Philosophy of Language (Alston)

WILLIAM P.
ALSTON
Philosophy
of
Language
UNOATIONS OF PHILOSOPHY SERIES

PRENTICE-HALL FOUNDATIONS OF PHILOSOPHY SERIES
Virgil Aldrich PHILOSOPHY Of ART
William Alston PHILOSOPHY OF LANGUAGE
Sl«ph«» Barker PHILOSOPHY OF MATHEMATICS
Roderick Chisholm THEORY OF KNOWLEDGE
William Dray PHILOSOPHY OF HISTORY
William Pironkona ETHICS
Carl Hempel PHILOSOPHY OF NATURAL SCIENCE
John Hick PHILOSOPHY OF RELIGION
Sidney Hook POLITICAL PHILOSOPHY
John ten* PHILOSOPHY OF EDUCATION
Richard Rudner PHILOSOPHY Of SOCIAL SCIENCE
Wesley Salmon LOGIC
Richard Taylor METAPHYSICS
Elizabeth and Monroe Beardsley, editors

Philosophy of Language
4
PHILOSOPHY
OF LANGUAGE
: * ' O - , 1 , . * > W ' * . ' i *s ' K t $ E i 1 E S

William P. Alston
University of Michigan
PRENTICE-HALL, INC.
© Copyright 1964
by PRENTICE-HALL, INC.
Englewood Cliffs, N. J.
All rights reserved. No part
of this book may be reproduced in any form,
by mimeograph or any other means,
without permission in writing from the publishers. Printed
in the United States of America.
Library of Congress
Catalog Card No.: 64-19006
PHILOSOPHY OF LANGUAGE, Alston
FOUNDATIONS Of PHILOSOPHY SERIES
C-66379
PRENTICE-HALL INTERNATIONAL, INC, London
PRENTICE-HALL OF AUSTRALIA, PTY., LTD., Sydney
PRENTICE-HALL OF CANADA, LTD., Toronto
PRENTICE-HALL OF INDIA (PRIVATE! LTD., New Delhi
PRENTICE-HALL OF JAPAN, INC., Tokyo
PRENTICE-HALL DE MEXICO, S.A., Mexico City

To Valerie
without whom not

foun datio n s
OF PHILOSOPHY
Many of the problems of philosophy are of such broad relevance

to human concerns, and so complex in their ramifications, that they
are, in one form or another, perennially present. Though in the course
of time they yield in part to philosophical inquiry, they may need to be
rethought by each age in the light of its broader scientific knowledge
and deepened ethical and religious experience. Better solutions are
found by more refined and rigorous methods. Thus, one who approaches
the study of philosophy in the hope of understanding the best of what
it affords will look for both fundamental issues and contemporary
achievements.
Written by a group of distinguished philosophers, the Founda¬
tions of Philosophy Series aims to exhibit some of the main problems
in the various fields of philosophy as they stand at the present stage of
philosophical history.
While certain fields are likely to be represented in most intro¬
ductory courses in philosophy, college classes differ widely in emphasis,
in method of instruction, and in rate of progress. Every instructor needs
freedom to change his course as his own philosophical interests, the
size and makeup of his classes, and the needs of his students vary from
year to year. The thirteen volumes in the Foundations of Philosophy
Series—each complete in itself, but complementing the others—offer a
new flexibility to the instructor, who can create his own textbook by
combining several volumes as he wishes, and can choose different com¬
binations at different times. Those volumes that are not used in an
introductory course will be found valuable, along with other texts or
collections of readings, for the more specialized upper-level courses.
ELIZABETH BEARDSLEY MONROE BEARDSLEY

*
. '
'
PREFACE
Though the philosophy of language might reasonably be thought

of as comprising anything that philosophers do when they think, qua
philosophers, about language, I have not attempted a survey of that
heterogeneous field of activity. Instead, I have presented the philosophy
of language in one of its guises, as an attempt to get clear about the
basic concepts we use in thinking about language. (This contrasts with
thinking of the philosophy of language as an attempt to exhibit lan¬
guage as one of the forms of the world spirit, or as an attempt to provide
an over-all synthesis of conclusions about language, arrived at in the
various social sciences.) As so conceived, the philosopher of language
tries to determine, for example, what language is and how it is related to
more or less analogous forms of activity (Chapter 3), what it is for a
linguistic expression to be meaningful (Chapter 4), what it is for a
linguistic expression to have a certain meaning (Chapters 1 and 2),
what it is for a linguistic expression to be vague or to be used metaphori¬
cally (Chapter 5). I have chosen to proceed primarily by discussing
these problems according to my lights, rather than by enumerating and
classifying alternative positions, though I hope that my discussion has
maintained contact with many of the significant ideas to be found in the
literature. With respect to what I take to be the central problem, “What
is it for a linguistic expression to have a certain meaning?”, I have de¬
voted almost all of the first chapter to a critical review of the most
prominent positions. I should also add that I have approached these
problems in a relatively informal manner, in contrast with logistically
minded theorists, like Carnap, who seek illumination by the construction
of formalized schemata for simplified languages.
XI
xii Preface
I am extremely indebted to the editors of this series, Monroe and

Elizabeth Beardsley, and to William Frankena and George Nakhnikian,
as well as to my wife, Valerie, all of whom read the manuscript in an early
and overinflated form and made many helpful suggestions. More gen¬
erally, I am indebted to all those students and colleagues with whom I
have discussed semantical topics over the years, particularly Richard and
Helen Cartwright, Paul Henle, Julius Moravcsik, Kenneth Pike, John
Searle, J. O. Urmson, Paul Ziff, and most particularly David Shwayder,
with whom ideas about language have been exchanged to the point that
only considerations of charity inhibit me from attributing the whole
production to him. Finally, I should like to express appreciation to Mrs.
Alice Gantt for prompt and expert typing.
WILLIAM P. ALSTON
CONTENTS
Introduction, 1 Sources of the philosopher’s concern with language: metaphysics, 1.

Logic, 3. Epistemology, 4. Reform of language, 5. Philosophy as
analysis, 6. Problems of the philosophy of language, 8.
Theories of Meaning, 10 The problem of meaning, 10. Types of theories of meaning, 11. The
referential theory, 12. Meaning and reference, 13. Do all meaningful
expressions refer to something? 14. Denotation and connotation, 16.
Meanings as a kind of entity, 19. The ideational theory, 22. Mean¬
ing as a function of situation and response, 25. Meaning as a function
of behavioral dispositions, 28. Summary of discussion of behavioral
theory, 30.
Meaning and the Use of Meaning as a function of use, 32. Types of linguistic action, 34.
Language, 32 Word meaning, 36. Analysis of illocutionary acts, 39. Rules of lan¬
guage, 41. Problems concerning synonymy, 44. Emotive meaning,
47. Problems about illocutionary acts, 48.
xiii
Language and Its Near The generic notion of a sign, 50. Regularity of correlation and regu-
Relationt, 50 larity of usage, 54. Icon, index, and symbol, 55. The notion of con¬
vention, 56. Icons, pure and impure, 58. Language as a system of
symbols, 59.
Empiricist Criteria ef Meaningless sentences, 62. Traditional form of an empiricist criterion,

Meaningfulness, 62 63. The semantic stratification of language, 67. Logical atomism,
68. Verifiability theory of meaning, 69. Deficiencies in usual formu¬
lations of verifiability criterion, 73. Problems in the formulation of the
verifiability criterion, 76. Verifiability criterion as description and as
proposal, 77. Arguments in support of verifiability criterion, 79. Final
assessment, 82.
Dimensions of What vagueness is, 84. Kinds of vagueness: degree and combination
Meaning, 84 of conditions, 87. Is absolute precision possible? 90. Precision through
quantiEcation, 91. Open texture, 93. Importance of the notion of
vagueness, 95. Metaphorical and other Egurative uses of expressions,
96. The nature of metaphor, 98. Basis of the literal-metaphorical
distinction, 99. Irreducible metaphors: God and inner feelings, 103.
For Further
Reading, 107
Index,111
I NTR DDUCTIDN
The philosophy of language is even less well-defined and less in

possession of a clear principle of unity than most other branches of
philosophy. The problems concerning language that are typically dealt
with by philosophers constitute a loosely knit collection, for which it is
difficult to find any clear criterion separating it from problems concern¬
ing language dealt with by grammarians, psychologists, and anthropol¬
ogists. We can get an initial sense of the range of this collection by
surveying the various points within philosophy at which a concern
with language emerges.
Sources of the First, consider the ways in which problems concerning language
philosopher’s crop up in the various branches of philosophy. Metaphysics is a
concern with part of philosophy roughly characterizable as an attempt to for-
language: mulate the most general and pervasive facts about the world,
metaphysics including an enumeration of the most basic categories to which
entities belong and some depiction of their interrelations. There
have always been philosophers who have tried to get at some of these
fundamental facts by considering the fundamental features of the lan¬
guage we use to talk about the world. In Book X of Plato’s Republic,
we find him saying, “Whenever a number of individuals have a common
name, we assume them to have also a corresponding idea or form.”
(596) To spell out this rather cryptic remark, Plato is calling our atten¬
tion to a pervasive feature of language, that a given common noun or
adjective, for example, ‘tree’ or ‘sharp,’ can be truly applied in the same
sense to a large number of different individual things; his position is
that this is possible only if there exists some one entity named by the
general term in question—treeness, sharpness—of which each of the
J
4
2 Introduction
individuals partakes. If this were not the case, it would be impossible

for the general term to be applied in the same sense to a number of
different individuals.
Again, we find Aristotle in his Metaphysics arguing as follows:
And so one might even raise the question whether the words ‘to walk,’
‘to be healthy,’ ‘to sit,’ imply that each of these things is existent, and
similarly in other cases of this sort; for none of them is either self-
subsistent or capable of being separated from substance, but rather, if
anything, it is that which walks or sits or is healthy that is an existent
thing. Now these are seen to be more real because there is something
definite which underlies them (i.e., the substance or individual) which
is implied in such a predicate; for we never use the word ‘good’ or ‘sit¬
ting’ without implying this. (Book Zeta, Chapter 1)
Here Aristotle starts from the fact that we do not use verbs except in
connection with subjects, that we do not go around saying ‘Sits/ ‘Walks/
etc., but rather, ‘He is sitting,” or ‘She is walking.’ From this fact he
concludes that substances, “things,” have an independent kind of
existence in a way that actions do not, that substances are more funda¬
mental ontologically than actions.
A more outre example can be found in the late nineteenth century'
German philosopher, Meinong, who started with the assumption that
every meaningful expression in a sentence (at least any meaningful
expression that has the function of referring to something) must have
a referent; otherwise, there would be nothing for it to mean. Hence,
when we have an obviously meaningful expression that refers to nothing
in the real world, for example, ‘the Fountain of Youth,’ in the sentence,
De Soto was searching for the Fountain of Youth,’ we must suppose
that it refers to a subsistent” entity, which does not exist but has some
other mode of being. This doctrine, as well as the Platonic position
presented above, is based on a confused assimilation of meaning and
reference, which we shall try to straighten out in the first chapter.
The assumption behind these patterns of metaphysical argumenta¬
tion has been made quite explicit in the twentieth century philosophical
movement known as logical atomism, the most distinguished exponents
of which have been Bertrand Russell and Ludwig Wittgenstein (in his
earlier period). In Russell’s series of articles, “The Philosophy of Logical
Atomism, he makes the principle quite explicit. ^
. . . m a logically correct symbolism there vail always be a certain

fundamental identity of structure between a fact and the symbol for
it; and . . . the complexity of the symbol corresponds verv closely with
the complexity of the facts symbolized by it.1
1956and Knowledge’ ed' R' C- Marsh (London: George Allen & Unwin, Ltd.,
Introduction 3
Note that this identity of structure is postulated to hold not between

any existing language and the basic metaphysical structure of the world,
but only between a “logically perfect language” and the metaphysical
structure. The assumption is that when we have devised such a language,
or have acquired at least a sketchy idea of what such a language would
be like, we will then be able to draw various conclusions concerning
the types of facts of which reality is made and the structure of each
of these facts. We will ascertain what different types of sentences we
have in that language for asserting facts, for example, simple subject-
predicate sentences like ‘This book is heavy’ and existential sentences
like ‘There is a cat on the porch’; and we will see how these various
types of sentences are logically related. This will tell us what the basic
sorts of facts are of which reality is made and how facts of these
various sorts are interrelated.
Logic Another branch of philosophy in which concern with language

becomes prominent is logic. Logic is the study of inference, more
precisely the attempt to devise criteria for separating valid from invalid
inferences. Since reasoning is carried on in language, the analysis of
inferences depends on an analysis of the statements that figure as prem¬
ises and conclusions. A study of logic reveals the fact that the validity
or invalidity of an inference depends on the forms of the statements that
make up the premises and conclusion, where by ‘form’ is meant the
kinds of terms the statements contain and the way in which these
terms are combined in the statement. Thus, of two inferences that
superficially look very much alike, one may be valid and the other
invalid because of a difference in the form of one or more of the
statements involved. Consider the following pair of inferences.
1. Joe Carpenter sells insurance in our town.
Joe Carpenter belongs to the First Methodist Church.
► Therefore, Joe Carpenter both sells insurance in our town and belongs
to the First Methodist Church.
2. Someone sells insurance in our town.
Someone belongs to the First Methodist Church.
► Therefore, someone both sells insurance in our town and belongs to
the First Methodist Church.
Now 1 is clearly a valid argument, and 2 is clearly invalid. Given the
facts that someone sells insurance in this town and that someone be¬
longs to the First Methodist Church, it does not at all follow that there
is anyone of whom both these things are true. Since one of these argu¬
ments is valid and the other invalid, it must be that despite superficial
grammatical similarities, a sentence like a. ‘Joe Carpenter sells insurance
in our town’ is of a very different logical form from a sentence like
*
4 Introduction
b. 'Someone sells insurance in our town.’ There are other indications of

this. Sentence b is equivalent to ‘There is someone who sells insurance
in our town’ and to ‘The class of persons who sell insurance in our town
is not empty,’ but we can find no such equivalents for sentence a. When
the premises and conclusion of inference 2 are put into one of these
forms, the argument loses its superficial resemblance to inference 1 and
does not look valid at all.
3. There is someone who sells insurance in our town.

There is someone who belongs to the First Methodist Church.
► Therefore, there is someone who both sells insurance in our town and
belongs to the First Methodist Church.
It is clear from such examples that an important part of logic consists

in a classification of statements in terms of their “logical” form (that
is, aspects of form that are relevant to the evaluation of inference).
And this classification in turn requires a classification of the types of
terms that enter into statements, for a difference in form quite often
rests upon a difference in the types of terms involved. In the preceding
example, the difference in logical form between sentences a and b rests
upon a fundamental difference between a proper name like ‘Joe Car¬
penter,’ which has the function of picking out a particular individual,
and a locution like ‘someone,’ which has quite a different function.
Epistemology The branch of philosophy known as Epistemology or Theory of

Knowledge becomes concerned with language at a number of
points, the most prominent of which is the problem of a priori knowl¬
edge. We have a priori” knowledge when we know something to be
the case without this knowledge being grounded on experience. It
seems that we have knowledge of this sort in mathematics, and perhaps
in other areas as well; and the fact that we do have such knowledge
has often seemed puzzling to philosophers. How is it that we are able
to know with certainty, apart from observation, measurement, etc., that
the angles of a Euclidian triangle all together equal 180 degrees’ and
that 8 plus 7 always and invariably equals 15? How can we be’ sure
that no experience will ever falsify these convictions? One answer that
has often been given is that in such cases what we are asserting is true
by definition, or true by virtue of the meanings of the terms involved
That is, it is part of what we mean by ‘8,’ 7,’ ‘15/ ‘plus,’ and ‘equals,’
that 8 plus 7 equals 15; and to deny this statement seriously would
involve changing the meaning of one or more of these terms. The
adequacy of this account of a priori knowledge is and has been the
subject of considerable controversy; but whether or not the position is
justified, it is clear that even in taking it seriously, we are inevitably led
Introduction 5
into questions concerning what it is for a term to have a certain

meaning and how a statement can be true by reason of the fact that
certain terms have the meaning they do.
Reform of There are also philosophical motives for concern with language,
language which have to do not with the problems of one or another branch
of philosophy but with kinds of activity into which philosophers
are typically led in many branches of the subject. One of these is the
reform of language. Thinkers in many fields are given to complaining
about the deficiencies of language, but philosophers have been more
preoccupied with this sort of problem than most, and for good reason.
Philosophy is a much more purely verbal activity than is a science
that collects facts about chemical reactions, social structures, or rock
formations. Verbal discussion is the philosopher’s laboratory, in which
he puts his ideas to the test. It is not surprising that the philosopher
should be especially sensitive to flaws in his major instrument. Philo¬
sophical complaints about language have taken many forms. There are
the philosophers of mystical intuition, such as Plotinus and Bergson,
who regard language as such to be unsuitable for the formulation of
fundamental truth. From this standpoint, one can really apprehend
truth only by some wordless union with reality; linguistic formulations
give us at best only more or less distorted perspectives. But more often
philosophers have not been willing to abjure talking, even in theory.
Complaints have usually been levelled against some current state or
condition of language, and the implication is that steps could be taken
to remedy this condition. These philosophers can usefully be divided
into two groups. There are those who hold that “ordinary language,”
the language of everyday discourse, is perfectly suitable for philosophical
purposes, and that the mischief lies in deviating from ordinary language
without really providing any way of attaching sense to the deviation.
We find examples of this sort of complaint here and there in the
history of philosophy, for example, in Locke’s complaints against
scholastic jargon; however, it is in our own day that such complaints
have become the basis of a philosophical movement, “ordinary lan¬
guage philosophy.” In its strongest form, as we find it in the later
work of Ludwig Wittgenstein, it maintains that all, or at least most, of
the problems of philosophy stem from the fact that philosophers have
misused certain crucial terms, such as ‘know,’ ‘see,’ ‘free,’ ‘true,’ and
‘reason.’ It is because philosophers have departed from the ordinary
use(s) of these terms without putting anything intelligible in their
place that they have fallen into insoluble puzzles over whether we can
know what other people are thinking or feeling, whether we ever really
directly see any physical object, whether anyone ever acts freely, and
<
6 Introduction
whether we ever have any reason to suppose that things will happen in
one way rather than another in the future. According to Wittgenstein,
the role of the philosopher who has seen this point is that of a therapist;
his job is to remove the conceptual cramps” into which we have
fallen.
Second, there are those who hold, by contrast, that the trouble
comes from the fact that ordinary language itself is inadequate for
philosophical purposes, by reason of its vagueness, inexplicitness, am-
biguity, context-dependence, and misleadingness. These philosophers,
such as Leibniz, Russell, and Carnap, see as their task the construction
of an artificial language, or at least the adumbration of such a language,
in which these defects will be remedied. As was pointed out previously,
this enterprise is sometimes enlivened by the conviction that from the
structure of such a language one can read off basic facts about the meta¬
physical structure of reality.
For our purposes, the main interest of these complaints and
schemes for reform lies in the way in which general conceptions of
language and meaning are involved in them. Even the mystical position
presupposes some notion of the nature of language; otherwise, one could
have no basis for holding language as such to be incapable of serving
as an adequate formulation of truth. The other positions necessarily
involve more positive conceptions of the conditions under which lan¬
guage is meaningful and is performing its functions adequately. Thus,
the verifiability criterion of meaningfulness, to which we shall devote
most of a chapter, grows out of a position of the sort last discussed.
Philosophy as The final point concerns the notion that the primary, if not the
analysis whole, job of philosophy is conceptual analysis. The analysis of
basic concepts has always been a major concern of philosophers.
In the Dialogues of Plato, Socrates is represented as spending a great
deal of his time asking questions like “What is justice?” and “What is
knowledge? Aristotle’s works are, in large part, concerned with at¬
tempts to arrive at adequate definitions of terms like 'cause,’ ‘good,’
motion, and know.’ Traditionally, it has been felt that however im¬
portant this activity was, it was still a preliminary to the ultimate tasks
of the philosopher—those, of arriving at an adequate conception of the
basic structure of the world and an adequate set of standards for human
conduct and social organization. But in our time there has been a grow¬
ing conviction that the method used in philosophy, which may be
briefly defined as armchair reflection unsupplemented by special obser¬
vation or experimentation, does not really suffice to yield any substantive
conclusions concerning the nature of the world or the conditions under
which life is lived well or ill, and that what it is fitted to produce is
Introduction 7
clarity and explicitness with respect to the basic concepts in terms of

which we think about the world and human life. This massive shift
in the center of gravity of philosophical activity is of particular relevance
to the philosophy of language because of an accompanying shift in the
conception of conceptual analysis itself. There are three importantly
different ways of formulating a problem in analytical philosophy,
whether we are dealing with causation, truth, knowledge, or moral
obligation. To take the problem of knowledge for our model, we can
say that 1. we are investigating the nature of knowledge, 2. we are
analyzing the concept of knowledge, or 3. we are trying to make
explicit what one is saying when he says that he knows something to be
the case. 1 and 2 are likely to be misleading methodologically. 1 suggests,
falsely, that the task is one of locating and inspecting some entity called
'knowledge/ an entity that exists and is what it is independent of our
thought and discourse. Unfortunately, no one has ever provided an
acceptable technique for locating and examining such entities. 2 is apt
to be misleading unless it is recognized to be simply an alternative form
of 3, for it suggests that the task is one of introspectively scrutinizing
something called a concept and discovering the parts of which it con¬
sists and the way they are put together. Again, it does not seem to be
possible to develop an objective technique for doing any such thing.
The realization has grown that even when a philosopher dealing with
knowledge formulates his problem as 1 or 2, what he really does, insofar
as his results have any value, is to reflect on various features of the use
of 'know’ and its cognates.
Thus to the extent that philosophy consists of conceptual analysis,
it is always concerned with language. And if it is either all or a large
part of the philosopher’s business to bring out features of the use or
meaning of various words and forms of statement, it is essential for
him to proceed on the basis of some general conception of the nature
of linguistic use and meaning. This becomes especially important when
analytical philosophers become involved in persistent disputes over what
a given word means, or over whether two expressions or forms of
expression have the same or different meaning. There are serious dis¬
agreements in analytical philosophy over whether 'I know that p’ means
the same as 'I believe that p, I have adequate grounds for this belief,
and p is the case’; whether ‘A is the cause of B’ means simply that A
and B are, in fact, regularly conjoined; whether 'feels sad’ means the
same in 'I feel sad’ and ‘He feels sad’; and whether any theoretical
statement in science can have the same meaning as some combination
of reports of observation. When such disputes are not settled by our
intuitive sense of what linguistic expressions mean, the philosopher is
forced to develop some explicit theory of what it is for a linguistic
*
8 Introduction
expression to have a certain meaning and of the conditions under which

two expressions will have the same meaning. Thus, insofar as philoso¬
phy is thought of primarily as conceptual analysis, the philosophy of
language occupies a central position in the theory of philosophical
method.
Problems of Having seen some of the points in the more central portions of
the philosophy at which one is naturally led to an explicit considera-
phiiosophy of tion of problems concerning language, we can proceed to a brief
language preliminary survey of these problems. As I pointed out earlier, it
would be unrealistic to expect a tight unity in this subject. But
if we can agree to regard conceptual analysis as the heart of philosophy,
we can give pride of place among these problems to the task of pro¬
viding an adequate analysis of the basic concepts we use in thinking
about language. Although there is no reason why a philosopher should
not put his analytical tools to work on any of the basic concepts dealing
with language, the tendency has been to concentrate on semantic con¬
cepts, for example, the concept of linguistic meaning and its cognates,
sameness of meaning, meaningfulness, etc. This has been partly because
many of the philosophical concerns enumerated in the first part of this
introduction naturally lead one to raise questions about the nature
of meaning, and partly because the fact that a given word has a certain
meaning is apt to appear mysterious in the way that often gives rise to
philosophical reflection. A large part of this book will be concerned
with the analysis of semantic concepts.
It would be misleading to suggest that the philosophy of language,
even as practiced by analytical philosophers, is restricted to conceptual
analysis, to clarifying the basic concepts dealing with language. There
are a number of other tasks which philosophers typically set themselves.
There is the classification of linguistic acts, of “uses” or “functions” of
language, of types of vagueness, of types of terms, of various sorts of
metaphor. There are discussions of the role of metaphor in extending
language; of the interrelations of language, thought, and culture; and
of the peculiarities of poetic, religious, and moral discourse. Proposals
are made for constructing artificial languages for various purposes.
There are detailed investigations of the peculiarities of particular sorts
of expressions, such as proper names and plural referring expressions,
and particular grammatical forms, such as the subject-predicate form!
Some of these problems lie in the borderland between philosophy and
more special disciplines, and all of them might be dealt with in one or
another of these disciplines. Thus, psychology might take on the job
of distinguishing between different sorts of linguistic behavior, and
descriptive linguistics could be expected to provide classifications of
Introduction 9
types of expressions. But if these problems belong in principle to the

more special disciplines, they belong to their foundations; and philoso¬
phy has traditionally had much traffic with high level problems in the
sciences, especially when these sciences are in early stages of construc¬
tion. I shall have something to say about some of these problems.
This book is written from a certain philosophical orientation—
that roughly indicated by the term analytical philosophy.’ There is
a great deal of philosophizing about language that is carried on from
very different standpoints, and there the problems take on quite differ¬
ent shapes. It is neither possible nor desirable in a volume of this size
to survey all the philosophical approaches to language. By way of com¬
pensation, I have included in the bibliography some suggestions for
reading in other approaches.
<
theories O F M EA N I N G
The problem
The present chapter is concerned with the nature of linguistic
of meaning
meaning. This is a problem of philosophical analysis, which is best
formulated as follows: “What are we saying about a linguistic
expression when we specify its meaning?” * That is, we are trying to
give an adequate characterization of one of the uses of ‘mean’ and its
cognates.
There are many other uses of mean/ some of which might be
confused with our sense. °
1. That is no mean accomplishment, (insignificant)

