the outcome of
lengthy discussions which cannot be undertaken at present.
There is another, somewhat vaguer, sense in which the
word "data" may be used in epistemology. If we are ana
lyzing some body of knowledge, say physics, with a view
to disentangling its truth and falsehood, we may give the
name "data" to all the beliefs which we entertain when
we enter on the inquiry. In this sense, "data" will be op
posed to "premises," and what we have hitherto called
data will be more fitly called premises. This usage is con
venient in some discussions, but for our present purposes
it seems not desirable.
It is a common error to suppose that it must be obvious
what are our data in any subject-matter to which we turn
our attention. This need by no means be the case, since
something may quite well be a datum without the fact of
its being a datum being also a datum. There is therefore
nothing surprising in the fact that it is often very difficult
to discover what our data really are.
I pass now to some methodological principles which
may be useful in our subsequent inquiries.
. The objects of acquaintance cannot be "illusory" or
"unreal" It is customary to speak of dreams and hal
lucinations as illusory, and to regard images as unreal in
some sense not applicable to the objects of normal sensa

This content downloaded from

154.161.142.91 on Mon, 22 Jul 2024 11:43:26 UTC
All use subject to https://about.jstor.org/terms
 THEORY OF KNOWLEDGE. 587
tions. No doubt there are important differences between
the objects commonly considered "real" and the objects
commonly considered "unreal." But for purely logical
reasons, these differences demand some entirely different
analysis. If an object is given in acquaintance, then that
object has a certain relation to the subject which is ac
quainted with it. But this would be meaningless if there
were no such object. An object of acquaintance is an
object to which it is possible to give a proper name, as op
posed to a description; it may become the "this" of at
tention. Of such an object, it is meaningless to suppose it
unreal. A proper name which names nothing is not a
proper name, but a meaningless noise. An acquaintance
which is acquainted with nothing is not an acquaintance,
but a mere absurdity. The" conception "unreal" is not
applicable to such immediate data, but only to described
entities. We may say "the present king of France is
unreal," meaning that there is no present king of France,
but not meaning "there is an entity which is at present
king of France and which is unreal." Unreality, in fact,
is a notion which owes its origin to grammatical forms.
These forms make it possible for the grammatical subject
of a sentence to be a description, which is not the name of
any constituent of the proposition expressed by the sen
tence. In such a case, if the description applies to nothing,
we may say that the subject is unreal; and if we fail to
realize that the unreal subject is merely grammatical, we
may come to think that there are entities which are unreal.
This, however, is a plain absurdity: given an entity, it is
absurd to inquire whether it is real or unreal: there the
entity is, and there is no more to be said. Thus the sup
position that some objects with which we are acquainted
are unreal is one which logic shows to be untenable.
The conclusion that no object of acquaintance can be
"unreal" forces upon us a certain attitude of respect

This content downloaded from

154.161.142.91 on Mon, 22 Jul 2024 11:43:26 UTC
All use subject to https://about.jstor.org/terms
588 THE MONIST.

towards dreams, hallucinations, and images. There are,

of course, important epistemological differences between
such objects and the objects of normal sensation, but these
differences belong to a much later and more complex part
of the subject. Their likeness to the objects of normal
sensation is much less realized, but is far more important
for our present purpose. Mankind has gradually selected
a kind of aristocracy among entities, which it calls the
"real" world. But, as Parmenides told Socrates, in phi
losophy we must not despise even the meanest things ; and
the entities which mankind has condemned as "unreal" are
full of interest and importance. It will be said : "No doubt
the things we see in dreams exist in a sense, but it is a
purely subjective sense: they exist only because we are
dreaming, and are merely parts of our mind." This may
or may not be true. I am not concerned at present with
the question whether it is true or false. I wish merely to
point out the great complexity of what is asserted, and
the immense system of knowledge which it presupposes.
It is asserted that the objects of dreams exist only because
we are dreaming ; this means, I suppose, that there is some
essential impossibility in such objects existing except when
some subject is aware of them. It is asserted that they
exist only in our mind; this means, I suppose, that not
more than one subject can possibly be aware of them. But
to know all this is to have an enormous knowledge of the
external world, and to see that it cannot harbor anywhere
such objects as I saw in my dream. And there is always
the fear that the same proposition might be maintained
regarding the things seen in waking life. If dreams are
to be given a lower status than waking life, as a source
of knowledge concerning the external world, this must be
done, not by condemning them ab initio, but by showing
that the world which is constructed on the assumption that
they and waking life have exactly the same immediate