2. He was so mean to me. (cruel)
3. I mean to help him if I can. (intend)
4. The pasage of this bill will mean the end of second class citizenship
tor vast areas of our population, (result in) V
5. Once again life has meaning for me. (significance)
6. What is the meaning of this? (explanation)
1. He just lost his job. That means that he will have to start writing
letters of application all over again, (implies)
rathe/ftanTw”6 ab°Ut pe°ple’ actions’ events’ or situations

rather than about words, phrases, or sentences. The cases in which we
apply, or seem to apply, ‘means' to a linguistic expression, but where
ean does not have the sense we are examining are rare; but it is here
that confusion is most likely to occur.
1 this is the canonical form. I shall

not hesitate to employ the other
forms What is linguistic meaning?” and “How
is the concept of linguistic mean-
mg to be analyzed?” as stylistic variants. (See p.7.
Theories of Meaning 11
8. Keep off the grass. This means you.
Here it seems plausible to regard 'this’ as referring to the sentence,

‘Keep off the grass.’ But it is clear that we are not saying what the
sentence means. An English-French phrase book would not contain the
entry: Keep off the grass—vous. This is the use in which ‘mean’ means
very much the same as ‘refer.’ It is more commonly used of people, as in
“Who do you mean?”—“I mean Susie.” But, as in 8, it can be used of
linguistic expressions. Again consider:
9. Lucky Strike means fine tobacco.
Here we are not talking about a linguistic expression, although it may

seem at first sight as if we are. We are not giving the meaning of the
phrase, ‘Lucky Strike.’ No dictionary could possibly contain such an
entry. (If Webster’s did contain this entry, the American Tobacco
Company would undoubtedly be delighted.) This is an example of the
same use we have in:
That look on his face means trouble.

When he begins complaining, that means he is getting better.
In all these cases, we are saying that one thing or event is a reliable
indication of the existence of another.
There is one sense in which we all know perfectly well what we are
saying when we say what a word means. We succeed in communicating
with each other by saying things like “ ‘Procrastinate’ means put things
off,” 2 “He doesn’t know what ‘suspicious’ means,” and so on. In gen¬
eral, we know how to support, attack, and test such statements, we
know when such statements are warranted and when they are not, we
know what practical implications accepting such a statement would
have, and so on. What we lack, in advance of a philosophical investiga¬
tion, is an explicit and coherent account of these abilities.
of The literature on this subject contains a bewildering diversity of

of approaches, conceptions, and theories, most of which can be
grouped into three types, which I shall call ‘referential,’ ‘idea¬
tional,’ and ‘behavioral.’ The referential theory identifies the
meaning of an expression with that to which it refers or with the ref¬
erential connection, the ideational theory with the ideas with which it
2 A word about the notation. We shall regularly italicize what follows ‘means’ in
‘E means. . . .’ (or what follows ‘is’ in ‘The meaning of E is. . . .’ ) This is intended
to reflect the fact that when expressions are put into this slot, they have a unique
kind of occurrence, for which we shall use the term ‘exhibit.’ (See p. 21.)
12 Theories of Meaning *
is associated, and the behavioral theory with the stimuli that evoke
its utterance and/or the responses that it in turn evokes. Each of these
kinds of theory exists in more forms than I shall have time to consider.
But I shall try to choose forms of each that will clearly exemplify its
basic features.
The The referential theory has been attractive to a great many theorists
referential because it seems to provide a simple answer that is readily as-
theory similable to natural ways of thinking about the problem of mean¬
ing. It has seemed to many that proper names have an ideally
transparent semantic structure. Here is the word ‘Fido’; there is the dog
the word names. Everything is out in the open; there is nothing hidden
or mysterious. Its having the meaning it has is simply constituted by
the fact that it is the name of that dog.3 It is both tempting and natural
to suppose that a similar account can be given for all meaningful ex¬
pressions. It is thought that every meaningful expression names some¬
thing or other, or at least stands to something or other in a relation
something like naming (designating, labelling, referring to, etc.). The
something or other referred to does not have to be a particular concrete,
observable thing like Fido. It could be a kind of thing (as with “com¬
mon names” like ‘dog’), a quality (‘perseverance’), a state of affairs
( anarchy ), a relationship ( owns ), and so on. But the supposition is
that for any meaningful expression, we can understand what it is for it
to have a certain meaning by noting that there is something or other to
which it refers. Words all have meaning, in the simple sense that they
are symbols that stand for something other than themselves.” 4
The referential theory exists in a more and a less naive version.
Both versions subscribe to the statement that for an expression to have
a meaning is for it to refer to something other than itself, but they
locate meaning in different areas of the situation of reference. The more
naive view is that the meaning of an expression is that to which the
expression refers;0 the more sophisticated view is that the meaning of
the expression is to be identified with the relation between the expres-
3 A more penetrating account of proper names would show that this is a singu¬
larly unfortunate model for an account of meaning. It is questionable whether proper
names can be correctly said to have meaning. They are not assigned meanings in
dictionaries. One who does not know what ‘Fido’ is the name of is not thereby
deficient in his grasp of English in the way he would be if he did not know what
og means. And the fact that ‘Fido’ is used in different circles as the name of
a great many different dogs does not show that it has a great many different
meanings or that it is a highly ambiguous word.
Press61903n)dpR47Se11’ FHncipleS of Mathem^s (London: Cambridge University
di*cult \° find a full-blown presentation of this version in the works of

reputable philosophers. But since it exercises an enormous influence on popular
thinking about language, it is worth-while to exhibit its defects.
sion and its referent, that the referential connection constitutes the
meaning.
Meaning and The first form of the theory can easily be shown to be inadequate
reference by virtue of the fact that two expressions can have different
meanings but the same referent. Russell’s classic example of this
point concerns 'Sir Walter Scott’ and 'the author of Waverley.’ These
two expressions refer to the same individual, since Scott is the author
of Waverley, but they do not have the same meaning. If they did, the
statement that Scott is the author of Waverley would be known to be
true just by knowing the meaning of the constituent terms. It is a funda¬
mental principle that whenever two referring expressions have the same
meaning, for example, 'my only uncle’ and 'the only brother either of
my parents has,’ then the identity statement with these terms as com¬
ponents, “My only uncle is the only brother either of my parents has,”
is necessarily true just by virtue of the meanings of these expressions.
But this is not the case with “Scott is the author of Waverley.” This
statement is a particularly good example because the identity of the au¬
thor of these novels was at first kept secret, so that many people could
understand the sentence ‘Scott is the author of Waverley’ (Scott was
already a famous poet) without knowing whether it was true. In gen¬
eral, anything to which we can refer can be referred to by many ex¬
pressions that do not have the same meaning at all, for example, John
F. Kennedy can be referred to as ‘the President of the U.S.A. in 1962,’
‘the U.S. President assassinated in Dallas.’ Such examples show that it
cannot simply be the fact that an expression refers to a certain object
that gives it the particular meaning it has.
The converse phenomenon—same meaning but different referents
—can be demonstrated, not for different expressions, but for different
utterances of the same expression. There is a class of terms, sometimes
called “indexical terms,” for example, ‘I,’ ‘you,’ ‘here,’ ‘this,’ which sys¬
tematically change their reference with changes in the conditions of
their utterance. When Jones utters the word ‘I,’ it refers to Jones; when
Smith utters it, it refers to Smith. But this fact doesn’t mean that ‘I’
has different meanings corresponding to these differences. If a word like
‘I’ had a distinguishable meaning for every person to whom it has ever
been used to refer, it would be the most ambiguous word in English.
Think of how many different meanings we would have to leam before
we could be said to have mastered the use of the word; in fact, every
time a new speaker of English learned to use the word, it would acquire
a new meaning. But this is fantastic. The word has a single meaning—
the speaker. And it is because it always has this meaning that its referent
systematically varies with variations in the conditions of utterance.
4
14 Theories of Meaning
Do all Because of these readily appreciable difficulties, the more careful

meaningful versions of the referential theory take the second alternative. Even
expressions though Russell, for example, often talks as if the meaning of an
refer to expression is what it stands for, we also find him saying things
something? like: When we ask what constitutes meaning, . . . we are asking,
not who is the individual meant, but what is the relation of the
word to the individual which makes the one mean the other.” 6 This
version cannot be disposed of by pointing out that reference and mean¬
ing do not always vary together. For it may be that although ‘Scott’ and
‘the author of Waverley’ refer to the same person, they are not related
to that referent in the same way, though it is difficult to say whether
they are until we have some account of the sort of relation in question.
But at this point a more fundamental difficulty comes into view. No sort
of referential theory will be adequate as a general account of meaning
unless it is true that all meaningful linguistic expressions do refer to
something. And if we take a careful look at the matter, it will be seen
that this is not the case.
First of all, there are conjunctions and other components of lan¬
guage, which serve an essentially connective function. Do words like
‘and,’ ‘if,’ ‘is,’ and ‘whereas’ refer to anything? It seems not. Referential
theorists usually reply to this objection by denying that “syncategore-
matic” 7 terms like these have meaning “in isolation,” or that they have
meaning in the primary sense in which nouns, adjectives, and verbs have
meaning. It is possible, of course, that in the end we may be driven
to the position that there is no one sense in which all linguistic units
to which we ordinarily assign meaning, have meaning. But to admit this
before we have made an earnest effort to find a single sense would be a
counsel of despair. It certainly seems that in saying “ ‘Procrastinate’
means put things off” and “ ‘If means provided that” we are saying
things which, in some important respect, are the same sort of thing;
we are, so to say, speaking in the same logical tone of voice. And we
should not lightly give up the attempt to spell out explicitly what is
common.
Moreover, the idea that every meaningful linguistic expression
refers to something encounters rough going even in those stretches of
language where the referential theorist feels most secure. Proponents of
this theory generally take it as obvious that nouns like ‘pencil,’ adjec-
6 Analysis of Mind (London: George Allen & Unwin, 1921), p. 191

nTsTh?.t?rm W3S intr°d,uced by medieval logicians to apply to’words like conjunc-
"isotron’-T? 35 Tv! s!;andingf°r an>’thing and so as not having mean¬
ing in isolation. These were the linguistic units that were left over after one had
gone through everything that could be assigned to Aristotle’s ten “Categories ” a
classjficahon of terms made by Aristotle. Thus, the remnants were terms that were
used only with (syn—categorematic) the categories.
tives like 'courageous/ and verbs like 'run' refer to something or other.
It is not always recognized that it is sometimes difficult to find a plausi¬
ble candidate for the referent. To what does ‘pencil’ refer? Not to any
particular pencil, for the word ‘pencil’ can be used in talking about any
pencil whatsoever. If saying what the word refers to is to bring out what
gives it its semantic status, what enables it to function as it does, then
we cannot limit its reference to any particular pencil or to any particular
group of pencils. The most plausible suggestion would be that it refers
to the class of pencils, that is, to the sum total of all those objects
correctly called “pencils.” Likewise, ‘courageous’ might be said to refer
to a certain quality of character, the quality of courage. And ‘run’ could
be said to refer to the class of all acts of running. It is to be noted that
in order to find anything that might conceivably be a referent for
words of these sorts (which make up the great bulk of our vocabulary),
we have had to bring in entities of a rather abstract sort—classes and
qualities. This should not disturb us, unless we are attached to the
groundless idea that words cannot be meaningful unless they refer to
concrete observable physical objects.
There is no doubt that ‘pencil’ is related in some important way
to the class of pencils, but does it refer to that class? One reason for
denying that it does is this. If we want to pick out the class of pencils as
what we are talking about, preparatory to going on to say something
about it, the word ‘pencil’ will not serve our purpose. If, for example,
we want to say that the class of pencils is very large, we will not succeed
in doing so by uttering the sentence “Pencil is very large.” The word
‘pencil’ simply will not do the job of referring to the class of pencils.
The same point can be made for adjectives and verbs. If we wanted to
pick out the quality of courage in order to say something about it, for
example, that it is all too rare in these times, we could not use the
adjective ‘courageous’ to do so. We would not say “Courageous is all
too rare in these times.” Again, I would not say that what I just did
belongs to the class of acts of running by uttering the sentence “What I
just did belongs to run.”
This point reflects the fact that referring is only one of the func¬
tions that linguistic expressions perform, a function assigned to some
sorts of expressions and not to others. What distinguishes referring from
other functions is the fact that it serves to make explicit what a given
bit of discourse is about. Thus:
W refers to x = df. W can be used in a sentence, S, to make it
explicit that S is about x.
There are other contexts in which expressions are used to refer to

something, for example, in lists, or in labels. But we may take the func-
16 Theories of Meaning 4
tion of making explicit what a sentence is about as the defining charac¬

teristic of referring. Types of expressions that normally have this func¬
tion include proper names like Winston Churchill’; abstract nouns like
courage; phrases combining a concrete noun or noun phrase with an
article or demonstrative, like ‘the pencil/ ‘this pencil/ ‘the pencil in my
pocket’; and concrete nouns in the plural, like ‘pencils/ and ‘dogs/ If
referring is one linguistic job among others and is assigned only to some
types of expressions, no account of meaning that presupposes that all
meaningful units refer to something can be correct. More specifically, it
cannot be the case that to say a word has a certain meaning is to say
it refers to something.
But perhaps it is just that ‘refer’ is an unfortunate term for what
the referential theorist really has in mind. In the preceding discussion,
we used the term stand for’ at one place; and there are other terms
that occur in expositions of this kind of theory, such as ‘designate,’
‘signify,’ ‘denote.’ Perhaps there is some more generic notion, such as
standing for,’ which is such that every meaningful linguistic unit stands
for something. Referring would then be only one species of this genus,
along with denoting, connoting, and any others there may be.
Denotation Is it the case that there is some one semantically important rela¬
and tion that each meaningful linguistic unit has to something or
connotation other? Now^ there is no doubt that expressions like ‘pencil’ and
‘courageous,’ which do not, in the strict sense, refer to anything,
stand m relations that are crucial for their meaning. Thus, ‘pencil,’
though it does not refer to the class of pencils, does denote that class;
which is simply to say that the class of pencils is the class of all those
things to which the word ‘pencil’ can be correctly applied. And, clearly,
it is crucial for its having the meaning it has that it denote this class
rather than some other. If it denoted another class, for example, the
class of chairs, it would not have the same meaning, and vice versa.
Again, although the adjective ‘courageous’ does not refer to the dispo¬
sition to remain steadfast in the face of danger, it does connote that dis¬
position, in the logician’s sense of ‘connote’; which is to say that the
possession of that disposition by someone is the necessary and sufficient
condition of the term ‘courageous’ being correctly applied to that
person. Thus, it would seem that many expressions that do not refer to
anything, nevertheless, do denote and/or connote something. Let us
pause for a moment to give explicit definitions of these terms as thev
are being used here. y
W denotes the class, C = df. C is the class of all those things

of which W can truly be asserted.
W connotes the property, P = df. The possession of P by some¬

thing is a necessary and sufficient condition of W being cor¬
rectly asserted of it (that is, of its belonging to the denotation
of W). 8
However, it is not at all clear that every non-referring expression

has a denotation and a connotation. Consider prepositions like ‘into,’
‘at,’ and ‘by.’ There is no doubt that each has a meaning, in most cases
a number of meanings. For example, one of the meanings of ‘at’ is
in the direction of; however, ‘at’ can hardly be said to refer, denote, or
connote. We can see that referring is out, by the same argument as that
given previously; we cannot use ‘at’ to pick out what we are going to talk
about in a given sentence. To see that denotation and connotation are
out as well, recall that ‘denote’ and ‘connote’ are defined in terms of an
expression being applied to or asserted of something; hence, we can
speak of an expression denoting or connoting only where it makes sense
to speak of it being applied to or asserted of something. If you try to
assert ‘at’ of something, you draw a blank. Having said that it is at,
you have said nothing. To make sense you must supplement this at least
to the extent of saying, for example, that it was thrown at the wall.
But then, what you have asserted of the thing in question is not ‘at,’
but rather ‘thrown at the wall.’ This phrase could be said to denote or
connote, but not the preposition itself.
It may be that at this point we are once more being victimized by
the poverty of the existing semantic terminology. It does seem plausible
to suppose that there is some distinguishable aspect of the situation we
are talking about when we utter sentences containing ‘at,’ to which the
word ‘at’ is related in some way not wholly unlike the way nouns or
adjectives are related to what they connote or denote. We might try to
formulate this aspect as direction toward something. We might then
try to introduce a term that would designate the relation a preposition
would have to such aspects of talked-about situations. But even if this
is possible for prepositions, it still seems that such parts of speech as
conjunctions and modal auxiliaries like ‘should,’ ‘would,’ and ‘might’
will resist such treatment. It seems impossible to identify independently
any aspect of talked-about situations to which ‘and,’ ‘if,’ or ‘should’ is
related in a way that is anything at all like reference, denotation, or
connotation. There are those who have been so hypnotized by the ref¬
erential theory as to insist, contrary to appearances, that such expres-
8 Note that this use of ‘denotation’ and ‘connotation’ (some logicians use ‘exten¬
sion’ and ‘intension’ instead) is very different from the literary use, in which denota¬
tion is something like the standard meaning of a word, whereas connotation com¬
prises the associations, which may well vary somewhat from person to person, to
which this meaning gives rise.
18 Theories of Meaning 4
sions do stand for something. It has been said that 'and’ stands for a
conjunctive function, or’ stands for a disjunctive function, etc. But this
view runs into the difficulty that there is no way of explaining what a
conjunctive function is, except by saying, for example, that it is what
we are asserting to hold between the fact that it is raining and the fact
that the sun is shining when we say “It is raining and the sun is shin-
ing. And this means that we cannot identify a “conjunctive function”
except by reference to the way we use 'and’ and equivalent expressions.
Thus, we have not really gotten at an independently specifiable referent
for 'and’ in the way we can for 'Winston Churchill.’ There we can
specify what it is this name stands for, namely, the prime minister of
Great Britain during the latter part of World War II, without having
to bring into the specification talk about the way the name is used.
In other words, to say that ‘and’ stands for a “conjunctive function” is
just to talk in a misleading way about the kind of function ‘and’ has in
sentences. No real extralinguistic reference has been demonstrated.
Thus, there seems to be no escape from the conclusion that such ex¬
pressions as conjunctions stand in no semantically interesting relations
to extralinguistic entities.
Over and above the point that not all meaningful expressions
stand for something m any sense of that term, there is a question as to
whether the various types of “standing for” we have been considering
ave anything important in common. Is there anything semantically
interesting that is common to referring, denoting, and connoting? If
there is not, then there is no one sense of ‘stand for’ in which all the
expressions that have these various relations to the extralinguistic stand
for something. And it seems that there is not. Of course we can say
hat what they all have in common is that they are all relationships that
1. hold between expressions and what the expressions are used to talk
about, and 2. are crucial for the meaning of the expressions. (The second
requirement is necessary because otherwise the fact that the word
pencil is very unlike the class of pencils would be a relation of the sort
in question.) But by bringing in requirement 2 we are making the
account viciously circular; since the generic notion of ‘standing for’
is brought in to give an account of the notion of meaning, we can
hardly bring the notion of meaning into an explanation of it. And unless
we do it seems impossible to find anything significant that is in com¬
mon to referring, denoting, and connoting. This leaves us with the
conclusion that (even if we forget about such items as conjunctions)
the principle, To say that a word has a certain meaning is to say that
it stands for something other than itself,” is either straightforwardly
false, or does not employ ‘stands for’ in any one sense. And this means
that we have failed to make explicit any single sense of the term ‘mean¬
ing in which all words have meaning.
The upshot of this discussion is that we cannot give a generally

adequate idea of what it is for a linguistic expression to have a certain
meaning by explaining this in terms of referring, or in terms of any
relation or set of relations like referring. The referential theory is based
on an important insight—that language is used to talk about things
outside (as well as inside) language, and that the suitability of an
expression for such talk is somehow crucial for its having the meaning
it has. But in the referential theory, this insight is ruined through over¬
simplification. The essential connection of language with “the world,”
with what is talked about, is represented as a piecemeal correlation of
meaningful linguistic units with distinguishable components of the
world. What the preceding discussion has shown is that the connection
is not so simple as that. Speech does not consist of producing a sequence
of labels, each of which is attached to something in “the world.” Some
of the meaningful components of the sentences we use to talk about
the world can be connected in semantically important ways to dis¬
tinguishable components of the world, but others cannot. Hence, we
must look elsewhere for an account of what it is for an expression to
have meaning, remembering that the account must be framed in such
a way as to give due weight to the fact that language is somehow con¬
nected with the world.
Meanings as If the referential theory of meaning is based on the fundamental

a kind of insight that language is used to talk about things, the ideational
entity and behavioral theories are based on an equally fundamental in¬
sight—that words have the meaning they do only because of what
human beings do when they use language. These theories focus on as¬
pects of what goes on in communication, in an effort to get at those
features of the use of language that give linguistic units the meanings
they have. These theories may or may not saddle themselves with the
assumption we found to be fatally involved in the referential theory—
that every meaningful linguistic unit stands for something, in some one
sense of ‘stands for.’ When they do make this assumption, as is often
the case, they bring in associations of ideas or stimulus-response connec¬
tions as ways of explicating this relation of standing for. Thus, one may
suppose a word stands for x by virtue of being associated with an idea
of x (ideational theory), or by virtue of having the potentiality of giving
rise to responses similar to those to which x gives rise (behavioral
theory). These theories, however, are not necessarily cast in this form;
and since the assumption just mentioned has already been amply crit¬
icized in connection with the referential theory, I shall, in considering
the ideational and behavioral theories, concentrate on problems that
would be there even if this assumption were not made.
Before examining these theories it would be well to note a certain
deficiency in the way the problem of meaning and theories of meaning

are often stated. More often than not, when people set out to clarify
the concept of meaning they do so by asking, “What sort of entity is
a meaning and how does an entity of this sort have to be related to a
linguistic expression in order to be the meaning of that expression?”
Theories of meaning are quite often expressed as answers to this kind
of question. Thus, the referential theory generally takes the form of an
identification of the meaning of E with that to which E refers, or alter¬
natively with the relation between E and its referent; the ideational
theory identifies the meaning of E with the idea(s) that give rise to it
and to which it gives rise; and behavioral theories typically identify the
meaning of an expression with the situation in which it is uttered,
with responses made to its utterance, or both. There is something funda¬
mentally wrong with this way of conceiving the problem. This can be
seen by noting that we run into absurdities as soon as we take seriously
the idea of identifying a meaning with anything otherwise specified
(that is, specified in terms that do not include ‘meaning’ or any of its
synonyms or near synonyms). No matter what sort of entity we try to
identify meanings with, we find many things that we would be pre¬
pared to say about an entity of that sort but would not be prepared to
say about a meaning, and vice versa. Since many things are true of one
but not true of the other, they cannot be identical. Suppose that, fol¬
lowing out the behavioral theory, we attempt to identify the meaning
of ‘Look out!’ with such activities as ducking, falling prone, and
fending. That the meaning is not identical with such activities can
be shown by the fact that although it is true that I sometimes en¬
gage in such activities, it can hardly be true that I sometimes engage in
the meaning of ‘Look out!’ It makes no sense to talk of engaging in a
meaning. Again it may be true that I have forgotten the meaning of
‘Look out!’ without its being true that I have forgotten the activities
of ducking, falling prone, and fending. Such examples show that mean¬
ings and activities belong to radically different categories; something
can be true of one without there even being any sense in the supposition
that it is true of the other. This same point will hold of anything else
with which we try to identify meanings. The cruder form of the refer¬
ential theory (in which the meaning is said to be the referent) is most
obviously in trouble in this respect, for anything whatsoever can be a
referent; at least, we cannot mention anything that is not a referent,
for we have referred to it in the act of mentioning it. This means that
the identification of the meaning of an expression with its referent could
be maintained only if anything could be true of a meaning that was
true of anything whatever; a random sample of referents will suffice to
show that this is not the case. For example, the phrase ‘the father of
pragmatism refers to C. S. Pierce. If the meaning of that phrase were

identical with its referent, we would have to be able to say, both in¬
telligibly and truly, that the meaning of ‘the father of pragmatism' was
married twice and that the meaning often wrote reviews for the Nation
Meanings, however, do not get married and they do not write reviews.
And so it goes.
(t ^he ™ral °f all this is that it is a basic mistake to suppose that
meanings are entities of a sort that are otherwise specifiable. If we
are to speak of meanings as a class of entities at all, we shall have to
recognize that they are so unique as not to admit of being characterized
m any other terms. The almost universal tendency to raise the problem
of meaning m this form may come from the supposition that in specify-
ing the meaning of a word, what we are doing is identifying the entity
that is so related to that word as to be its meaning. That is, it is very
natural to regard: ’ y
1. The meaning of ‘procrastinate’ is put things off.

as having the same logical form as:
2. The capital of France is Paris.
and, consequently, to think that just as in 2 we are specifying the entity

so related to France as to be its capital, so in 1 we are specifying the
entity so related to ‘procrastinate’ as to be its meaning. The simplest
way of seeing that this is not what we are doing is to note that, in
general, what follows is in statements like 1 is not a specification of
any entity whatsoever. This is true of 1; ‘put things off’ is not a phrase
that has the function of specifying a certain entity of which we might
then go on to ask and answer certain questions. The generalization is
much more obviously true of:
3. The meaning of ‘if’ is provided that.
Here it is perfectly clear that there is no such entity as ‘provided that.’

This is so, not because no such thing happens to exist, but because it
makes no sense to suppose that it does, for the phrase ‘provided that’
simply does not have the function of designating some entity that may
or may not exist.
Then what are we doing when we say what a word means? What
we are doing is exhibiting another expression that we are claiming has
at least approximately the same use as the one whose meaning we are
specifying.9 The primary reason for saying things like 1 and 3 is to
9 For some qualifications to this conclusion see M. Scriven, “Definitions, Explana¬

tions, and Theories,” Minnesota Studies in the Philosophy of Science, Vol. II
(Minneapolis: University of Minnesota Press, 1958).
help someone learn how to use the expression whose meaning we are
specifying; when we provide a specification of meaning, we seek to
accomplish this end by telling the person that this expression is used
in the same way as another one that, we suppose, the person already
knows how to use. Thus, 1 is roughly equivalent to “Use ‘procrastinate’
in the way you are accustomed to use ‘put things off,’ and you will be
all right.” We will be misled by superficial grammatical similarities if
we suppose that what we are really doing is picking out a particular
example of a special kind of entity called “meanings.” 10
If this account of meaning-statements is accurate, the problem of
meaning should be formulated as follows: “How must one expression
be related to another in order that the one can be exhibited in a specifi¬
cation of the meaning of the other?” If we can agree to use the term
‘have the same use’ as a label for that relationship, whatever it may
turn out to be in detail, then the crucial question can be stated: “What
is it for two expressions to have the same use?” And since whenever
can be exhibited in a specification of the meaning of E2, E1 and E2
would be said to have at least approximately the same meaning, to be
at least approximately synonymous, we can formulate what is essentially
the same question by asking, “What is it for two expressions to be
synonymous?”
This point concerning the right way to raise the problem of mean¬
ing has absolutely no implications as to what kind of theory is or is not
adequate; any of the standard types of theory can be formulated as an
answer to this question. Thus, the referential theory can be stated by
saying that two expressions have the same use if and only if they refer
to the same object (or perhaps refer to the same object in the same
way). The ideational theory would be that two expressions have the
same use if and only if they are associated with the same idea(s); and
the behavioral theory would hold that two expressions have the same
use if and only if they are involved in the same stimulus-response
connections. Henceforth, I shall proceed as if the theories were in this
form, even when they are not explicitly so put.
The ideational The classic statement of the ideational theory was given by the
theory seventeenth century British philosopher, John Locke, in his Essay
Concerning Human Understanding, section 1, Chapter 2, Book
III. “The use, then, of words is to be sensible marks of ideas; and the
ideas they stand for are their proper and immediate signification.” This
kind of theory is in the background whenever people think of lan¬
guage as a “means or instrument for the communication of thought,”
10 For a more extended presentation of this point, see W. P. Alston, “The Quest
for Meanings,” Mind, Vol. LXXII (Jan. 1963).
or as a “physical and external representation of an internal state,” or

when people define a sentence as a “chain of words expressing a com¬
plete thought.” The picture of communication involved is set forth
with great clarity by Locke in the passage immediately preceding the
sentence just quoted.
Man, though he has great variety of thoughts, and such, from which
others, as well as himself, might receive profit and delight; yet they
are all within his own breast, invisible and hidden from others, nor can
of themselves be made to appear. The comfort and advantage of
society not being to be had without communication of thoughts, it
was necessary that man should find out some external sensible signs,
whereof those invisible ideas, which his thoughts are made up for’
might be made known to others. . . . Thus we may conceive how words
which were by nature so well adapted to that purpose, come to be
made use of by men, as the signs of their ideas; not by any natural
connexion that there is between particular articulate sounds and cer¬
tain ideas, for then there would be but one language amongst all men;
but by a voluntary imposition, whereby such a word is made arbitrarily
the mark of such an idea.
According to this theory, what gives a linguistic expression a cer¬

tain meaning is the fact that it is regularly used in communication as
the “mark” of a certain idea; the ideas with which we do our thinking
have an existence and a function that is independent of language. If
each of us were content to keep all his thoughts to himself, language
could have been dispensed with; it is only because we feel a need to
convey our thoughts to each other that we have to make use of publicly
observable indications of the purely private ideas that are coursing
through our minds. A linguistic expression gets its meaning by being
used as such an indication.
Let us see what would have to be the case for this theory to work.
For each linguistic expression, or rather for each distinguishable sense
of a linguistic expression, there would have to be an idea such that
when any expression is used in that sense, it is used as an indication
of the presence of that idea. This presumably means that whenever an
expression is used in that sense, 1. the idea must be present in the
mind of the speaker, and 2. the speaker must be producing the ex¬
pression in order to get his audience to realize that the idea in question
is in his mind at that time.11 Finally, 3. insofar as communication is
11 On a less stringent version of the theory, there can be occasions on which the
expression is used in that sense without the idea being consciously in the mind of
the speaker, provided that the speaker could call up the idea if any question arose
as to what he had meant. These occasions would be those on which we are using
words “automatically,” “unthinkingly,” without having our minds on what we are
saying. But even this form of the theory maintains that the above conditions are
satisfied whenever we say anything with our minds on what we are saying; and,
successful, the expression would have to call up the same idea in the
mind of the hearer, with analogous qualifications as to an “unthinking”
grasp of what was being said that might hold on some, though not on
all, occasions.
These conditions are not in fact satisfied. Take a sentence at
random, for example, “When in the course of human events, it be¬
comes necessary for one people to . . . and utter it with your mind
on what you are saying; then, ask yourself whether there was a distin¬
guishable idea in your mind corresponding to each of the meaningful
linguistic units of the sentence. Can you discern an idea of 'when/
‘in,’ ‘course,’ ‘becomes,’ etc., swimming into your ken as each word is
pronounced? In the unlikely event that you can, can you recognize
the idea that accompanies ‘when’ as the same idea that puts in an
appearance whenever you utter ‘when’ in that sense? Do you have a
firm enough grip on the idea to call it up, or at least know what it
would be like to call it up, without the word being present? In other
words, is it something that is identifiable and producible apart from
the word? Do you ever catch the idea of ‘when’ appearing when you
utter other words—‘until,’ ‘rheostat,’ or ‘epigraphy’?
What is disturbing about these questions is not that they have
one answer rather than another, but that we do not know how to go
about answering them. What are we supposed to look for by way of
an idea of ‘when’? How can we tell whether we have it in mind or not?
Just what am I supposed to try for when I try to call it up out of con¬
text? The real difficulty is that we are unable to spot “ideas” as we
would have to in order to test the ideational theory.
There is, to be sure, a sense of ‘idea’ in which it is not completely
implausible to say that ideas are involved in any intelligible bit of
speech. This is the sense ‘idea’ has in such expressions as “I get the
idea,” “I have no idea what you are saying,” and “He isn’t getting his
ideas across.” In that sense of the term, I don’t understand what some¬
one is saying unless I get the idea. But that is because the phrase ‘get
the idea’ would have to be explained as equivalent to ‘see what the
speaker meant by his utterance’ or ‘know what the speaker is saying.’
‘Idea’ in this sense is derivative from such notions as ‘meaning’ and
‘understanding,’ and so can provide no basis for an explication of mean¬
ing. If we are to have an explication of meaning in terms of ideas, we
must be using ‘idea’ so that the presence or absence of an idea is de-
furthermore, it maintains that this is the primary kind of situation, from which the
automatic use of words is derivative. That is, a given person could not meaningfully
use a given word without having the corresponding idea in his mind, unless he fairly
often produced it with the conscious intention of making it known that a certain
idea was consciously in his mind.
cidable independent of determining in what senses words are being

used. Ideas would have to be introspectively discriminable items in
consciousness. Locke was trying to satisfy this requirement when he
ook idea to mean something like ‘sensation or mental image.’ But the
more we push ‘idea’ in the direction of such identifiability, the clearer
it becomes that words are not related to ideas in the way required
by the theory. 1 ^
The ideational theory will not work even for words that have an
obvious connection with mental images, for example, ‘dog,’ ‘stove,’ and
book. A little introspection should be sufficient to convince the reader
that insofar as his use of the word ‘dog’ is accompanied by mental
imagery, it is by no means the case that the mental image is the same
on each occasion the word is used in the same sense. At one time it
may be the image of a collie, on another the image of a beagle, on one
occasion the image of a dog sitting, on another the image of a dog
standing, and so on. Of course a determined defender of the theory
could claim that this fact is sufficient to show that the word is not being
used in quite the same sense on these different occasions; however, if
one takes this way out, he has simply lost contact with the concept of
meaning he set out to explicate. For it is perfectly clear that such differ¬
ences in mental imagery need not, and undoubtedly will not, be re¬
flected in any difference in what one is saying. Conversely, one can
have indistinguishable mental images accompanying different words
with quite different meanings. Thus, the image of a sleeping beagle
might accompany the utterance of the words ‘beagle,’ ‘hound,’ ‘dog,’
‘mammal,’ ‘animal,’ ‘organism,’ ‘sports,’ ‘hunting,’ and so on. Clearly,
this example does nothing to show that all these words were being used
in the same sense.
Meaning as a The behavioral theory of meaning also concentrates on what is

function of involved in using language in communication, but it differs from
situation and the ideational theory in focusing on publicly observable aspects
response of the communication situation. There are several reasons for the
attractiveness of such a project. One deficiency of the ideational
theory that was not explicitly mentioned above stems from the fact
that we do not look for ideas in the minds of speakers and listeners in
order to settle questions about what a word means in the language or
about the sense in which a speaker used a term on a given occasion.
If I am not sure as to the exact sense in which you used the word
‘normal’ in something you just said, it would be absurd for me to try
to find out by asking you to carefully introspect and tell me what
imagery accompanied your utterance of the word. It is not at all clear
just what we do look for in settling such questions; however, the fact
that there is a broad consensus concerning what various words mean

strongly suggests that meaning is a function of aspects of the com¬
munication situation that are open to public inspection. Moreover, the
success of psychology in explaining some aspects of human behavior
in terms of regular connections between observable stimuli and re¬
sponses naturally gives rise to a hope of being able to give the same
kind of treatment of verbal behavior.
The simplest forms of the behavioral theory are to be found in
the writings of linguists who, not surprisingly, take over ideas from be-
haviorally minded psychologists with little awareness of the complexities
involved. Thus, Leonard Bloomfield says that the “. . . meaning of a
linguistic form . . .” is “. . . the situation in which the speaker utters
it and the response which it calls forth in the hearer.” 12 The crudity
of this definition is mitigated by the qualifying statement: “We must
discriminate between the non-distinctive features of the situation . . .
and the distinctive, or linguistic meaning (the semantic features)
which are common to all the situations that call forth the utterance
of the linguistic form.” 13 The requirements for the adequacy of this
kind of theory are similar to those for the ideational theory. If it is to
work, there must be features that are common and peculiar to all the
situations in which a given expression is uttered in a given sense, and
there must be features common and peculiar to all the responses that are
made to the utterance of a given expression in a given sense. (Further¬
more, these common elements must be actually employed as criteria for
assigning the sense in question to that expression. But if the preceding
requirements are not satisfied, we do not have to worry about this one.)
Again, this seems not to be the case. Certainly, for a single word like
'shirt,' there is nothing of interest that is common and peculiar to the
situations in which the following utterances are made or to the responses
they call forth.
Bring me my shirt.
This shirt is frayed.
I need a new shirt.
Shirts were rarely worn before the fourteenth century.
What a lovely shirt!
Do you wear a size 15 shirt?
Perhaps we would do better to start with sentences, leaving subsenten-

tial units for later treatment. But even here the prospects are discourag¬
ing.
1. Bring me another cup of coffee, please.

2. My shirt is torn.
12 Language (London: George Allen & Unwin, Ltd., 1935), p. 139.
13 Ibid., p. 141.
3. What a superb steak!
If we abandon counsels of perfection and overlook a few exceptions,

we can find situational features that are common to most of the occa¬
sions on which each of the preceding sentences are uttered. For example
(using S as an abbreviation for the speaker’ and H as an abbreviation
for 'the hearer’):
1. S has recently had a cup of coffee.

H is in a position to bring S a cup of coffee.
2. S possesses a shirt.
S is doing something to call H’s attention to one of his shirts.
3. S is doing something to call H’s attention to some particular steak.
S is enthusiastic about the steak.
Items like these, however, will not do the job, for several reasons.
First, these uniformities hold equally for quite different sentences
with quite different meanings. Thus, the situational features listed for
sentence 1 hold just as often for the sentence 'No more coffee, please,’
and the situational features for sentence 2 are equally correlated with
the sentence 'Bring me my torn shirt.’
Second, we have been considering the favorable cases. With
declarative sentences that have to do with states of affairs remote from
the situation of utterance, we are hard pressed to find any common
situational features, at least any that seem likely to have an important
bearing on the meaning of the sentence. Consider the sentences:
4. The disarmament conference is about to collapse.
5. Mozart wrote ldomeneo at the age of 25.
6. Affirming the consequent produces a fallacious argument.
Each of these sentences can be uttered in a wide variety of situations,
and there is little or nothing of relevance that they have in common.
There are certain temporal limitations for 4 and 5. Thus, 4 is usually
uttered while a disarmament conference is going on, and 5 is uttered
only after 1781. But it is obvious that these uniformities are radically
insufficient to distinguish these sentences from many others with quite
different meanings.
Third, in all these cases we are going to have great difficulty in
finding any interesting features common to the overt responses made
to the utterance of the sentences. Imperatives look most promising in
this respect, for they clearly call for a specific response from the hearer.
But in what proportion of the cases in which an imperative is heard
and understood is the standard compliance forthcoming? Think of the
variety of responses that a parent’s “Come in now” will elicit.
No response. Whatever activity was in progress proceeds as if the

utterance had not been made.
Explicit refusal to comply.

Demand for justification.
Criticism of the parent for issuing the command.
Justification of noncompliance.
Plea for mercy.
Change of subject.
Running in the opposite direction.
Compliance.
If the response last mentioned were made in a substantial proportion

of cases, the life of a parent would be much easier. And imperatives
constitute the favorable case. With assertions, it is much more difficult
to suggest even a plausible candidate for a common response.
Meaning as a Psychologists and psychologically oriented philosophers who have

function of taken this kind of approach to meaning have tried to develop more
behavioral sophisticated accounts. Interestingly enough, they have pretty
dispositions much confined themselves to responses to linguistic utterances
and have said little about the situation of utterance as a determi¬
nant of meaning. Presumably this is so because they have generally
started with natural signs, where there is no intentional production of
the sign. In concentrating on responses, men like the philosopher,
Charles Morris, and the psychologist, Charles Osgood,14 recognize that
having a certain meaning cannot be simply identified with regularly
evoking a certain overt response, for, as noted previously, a. we can
have meaningful utterances that evoke no response at all, and b. where
there are overt responses, they can vary widely among themselves with¬
out any variation in meaning. Something more elaborate is called for.
This, Morris tries to provide in the concept of a disposition to re¬
spond.15 To say that someone has a disposition to make a certain
response, R, is simply to say that there are conditions, C, under which
he will do R. It is to assert a certain hypothetical proposition of him—
'if C, then R.’ Now, even though there is no one response that is uni¬
versally, or even generally, elicited by the utterance of “Come in now,”
it may be that this utterance regularly produces a disposition to come
in if the hearer has a strong inclination to obey the speaker. In other
words, it may be that this utterance brings it about that a certain hypo¬
thetical proposition comes to be true, namely, that if the hearer is
generally inclined to obey the speaker, he will come into the house.
This is what Morris is banking on. If this is generally the case, then
14 See his Method and Theory in Experimental Psychology (New York: Oxford
University Press, 1953), Chap. 16.
15 See his Signs, Language, and Behavior (Englewood Cliffs, N. J.: Prentice-Hall,
Inc., 1946), especially Chap. 1.
we can specify something of a behavioral character that is common to

all utterances of a given sentence even though there be nothing overt
that is common.
Unfortunately, this version of the behavioral theory fares almost
as badly as the simpler one. There are several reasons why it breaks
down.
1. If we try to specify the particular disposition(s), the regular
production of which would give a particular sentence the meaning
it has, we can think of some plausible candidates for certain kinds of
sentences, for example, imperatives like “Come in now,” and declara¬
tive sentences that have to do with matters of fact that have a direct
bearing on the hearer s future conduct. As an example of the latter,
consider Your son is ill.’ We might think of this sentence as regularly
producing a disposition to go where the hearer believes his son to be
if he has a great deal of concern for him. But when we consider utter¬
ances having to do with matters more remote from practical concerns
of the moment, things do not go so well. What semantically important
dispositions are produced by historical utterances like 'Mozart wrote
Idomeneo at the age of 25’? It may be said that this produces a dis¬
position to say ‘25’ if one is asked “At what age did Mozart write
Idomeneo? But if this is the only kind of disposition we can specify,
we are in trouble. For presumably the meaning of a sentence has some¬
thing to do with its connection with the sorts of things it is used to
talk about; consequently, the meaning of a sentence that is not about
language can hardly be a function of purely linguistic dispositions.
2. In fact, it is only in one sort of case that an utterance like
‘Your son is ill’ will produce a disposition to go to where one believes
one’s son to be if one has a great deal of concern for one’s son. This
kind of case is one in which the hearer believes that the speaker is pro¬
viding correct information and the hearer has not previously acquired
that information. If I do not believe you when you say “Your son is
ill,” your utterance will certainly not produce any such disposition. And
if I were already aware of the fact, no such disposition will be produced
by your utterance. I will already have been so disposed if I am go¬
ing to be.
3. Even in the ideal case there are problems. For one thing, we
may have to make a given disposition extremely, perhaps indefinitely,
complicated before there is any plausibility in supposing that it is
regularly produced by the utterance of a certain sentence. Thus, even
if I understand and believe someone when he tells me that my son is
ill, and even if I were not previously aware of this, this utterance will
not lead me to go to him, even given concern on my part, if I am in
jail and cannot escape, or if I am at the crucial stage of some enormous
business transaction that will affect my financial situation for years (and
I have not been told that he is seriously ill), or if I have very strong
religious scruples against travelling on a certain day, and so on. With a
little ingenuity, we could keep going indefinitely in listing factors that,
if present, would prevent the antecedent from giving rise to the con¬
sequent. Of course, for each of these possible interferences we could
save the claim that the utterance generally produces the disposition
in question by stipulating the absence of this interfering factor in the
antecedent. Thus, the utterance of Tour son is ill’ will produce in a
hearer a disposition to go to his son if he has a great deal of concern
for him, if he is not physically prevented from doing so, if he has no
religious scruples against doing what is necessary for accomplishing
this, . . . However, it is not at all clear that we can ever complete
this list.
Thus far in my discussion of Morris, I have ignored the fact that
he has saddled himself with the assumption that every meaningful
expression is a “sign” of something. Despite the indefensibility of this
assumption, we can see why a behavioral response theory needs it. For
if we were to allow any disposition produced to have a bearing on the
meaning of the sentence, we would be dragging in things that have
nothing to do with meaning. Suppose that uttering 'The sun is 97,-
000,000 miles from the earth’ produces a disposition to open one’s
mouth in amazement if one were previously unaware of this. It is ob¬
vious that this disposition-production has nothing to do with the
meaning of the sentence. We can imagine a lot of other sentences with
widely different meanings having the same effect, for example, ‘The
pyramids are several thousand years old.’ With the assumption that
every expression is a “sign” of something, we can limit the relevant
dispositions to those that are dispositions to the kinds of responses
that are involved in some important way with the object. No doubt it
is extremely difficult to see just what such responses would be in the
case of “objects” like the distance of the sun from the earth and the age
of the pyramids. In fact, the difficulty of deciding which “responses”
are to be called relevant to a given “object” constitutes one of the main
weaknesses in Morris’ theory.
Summary of Whether we consider relatively crude versions that regard mean-

discussion ing as a function of common features of situations in which ex-
of behavioral pressions are uttered and responses made to those utterances, or
theory relatively sophisticated versions in terms of dispositions to response
produced by utterances, we will be unable to find situation and
response features that are distributed in the way the theory requires.
Meaning simply does not vary directly with the kinds of factors high¬
lighted in these theories.
Like the others, the behavioral theory is based on an insight that

it perverts through oversimplification. Just as the meaningful use of
language has something to do with reference to the “world,” and just
as in some way we do express and communicate our thoughts in using
language, so it is also a significant fact that units of language get their
meaning through being used by people, through the fact that they are
involved in various sorts of behavior. Behavioral theories err in con¬
ceiving this behavioral involvement in oversimplified terms. They sup¬
pose that a word or sentence has a certain meaning by virtue of being
involved, as response and/or stimulus, in stimulus-response connections
that are basically similar, except for complexity, to a simple reflex like
the knee jerk. Unfortunately, no such connections have ever been found,
except for those that are obviously not determinative of meaning, like
the fact that a sudden loud utterance of “Look out!” typically elicits
a start. Some more adequate characterization of linguistic behavior is
called for, one that will get at the units of behavior that are crucial
for the meaning. We shall take up the search for such units in the next
chapter.
4
M E AN I N G
AND THE USE OF LANGUAGE
Meaning as It is a striking fact that in “behavioral” theories of meaning almost

a function no attention is given to what speakers are doing when they use
of use language. As I pointed out in the discussion of this kind of theory
in Chapter 1, many behavioral theorists try to construe meaning
solely in terms of the hearer’s response. Even when something on the
speaker s side is brought in, as with Bloomfield, it is something about
the situation in which the speaker operates rather than anything about
what the speaker is doing in that situation. What makes this choice of
focus surprising is the fact that although the hearer may or may not
respond to what is being said, it will always be true that the speaker is
doing something (otherwise there is no linguistic transaction to con¬
sider). If we are proceeding on the assumption that the meaning of an
expression is somehow a function of what members of the linguistic
community do with the expression, the activity of the speaker would
seem to be a more promising place to look. Perhaps theorists have been
discouraged from taking this tack because they have supposed that all
we could say about the speaker’s activity is that it consists of the pro¬
duction of a certain sentence. And because whatever gives that sentence
the meaning it has will have to be something beyond the bare fact
of its production (otherwise it would be impossible to have the same
sentence used in different senses on different occasions), it might seem
that in canvassing the speaker’s activity we have only presented the
problem of meaning without doing anything to contribute to its solu¬
tion. But as we shall see, it would be a great mistake to suppose that
this is all that can be said about the speaker’s activity. Let us explore
the possibility of exhibiting the meaning of a linguistic expression as a
Meaning and the Use of Language 33
function of the way in which it is used by speakers of the language.1