This content downloaded from

154.161.142.91 on Mon, 22 Jul 2024 11:43:26 UTC
All use subject to https://about.jstor.org/terms
 THEORY OF KNOWLEDGE. 589
reality is one in which dream-objects are less intimately
related to other entities than are the objects of our waking
sensations. In fact, what is called the unreality of an
immediate object must always be really the unreality of
some other object inferred from the immediate object and
described by reference to it. This conclusion, we shall
find, is of great importance in all problems concerning our
knowledge of the external world.
2. The possibility of error in any cognitive occurrence
shows that the occurrence is not an instance of a dual rela
tion. This maxim is closely connected with our first maxim,
and equally has a purely logical origin. We saw that when
we are acquainted with an object, there certainly is such
an object, and the possibility of error is logically excluded.
Exactly the same argument would apply to any two-term
relation. But when the occurrence is one which unites the
subject with several objects in a single fact, the situation
is different. In this case, the several objects may or may
not be themselves the constituents of a certain related fact ;
thus a certain described fact, namely the fact composed, in
some specified manner, of the objects of the occurrence in
question, may occur in certain cases and not in others.
The further explanation and exemplification of this maxim
will come better in the account of belief. For the present,
its chief importance is negative: where error is possible,
something not acquaintance or attention or any other two
term relation is involved. This applies, for instance, to the
case of memory, in so far as memory is fallible.
3. The epistemological order of deduction involves both
logical and psychological considerations. It is obvious
that any order of valid deduction must involve logical
considerations, since the propositions deduced must follow
logically from their premises. But given a certain body
of propositions of which some are deducible from others,
there will be many ways of selecting some among them as

This content downloaded from

154.161.142.91 on Mon, 22 Jul 2024 11:43:26 UTC
All use subject to https://about.jstor.org/terms
 590 THE MONIST.

the premises from which the others are to be inferred.

Any such set of premises may be called a set of logical
premises for that body of propositions. In a purely logical
problem, in which we do not consider the question whether
any of the body of propositions are known to be true, con
ditions of elegance suggest that a set of logical premises
should be such that no one follows from the others, and
should consist of the smallest possible number of the sim
plest possible propositions. In epistemology, however, our
premises have to fulfil other requisites. We are concerned
not merely with truth, but with knowledge; hence our
premises should be such as can be known to be true without
being deduced from any others of the body of propositions
in question. For this reason, a certain artificial na?vet? is
required in beginning epistemology : we must avoid assum
ing many things which we firmly believe to be true, but
which can only be reached by a process of inference. On
the other hand, in order to minimize the risk of error,
which always exists in regard to premises, it is wise to in
clude whatever appears to be known without proof, even
if it could be deduced from other propositions so known.
Thus the set of epistemological premises for any body of
propositions consists of all those relevant propositions which
appear to be known otherwise than by inference, and from
which the given body of propositions can be deduced. If
some of the epistemological premises can be deduced from
others, that tends to confirm our belief in the premises, and
does not afford a reason for excluding those that can be
deduced. We may call a body of propositions epistemo
logically self-contained when it contains all its own episte
mological premises. The propositions constituting our
epistemology must, of course, be epistemologically self
contained? except that some of its premises belong to
logic; but those that belong to logic ought to be stated.
Since the question as to what appears to be known without