This enterprise will have a chance of getting off the ground only
if we start with sentences. For a sentence is the smallest linguistic unit
that can be used to perform a complete action that is distinctively
linguistic.2 (If our concepts of linguistic behavior were more highly
developed, we could define a sentence in this way.) Of course our
ultimate interest is in elucidating the notion of the meaning of a word,
for talk about the meaning of words is much more common, and much
more important, than talk about the meaning of sentences. This is so
because the primary use of specifications of meaning is in helping some¬
one acquire or extend his mastery of a language. Specifications of
meaning cannot be the only means employed, for before we can use
them for this purpose, our pupil must already know enough of the
language to be able to understand the specifications. By the time he
has come this far, it is obvious that the most economical way of pro¬
ceeding is to give him the meaning of individual words, and to let him
use his already acquired practical mastery of sentence structure to put
these words together with others in various sentences, rather than to
tell him the meaning of sentences one by one. In fact, strictly speaking,
the latter is impossible, because no limit can be put on the number
of sentences in a language (for example, compound sentences can be
made longer and longer without ever reaching a definite stopping
place), whereas there is a finite number of words in a language. But if
we are to see how the semantic status of an expression is a function of
what speakers do with it, we must begin with expressions that can
themselves be used to perform complete actions, and then try to specify
component parts of these actions that are performed by component
parts of the complete sentence.
1 The idea that meaning is a function of use was forcefully stated in Ludwig
Wittgenstein’s Philosophical Investigations, tr. G.E.M. Anscombe (Oxford: B.
Blackwell, 1953). Although many philosophers influenced by Wittgenstein have
made use of this idea in discussing the meaning of particular expressions, virtually
nothing has been done by way of going beyond Wittgenstein’s cryptic remarks to
an explicit analysis of semantic concepts. The theory briefly presented in this chapter
represents pioneer work.
2 This thesis needs qualification. For one thing, there are so-called one-word
sentences like ‘Fire!’ But an adequate linguistic analysis would distinguish the word
‘fire’ from the one-word sentence ‘Fire!’, and would thereby relieve us of the embar¬
rassment of having to recognize that the word ‘fire’ can be used by itself to report a
fire. For another thing, any word can be used alone to answer a question. I can use
the single word ‘salt’ all by itself to answer the question, “What’s that on the table?”
In this case, it is plausible to say that the preceding context permits ‘salt’ to func¬
tion as an elliptical substitute for the sentence “That is salt on the table.” Without
a special linguistic context the single word ‘salt’ could not be so used. We can,
therefore, state the above thesis more adequately by saying: “To perform a complete
linguistic action we must utter a sentence or some expression which in that context
is elliptical for a sentence.”
34 Meaning and the Use of Language
At this point, let us recall our previous conclusion that what we

are doing when we say what an expression, Eu means is to exhibit
another expression, E2, which, we claim, is used in the same way as E1.
We can embody this conclusion in a definition as follows:
1. E1 means E2 = df. E1 is used in the same way as E2.3
This means that the fundamental question is: “What is it for two ex¬
pressions to be used in the same way?” (“Wliat is it for two expressions
to have the same meaning?”). If we can answer that question, then we
will be in a position to make explicit what we are saying when we say
what an expression means. The answer (for sentences) suggested by
the preceding considerations is that two sentences have the same mean¬
ing if they are used to do the same thing. But before we can take this
suggestion seriously, we must make some restrictions on the class of
things done that we will be considering. One thing a person does
when he utters Sj is to utter Sx. But this sort of action will always be
different when two different sentences are uttered; this cannot be the
kind of action such that if two sentences are used to perform the same
action they have the same meaning. And if we go beyond the mere
sentence utterance to things that one can do in or by uttering a sentence,
the suggestion seems to work for some things but not for others. Thus'
I might impress someone by uttering either *1 have just been to dinner
at the White House’ or ‘Toynbee just asked me to write a preface to
his latest book’; however, this fact does nothing to show that these
two sentences have anything like the same meaning. On the other hand,
the sameness of meaning of ‘Quelle heure est-il?’ and ‘What time is
it?’ seems to be due to the fact that they can both be used to ask the
same question.
Types Of What we need is a classification of the different sorts of actions

linguistic that involve the use of sentences. In general, when a person utters
action a sentence, we can distinguish three sorts of actions that he is
performing. 1. He utters a certain sentence, for example, ‘Would
you please open the door?’. 2. He brings about one or more results of
this utterance, for example, he gets the hearer to open the door, he
3 All the definitions of this form are to be taken with the presupposition that the
person to whom the specification of meaning is addressed already knows how to use
is2- Otherwise, we would not get an equivalence between ‘Ex means E2’ and ‘Ei is
used in the same way as E2.’ For, in general, it is possible to tell someone that two
expressions have the same use without telling him what either of them means. I
w o know Japanese, could inform you that a certain expression in Japanese is used in
the same way as another expression in Japanese; and if I realized that you were
completely ignorant of Japanese, I would clearly not be telling you what either of
these expressions means. But if we add the proviso that the addressee already knows
how to use E2 (and that the speaker realizes this), then we will get an equivalence
wtaE?7J1 8 y°“ ‘hat E‘ h3S 'he same ““ as E* is ‘elli"S y°“

irritates the hearer, he distracts someone who is reading. 3. He does

something that falls between actions 1 and 2, for example, he asks
someone to open the door. The reason for saying that 3 falls between
1 and 2 is this. Unlike 1, it is not simply the utterance of a certain
sentence. No matter what sentence is specified (for example, 'Would
you open that door?’), it is conceivable that one might utter that sen¬
tence without asking anyone to open any door; one might, for instance,
be giving an example or testing his voice. Action 3 does not go beyond
the utterance of a sentence by essentially involving a certain effect,
as in the case of 2. There is no particular kind of effect that the utter¬
ance must have if the speaker is to be said to have asked someone to
open a door. His utterance may have the effect of getting the hearer
to open the door; it may arouse amusement, scorn, terror, or incredulity;
or it may produce no effect at all. In all these cases, it could be true
that the speaker asked someone to open the door. I am not saying that
an action of this category normally has no effect. I am saying, rather,
that the truth of the claim that an action of this sort has been per¬
formed does not depend on the production of any particular sort of
effect. We shall return to the question of the way this kind of action
goes beyond the mere utterance of a sentence.
Borrowing some terminology from John Austin,41 shall call actions
of these three sorts 1. locutionary, 2. perlocutionary, and 3. illocutionary.
The distinction between illocutionary and perlocutionary acts is of
crucial importance for our purpose. We can find many verbs and verb
phrases that stand for actions of one or another of these sorts.
Illocutionary Perlocutionary
report bring x to learn that. . . .
announce persuade
predict deceive
admit encourage
opine irritate
ask frighten
reprimand amuse
request get x to do. . . .
suggest inspire
order impress
propose distract
express get x to think about. . . .
congratulate relieve tension
promise embarrass
thank attract attention
exhort bore
4 See his How to do Things with Words (London: Oxford University Press,
1962), lecture viii, ff.
There are two main distinctions between the two categories. 1. As

already noted, perlocutionary, but not illocutionary, acts essentially in¬
volve the production of some effect. To say that I distracted or em¬
barrassed or persuaded someone is to say that what I did had a certain
kind of effect on the individual. But I can be said to have reprimanded
someone or to have made a certain announcement, prediction, or pro¬
posal no matter what effect I had on anyone, if, indeed, I had any
effect at all. 2. An illocutionary act, unlike a perlocutionary act, requires
a locutionary act as a base. I can bring you to learn that my battery is
dead by maneuvering you into trying to start the car yourself, and I
can get you to pass the salt by simply looking around for it. But there
is no way in which I can report that my battery is dead or request that
you pass the salt without uttering a sentence or using some equivalent
conventional device, for example, waving a flag according to some pre¬
arranged signal. 3. Another noteworthy difference between the two
categories is that an illocutionary act can be a means to a perlocutionary
act, but not vice versa. I can request that you pass the salt in order
to get you to pass the salt and/or in order to irritate, distract, or amuse
you. But I could hardly amuse you in order to request that you pass
the salt, or get you to know that my battery is dead in order to report
that my battery is dead.
To return to the problem of sentence meaning, the first pair of
sentences presented on page 34 showed that for two sentences to have
the same meaning it is not sufficient that they be commonly used to
perform the same perlocutionary act (have the same perlocutionary-act
potential). The second pair of sentences suggested, to use the termi-
nology just developed, that the fact that two sentences are commonly
used to perform the same illocutionary act (have the same illocutionary-
act potential) is sufficient to give them the same meaning. A wider
survey will reinforce the impression that sameness of illocutionary-act
potential is what constitutes sameness of meaning for sentences. ‘Das
is gut’ and ‘That’s good’ are both used to positively evaluate something.
In the cases in which ‘Can you reach the salt?’ and ‘Please pass the salt’
mean the same, they are both used to make the same request. ‘That’s
my paternal grandmother’ and ‘That’s my father’s mother’ are both
used to identify a person in the same way. And so on. Thus, we can
provide the following elucidation for specifications of sentence meaning.
S1 means S2 = df. S1 and S2 have the same illocutionary-act potential.
Word How are we to extend this account to words and other subsentential
meaning units? (For the sake of brevity, we shall henceforth use ‘word’ to
cover all meaningful components of sentences.) A single word is
not itself used to perform an illocutionary act. But perhaps we can
think of each word within a sentence making some distinctive con-
tribution to the illocutionary-act potential of the sentence, in such a

way that the omission of the word or its replacement with a non-
synonymous word would bring about a change in the potential of the
sentence. Thus, if we change ‘Please pass the salt’ to ‘Please pass the
sugar’ or to ‘Please dissolve the salt,’ we have not made the same re¬
quest; and if we change the sentence to ‘That’s the salt,’ we have not
made any request at all. Likewise, if we change ‘That’s good’ to ‘That’s
unfortunate,’ we have not made the same evaluation. Thus, it would
seem plausible to think of two words as having the same meaning if and
only if they make the same contribution to the illocutionary-act poten¬
tials of the sentences in which they occur; and whether or not they do
can be tested by determining whether replacing one with the other
would bring about any change in the illocutionary-act potentials of the
sentences in which the replacements are carried out. For example, the
synonymy of ‘procrastinate’ and ‘put things off’ would consist in the
fact that they make the same contribution, which in turn would be
evinced by the fact that ‘You’re always procrastinating’ would normally
be used to make just the same complaint as ‘You’re always putting
things off, I never procrastinate would be used to make just the same
claim as ‘I never put things off’; ‘Never procrastinate’ and ‘Never put
things off would be used to make just the same injunction; and so on.
This suggests the following elucidation of word meaning.
Wj means W2 = df. and W2 can be substituted for each

other in a wide range of sentences without altering the illocu¬
tionary act potentials of those sentences.
Thus far I have been greatly oversimplifying the situation by talk¬

ing as if each expression has a single meaning. In general, this is not
the case. ‘Can you reach the salt?’ sometimes means please pass the
salt, sometimes is your reach long enough to enable you to touch the
salt?, and sometimes show me whether you can touch the salt. ‘Run’
has a great many meanings—move rapidly, flee, operate, extend, chase,
etc. When there is more than one distinguishable sense, as there usually
is, the most fundamental statement of meaning is “A meaning of E, is
E2.” In such cases, we are only speaking loosely when we speak of the
meaning of E,. We engage in such loose talk when there is one mean¬
ing that is much more prominent than any of the others; thus, we
might say without qualification that ‘ill’ means sick, even though in
some contexts it means unfavorable, as in ‘bird of ill omen.’ Or we
might use the phrase ‘the meaning’ when the context makes it plain
which of the meanings is in question. Specifications of a meaning of
an expression would be elucidated according to this theory as follows:
A meaning of E1 is E, = df. Sometimes Et has the use that E2

usually has.
It has to be “usually” rather than “sometimes” for E2 because the point

of the meaning specification is to make clear the sort of use Ex is being
said to have sometimes; to do this we need to find an expression that is
clearly connected with this use. If E, only exceptionally had this use, its
exhibition would not clearly identify it. Expanding this schema for
sentences, we get:
A meaning of S1 is S2 = df. Sometimes S1 is used to perform the

illocutionary act(s) that S2 is usually used to perform.
The expansion of the formula for words is:
A meaning of Wj is W2 = df. In most sentences in which W2

occurs, Wj can be substituted for it without changing the illocu¬
tionary-act potential of the sentence.
The requirement that we choose an E, that usually has the meaning

we wish to specify for E± is reflected here in the specification of most
of the sentences in which W2 occurs. To say that sometimes (but
not necessarily usually) has the use that W2 usually has, is to say that
Wj can be substituted for W2 in most sentences containing W2, but
not necessarily vice versa. Thus, it would be quite in order to say that one
meaning of ‘case’ is example. For ‘example’ predominantly has the use in
question. But it would not be illuminating to turn this around and say
that one of the meanings of ‘example’ is case. For ‘case’ is too strongly
associated with other senses, for example, box. (Of course, a given per¬
son’s linguistic background might be such that this would be an effective
way of telling him what ‘example’ means; but this would not be the
standard way to proceed.)
It will undoubtedly not have escaped the reader’s notice that there
is an important difference between our treatments of sentence meaning
and word meaning. With respect to words, we have thus far only given
an account of what it is for two words to have the same use; we have
not also provided a way of characterizing that use which a given pair
of words have in common. Two words have the same use when they
are intersubstitutable in a certain way; but this criterion of intersub¬
stitutability does not itself provide us with any characterization of the
use the words both have. With sentences, on the other hand, we have
done both of these jobs. Two sentences have the same use to the extent
that they have the same illocutionary-act potentials; in specifying the
illocutionary acts in question, we have specified the use(s) that each
of the sentences has. This can be done for each sentence without bring¬
ing in the claim that another sentence has the same use. The notion of
sameness of use is sufficient, as we have seen, for the elucidation of
specifications of meaning; that is, if we can say what the conditions
are under which two expressions have the same use, then we will be
in a position to make explicit what it is we are saying when we say
what an expression means. But there are other contexts in which the
concept of meaning occurs, for example, talk about an expression hav¬
ing a meaning (without specifying what its meaning is) and talk about
learning what an expression means and knowing what an expression
means. A complete account of meaning would involve an analysis of
these notions as well. For such an analysis, the notion of sameness of
use does not suffice. For an expression to have a meaning, it is by no
means necessary that it have even approximately the same use as some
other expression; there are many meaningful expressions for which no
approximate synonym can be found, for example, 'is’ and 'and.’ By the
same token, to learn or know what an expression means is not to leam
or know that it has the same use as some other expression. Even for
those expressions that have synonyms, one can know what the ex¬
pression means without knowing that so-and-so is a synonym of it.
(This reveals a gulf between knowing what an expression means and
being able to say what an expression means, for the latter capacity re¬
quires the ability to specify a synonym.) Where we are able to charac¬
terize the use of an expression, as with sentences, we can provide further
elucidations. Thus, a sentence has a meaning if and only if it has
illocutionary-act potential; and to know what a sentence means is to
know what its illocutionary-act potential is—in the practical, know-how
sense of being prepared to use it to perform certain illocutionary acts
and not others and of being able to recognize misuses—not necessarily
in the theoretical sense of being able to say what its potential is. In
order to do these jobs for words, we will have to develop some way of
characterizing uses of words. It is to be hoped that progress will be
made along this line in the near future.
Analysis of To the extent that this analysis is, or can be made to be, adequate,
illocutionary it has the great merit of showing just how the fact that a linguistic
acts expression has the meaning it has is a function of what users of
the language do with that expression. This result has been achieved
by concentrating on the appropriate unit of linguistic behavior, the
illocutionary act. If this is the line along which meaning should be
analyzed, then the concept of an illocutionary act is the most funda¬
mental concept in semantics and, hence, in the philosophy of language.
Thus far we have taken this concept for granted, relying on the fact
that we have a large battery of terms in common use that stand for
actions of this sort. No doubt, for all practical purposes we are able to
tell well enough when someone is making a certain prediction, a certain
promise, or a certain suggestion; and we are able to tell when the
same illocutionary act and when different illocutionary acts have been
performed on two different occasions. As with anything else, there are
different levels of generality at which an illocutionary act can be speci¬
fied; what someone did on a given occasion could be reported as making
a request: requesting someone to open a door, requesting someone to
open that door, requesting Jones to open that door, etc. But given a
particular level of generality, we can handle the concepts fairly well.
However, we saw earlier that when decisions about meaning become of
theoretical importance, our unformulated capacity to wield the concept
of meaning often falters and explicit criteria are required. When the
notion of an illocutionary act becomes fundamental to the concept of
meaning, these difficult questions of meaning will be seen to turn on
questions of the sameness and difference of illocutionary acts; and in
some of these cases, we shall again need explicit criteria. In the Intro¬
duction to this book, we saw that philosophers find it hard to agree
on whether 1. 'I know that p’ means the same as 2. ‘I believe that p, I
have adequate grounds for this belief, and it is the case that p! On the
analysis of meaning I have presented, this issue rests upon the question
of whether 1 and 2 have the same illocutionary-act potential. But this
question will have no more obvious an answer than the original one. If
I say “I know that p,” and you say, “I believe that p, I have adequate
grounds for this belief, and it is the case that p” are we or are we not
performing the same illocutionary act? Our native ability to handle
illocutionary-act terms does not suffice here. We need an explicit ac¬
count of what it is to perform a given illocutionary act.
Let us tackle this problem by taking a particular illocutionary act,
namely, requesting that someone open a door, and asking what is in¬
volved in performing this act besides uttering a certain sentence or
sentence-surrogate. We have already seen that effects on the hearer are
not essentially involved. Perhaps it has something to do with the situa¬
tion in which the sentence is uttered. Indeed, there do seem to be cer¬
tain conditions that are related in some important way to this sentence.
1. There is a particular door that is singled out by something in the

context. °
2. That door is not already open.
3. It is possible for H (the hearer) to open that door.
4. S (the speaker) has some interest in getting H to open that door.
That these conditions are important can be seen from the fact that if
any of them is not satisfied, something has gone wrong with the request
If 1 or 2 does not hold, there is nothing that anyone could do to comply
wi the request. If 3 does not hold, it would be pointless to make the
request o that person. If 4 fails, we have an insincere request. It is
clear, however, that these conditions are not, as they stand, necessary
conditions of the performance of that act, as, for example, the fact
that a certain door is not already open is a necessary condition of open¬
ing it. If the door is already open, it is logically impossible that I should
open it now. But it is not logically impossible that I should ask you
to open it. I might have been under the mistaken impression that the
door was closed (it was closed the last time I looked). In such a case,
you would not deny that I had made the request in question. You
would not reply, “You’re not asking me to do anything,” but rather,
“What a silly thing to ask me to do!” or “How can I? The door is
already open.” These replies clearly imply that I did make the request.
In a similar manner, the other conditions can also be shown to be
unnecessary for the performance of the illocutionary act. For example,
an insincere request is still a request.
One thing that is incompatible with the supposition that S asked
H to open the front door is for S to (sincerely) reply to H’s retort, “But
the front door is already open,” with “What’s that got to do with it?”
That is, if he is making that request, then he will recognize that a
complaint alleging the nonsatisfaction of one of our four conditions is a
pertinent complaint. (That is not to say that he has to admit that it
is a justified complaint. He may maintain that the condition in question
is, in fact, satisfied. But in so arguing, he is tacitly admitting that the
complaint is pertinent.) We can put this in a less backhanded way by
saying that in making that request S takes responsibility for the satis¬
faction of our four conditions. This is something like the sense of ‘re¬
sponsibility’ in which an administrator is responsible for the efficient
functioning of the departments in his charge. Responsibility for x being
the case is essentially connected with the possibility of being called to
account if x is not the case, and such a possibility can be taken as an
indication of responsibility.
Rules of There is an alternative way of putting this point, which is of

language considerable interest in that it reveals important affinities between
illocutionary acts and various forms of nonlinguistic activity, espe¬
cially moves in games. If we set out to analyze the concept of a serve
in tennis, the problems we encounter will be very similar to those we
have just discussed. To serve is not just to make certain physical move¬
ments, even given certain external circumstances. (I can be standing at
the baseline of a tennis court, swinging a racket so that it makes con¬
tact with a ball in such a way as to propel it into the diagonally opposite
forecourt, without it being the case that I am serving. I may just be
practicing.) Nor are any specific effects required. A shot can have widely
varying effects—it can inspire one’s opponent with fear, despair, exulta-
42 Meaning and the Use of Language <
tion, contempt, or boredom; these variations, however, do not keep it

from being true that one was serving in all these cases. Then what does
change when, after a few practice shots, I call to my opponent, “All
right, this is it,” and then proceed to serve? The new element in the
situation, I suggest, is my readiness to countenance certain sorts of
complaints, for example, that I stepped on the baseline, hit the ball
when my opponent was not ready, or was standing on the wrong side
of the court. I actually serve when, in hitting the ball in a certain kind
of environment, I take responsibility for the holding of certain con¬
ditions, for example, that neither foot touches the ground in front of
the baseline before the racket touches the ball.
In games, when such complaints are made, the plaintiff is said
to be charging the other player with a violation of the rules. And he
can, if necessary, back up his charges by referring to a list of rules for
the game. In this area, the practice of making and accepting com¬
plaints has been “formalized” by the explicit specification of a set of
conditions for the satisfaction of which a player takes responsibility at
a certain stage of the game, and for the nonsatisfaction of which he
will be taken to task. Thus, one can reformulate the above point about
serving in terms of rules. One is serving only if in hitting the ball he
recognizes that certain rules apply to what he is doing. There is no
reason why we should not use the same terminology for illocutionary
acts. Our point about asking someone to open a door would then read
as follows. In order that S can be said to have asked H to open a door,
S must utter an appropriate sentence, s, and recognize that the follow¬
ing rules govern his utterance:
s is not t0 be uttered in that sort of context unless the following con¬

ditions hold:
1. There is a particular door that is singled out by something in
the context.
2. That door is not already open.
3. It is possible for H to open that door.
4. S has some interest in getting H to open that door.
If the definitions given earlier in this chapter are adequate, it is the

rules which are constitutive of illocutionary acts that are crucial for
meaning. For according to these definitions, meaning is a function of.
illocutionary-act potential.
A complete analysis of even the relatively simple illocutionary act
of asking someone to open a door would be beyond the scope of this
volume. But I have said enough to bring out at least one crucial point.
What is required for a given illocutionary act, in addition to the utter¬
ance of an appropriate sentence, is not that certain environmental
conditions actually hold or even that the speaker believe them to hold,
but only that he take responsibility for their holding. In other words,
what is required is that he recognize that what he is doing is governed
by rules requiring that the conditions hold. Thus, the conditions are
involved in the act in a rather subtle way, one that is easily missed.
Having seen this point, we can use our sample act as a schema for the
analysis of any illocutionary act. To get a list of conditions for which
S takes responsibility in performing a given illocutionary act, the fol¬
lowing rule of thumb can be employed. Ask yourself what conditions
are such that if S were to admit overtly that one of these conditions
did not hold, it would be impossible for him, at that time, to perform
the act. (This is logical, not psychological, impossibility. That is, given
this admission, one would not say that he was performing the act.)
Thus, if someone says, “I know that that door is already open, but
would you please open it?” and if he is using ‘I know that that door
is already open’ in the usual way, he can’t be asking you to open that
door. He may be making a joke or testing your reactions to absurd
utterances, but he is not asking you to open a door. If we apply this
test to several different illocutionary acts, we come out with the follow¬
ing lists of conditions.
Advising H to take chemistry.
1. H is not now taking chemistry (or at least not taking a certain
chemistry course singled out by something in the context).
2. It is possible for H to take chemistry.
3. S believes it would be good for H to take chemistry.
Telling H one’s battery is dead.
1. S has a battery.
2. If S has more than one battery, something in the context singles
out one of them.
3. This battery has lost its electric potential.
Expressing enthusiasm for Jones’ plan.
1. Something in the context singles out a certain person named
‘Jones.’
2. This person has put forward a plan.
3. S is enthusiastic about this plan.
Promising to read H’s paper by tomorrow.
1. There is a certain paper of H’s that is singled out by something
in the context.
2. S has not yet read this paper.
3. It is possible for S to read this paper by tomorrow.
4. S intends to read this paper by tomorrow.
Putting the point in this way shows us how the rule-governed

character of language is crucial for semantics. Linguistic activity is sub-
ject to more obvious sorts of rules, which are not so intimately related
to meaning, and in some cases, not related at all. Linguistic behavior,
like most other forms of behavior, is subject to moral rules and rules
of etiquette. However, the fact that it would be impolite in certain
circumstances to say Your false teeth are loose” can play no part in
determining the meaning of that sentence. (Many other sentences with
very different meanings, for example, “The food is tasteless” would be
impolite in just the same way.) Again, there are grammatical rules that
govern the way words can be put together to form sentences. But al¬
though the fact that ‘desk’ can be inserted in the blank in the sentence,
‘I’ve just bought a-,’ whereas ‘if,’ ‘into,’ ‘scribble,’ and ‘lovely’
cannot, tells us something about what ‘desk’ means, it does not tell us
very much. It does not distinguish ‘desk’ from many other expressions
with different meanings that could be put into that slot, for example
house,’ ‘dog,’ and ‘share.’ ^ ’
Problems To help give a more concrete idea of the theory of meaning here
concerning presented, I shall provide an indication, however sketchy, of the
synonymy way in which it would be applied to a particular problem concern¬
ing meaning. For this purpose, I shall take the problem of sy¬
nonymy. ■’
It is often said that it is impossible to find a pair of words that
are exact synonyms. This impossibility, or at least great difficulty, is
reflected in the definitions given earlier in this chapter. According to
my theory, two words are synonymous to the extent that they are inter-
substitutable in sentences without altering the illocutionary-act poten¬
tials of the sentences. Perfectly synonymous words would be so inter-
substitutable m every sentence. It was because of the difficulty of
establishing complete synonymy that I defined ‘W, means W„’ so as to
require only that W, and W2 be intersubstitutable in most "sentences
without altering the illocutionary-act potential. Now I want to look
more carefully at the factors that prevent complete synonymy.
, The main r^son why it is so difficult to find exact synonyms is
that practically all words have more than one meaning. The more
meanings a given word has, the more unlikely it is that another word
wfll have exactly the same range of meanings over the same range of
contexts. Thus, although ‘sick’ and ‘ill’ share the meaning not well in
many contexts each has other meanings that are not shared by the
o er Thus, ill, but not ‘sick,’ can mean unfavorable, as in ‘bird
i omen; and^ sick, but not ‘ill,’ can mean tired, as in ‘I’m
sk: o oing t at. Insofar as restrictions on synonymy are due to a
understand.mCldenCe °f demarcated senses> th^ ™ easy to

But even if we restrict ourselves to those contexts within which

a pair of terms seem to have exactly the same meaning, we are not at
the end of our troubles! For even in such contexts, there are various
differences that attach to the use of terms; and if these differences are
differences in meaning, we may well be left with the uncomfortable
conclusion that no two terms have exactly the same meaning in a given
context. Let us survey some of these differences and then tackle the
question as to which, if any, are to be called differences in meaning.
1. The social environment in which the utterance of a word is
appropriate. We get something of this difference with ‘sick’ and ‘ill,’
the latter being more suited to polite discourse. A stronger contrast
of this sort is found in ‘sweat’ and ‘perspire.’ Almost any two “syno¬
nyms” will exhibit this difference to some extent.
2. Associations. Any two words will show this sort of difference,
but with many pairs it is not easy to give an adequate formulation.
Consider ‘earth’ and ‘ground.’ ‘Earth’ conjures up all sorts of associa¬
tions—earth mother, fertility, earthy qualities in people, the source of
our being—that are lacking for ‘ground.’ Try rewriting Keats’ lines,
“Oh, for a draught of vintage that hath cool’d a long age in the deep-
delved earth,” as “Oh, for a drink of wine that has been reduced in
temperature over a long period in ground with deep furrows in it.”
Such a paraphrase exhibits a number of differences in associations
between near synonyms.
3. “Emotive force.” We can find pairs of words that seem to be
synonymous apart from the fact that one carries a certain attitude or
evaluation while the other carries a different one or none at all. Ex¬
amples are ‘stool pigeon’ and ‘informant for the police,’ ‘inexpensive’
and ‘cheap,’ and ‘office-seeker’ and ‘candidate for office.’ We must
distinguish cases where the attitudinal force seems to be the only
possible difference in meaning from other cases in which this difference
goes along with others. Good examples of the latter are to be found in
the so-called “emotive conjugations” inaugurated by Bertrand Russell.
I am firm, you are obstinate, he is pig-headed.

I am righteously indignant, you are annoyed, he is making a fuss over
nothing.
I have reconsidered, you have changed your mind, he has gone back
on his word.
It is clear that ‘gone back on his word’ simply does not mean the same
as ‘reconsidered,’ quite apart from the attitude conveyed, and likewise
with ‘righteously indignant’ and ‘making a fuss about nothing.’
If differences like these are differences in meaning, then we do
not have exact equivalence in meaning between any two words, even if
Meaning and the Use of Language
we restrict ourselves to one sense for each word. (In fact, this conclusion
would follow even if only 2 is a difference in meaning.) But are they
differences in meaning? Plausible arguments can be marshalled on both
sides. On the negative side, it can be pointed out that in telling someone
what earth means, we do not go into an account of the associations
typically aroused by it; and in defining ‘stool pigeon,’ we do not go into
the fact that to call someone a stool pigeon is to insult him. But against
this view, it can be argued that the specifications of meaning that we
give in everyday life, and even in dictionaries, are rather crude affairs
and cannot safely be taken as a guide in theoretical matters. Moreover
if the concept of linguistic meaning is such that to know the meaning
of a word is to be able to use it correctly, then a theoretically complete
specification of meaning would have to include anything relevant for
such guidance. And it would seem that if we do not realize that ‘stool
pigeon’ is a term of abuse or that ‘sweat’ is a relatively vulgar term, we
are not going to use it correctly.
On the account of meaning we have presented, these questions
would be resolved by determining whether, for example, ‘I’m sweating’
has the same illocutionary-act potential as ‘I’m perspiring,’ whether ‘it
has been cooled in the deep-delved earth’ has the same illocutionary-act
potential as ‘it has been reduced in temperature in ground with deep
furrows m it,’ and whether ‘he is a stool pigeon’ has the same potential
as he is an informant for the police.’ And given the account of illocu¬
tionary acts just presented, this would be determined in turn, by de¬
termining whether in uttering ‘I’m sweating,’ in normal circumstances,
1 would be taking responsibility for any conditions other than those
for which I would be taking responsibility in uttering ‘I’m perspiring ’ or
vice versa; and so for the other pairs of sentences.
Proceeding in this way, it seems clear to me that 2 does not involve
a difference in meaning. I cannot see that in saying ‘It came from the
earth I am taking responsibility for any conditions over and above
those for which I am taking responsibility in saying ‘It came out of
e ground. The fact that two words will normally call up different
sorts of associations seems to be a fact over and above anything I am
disposed to allow myself to be called to account for. It is not as if
I will recognize the hearer’s right to complain if ‘earth’ does not evoke
nc Poetlc associations m his mind. With respect to 1, I am inclined to
give the same verdict although this point is arguable. It is true that
may be taken to task for using the word ‘sweat’ at a ladies’ tea and
I may recognize the justice of this. But this is not to say that the
restriction of social context has any implications for what is being said-
we have already noted that linguistic behavior, along with behavior of
other sorts, is governed by rules that have no semantic importance. For
a condition to be involved in the nature of an illocutionary act, it has

to be such that if one were to overtly admit that the condition does
not hold, he could not then be taken to be performing that illocutionary
act. And it does not seem that one takes responsibility in this way for
being in a certain kind of social context when one says “I’m sweating.”
That is, if one says “I know that I am at a D.A.R. tea, but still I’m
sweating,” he could be saying just the same thing by uttering ‘I’m
sweating’ in this context as he would be saying by uttering that sentence
on the squash court. The social context restriction affects not what is
said but how it is said.
Emotive With respect to 3, we seem to be in a different case, although we

meaning must not forget to distinguish between expressing a certain atti¬
tude or feeling and evoking that attitude or feeling in the hearer.
The mere fact that ‘He’s a stool pigeon,’ unlike ‘He’s an informant for
the police,’ tends to elicit unfavorable attitudes toward the person to
whom the term is being applied does not suffice to show any difference
of meaning between the two. This is a difference in perlocutionary-act
potential. But there is a difference in meaning if in saying ‘He’s a stool
pigeon,’ one is taking responsibility for having an unfavorable attitude
toward him and is not taking any such responsibility in saying ‘He’s an
informant for the police.’ In other words, there is a difference in mean¬
ing if, having said ‘He is a stool pigeon,’ I am prepared to recognize that
a response like “What’s wrong with what he is doing?” is not out of
order. And it does seem that there is such a difference in the cases cited.
Thus, our theory leads us to recognize the legitimacy of the term ‘emo¬
tive meaning’ in some of its applications. We can distinguish between
the “emotive meaning” and the “cognitive meaning” of a sentence
insofar as we can distinguish, within the class of conditions for which
a speaker would take responsibility in uttering the sentence, between
those that have to do with the feelings and attitudes of the speaker
and those that have to do with other things. Thus, we might list the
following conditions for ‘He’s a stool pigeon.’
1. Some particular male person is singled out by the context.

2. This person is an informant for a police organization.
3. S has an unfavorable attitude toward this sort of activity.
We can say that 1 and 2 contribute to the “cognitive meaning” of the

sentence and 3 to the “emotive meaning.” But we would not be justified
in speaking of the emotive meaning of ‘communist’ just on the grounds
that it typically evokes unfavorable reactions, apart from any regular
practice of using it in such a way as to take responsibility for the
existence of unfavorable attitudes in the utterance of it.
It is worthy of note that by and large our approach preserves and

provides justification for the common distinction between what is said,
the way it is said, and the effects that saying it has, along with the
parallel distinction between what is meant, the way of expressing what
is meant, and the effects that this expression has. As these distinctions
are ordinarily made, differences in social context would be said to be¬
long to the second category, differences in association to the third, and
differences in emotive force to the first and third. Everyone would agree
that these distinctions will have to be made at some point; no one
would suppose that the fact that someone spoke loudly, aggressively, or
in a Brooklyn accent would have to do with what he said rather than
with how he said it. And if my telling you that it is raining leads you
to weep, it would be universally agreed that this is a fact about the
effects of what I said rather than about what I said. One merit of the
illocutionary-act account is that it justifies drawing these distinctions in
just about the place they are ordinarily drawn.
Problems The nature and variety of illocutionary acts is of interest for the phi-
about losophy of language not only because of their crucial place in the
Illocutionary analysis of meaning, but for other reasons as well. In virtually every
acts branch of philosophy, the analysis of one or another sort of illocu¬
tionary act sometimes takes over the center of the stage. In logic
and epistemology, it often becomes important to get clear as to what it is
to make a statement or assertion and as to the conditions under which
we have the same statement or assertion made on two occasions. For ex¬
ample, much of the discussion of the nature of truth hangs on whether
in saying 1. 'It's true that caviar is expensive,’ I am making just the
same statement (if I am making a statement at all) as in saying 2
Caviar is expensive.’ The defenders of the correspondence theory of
truth, according to which the truth of a statement consists in its cor¬
respondence with the facts, hold that in 1, we are not making a state¬
ment about caviar at all, but quite a different statement about the
statement made in 2. Some of the critics of this theory maintain that in
1 we are not making a statement at all, but rather performing some
other kind of illocutionary act, such as endorsing, conceding, or admit-
mg w at someone else has said. Other critics maintain that truth will
lose its mysterious aura once we realize that sentence 1 is just a more
emphatic way of making the very same statement that is made in 2.
The analysis of illocutionary acts also becomes of crucial im¬
portance in ethics. A great deal of ethical theory is concerned with get-
mg clear as to what we are doing when we make moral judgments.
We must be clear about this if we are to know what considerations are
appropriate for supporting and criticizing such judgments. Actually
‘moral judgment’ is a blanket term, covering a loosely organized group

of types of illocutionary acts—reprimands, behests, injunctions, exhorta¬
tions, imputations of obligation, etc. The various positions in ethical
theory can be most profitably distinguished by the different positions
they take on the nature of such illocutionary acts. Thus, naturalists in
ethics hold that in telling someone that he ought to do something,
one is making a special kind of statement of empirical fact. Naturalists
differ as to the content of the statement; one version is that such state¬
ments say something about the consequences of the action in question
for human welfare. Emotivists, on the other hand, tend to assimilate
imputations of obligation to expressions of feelings and attitudes. It
seems that this problem could be discussed more effectively if an ade¬
quate method for analyzing illocutionary acts were available.
4
LANGUAGE
AND ITS NEAR RELATIONS
Thus far I have been taking the concept of language for granted.
It is high time I undertook to give an explicit account of the nature
of language and of what distinguishes it from more or less similar
matters. This task is best approached by looking at the relation between
linguistic elements on the one hand and various more or less similar
items, such as signs, signals, diagrams, pictures, and religious symbols,
on the other.
The generic Many theorists have supposed that items of all these sorts can be
notion of usefully grouped together under the heading of “signs.” Words
a sign (and other linguistic units) would then be one sub class of this
genus; language would be made up of one particular kind of sign.
Thus, the following facts would all be regarded as cases of “sign-func¬
tioning.”
1. Boulders of this sort are a sign of glacial activity.

2. A hum like that indicates a loose connection in the wiring.
3. That expression on his face means trouble.
4. That is a sample of forest green wall paint.
5. This is a diagram of an 80-watt power amplifier.
6. In early Christian art, a ship symbolized the church.
7. When the umpire moves his hands horizontally, palm down, that
means safe.
8. A red light means stop.
9. Four bells indicate fire.
10. ‘Plume’ denotes pens.
11. ‘Oculist’ denotes eye doctors.
12. Equiangular’ connotes the property of having all angles equal to
each other.
50
Language and Its Near Relations 51
13. ‘Uncle Sam’ is a nickname for the U.S.A.

14. ‘Pinochle’ is the name of a game.
Before we can accept the idea that words are to be regarded as

one kind of “sign,” it must be shown that all the facts on this list have
something important in common. That is, it must be shown that there
is some one sense of ‘sign’ that applies throughout the list. It is not
obvious that there is. Note that as ‘sign’ is ordinarily used, it does not
apply so widely. What would it mean to say “ ‘Plume’ is a sign of pens”
or “This diagram is a sign of an 80-watt amplifier”? (And the same is
true of ‘symbol,’ ‘signal,’ or any other semiotic term, which one might
try to apply generally to all items on the list. For example, a hum is not
a symbol of a loose connection in the wiring, nor is ‘oculist’ a signal of
eye doctors.) Thus, the general sign theorist must be using ‘sign’ in a
technical sense and it is up to him to tell us what that is. Peirce’s
definition may be taken as typical. “A sign ... is something that stands
to somebody for something in some respect or capacity.” 1 In this
definition, the weight is being placed on the term ‘stand for.’ (Other
definitions rest on similar terms like ‘represent.’) Is there any sense of
‘stand for’ in which in each case we have one thing standing for another?
Perhaps the most plausible move is to define ‘stand for’ as call
to mind. Thus, the claim would be that what makes each entry on the
list a case of sign-functioning is that, in each case, a part of what we are
saying is that one thing calls the other to mind. (The rest of what
we are saying would he in what distinguishes one kind of sign from
another.) But this suggestion will not survive scrutiny. It is clear that
if boulders of a certain kind are a sign of glacial activity, they were
a sign of glacial activity before anyone realized this. In fact, they would
still be signs of glacial activity, even if no one should ever realize this.
(This reasoning may be applied to cases 2 and 3 as well.) That means
that the boulders would still be a sign of glacial activity even if they
never called glacial activity to mind for anyone. When “general sign
theorists” treat cases like 1 through 3 (signs in the ordinary sense of the
term), in effect, they make use of the notion of x being taken as a
sign of y, rather than the notion of x being a sign of y. Calling to mind
may be essentially involved in the former notion, but these two notions
are quite different. We have already shown that we can have the latter
without the former. And superstitions show that we can have the former
without the latter. Black cats are often taken as a sign of bad luck,
but that does not show that they are a sign of bad luck.
1 C. S. Peirce, Collected Papers (Cambridge, Mass.: Harvard University Press, 6

vols., 1931-35), Vol. II, paragraph 228. All future references to material in these
volumes will be given by specifying the volume followed by a period, followed by
the paragraph number, as: 2.228.
52 Language and Its Near Relations 4
With respect to the other items on the list, calling to mind seems
to be involved in some way. It seems clear that something could not
be a diagram of an amplifier unless it sometimes led someone to think
of an amplifier; and it seems that ‘Uncle Sam' could not be a nickname
of the U.S.A. unless it sometimes happened that a presentation of this
name evoked a thought of the U.S.A. Even so, it does not seem to be
the case that each of these “signs” calls to mind the appropriate object
on each occasion on which it is performing its normal function. Is it
always the case, when a ship is functioning as a symbol of the church,
that seeing the ship in a painting brings the church before the mind?
And is it always the case, when one understands an utterance involving
the word ‘oculist,’ that the idea of oculist pops into one’s consciousness
as a result of hearing the word? The considerations we brought up in
Chapter 1 in connection with the ideational theory of meaning are
relevant here. As pointed out there, it seems impossible to verify the
proposition that such ideational effects take place consistently. And
insofar as they do not, it cannot be claimed that “sign-functioning,”
even for these kinds of items, consists in x calling y to mind.
Of course, one might modify the definition of ‘x stands for y’ to
read x calls y to mind given certain appropriate conditions. But the
trouble with this suggestion is that any x will call any y to mind, given
appropriate circumstances. In a way, this is all right. We would want
to frame a general definition of ‘sign’ in such a way that anything could
be a sign of anything else. But as terms like ‘sign of,’ ‘symbol of,’
‘means,’ ‘indicates,’ and ‘diagram of,’ are actually used, their force is
much stronger. To say that x is an indication of y is not just to say
that one could condition a person so that presenting x to him would
bring y before his consciousness. There is a distinction between x ac¬
tually being an indication of y, and it merely being possible that x is
an indication of y. It is no accident that these terms have this force.
If it is true that we could set up an association between any x and any
y, then to say that two things are so related that an association could
be set up between them would be to say nothing about those terms.
(That is, it is to say nothing, though it is to say something about the
associative process.)
Finally, there is the point that calling to mind extends more widely
than sign-functioning. Wherever there is any kind of ideational asso¬
ciation, we have a person so conditioned that x calls y to mind. Thus,
my childhood experiences might have been such that every time I see
an apple tree, it brings to mind my grandparents’ house in the country.
But it does not seem that this kind of phenomenon has important
affinities with the items on our list. Of course, having cut ourselves loose
from any ordinary sense of ‘sign,’ we can, if we wish, count this as an
example of sign-functioning”; but if we do, one might well wonder