This content downloaded from

154.161.142.91 on Mon, 22 Jul 2024 11:43:26 UTC
All use subject to https://about.jstor.org/terms
 THEORY OF KNOWLEDGE. 59I
inference involves psychology, the epistemological order is
partly determined by psychological considerations. The
real business of epistemology is largely the discovery of its
own epistemological premises; the deductions from them
will of course proceed according to the rules of logic, and
may therefore be left to the logician.
4. A knowledge of physics and physiology must not be
assumed in theory of knowledge. This maxim follows
from the preceding account of the epistemological order.
Physics and physiology belong to our knowledge of what is
called the external world; our knowledge of them is ob
viously dependent upon sensation, and is obtained by meth
ods which the theory of knowledge must investigate. Thus
however firmly we may be convinced of the truth of phys
ics and physiology, we must only use this conviction as a
means of testing our epistemology. not as affording prem
ises upon which our epistemology may build from the start.
To take a crude illustration : if we are considering whether
or how the sense of sight gives knowledge of physical
objects, we must not assume that we know all about the
retina, for the retina is a physical object of which we obtain
knowledge by seeing it. Thus to assume that we know this
or that about the retina is to assume that we have already
solved the epistemological problem of the physical knowl
edge to be derived through sight.
It is to be observed, however, that it would not be a
logical fallacy to assume knowledge of the organs of sense
for a purely sceptical purpose, i. e., in order to show that
no knowledge of physical objects is to be derived from the
senses. If it could be shown that the hypothesis that we
obtain such knowledge through the senses leads to a theory
of the sense-organs which shows that they cannot give
such knowledge, then the hypothesis with which we started
would have refuted itself, and would therefore be false.
But although some such position is sometimes advanced.

This content downloaded from

154.161.142.91 on Mon, 22 Jul 2024 11:43:26 UTC
All use subject to https://about.jstor.org/terms
 592 THE MONIST.

and is not logically unsound, it is hardly possible that it

should be true in substance. The hypothesis that we ob
tain knowledge of physical objects through the senses is
one which is capable of a multitude of forms, according to
the view we adopt as to the nature of physical objects. If
one view on this point is self-refuting, another may not be,
and we can never be sure that we have tried all possible
views. Thus a dogmatic scepticism as to the knowledge of
physical objects derived from the senses can never be war
ranted by such arguments as we have been considering.
Nevertheless, the possibility is one which it is important to
remember, since it affords a test to which any theory as
to our knowledge of the external world may be submitted.
It is no doubt the case that our sensations are func
tions of the sense-organs and the nervous system, but this
is not the primitive knowledge?it is a scientific inference,
and cannot form part of the epistemological premises of
epistemology. In epistemology, it is important to reduce
our problems to what is actually experienced. Take, for
example, such a simple fact as that we no longer see objects
when we shut our eyes. A number of immediate expe
riences have to be combined in order to reach this result.
We can know from our own experience that visual objects
disappear when we have certain muscular and tactile sen
sations, which we learn to locate in our eyelids. By shut
ting one eye in front of a looking-glass we may learn
(assuming what we know of looking-glasses) what is the
visual appearance of the eyelid when we have the muscular
sensation in question. By seeing others shut their eyes,
and learning that then they no longer see, we can learn
without looking-glasses what shutting the eyes looks like.
But all this is a complicated process, learned, no doubt, at
an early age, but not known without learning. Thus the
dependence of sight upon the eyes is not the sort of thing
to be assumed in epistemology. And what is true of this

This content downloaded from

154.161.142.91 on Mon, 22 Jul 2024 11:43:26 UTC
All use subject to https://about.jstor.org/terms
 theory of knowledge. 593
very simple physiological fact is true a fortiori of the
complicated and difficult knowledge concerning the brain
which is sometimes thought relevant to the foundations of
epistemology. All such knowledge is excluded by the
rule which must guide us throughout our investigation:
Seek always for what is obvious, and accept nothing else
except as the result of an inference from something ob
vious which has been found previously. With this rule,
a great simplification is effected, and vast masses of erudi
tion can be swept aside as irrelevant.
Bertrand Russell.
Cambridge, England.

This content downloaded from

154.161.142.91 on Mon, 22 Jul 2024 11:43:26 UTC
All use subject to https://about.jstor.org/terms