whether the generic notion of 'sign’ contains anything that is im¬
portant for the kinds of cases we were originally trying to understand.
Other attempts to formulate a generic feature of “signs” are sub¬
ject to similar criticisms. In Peirce’s writing, and more explicitly in
Morris’, we find an account of 'stand for’ in terms of 'taking account
of, where taking account of’ is conceived in behavioral rather than
ideational terms. That is, x will be said to be a sign of y for A to the
extent that on being presented with x, A is led to take account of y.
If we try to think this through, we will discover difficulties very similar
to those we encountered in connection with behavioral theories of
meaning. Such an account will not work even for natural signs of objects
or events that have a bearing on our activity in the immediate future.
One does not actually have to do anything to prepare for rain in order
to take dark clouds as a sign of impending rain. When we go beyond
signs, ordinarily so called, to other items on the list, it is even less
plausible to suppose that any behavioral “taking account” of the object
is always, or even typically, present. To recognize a diagram as a dia¬
gram of an 80-watt amplifier is not to do something that would consti¬
tute “taking account” of such an object. It is not as if one would natu¬
rally expect to find an 80-watt amplifier in the immediate vicinity.
Still less does hearing the word ‘oculist’ typically lead one to get set for
the aproach of an oculist (the announcing use of the word is not sta¬
tistically predominant), nor does seeing a ship in an early Christian
painting lead one to “take account” of the church (whatever that
would be).
At this point, I propose that we abandon the attempt to provide
a justification for our intuitive sense that all the facts on our list have
something important in common. The review of possibilities we have
just completed gives little ground for hope that we will ever be able to
find a single sense of ‘stand for’ that is applicable to every case. If there
is a point in classing them together, it is not because of any feature
common to them all, but because of a “family resemblance” between
them in respect to several different features, none of which is shared
by all.2 Whether or not there is anything to this notion, it will be in
line with our ultimate goal of understanding the nature of language, and
its similarities with and differences from closely related matters, to turn
now to the task of bringing out the major differences between sub
classes of our original list.
2 For the notion of ‘family resemblance’ among the things to which a term is
applied, see L. Wittgenstein, Philosophical Investigations, tr. G.E.M. Anscombe
(Oxford: Basil Blackwell, 1953).
54 Language and Its Near Relations
Regularity of Given any list, there are various ways in which it can be sub-
correlation divided. With this one, we could, for example, discriminate be-
and regularity tween man-made signs, which would include 3 through 14, and
of usage those that exist apart from human contrivance, 1 and 2. Or we
could distinguish between those that fit into an elaborate system
of “signs,” 10 through 14, and those that do not, 1 through 9. We will
get the most penetrating first division by considering the kind of
justification that would be given for each statement on our list. With¬
out attempting a complete account of the justification in each case, we
can note some important differences. Cases 1 through 3 and 7 through
14 are distinguished in the following way. A statement in the first group
would be justified by claiming that, in fact, x and y3 are always or
generally correlated in a certain way. Thus, 1 is justified by showing
that, in fact, wherever one finds boulders like this, glacial activity has
gone on in the past; 3 is justified by showing that, generally, when an
expression like this has appeared on the face of this person, he has
caused trouble in the near future. The correlation always involves a
more or less definite spatio-temporal relationship between the x and y,
but this differs from case to case. In 1, it is spatial identity, with y
being before x. In 2, it is both spatial and temporal identity (insofar
as hums can be precisely located spatially). In 3, y is after x in time,
and they are spatially related not by being in the same spatial location
but by being connected with the same organism. By contrast, a state¬
ment in the 7 through 14 group is justified by showing that there is
something about the way in which the “sign” is used that makes it re¬
lated to something else in the way specified. It is clear that there is
nothing about a certain gesture, apart from rules governing the umpiring
of baseball games, that makes it an indication that the runner is safe,
any more than there is anything apart from the rules governing the
English language that brings it about that 'oculist’ denotes eye doctors.
Moreover, for each group, the kind of justification relevant for the
other group is not involved. We have already seen that it is neither
necessary nor sufficient for boulders of a certain sort being a sign of
glacial activity that anyone respond to them, much less use them, in
a certain way. It is equally true, although less obvious, that correlations
are not essentially involved in the justification of 7 through 14. No
doubt, there will often be some rough correlations for “signs” like these.
Unless it were quite often true, in the community in question, that
there was a fire when four bells were sounded, then that signal would
no longer be used as an indication of fire. Again, if umpires did not by
3 Using these variables for whatever fits into the appropriate slots in the schema
of the form, ‘x is a sign of y,’ which we are taking as the form of all the statements
on our list, despite the fact that we have been unable to find any sense of 'sign' in
which all the statements on the list do assert that one thing is a sign of another.
and large use the signal specified when and only when a runner was
safe, things would break down. But the relation is quite indirect. To
say that the ringing of the bell or the gesture has the significance it has
is not to say that such a correlation holds. One way to see this is to
note that with cases 1 through 3, where a correlation is what is being
asserted, if the correlation is not universal but only holds for the most
part, we are to qualify the assertion of the sign-relation. Thus, if a cer¬
tain expression on his face is only followed by trouble most of the time,
we should not make the unqualified statement, 3, but a qualified one,
such as “That expression on his face usually means trouble,” or “That
expression is a fairly reliable sign of trouble.” However, the fact that
sometimes the gesture specified in 7 is given when the runner did not
touch the bag before being tagged with the ball by an opposing player
(after all, umpires are fallible and perhaps even occasionally dishonest)
is no ground for qualifying 7 by saying that this gesture sometimes
(usually) means safe. And still less is the fact that oculist’ is often ut¬
tered when there are no eye doctors around (or in any predictable spatio-
temporal relation to the utterance) any reason for altering 11 to “ ‘Ocu¬
list’ sometimes denotes eye doctors,” or “ ‘Oculist’ denotes eye doctors
to a certain extent.” That is, even though correlations with the “ob¬
ject” may be indirectly involved, they are not crucial for what is being
said when we say what a linguistic expression or signal means, denotes,
or indicates.
Icon, Index Peirce has made popular a threefold distinction of “signs” into
and symbol icon, index, and symbol.4
Icon—a sign which refers to the Object that it denotes merely by
virtue of characters of its own. . . . (2.247)
Index—a sign which refers to the Object that it denotes by virtue of
being really affected by that Object. (2.248)
Symbol—a sign which is constituted a sign merely or mainly by the
fact that it is used and understood as such. . . . (2.307)
This distinction, in terms of that in virtue of which the sign is a
sign of something, is very similar to the distinction I have drawn in
terms of the kind of justification that could be given for different state¬
ments on our list, except that in my version, one is not committed to
to the assumption that there is some one sense of ‘sign’ in which, in all
these cases, we have one thing functioning as a sign of another. It
should be clear that the two classes we have so far demarcated are very
close to Peirce’s index and symbol. We have extended the notion of
“being really affected by” to cover any sort of de facto correlation, but
4 For a good discussion of Peirce’s trichotomy see A. W. Burks, “Icon, Index, and
Symbol,” Philosophy and Phenomenological Research, Vol. IX, June, 1949.
otherwise there is little difference. Henceforth, I shall refer to “natural

signs,” as in cases 1 through 3, as indices, and items like 7 through 14
as symbols.
There are other interesting distinctions between indices and sym¬
bols, although they are often overstated. Thus, it would be a mistake
to say that symbols are used in communication and indices are not.
It is perfectly possible for indices to be used in communication, as when
I bare my chest to show you that I have been shot, or as when a ma¬
rooned sailor keeps a fire going in the hope that someone in a passing
airplane or ship will see it and realize that someone is on the island.
What is a decisive difference is that the status of indices, unlike that of
symbols, does not depend on their being used in communication. This
has the interesting implication that even when x is being used in com¬
munication as an index of y, an interpreter can correctly take it to be
an index of y without realizing that it was produced or exhibited for
purposes of communication. Thus, a passing aviator can quite rea¬
sonably and quite correctly take the column of smoke as an indication
of human habitation without realizing, or even hypothesizing, that the
smoke was produced in order to communicate that idea. Contrast this
state of affairs with one in which the marooned sailor is sending up
smoke signals according to some widely used code. In this case, the
aviator could not take a certain pattern of smoke to mean I have no
food without supposing that it was produced with an intent to com¬
municate. Of course, he might notice that the smoke he took to be
produced naturally exhibited that kind of pattern; just as people may
find marks on stones, which are very much like Phoenician inscriptions,
but which they take to be due to weathering. In such a case, however,
the aviator would not regard the smoke as meaning, I have no food.
Instead, he would say that the smoke looks like signals that have that
meaning.5 With respect to some indices, for example, yawns, the very
reverse is true. If we believe that a given yawn was deliberately produced
in order to make the spectators think that the person is sleepy, then
we will not take it to be a reliable indication of sleepiness.
The notion of It is commonly said that symbols (in Peirce’s sense) are distin-
convention guished from other “signs” by the fact that their significance is
conventional. I have largely avoided this term because it generally
carries unjustified, and very probably untrue, assumptions about the
origins of languages.
5 This point is also brought out by the fact that before we can identify a given
sound pattern as a word, for example, we must locate it in one or another language.
The sound pattern ‘link’ constitutes one word in English, another in German. No
such problem is involved in the identification of indices.
. . . After one person or group decided to use this to stand for that
other people decided to do the same thing, and the practice spread!
that is, these symbols were adopted by common convention. . . ,6
A symbol, such as a word, designates a referent by agreement or con¬

vention. Human decisions are thus required in order to establish the
meaning of symbols and such decisions are arbitrary ones. . . . Names
arise as a result of human agreements, or stipulations.7
I think these passages are typical of the less guarded remarks on

this subject to be found in the literature. But on reflection, we can see
that language, as such, could not have originated by having decisions
adopted by “common convention.” As Russell has said, “We can hardly
suppose a parliament of hitherto speechless elders meeting together and
agreeing to call a cow a cow and a wolf a wolf.” 8 By the nature of the
case, making agreements and conventions presupposes that people al¬
ready have a language in which to carry on these activities. No one
knows how language originated, but at least we can be certain that it was
m no such way as this. However this does not show that the words in
languages spoken today did not get their meanings by convention—for
any such word acquired its meaning after a language was already being
spoken in the community in question—but all the evidence is against it.
We know very little about the mechanisms by which new words come
into being and old words change their meaning, but what we do know
about it indicates that conscious decision and deliberately adopted con¬
ventions have very little part to play. There are cases in which new
senses of words are explicitly proposed, as when Peirce proposes to use
‘icon’ to mean sign which refers to its object merely by virtue of charac¬
ters of its own. And there are cases in which a word, or a meaning of a
word, is adopted by convention, as in the stabilization of scientific
terminology by scientific congresses. But these are exceptional cases,
pretty much confined to technical terminology. For the rest, semantic
change seems to be largely an unconscious affair, a matter of habits get¬
ting established without anyone or any group trying to establish them.
Like the social contract theory in political science, the idea that
words get their meaning by convention is a myth if taken literally.
But like the social contract theory, it may be an embodiment, in mythi¬
cal form, of important truths that could be stated in more sober terms.
It is our position that this truth is best stated in terms of the notions
of rules. That is, what really demarcates symbols is the fact that they
6 J- Hospers, An Introduction to Philosophical Analysis (Englewood Cliffs N T •

Prentice-Hall, Inc., 1953), p. 2. ’
7 L. Ruby, Logic (Philadelphia: J. B. Lippincott, 1950), p. 20.
8B. Russell, The Analysis of Mind (London: George Allen & Unwin 19211
„ inn ° ’ >•
have what meaning they have by virtue of the fact that for each there
are rules in force, in some community, that govern their use. It is the
existence of such rules that is behind the fact that they are “used in
a certain way,” in the sense of this phrase that is relevant here. In
Chapter 2, we made a sketchy beginning at indicating just how it is
that the meaning of linguistic expressions is a function of the way cer¬
tain kinds of rules govern their employment. Henceforth, we shall feel
free to use the term ‘conventional’ purged of misleading associations,
as shorthand for “on the basis of rules.”
icons, pure We have not yet identified Peirce’s icons on our list. Can cases
and impure 4 through 6 be so identified? Before answering this question, we
shall have to get clearer as to the nature of an icon. As we saw,
Peirce defined an icon as a sign that signifies its object merely by virtue of
its intrinsic characteristics, rather than by a causal or a “conventional”
connection. It seems clear that x can signify y on the basis of its (x’s)
characteristics only if it is similar to y with respect to these character¬
istics. Hence, we can also define an icon as a sign that signifies an object
by virtue of similarity to the object. It seems clear that similarity plays
a crucial role in 4 through 6. A paint sample can play its role only if it is
the same color as the paint of which it is a sample. The similarity in¬
volved in 5 and 6 is more abstract. In 5, it is a structural similarity
between the spatial relations that hold between parts of the amplifier
and the spatial relations that hold between corresponding elements
of the diagram. That is, by looking at the relative position of two
elements of the diagram, we can tell something about the relative posi¬
tion of the corresponding parts of the amplifier. In 6, a ship is fitted to
be a symbol of the church because of a similarity in function. Just as a
ship protects voyagers from the water and conveys them to their destina¬
tion, so the church, it is believed, protects men from the snares of the
world and conveys them to their ultimate destination. However, it is
equally clear that conventions are involved in all these cases. This is
most evident in 5, where we have to arbitrarily set up a correlation
between elements of the diagram and elements of the amplifier before
the structural similarity comes into play. This correlation can be es¬
tablished by having circles, for example, represent tubes and lines repre¬
sent wires, etc., by verbally labelling various items on the diagram, or
by a combination of these techniques. In 6, the ship is well suited to
symbolize the church by reason of the similarity just mentioned; how¬
ever, it is an artistic convention that the church is symbolized in this
way rather than by other objects that would be equally well suited by
reason of similarity, for example, a fortress. In 4, the main point is that
there is a convention that selects from among the various characteristics
of the sample those with respect to which it is functioning as a sample.

No one expects the paint to have the same texture as the sample or to
be of the same shape or weight. There is a convention operative accord¬
ing to which it is the color of the sample that is crucial. Thus, all of
these cases are mixtures of icons and symbols.
The question arises as to whether there could be such a thing as
a pure icon. It would seem that so long as we have pure similarity and
nothing else involved, the “sign” would not have the typical use of icons,
namely, that of helping someone acquire certain kinds of information
about something. For if we just present someone with an object and
leave it at that (perhaps making it clear that he is to take it as similar
in some way to something), he has no way of knowing either to what
he is to take it as similar or in what respects he is to take it as similar.
It would seem that by any such procedure, we could not accomplish
what is typically accomplished by diagrams, maps, samples, and photo¬
graphs. The process is too uncontrolled. Of course, icons are not always
used to convey information. Representation of the church as a ship
in a painting does not have this purpose. But even here, we need a
convention operative to make clear what the ship is intended to sym¬
bolize. We do get similarity operative all alone in the association of
ideas; but as we saw earlier, it is doubtful that this should be counted
as “sign-functioning” at all, even on the most liberal criteria. Perhaps the
most plausible candidates for pure icons are to be found in primitive
religious symbolism and in dream symbolism. In a primitive religion,
ritual acts may focus around a bull, a mountain, or a sacred fire, without
there being an explicit account of what gives these objects the signifi¬
cance they are felt to have. It might be suggested that what is hap¬
pening here is that the bull, for example, is being treated as sacred be¬
cause of certain characteristics it embodies to a marked degree, such as
virility, even though none of the worshippers has formulated this as
such. The bull could then be said to function as an icon of virility
(or of other virile things), without this functioning being tied down
by any conventions or rules. Again, in a dream, mountains may be
functioning as a symbol of the dreamer’s mother, solely on the basis
of some similarity that the dreamer sees (without necessarily being
aware that he sees it). But in areas such as these, it is notoriously diffi¬
cult to get a clear idea of just what is going on.
Language as It should be clear that language belongs somewhere within the

a system category of symbols, in Peirce’s sense of the term. Language is
of symbols often defined as a system of symbols, and this can be accepted as
a summary statement. However, it will have to be elaborated
before it is very informative.
First, let us try to get a more concrete idea of the sense in which
a system is involved. 1. The elements of language, such as words, are
combinable in some ways and not in others; and the meaning of the
combination is a determinate function of the meanings of the con¬
stituents and their mode of combination. (We can have ‘Come in
now’ but not ‘Now in come’ or ‘Come although now.’) 2. Each constit¬
uent of a sentence can be replaced by certain words and not by others.
This is partly just another way of saying what was said by 1. (Thus, ‘in’
in ‘Come in now’ can be replaced by ‘over’ or ‘through,’ but not by
‘bookcase’ or ‘impossible.’) 3. A new sentence can be constructed by
transforming an old sentence in a certain kind of way, with a certain
kind of alteration of meaning always attaching to a certain kind of
transformation. Thus, the kind of transformation involved in going
from Brooks resolved the problem’ to the ‘The problem was resolved
by Brooks’ carries with it a certain kind of meaning relation between
the two.
The notion of language as a system of symbols will be misleading
if we suppose that each of the symbols that enters into the system is
what it is independent of its involvement in the system, so that it could
be just the same symbol if it were in no system at all. A word is identified
only through an analysis of the speech that is carried on in a certain
community. We are so accustomed to the rather rudimentary analysis
of our speech, which is involved in our writing system, that we are likely
to think of it as an immediately obvious feature of the nature of things.
In fact, the concept of a word represents a certain way of analyzing
utterances into repeatable segments or segment-types; what is to count
as two utterances of the same word, rather than utterances of two
different words, is always more or less a question as to what decision
will give us the most useful way of representing the language. Are ‘is’
and ‘am’ two words or two forms of the same word? How about ‘wave’
as a noun and ‘wave’ as a verb? Or ‘ox’ and ‘oxen’? If we say that the
first and third examples involve two different forms of the same word,
while the second involves two different words, it is abundantly clear
that the decision is not solely on similarity in sound pattern. This point
can also be seen from the fact that in dealing with different dialects
we will count ‘aw’ (Cockney) and ‘high’ (“standard” English) as the
same word, even though the latter is much more similar in sound to the
“different” word “nigh” than it is to the former. Thus, the elements
making up the system that constitutes a language are not items that
might have been what they are apart from any such system.
The above remarks bring out more than one way in which lan¬
guage is abstract. In this connection, we should keep in mind the often
repeated, but seldom consistently observed, distinction between lan-
guage and speech. Speech comprises the totality of verbal behavior that
goes on in a community; whereas language is the abstract system of
identifiable elements and the rules of their combinations, which is
exemplified in this behavior and which is discovered by an analysis of
the behavior. Not only the system as a whole, but also each element
thereof, is an abstraction from concrete behavior. (This is a conse¬
quence of the fact that the element cannot be identified apart from
an analysis of the system.) We have just noted briefly the impossibility
of identifying a word with a certain sensibly recognizable combination
of sounds. A word is a certain disjunction of sound patterns, for exam¬
ple, 'aw’ and ‘high,’ such that whenever one of these is exemplified
(perhaps with further restrictions as to the linguistic environment in
which the exemplification takes place), we will say that we have an
example of that word. Thus, a word is more abstract than a melody, for
example. It has the same degree of abstractness as a type of melody.
It is still more impossible to identify a language with a series of events
or aggregate of verbal behaviors. Every time I speak I add to the sum
total of verbal behavior that has gone on in English speaking communi¬
ties, but I do not thereby add to the English language. It is also note¬
worthy that the English language is something that might change over
a period of time, whereas a sum total of acts of speech is not the kind
of entity that can either change or remain unchanged; it is something
to which new components may or may not be added.
4
EMPIRICIST CRITERIA
OF MEANINGFULNESS
Meaningless It is an interesting and important fact about language that it is

sentences possible to construct expressions that at first glance look to be
perfectly in order and yet are unintelligible. Uncontroversial ex¬
amples of this phenomenon are such sentences as ‘Saturday is in bed/
‘Quadruplicity drinks procrastination/ and ‘My dream was three times
as large as yours.’ Although all of the words that make up these sentences
are quite intelligible and although none of the sentences violate any
commonly recognized rules of grammar (they are not at all like ‘And
the into when’), they do not make sense. We simply cannot understand
what it would be for Saturday to be in bed (rather than somewhere
else), or for two dreams to be comparable in respect of size. Now, if all
the examples of grammatically impeccable but unintelligible sentences
were so obvious, they would never have attracted the attention of
philosophers, although they might have become of interest to linguists
who were seeking to formulate the basic principles that govern the for¬
mation of intelligible sentences. It has seemed to many philosophers,
however, that some of the sentences regularly employed by their col¬
leagues suffer from the same defect, although less obviously. These
sentences include the following:
1. The physical universe depends for its existence on an omnipotent

spiritual being.
2. Properties have a mode of existence that is independent of their
being exemplified.
3. A human being is made up of two substances—one material and
one immaterial.
62
Empiricist Criteria of Meaningfulness 63
4. Physical objects are not just made up of properties; in addition, there

is something (a “substratum”) that has these properties.
5. It is possible that no human beings other than myself are really
conscious; they may all be very intricate machines.
6. It is possible that the world came into existence five minutes ago,
complete with records, memories, geological strata, etc., just as if
it had existed for millions or billions of years.
7. Moral standards have an objective existence.
Each of these sentences plays an important role in one or another

branch of philosophy—philosophy of religion (1), metaphysics (2
through 4), epistemology (investigations into the foundations of one
or another sort of knowledge) (5 through 6), ethics (7).1 Each of them
has been put forward as though it were obviously intelligible, and ex¬
perienced readers of philosophy have generally supposed that they
understood what was being said when such sentences were employed.
Yet the persistence of centuries-old disputes over such matters, the lack
of any prospect of the issues ever being definitively settled, and the con¬
sequent doubt that parties to such disputes really understand each other
have led some philosophers to question the meaningfulness of such
sentences.
If one wishes to maintain that sentences that have been taken as
meaningful for centuries are, in fact, meaningless, he will have to give
arguments in support of his claim. And such arguments will have to
proceed on the basis of some account of what is required for meaning¬
fulness. Hence, the philosophers who have taken this line have felt the
need to formulate criteria of meaningfulness. These criteria have gen¬
erally laid down some kind of connection with sensory experience as a
necessary condition of meaningfulness. In principle, we could have
criteria that require quite different things, for example, coherence with
other expressions in a system; but in fact, the formulations that have
attracted attention have all been of an empiricist sort.
Traditional We can begin by considering the kind of empiricist criterion de¬

form of an veloped by the British empiricists, Locke, Berkeley, and Hume,
empiricist Without going into the details of the views of any one of these
criterion philosophers, we can present a composite view as follows. A word
gets a meaning by becoming associated with a certain idea in
such a way that the occurrence of the idea in the mind will set off (or
tend to set off) the utterance of the word, and hearing the word will
1 We could find examples from other fields, for example, psychology—“When a

person is dreaming, he is carrying on trains of thought of which he is unconscious
and which are hidden by the things of which he is conscious”; physics—“Everything
in the universe is continually drifting further apart”; literary criticism—“A sense of
64 Empiricist Criteria of Meaningfulness
tend to bring about the appearance of the idea in the mind of the
hearer. (See Chapter 1 for Locke’s theory of meaning.) All ideas are
copies or transmutations of copies of sense impressions. Therefore, a
word can have a meaning only if an association has been set up between
it and an idea that was derived from sense experience. In this way, all
meaning is necessarily derived from sense experience. The British em¬
piricists used this criterion to justify branding certain philosophical,
theological, and scientific locutions as meaningless. Berkeley’s celebrated
rejection of material substance is a good example. Berkeley surveyed
various terms that were used to explain what a material substance is,
in distinction from its sensible properties, and how the substance is
related to the properties. For example, it was said that the properties
inhere in the substance, that the substance stands under or supports the
properties. He then argued that insofar as these words are meaningful,
that is, insofar as they can be given meaning in terms of sensory ideas,
they designate sensible relations between things rather than anything
that exists over and above sensible properties and relations. Thus, insofar
as we use these words meaningfully, we are still within the circle of
what can be perceived by sense and have not really succeeded in talking
about something itself unperceivable, which stands in a certain relation
to what can be perceived.2 Hume extended this critique to all sub¬
stance, mental as well as material, thus rejecting such terms as ‘self,’ as
used by philosophers.
As the reference to Locke s theory of meaning indicates, this par¬
ticular criterion of meaningfulness was tied closely to a certain theory
of meaning. One would not put forward a criterion in the way these
men did unless he accepted an ideational theory of meaning. But the
real thrust of the criterion can be preserved when restated in terms of
other theories. This thrust is constituted by the stipulation that sense
experience play an essential role in the acquisition of meaning by a
given expression. In terms of the behavioral theory, the requirement
would be that the stimulus-response bonds, which are crucial for mean¬
ing are acquired through repeated experience of the coincidence of
such stimuli and such responses and/or other factors strengthening the
• ccording to this view, the habit of uttering an expression in a
certain kind of situation is acquired through repeatedly hearing the ex¬
pression m that kind of situation, and also, perhaps, through having
reawakening powers and of limitless horizons was evident in Elizabethan drama ”

Sentences of these sorts, and many others, have been pronounced meaningless bv
philosophers. But for this discussion we shall concentrate on supposed examples of
nonsense from philosophy itself. examples ot
Alciphmn^’ Berkdey latCr went bey°nd this restrictive view. See Book VII of his
one’s utterance of the expression in that kind of situation regularly

rewarded. On the referential theory, the requirement would be that
an expression acquires the capacity to refer to a certain kind of thing
through being paired with that thing in experience. It is this latter form
of the empiricist criterion that rests on the notion of an “ostensive
definition.” To define a word ostensively is to get someone to realize
what the word means by pointing to an example of that to which it
refers (denotes, names, . . .), or to otherwise see to it that the person’s
attention is directed to such an object while the word is being uttered.
In the immediately ensuing discussion, we shall concentrate on this
version of the criterion, according to which ostensive definition is neces¬
sary for words to acquire meaning. We do this partly because it is this
form of the theory that has had the most pervasive influence in our
day (as in the “general semanticist's” view that we are not speaking
intelligibly unless we are referring to something that can be “kicked”),
and partly because it will enable us to avoid the extra complexities in
which, as we saw in Chapter 1, the ideational theory entangles us. In¬
stead of ‘refer,’ we shall use ‘stand for’ in a deliberately ambiguous way
to cover any sort of semantically interesting relation in which expres¬
sions stand to what they are used to talk about. In this way, we can
ensure that our discussion will have relevance to at least the great ma¬
jority of meaningful expressions.
There are various reasons why an empiricist criterion has seemed
to be acceptable, or even necessary. Perhaps the most powerful is this.
Assuming that in some way meaningfulness depends on expressions
being connected with aspects of the extralinguistic world about which
they are used to talk, how is such connection possible? A given sound
pattern is not related to one aspect of the world rather than another
by virtue of its intrinsic characteristics; and we can hardly suppose that
such linkages are innate to the human mind. (If they were, all men
would speak the same language.) The only alternative seems to be that
they are established by experience, through repeated pairings of the
expression with what it stands for in the learner’s experience.
Another argument is this. How could I have any reason to suppose
that anyone else attaches the same meaning to a given expression that
I do? Of course, we could each produce a verbal definition of the ex¬
pression, but that will yield the desired conclusion only if we assume
we are both using the words in the definition in the same way (and
also that we understand the sentence form ‘Give a definition of . . .’
in the same way). And the question whether that assumption is true
is a question of exactly the sort we set out to answer. It would seem
that we can break out of this circle only if, at some points, we can test
the hypothesis of common meaning without relying on community of
meaning for other expressions. And how could we make such a test
except by investigating the way the expression is or is not paired with
experienced objects in the verbal activity of each of us? This means
that such tests are possible only if it is necessary for meaningfulness
(of at least some expressions) that such pairings exist.3
Empiricist criteria of the sort we are considering are usually stated
as genetic theories about the way people leam what words mean or the
way words acquire meaning. This is, in part, a reflection of the fact
that in British empiricism of the seventeenth and eighteenth centuries,
epistemology and semantics were not really separated from psychology.
The separation is by no means complete today, but now we are well
aware of the dangers of seeking answers to questions of fact, including
psychological fact, by the traditional armchair methods of philosophy-
reflection and clarification. If we really want to find out how people
learn the meanings of words and what mechanisms are involved in such
learning, there is no substitute for careful observation of the process
itself; it is ill-advised to rest theories about this on a priori considera¬
tions, such as we have in the preceding arguments. Fortunately, it is
not necessary to give these criteria a genetic form. In general, it is possi¬
ble to replace any empiricist genetic account with a parallel statement
of what must be the case for an expression to have a meaning for some¬
one at a certain time—no matter how it acquired the meaning. Thus,
in place of the Lockean genetic account, we can propose the following:
in order for an expression to be meaningful in my current use of it,
it is necessary that there be a tendency for the word to elicit in me a
certain idea and vice versa. The formulation in terms of ostensive defi¬
nition seems to be more wedded to the genetic form, but it can be
restated without losing its empiricist force: a word can have a meaning
for someone only if he is able to pick out its referent” in his experience.
This means that we have shifted from the genetic requirement that a
word have acquired its meaning by way of an ostensive definition to
the requirement that it be possible to give an ostensive definition. Since
genetic formulations are thus easily convertible, I shall continue to
make use of them for the sake of easy intelligibility. (The first argu¬
ment just given for an empiricist criterion, which, as stated, supports a
genetic criterion, could also be reformulated along similar lines.)
3 Note that this argument is not available to “ideational” theorists like Locke
and Hume, who hold that meaning is essentially a matter of intramental associations,
ror these philosophers, it is quite conceivable, though in fact not the case that
everyone should have a private language that he uses only in (silently) talking to
himself, and that the words in such a language would have meaning in just the
way words have meaning in the public languages that actually exist. Thus for
Locke and Hume, it would not be the case that the conditions of meanings being
publicly shared are ipso facto conditions of words having meaning.
The semantic Complications begin to emerge when we note that it cannot be

stratification the case that every meaningful expression in the language gets its
of language meaning through direct confrontation with an experienced refer¬
ent. This account seems plausible for common nouns denoting
observable physical objects—‘tree/ ‘house/ ‘cloud’; adjectives connoting
directly observable properties—‘blue/ ‘round/ ‘shiny’; and verbs that
are concerned with directly observable activities—‘walk/ ‘speak/ ‘wave.’
However, there are many other words belonging to these grammatical
classes, whose meaningfulness would not be questioned by any but the
most hardy empiricists and which could not possibly get hooked up
with their extralinguistic objects in this way, because the kind of thing,
property, or activity involved is not directly observable. I am thinking
of such words as ‘society,’ ‘conscientious/ ‘intelligent,’ ‘neurosis,’ ‘lan¬
guage,’ ‘education/ ‘brilliant,’ ‘manage,’ ‘pray/ ‘prosper.’ One cannot
teach someone what the word ‘prosper’ means by pointing to someone
prospering while uttering the word in the way one can teach someone
what ‘run’ means by (repeatedly) pointing to someone running while
uttering the word. Of course, one can observe instances in these cases.
One can watch someone praying or (engaged in) managing a business,
one can see a neurotic or an intelligent or a conscientious person and
can even observe him doing something that displays his intelligence
or conscientiousness or is a symptom of his neurosis. But the meanings
of these words are such that it is not the directly observable features of
such objects or occurrences that are crucial for the application of the
terms in the way it is the directly observable features that are crucial
for the application of terms like ‘run’ or ‘shiny.’ (One might cavil even
at ‘run,’ if he thought that some intention on the part of the agent is
necessarily involved in running.) To say that a person is conscientious
is not to say that he exhibits certain directly observable features, which
could be noted in a single observation; and to say that he is acting in
a conscientious manner is not to say that his behavior exhibits certain
directly observable features. One can’t learn what ‘conscientious’ means
by having conscientious people or conscientious behavior pointed out
to one, because one could not see these people as conscientious unless
one realized a lot of things about the people,4 the realization of which
requires having a great deal of language already under one’s belt.
Thus, in all forms of empiricism except the crudest, language is
divided up into semantic levels or strata. The fundamental level is made
up of words that have their meaning by virtue of association with di-
4 For example, that they had taken on certain obligations or duties, that they
are now engaged in carrying out one or more of these, and that this present per¬
formance is only one manifestation of a fixed habit of promptly carrying out such
tasks whenever they are committed to do so.
rectly experienced items; the principle then states that in order for
other words to have meaning they must be definable in terms of words
on this first level plus, perhaps, other words that have already been so
defined. Some words get their meaning from experience more directly
than others; but, directly or indirectly, experience is the source of mean¬
ing for all. This is the simplest version of an empiricist theory that is
at all plausible.
It is a defect in such a theory that no one has ever made a plausi¬
ble case for the possibility of defining all meaningful words in the
language in terms of the lowest level. The most strenuous efforts have
been made with theoretical terms in science; and even for such rela¬
tively low-level terms as ‘electric charge,’ ‘specific gravity,’ ‘habit,’ and
‘intelligence,’ empiricists have by now admitted that such definitions
cannot be provided. Rather than go into the complexities of this issue,
I shall focus on another difficulty that is both easier to exhibit in a
short space and more revealing semantically. If, with Locke, we think
of the basic level as containing word-sized units, then we are never
going to get sentence-sized units into the language at all, which means
that we are never going to be able to say anything; and in that case,
there is no reason to say that we are dealing with a language. In order
to understand and be able to use a sentence, one must not only know
the meanings of nouns, verbs, and adjectives, one must also understand
the significance of the syntactical form of the sentence; and for many
sentences, one must understand various kinds of words that serve to
connect nouns, adjectives, and verbs into sentences so as to affect the
meaning of the sentence as a whole. One must be able to distinguish
semantically between ‘John hit Jim,’ ‘Jim hit John,’ ‘Did John hit Jim?’
John, hit Jim! and John, please don t hit Jim.’ This means that before
one can engage in conversation one must be able to handle and under¬
stand such factors as word order; “auxiliaries” like ‘do,’ ‘shall,’ and ‘is’;
and connectives like ‘is,’ ‘that,’ and ‘and.’ These elements can neither
get their meaning by association with distinguishable items in experience
nor be defined in terms of items that can. Where could we look in our
sense perception for the object of word-order patterns, pauses, or words
like ‘is’ and ‘that’? And as for defining these elements in terms of words
like ‘blue’ and ‘table,’ the prospect has seemed so remote that no one
has so much as attempted it.
Logical Thus, both from the analysis of what is involved in knowing a

tomism language and from a consideration of what seems to go on when
one is learning a (first) language, it would appear that if we are
to understand how language is based on experience, we must people our
lowest level where meaning is based on direct confrontation with ex-
perience—with sentence-sized units, not word-sized units. We have this

form of the theory in Bertrand Russell’s “Philosophy of Logical Atom¬
ism, where the logical atoms are the sentences that could be used
to report a single observation, for example, 'This is red’ and ‘This book
is on top of the table. We may call such sentences “observation sen¬
tences.” (Russell wants to restrict the basic sentences to those that
simply report the speaker’s own sense experience, for example, ‘I am
aware of a round, bluish visual datum,’ excluding those that make claims
about publicly existing physical objects, for example, ‘There is a blue
saucer on the table.’ But for our purposes, we shall restrict attention to
sentences about physical objects.5 6) When the theory is stated in this
way, we do not have the difficulties just considered. We have various
sorts of syntactical structures, connectives, and auxiliaries built into the
components of our empirical base, and we are not faced with the ob¬
viously impossible task of defining all such elements on the basis of
(some) nouns, adjectives, and verbs.
Verifiability When traditional empiricism develops into this form, it is very

theory of close to the kind of empiricist criterion that has been most promi-
meaning nent in the past few decades, the “Verifiability Theory of Mean¬
ing.” This view was initially put forward by members of the
“Vienna Circle,” 7 a group of philosophers, mathematicians, and scien¬
tists that gathered around Moritz Schlick in Vienna in the 1920’s. They,
and those heavily influenced by them, are called “logical positivists.”
These men were concerned with the logic of mathematics and science
and with giving philosophy a scientific orientation. They felt that phi¬
losophy in the past had been largely given over to useless controversy
over metaphysical and normative problems that were, in principle, in¬
soluble.8 Like Hume, they felt that such controversies were fruitless
because the participants were not making sense. It was in order to nail
down this conclusion that they first introduced the principle that in
order for one to be talking sense, he must be able to specify the way
in which what he says can be empirically verified; in other words, it
must be possible to specify what observations would count for or against
its truth.9
5 In Logic and Knowledge, ed. R. C. Marsh (London: George Allen and Unwin
1956).
6 This issue is discussed on p. 76.
7 There were earlier foreshadowings, especially C. S. Peirce’s “pragmatic theory
of meaning.”
8 See the list at the beginning of this chapter for examples of propositions, con¬
troversy over which the positivists considered useless.
9 Logic and mathematics were excluded from these strictures on the ground that
they are made up of “analytic” propositions. In calling “2 -f- 2 = 4” an analytic
proposition, one is saying that, like “All bachelors are unmarried,” it is true just by
Let us not neglect the distinction between verifiability and verifi-

cation. In laying down verifiability as a condition of meaningfulness,
the positivist is not saying that only sentences that have been verified
are meaningful. Such a statement would be absurd; it would imply, for
example, that we could not understand a statement until after we had
established that it was true. The positivist recognizes that there are
perfectly meaningful statements that have not yet been tested, and
even meaningful statements that we are not now in a position to test.
In requiring verifiability, the positivist is simply requiring that it be
possible to specify what an empirical test would consist in; he is not
requiring that the test have been carried out. Verifiability is possibility
of verification. Furthermore, this does not have to be physical possi¬
bility or technological possibility. The positivist recognizes that there
are perfectly meaningful statements that we are not, in fact, in any
position to test. The standard example of this has been ‘There are
mountains on the other side of the moon,’ but recent technological
developments are forcing philosophers to change their examples. Per¬
haps ‘There is life in other galaxies’ will do for the foreseeable future.
We have at least a rough idea of what sorts of observations would
count for or against the truth of this statement, even though we are
quite unable to get into position to make those observations. So long
as we can give an intelligible specification of what observations would
establish the truth or falsity of the statement, we have satisfied the
criterion.
We should also note the specially wide sense in which the posi¬
tivists use ‘verifiability.’ In this use, it is really equivalent to the dis¬
junction ‘verifiable or falsifiable,’ that is, ‘capable of being established
as true or false.’ Thus, the requirement is really that the sentence in
question be capable of an empirical test. This is an important point,
because the language used might give the impression that the positivist
will recognize as meaningful only statements that can be shown to be
true. (And how would we know in advance which these are?)
It may seem that the verifiability criterion is quite different from
the ones so far considered in that it does not involve any stratification
of language. Here, we apply the same test to any sentence; it is mean¬
ingful if and only if it is empirically testable. But as soon as we probe
virtue of the meanings of the terms in which it is formulated; hence, it should not
be interpreted as making any claim to say something “about the world.” Therefore,
logic and mathematics were excused from the verifiability requirement Although
the status of logic and mathematics, and the concept of an analytic proposition are
high y controversial matters, we shall not have time to go into this aspect of the
problem In the ensuing discussion, we shall tacitly assume that these matters can
be satisfactorily handled. For further discussion of this topic see Stephen C Barker
Philosophy of Mathematics, in this series.
the notion of empirical testability, this can be seen to be an oversim¬

plification. Specifying an empirical test involves specifying a way in
which a statement could be supported or attacked by the carrying out
of certain observations. Thus, the statement ‘Sam has measles’ could be
supported by observing that Sam has spots all over his body. This means
that we must separate certain sentences in the language that could be
used to report observations, “observation sentences” like ‘Sam has spots
all over his body/ which are meaningful just by the fact that they have
this function. We then say that a sentence that is not an observation
sentence, like Sam has measles, is meaningful if we can specify certain
observation sentences (not necessarily true ones and not necessarily
ones that we know to be true or false) to which it is so related that
each of those observation sentences could be used as positive or negative
evidence. Once we put the matter in this way, we see that the verifi¬
ability criterion presupposes the same kind of stratification as the others.
A basic level is singled out, the members of which have meaning in a
specially direct way; other elements of the language have meaning only
if they have the right kind of connection with the basic level. In fact,
logical atomism and the verifiability theory are virtually the same theory
stated in different ways. They sound different because the verifiability
theory looks down from the nonobservation sentences, asking how they
can be verified; while logical atomism looks up from the observation
sentences, asking what else can be explained in their terms. It is, how¬
ever, the same topography whichever perspective we take.
The earliest forms of the verifiability criterion required complete
verifiability. That is, a sentence could not be regarded as meaningful
unless it were possible to specify a way in which it could be conclusively
shown by empirical evidence to be true or false. It soon became obvious
that this requirement was much too strong. It would exclude, for ex¬
ample, all unrestricted generalizations. If we take something so simple
as “All lemons are yellow,” it is clear that we cannot specify any finite
set of observations that are such that having carried them out we can
then be absolutely certain that this statement is true. Of course, a state¬
ment of this form could be conclusively falsified by a single observation
(of a red lemon), assuming that there is no question of whether we
do in fact have a lemon before us and that it is red. (These assumptions
are together equivalent to the assumption that ‘This lemon is red’ is an
observation sentence.) Just the opposite is true of the contradictory of
this statement, namely, ‘There is a lemon that is not yellow.’ This
statement can be verified by a single observation, but no matter how
many yellow lemons we had observed we still would not have conclu¬
sively falsified ‘There is a lemon that is not yellow.’ Thus, we can find
an unlimited number of simple generalizations for which we can specify
no conclusive empirical proof, and an unlimited number for which

we can specify no conclusive empirical disproof. This will be the case
whenever we are dealing with “open-ended” classes like lemons, that is,
classes that are so specified that no definite limit can be put on their
membership, as contrasted with “closed” classes like the books in this
room at this moment. Again, there are many apparently “singular”
propositions that conceal unlimited generalizations of one sort or an¬
other, for example, Jones is conscientious.’ To say of a particular person
that he is conscientious is to say that he would react in a certain kind
of way to a certain kind of situation, whenever or wherever it arose.
(It might be more accurate to make this a statistical statement or a
statement of tendency that he would react in a certain way in most
instances, or that he would always tend to so react in such situations.
These versions would give rise to analogous problems, but in a more
complicated way; hence, I will make the point with respect to the
simpler interpretation.) It is to say that whenever he recognizes himself
to have an obligation to perform a certain task, he does what he can
to carry it out. But again, no matter in how many cases we have seen
this to hold, we cannot be certain that the unrestricted generalization
is true.
Thus, to require complete verifiability or falsifiability would be to
throw out the baby with the bath water. Positivists soon came to modify
the criterion so that it requires only the specification of observations
that would count for or against the statement, which would serve to
confirm or disconfirm it to a certain extent. “Confirmability criterion
of meaningfulness” would be a more apt title for this revised version
Positivistically oriented critics of theology still take religious believers
to task for not being able to say what observable happenings would
count decisively for or against the existence of God; but in doing so
they are either lagging behind developments in the movement or un-
warrantedly making more stringent demands on theology than on
The verifiability version of empiricism has suffered more than the

others from a confusion between criterion of meaningfulness and theory
of meaning. The common label, “Verifiability Theory of Meaning ”
would suggest that it is the latter that is being put forward and some
ot the more common formulations, for example, “The meaning of a
proposition is the method of its verification,” seem to bear this out.
But a careful examination of the literature and a just appreciation of
the predominantly polemic aims of the movement will show that what
the positivists have really been concerned with is a criterion of mean¬
ingfulness. There has been little attempt to work out the notion that
m specifying a method of verification for a sentence we are thereby
giving its meaning. Such attempts as have been made are either un¬
promising or disastrous. A prime example of the latter is the suggestion
that the meaning of an historical statement consists of certain investiga¬
tions that we might carry out in the future in order to test it. That is,
when we are talking about the past, we are really talking about the
future! We will be quite in order if we concentrate on the “verifiability
theory” as a criterion of meaningfulness.
Deficiencies As usually stated, the verifiability theory exhibits some glaring de¬
in usual ficiencies that will have to be remedied before it can be taken
formulations seriously. First, note that it is not a sentence that can be said to
t>f verifiability be true or false (verified or falsified), but rather an assertion or
criterion statement that one makes by uttering the sentence. If we try to
assign truth values to sentences, we run into hopeless dilemmas.
Is the sentence 'I am hungry’ true or false? On one occasion, a speaker
might say something true, by uttering that sentence and on another
occasion a speaker might say something false by uttering it. If we re¬
garded the sentence as the bearer of truth value, we would have to think
of it as constantly oscillating between truth and falsity, or even as being
both true and false at the same time (if at the same time one speaker
said truly that he was hungry and another speaker said falsely that he
was hungry). But it is the sentence, not an assertion or statement made
by uttering a sentence, that either has or does not have a meaning.
Once we admit that a statement or assertion has been made, we have
already granted meaningfulness. Meaning (in the sense in which we
are concerned with it) is not something that a statement might or might
not have. To avoid this difficulty, we shall have to revise the criterion
as follows: a sentence has meaning only if it can be used to make an
assertion, and it can be used to make an assertion only if it is possible
to specify some way of verifying or falsifying the assertion. Stated this
way, the criterion looks much closer to our analysis of meaning in terms
of illocutionary-act potentials.
Second, even as so revised, this could not possibly be a general
criterion of meaningfulness, not even for sentences. There are many
sentences in the language that are obviously meaningful but that
just as obviously are not usable for making assertions. These include,
for example, interrogative sentences like ‘Where is the butter? im¬
perative sentences like ‘Please go out quietly, and interjections like
‘Splendid!’ Such sentences are used to ask questions, make requests,
or express feelings and attitudes. When we are using sentences in these
ways, questions of truth and falsity do not arise; consequently criteria
of meaningfulness in terms of verifiability have no application. It would
seem that what we have is not really a criterion of meaningfulness at all,
but rather a criterion of the usability of a sentence for performing a

certain kind of illocutionary act.
Positivists have tried to take account of this point by construing
their criterion as a criterion of a certain kind of meaningfulness, variously
called “cognitive meaning,” “factual meaning,” and “literal meaning,”
which is then distinguished from something called “emotive meaning”
or “expressive meaning.” But apart from the absurdity of lumping the
indefinitely various uses of sentences other than the assertive under the
heading “emotive” or “expressive,” one can question the significance
of distinguishing different kinds of meaning in this way. If we are to
say that an imperative sentence has a different kind of meaning from a
declarative sentence on the ground that it is used for a different kind
of illocutionary act, how far are we going to carry this? Are we going
to say that 'Shut the door!' has a different kind of meaning from 'Please
shut the door’ on the grounds that one is used to issue an order and
the other to make a request? And would we want to say that ‘I was out
late last night’ has two different kinds of meaning on the grounds that
it can be used to make an admission as well as to simply inform someone
of a fact? If one is undeterred by this indefinite multiplication of “kinds
of meaning,” he may still well consider the fact that there is no reason
to think that sentences of these different sorts have meaning, or get
their meaning, in different ways. In each case, to have a certain meaning
is to be usable for the performance of a certain illocutionary act; the
differences simply come from the differences in the illocutionary acts
involved. In each case, what it is to have a certain meaning(s) is the
same. However, it does clearly make sense to speak of different kinds of
meaning for sentences on the one hand and sentence-components like
words on the other, for the account of what it is to have a certain
meaning is different in the two cases. And still more can we distinguish
the kind of meaning that a sentence and a knock in an engine have.
Whether or not one can properly speak of “cognitive meaning,”
the existing formulations of the impiricist meaning criteria are restricted
to only one segment of language. This may not disturb those who are
interested only in finding ways of ruling out supposed assertions that
they find objectionable and in restricting scientific and philosophical
discussion to questions that can be settled empirically. But to some¬
one who is interested in the philosophy of language for its own sake,
the restriction is disappointing. After all, it does seem that experience
is and must be crucial for the way one attaches meanings to words, not
just in making assertions, but in linguistic activity of all sorts; and it
ought to be possible to find a criterion of meaningfulness that would
adequately reflect this fact. The arbitrary character of the restriction
can be brought out in this way. Let us take 1. 'The Holy Ghost de¬
scended upon us’ as an example of a sentence to which a positivist
would object as not being capable of being used to make an assertion
that could be empirically tested. In this respect, he would contrast it
with, for example, 2. ‘John came down out of the tree.’ Now it seems
that any reason for regarding 1 as semantically defective would equally
be a reason for regarding 3. ‘Come, Holy Ghost, descend upon us’ as
semantically defective in contrast with 4. ‘John, come down out of that
tree.’ Roughly speaking, if 1 is defective in contrast with 2, because
we don’t know what empirical observations would count as verifying
it, by the same token, 3 should count as defective by contrast with 4
because we don’t know what empirically observable states of affairs
would count as compliance with the request. But as the verifiability
criterion is usually stated, 1, 3, and 4 are all said to be lacking in “cog¬
nitive meaning,” and as such, they are all lumped under the heading
of “emotive meaning” or “expressive meaning.” A theory that can make
no finer distinctions than this is badly in need of supplementation.
Fortunately, the materials for such supplementation are ready at
hand. According to the theory presented in Chapters 1 and 2, the
meaning of a sentence is a function of its utterance’s being governed
by a rule that stipulates that that sentence is not to be uttered in a
given kind of context unless certain conditions hold. This is a general
account applying to sentences of all sorts that are used to perform illo¬
cutionary acts of all sorts. If this theory is acceptable, we have a way of
generalizing the verifiability requirement so that it covers all sorts of
speech. No matter what kind of illocutionary act a sentence is used to
perform, the claim that a given condition required for the utterance
of that sentence holds is an assertion that can be evaluated as true or
false. Hence, we can give an unrestricted formulation of the verifiability
criterion as follows: A sentence is meaningful only if its utterance is
governed by at least one rule that requires that certain conditions hold
such that for each of those conditions the claim that the condition
holds is empirically confirmable or disconfirmable. Using this criterion,
3 would be ruled out in just the same way as 1 (if, indeed, either would
be). For although 3 is not itself used to make an assertion that could
be empirically confirmed or disconfirmed, its utterance is subject to a
rule that requires the holding of such conditions as that there be some
entity called the Holy Ghost, that it be possible for this entity to de¬
scend or enter into the human spirit, etc.; just the same diEculties that
attach to specifying any empirical test for 1 attach to the task of specify¬
ing any way in which the claim that one of these conditions hold can
be empirically tested.
Problems Now that we have the verifiability criterion in a more acceptable

in the form, we can concentrate on some problems regarding its con¬
formulation tent.10 First, there are controversies as to just what should count
of the as an observation sentence. On the one hand, there are reasons
verifiability for holding that no statement that makes a claim about an objec¬
criterion tive physical object, event, or state of affairs can be conclusively
established by one observation or any finite number of observations.
The main reason for this conclusion is the fact that any such statement
has an indefinite number of consequences. For example, the statement,
There is a black telephone on the desk in front of me,” implies that
this object was constructed for the purpose of telephonic communica¬
tion, that it contains certain kinds of equipment within it, that it will
look like a telephone to most normal observers, that it will not vanish
into thin air, etc. Whether or not this list of implications can be in¬
definitely extended, it is clear that we cannot by a single observation
assure ourselves that all these implications are true. But if any one of
the implications were false, then there would not be a telephone on
the desk in front of me. Therefore, I cannot make sure that this state¬
ment is not false (that is, make sure that it is true) by a single ob¬
servation. Therefore, no statement about objective, physical matters of
fact can be an observation statement. This line of thought naturally
leads to the position that the ultimate pieces of empirical evidence are
phenomenal statements,” each of which is restricted to the claim that
the observer’s sensory experience was characterized in such and such a
way, for example, I seemed to see a black, telephone-shaped object.”
Purely subjective statements of this sort can be known with certainty
on the basis of a single observation. They do not give rise to a multi¬
plicity of independently testable implications. But having reached this
rock bottom of certainty, it is difficult to get beyond it. Attempts to
show how hypotheses concerning the physical world can be confirmed
or discontinued by reference exclusively to "phenomenal data” have
been unconvincing. Hence, many positivists have taken the position
that whatever may be the case with "theoretical certaintv,” ordinarv
judgments of sense perception like "there is a telephone on my desk”
have sufficient certainty to be taken as the ultimate evidence on which
scientific testing rests. But even here, there are problems. Just what sorts
of objective statements can be warrantably taken as conclusively estab¬
lished on the basis of a single observation? This problem becomes of
practical importance in a science like psychology, where different schools
differ in what they will recognize as empirical evidence. Psychoanalyti-
10
'Since we have seen how to extend the criterion to all sorts of sentences we
,a , continue to discuss it in the more familiar form in which it is specifically ap¬
plied to assertive sentences. F ^ dr
cally oriented clinical psychologists will take “He was very defensive,”
“He was extremely hostile,” or “He tried hard to reassure me” as basic
empirical data; whereas more hard-boiled “stimulus-response” psycholo¬
gists will claim that these statements are themselves hypotheses that
should, in principle, be tested in terms of such data as “He made very
jerky movements,” “His face was contorted,” or “He uttered the sen¬
tence, ‘Don’t feel bad about it.’ ”
Second, there are problems over the kind of logical relations data
must have to an assertion in order to count for or against it. These
problems arise because of the fact that no nonobservation statement
logically implies any observation statement by itself, but only in con¬
junction with other statements. For example, the nonobservation state¬
ment, “Ernest has intense unconscious hostility toward his father,”
will not by itself imply any statement reporting what would normally
be taken to be a manifestation of such hostility, for example, “Ernest
flared up at Mr. Jones.” The latter will follow from the former only
in conjunction with other premises, for example, “The repression is
not so severe as to permit no expression,” “Mr. Jones is perceived as
sufficiently similar to Ernest’s father to permit a displacement of the
hostility onto him,” and “The hostility has not all been worked off in
other ways.” Thus, the presence or absence of a given piece of data
counts not just for or against one particular hypothesis, but rather for
or against the whole body of premises used in deriving it. This makes
the logic of confirmation rather complicated. Positivists have become
aware that this makes it very difficult to exclude unwanted metaphysical
statements. For such a statement can always be added to the premises
yielding a given piece of empirical data in a way that makes it seem to
be among the assertions that the data count for or against. Attempts
to specify the logical relations involved so as to exclude this kind of
maneuver have so far been unsuccessful.11
Verifiability Let us suppose that we have the kinks out of the notion of an
criterion as observation sentence and that we can specify the way in which a
description supposed assertion must be related to certain observation sentences
and as in order that the evidence formulated in those observation sen¬
proposal tences can be said to count for or against that assertion. We can
finally come to the question: What can be said for or against
accepting the verifiability criterion?
It is sometimes recommended on the grounds that it is simply a
formulation of a criterion that, in fact, we always use in deciding what
11 For an interesting account of these attempts, see C. G. Hempel, “Problems

and Changes in the Empiricist Criterion of Meaning,” in Semantics and the Philos¬
ophy of Language, ed. L. Linski (Urbana, Illinois: University of Illinois Press, 1952).
does or does not make sense. However, unless the reference class of sen¬
tences (or the extension of ‘we’) is restricted in a question-begging
way, this is palpably false. Positivists would never have made such a
fuss over the criterion in the first place if it were not for the fact that
the use of sentences that violate it is so widespread. Nor is this use
restricted to professional philosophers. Utterances like ‘God created the
heavens and the earth,’ which positivists take to be unverifiable, figure
heavily in the discourse of the man in the street. It may be claimed
that when people say things like this they are really confused, in that
they are violating standards of meaningfulness to which they are firmly
attached, and that by reflection on what they are doing, they could
come to see that what they are doing is meaningless by their own stand¬
ards. But nothing has been done to show that this is actually the case.
Most positivists have represented the theory as a proposal as to
how the class of meaningful sentences should be delimited, rather than
as an account of how it is, in fact, delimited. It may seem that once
it takes on this form, the theory loses all pretension to being a criterion
of meaningfulness. For it would seem that a given sentence is or is not
meaningful (in a certain language community), whatever proposals
we might make. Wouldn’t proposing that certain sentences not be
classed as meaningful be like proposing that certain bottles of milk not
be classed as sour? If they are sour, then no proposal that we make is
going to alter the matter. Of course, we could decide to change the
meaning of the word ‘meaningful.’ But that would not make sentences
'that were meaningful in the usual sense not meaningful in the usual
sense. And presumably it is meaningfulness in the usual sense in which
we are interested, not meaningfulness in some sense that some group
of philosophers sees fit to give it. However, this would be an insensitive
way of viewing the situation. If we remember that ‘meaningful,’ like
many terms, is markedly vague (see Chapter 5), we may realize that a
proposal does not necessarily involve introducing a completely new
sense. In fact, it is not clear on the face of it in every case just what
we should regard as making sense. ‘There is a telephone on the desk
in front of me’ clearly does make sense, and ‘Green ideas sleep furiously’
clearly does not. But if we simply reflect on the way we handle these
clear cases, it is not easy to make explicit the principles on which we
separate the sheep from the goats; it is still less easy to decide how we
should, in terms of such principles, dispose of the kind of sentence illus¬
trated by the list at the beginning of this chapter. Let us follow the
theory of Chapters 1 and 2 and say that a sentence is meaningful if
and only if it is usable for the performance of one or more illocutionary
acts, and it is so usable if and only if its utterance is subject to certain
kinds of rules. Having gone so far, we have still left certain questions
dangling, in particular the question as to whether a sentence could

be said to be governed by certain rules if the supposed rules stipulate
conditions such that there is no empirical way of telling whether or not
they hold. Thus, we may suppose that ‘Every property inheres in a
substratum is subject to a rule that stipulates, among other things, that
there are such entities as substrata. Let us grant that there is no em¬
pirical way of telling whether or not this is so. In this case, is the sen¬
tence really usable for performing an illocutionary act? Is it meaningful?
It is not clear what we should say. We might well be pulled in both
directions by our dispositions with respect to such words as ‘meaning¬
ful/ Hence, there is a legitimate place for proposals as to how ‘mean¬
ingful’ can be made more precise.
Arguments in What can be said by way of recommendation of this particular

support Of proposal? For one thing, the positivist can point to what happens
verifiability when new theoretical terms are introduced into science. Consider
criterion the first introduction of the notion of unconscious desires, fears,
guilt, etc. If we simply put the term ‘hatred of one’s father’ to¬
gether with ‘unconscious’ and let it go at that, it is hard to know what
to make of the sentence, ‘Ernest has an unconscious hatred of his
father.’ We know what to make of the statement that x hates y where
there isn’t this kind of qualification, and we know what it is to be
unconscious of, say, the furniture in the room. But if all we have done
is to take these terms with their established senses and combine them
in this way, we have not given a sufficient indication of how this is to
be taken. The most we have is a suggestion that the person in question
might be thought of as hating his father (even though he really doesn’t);
we have what John Wisdom calls the expression of a “picture prefer¬
ence.” In order to make clear what assertive force the sentence has, we
would have to begin to specify how unconscious hatred would be mani¬
fested in observable behavior. It should not be expected that this would
take the simple form of a hypothetical statement, specifying a certain
kind of behavior as always contingent on unconscious hatred. We have
to keep in mind the point made previously to the effect that a scien¬
tific hypothesis will entail a given observation report only in conjunc¬
tion with other hypotheses and statements of observable conditions.
Thus, when we begin to spell out connections, of the sort sketched on
page 77, between the existence of unconscious hatred, degree of re¬
pression, relevant associations, and so forth on the one hand and ob¬
servable behavior such as rudeness to an employer on the other, we
have made a beginning at giving an assertive force to a sentence like
‘Ernest has an unconscious hatred of his father.’
A similar point emerges from a consideration of what happens
to an ordinary statement of fact if one begins to qualify it in such a

way as to render it completely immune to any empirical discontinua¬
tion. Consider the oft-cited parable of the gardener.
Once upon a time two explorers came upon a clearing in the jungle.
In the clearing were growing many flowers and many weeds. One
explorer says, “Some gardener must tend this plot.” The other dis¬
agrees, ... So they pitch their tents and set a watch. No gardener is
ever seen. “But perhaps he is an invisible gardener.” So they set up
a barbed-wire fence. They electrify it. . . . But no shrieks ever suggest
that some intruder has received a shock. No movements of the wire
ever betray an invisible climber. ... At last the Sceptic despairs, “But
what remains of your original assertion? Just how does what you call
an invisible, intangible, eternally elusive gardener differ from an im¬
aginary gardener or even from no gardener at all?” 12
As this story suggests, what starts out as a genuine assertion will be

reduced to a mere picture preference if we remove any possibility of
putting it to an empirical test.
The question is: to what extent do these considerations show
that sentences like those on our initial list do not make sense? 13 Let
us consider a metaphysical philosopher who holds that properties like
roundness and intelligence have a non-spatio-temporal mode of existence
that is independent of their exemplifications. (There would still be
such a thing as roundness even if there were no round things.) Call
him a Platonist. Our Platonist might make the following reply to
the above arguments for the verifiability criterion.
“It is quite true that a supposed scientific principle, which is in
principle untestable, is thereby debarred from consideration. But this
is so because scientific hypotheses are constructed for the specific pur¬
pose of explaining and predicting observable phenomena; consequently,
if a supposed scientific hypothesis has no connection with observation
statements, it is not really a scientific hypothesis at all. Likewise, a
supposed assertion about the physical world cannot be taken seriously
unless it is possible to specify empirical tests. A physical state of affairs
could not be objectively real unless it manifested itself in some way in
sense experience. But this is because of what it is to be physical rather
than because of what it is to make an assertion. We would not call
anything physical’ unless its existence would make some possible
12 A. Flew, “Theology and Falsification,” in New Essays in Philosophical Theology

eds. A. Flew and A. Macintyre (London: S.C.M. Press, 1955), p. 96. The idea is
taken from J. Wisdom, “Gods,” in Logic and Language, First Series, ed. A. Flew
(Oxford: Basil Blackwell, 1952).
13 We are assuming, for purposes of this discussion, that none of those sentences
are empirically testable to any extent. This assumption is quite often made but it
has been challenged, especially in the case of sentence 1.
difference in the course of sense experience; hence, we would not take

any utterance to be a genuine claim about the physical world unless
it were possible to specify some empirical test of it. But it would be
arbitrary in the extreme to extend these restrictions to assertions in
general. In putting forward my metaphysical thesis, I am not making
any claim about physical objects, events, or states of affairs, nor am I
putting forward a hypothesis that has as its purpose the explanation
or prediction of such matters. How could one reasonably hold these
assertions, which are concerned with trans-physical matters, subject to
the same restrictions? If there are such entities as properties existing
independent of their exemplifications and an omnipotent spiritual cre¬
ator of the physical universe, there would be no reason to expect them
to manifest themselves in the details of our sense experience. (To sup¬
pose that we could formulate empirical tests for the thesis of the exist¬
ence of God is to suppose that we could discern fixed regularities in
the way in which God acts in the world; and a theist might well re¬
gard this idea as blasphemous.) Again, I could not expect my sense
experience ro be any different whether other people are really conscious,
or whether they are simply intricate machines (provided the latter
hypothesis allows for them being very ingeniously constructed). To
adopt the verifiability criterion is to rule out even wondering whether
such things are so; and it would seem that any principle that would
prevent our recognizing the fact that a certain sort of thing exists is
unreasonable. Thus, to show that a certain supposed assertion cannot
be empirically tested is not to show that it is not an assertion; it is
simply to show that it is a very different kind of assertion from scientific
hypotheses and claims as to the nature of the physical world. And it
is hardly surprising that metaphysics and theology should turn out to
be very different from science.”
It is hard to see what the positivist could say in rejoinder. He
might say that the last point is not valid because one cannot talk in¬
telligibly about the sorts of entities the metaphysician is claiming the
positivists arbitrarily bars his talking about. But this claim could be
supported only by recourse to the verifiability criterion, which is the
very point at issue. Again, the positivist might ask the metaphysician
how he proposes to determine, or even make a start at determining,
whether his supposed assertions are true. Even if the metaphysician
should admit that there is no way of showing that any one of these
assertions is true, that in itself would not force him to admit to talking
nonsense. That charge can be made to stick only if we can employ a
wider version of a verifiability criterion, according to which assertive
force is present only if there is some way, empirical or otherwise, of
showing that what one says is true or false. But in any event, meta-
physicians and theologians are rarely prepared to make any such ad¬
mission. They usually think that there is some nonempirical way of
showing that one or another metaphysical or theological position is
correct. Thus, the Platonist metaphysician thinks that the objective
existence of properties apart from their exemplifications can be estab¬
lished by a sort of intellectual nonsensory intuition of such entities,
or perhaps by showing that their existence is a necessary presupposition
of the use of language.14 The positivist may maintain that the only
way of really establishing any claim about matters of fact is the em¬
pirical way, but it is just as difficult to see how one can establish that
claim as it is to see how one can establish the empirical verifiability
criterion of meaningfulness. At this point, we may be at such a basic
stratum of philosophical conviction that no basis can be found for
argument. These are the convictions in terms of which arguments are
given for lesser convictions.
Final But perhaps something can be done to adjudicate the dispute. Re¬
assessment call the point made earlier that ‘meaningful’ is not a term with
clearly defined boundaries. It is quite clear that ‘My car is in the
garage’ does make sense, and equally clear that ‘Quadruplicity drinks
procrastination’ does not make sense. But in between there is a border
region where there are considerations that might well incline us in
either direction. The positivist can at least claim that a sentence like
‘Properties exist independent of their exemplifications,’ which looks
all right except for the absence of possible empirical tests, is defective
in important respects. If it doesn’t lead us, under any conceivable
circumstances, to expect one sort of thing rather than another and if
there is no way in which empirical investigation can throw any light
whatsoever on its truth or falsity, then it is certainly not performing
many of the functions we expect more typical assertions to perform; it
is not a profitable subject of investigation in the way in which many
assertions are. These considerations may or may not lead us to deny
that the sentence makes sense. This will depend on the weight we give
these sorts of considerations relative to others, for example, the meaning¬
fulness of the components of the sentence, the correctness of the gram¬
matical form, and the extent to which the sentence is logically related
to a number of other (nonobservation) sentences, which also figure in
discourse of this kind, for example, ‘Not everything that exists is in
space and time,’ ‘Nothing exists but particulars,’ and ‘Properties exist
only in their exemplifications.’ The important point is not that we
should agree to tighten up the vague term ‘make sense’ in one direction
14 See the quotation from Plato in the Introduction to this book.

rather than another, but that we should clearly see in just what ways a
sentence that is not empirically testable is defective and in just what
ways it is not. Having seen this, we shall know how it can and how it
cannot be employed. And if we know that, labelling it ‘meaningless’ is
of merely ritual significance.
DIMENSIONS OF MEANING
Insofar as we achieve simple, clear-cut specifications of the mean¬

ings of words, we inevitably gloss over certain complexities. We have
already seen how the multivocality of expressions (having more than
one meaning) complicates talk about meaning. This is a relatively easy
factor of which to take account, for we can simply give separate specifi¬
cations of the various meanings of an expression and say something
about the conditions in which it will be used in one or another sense. A
semantic feature that is more difficult to handle is vagueness. Let us
first get clear as to what vagueness is and then see how it complicates
the semantic picture.
what A term is said to be vague if there are cases in which there is no

vagueness is definite answer as to whether the term applies. 'Middle-aged’ is
vague in this sense. At age 5 and at age 80, one is not middle-aged;
at age 50, one is. But what about 39, 41, or 60? There seem to be bands
on either side of the clear cases of middle age where it is not clear what
we should say. To say that there is no definite answer is not to say that
we have not yet been able to give a definite answer because of insuf¬
ficient evidence. The preceding point about 'middle-aged’ is to be
sharply distinguished from the fact that we are not able to say whether
the term 'inhabited planet' is to be applied to Mars. In this case, we
know reasonably well what kind of observations would lead to a positive
or negative answer; it is just that, at present, we are not in a position
to make those observations. But when we are unable to say whether a
41-year-old man is middle-aged, it is not because we have not yet made
certain observations that would settle the question. It is not as if we
84
Dimensions of Meaning 85
could decide this point by settling certain questions about the average
blood pressure or metabolic rate of 41-year-old men. We have no idea
what would definitely settle the question. It is not that we have not
succeeded in finding the answer; there is no answer. This shows that
the situation is due to an aspect of the meaning of the term, rather than
to the current state of our knowledge.
The word Vague' is commonly used very loosely (there is no
inherent reason why Vague’ should be used loosely or even vaguely) to
apply to any kind of looseness, indeterminacy, or lack of clarity. If we
leave it in this condition, we shall run the risk of missing important
distinctions. For example, we should distinguish vagueness, as just de¬
fined, from lack of specificity. If someone says, “We must take steps
to meet this emergency,” or if an advertisement reads, “It’s the hidden
quality that spells true value,” people are likely to respond with “That’s
a very vague statement or Can t you be less vague?”. However, the
main difficulty here is not vagueness but lack of specificity. It is not
that the word steps is vague in that there are cases when it is not clear
whether something should or should not be called a step; and it is not
that there are cases where it can t be decided whether something is or is
not a quality or is or is not hidden. (I am not denying that the words
‘steps,’ ‘hidden,’ and ‘quality’ are vague to some extent. I am saying
that it is not the vagueness that attaches to these words that is primarily
responsible for the insufficient determinateness of these statements.)
The trouble lies in lack of specificity, in simply using the very general
term steps instead of spelling out some specific steps, and in using the
very general term ‘quality’ instead of saying specifically which quality.
We will be most likely to keep this distinction in mind if we restrict
Vague’ to the definition given above. Of course, both vagueness and
lack of specificity can attach in an important way to the same utterance,
as in the advertisement, “Cash loans. Simple requirements.” This is
deficient both because of the failure to be specific as to what the require¬
ments are and because of the vagueness of the term ‘simple.’ (Just how
simple is simple?)
A more serious confusion is that between vagueness and meta-
phoricity. (People talk of “vague and metaphorical language.”) We
shall go into metaphor later in this chapter.
Another confusion that has infected many theoretical discussions
is that between vagueness as a semantic feature of a term, which is
what is specified by the above definition, and vagueness as an un¬
desirable feature of a certain piece of discourse. This distinction is
necessary because of the fact that vagueness in the first sense is not
always undesirable. There are contexts in which we are much better
off using a term that is vague in a certain respect than using terms that
86 Dimensions of Meaning
lack this kind of vagueness. One such context is diplomacy. Suppose

the American ambassador to the U.S.S.R. is instructed to say "My
government will strongly oppose any interference in the internal affairs
of Hungary.” This is vague because of the vagueness of the adverb
‘strongly.’ Just what constitutes strong opposition? Simply expressing
disapproval in a press conference would clearly not be strong opposition,
and declaring war clearly would be. But where in between is the line
to be drawn? Is pressing for a UN resolution strong opposition? How
about an economic embargo, open subsidization of anti-Russian ele¬
ments in Hungary, or sending military “advisers”? The interesting fact
is that there would be grave disadvantages in removing this vagueness.
To commit ourselves definitely to opposition of a given degree of strength
would sharply limit the range of alternatives open to us in a situation
in which there may be a real strategic advantage in 1. keeping the
opponent guessing and 2. choosing an alternative in the light of day-to-
day changes in the situation. We need vague terms for situations like
this.
There are also theoretical advantages to vagueness. Often our
knowledge is such that we cannot formulate what we know in terms that
are maximally precise without falsifying the statement or going far
beyond the evidence. Thus, we have reason to think that city life im¬
poses much more psychological strain on people than country life.
However, in order to formulate this bit of knowledge, we have to use
the vague term ‘city.’ (This word may receive precise definitions in
certain legal contexts, but as ordinarly used it is not quite clear whether
a community of 10,000 inhabitants is to be called a city.) This term
can easily be made more precise by stipulating some minimum number
of inhabitants, for example, 50,000. But as soon as we do so, we can
no longer make the statement with any assurance. There is no precise
population cutoff point such that there is a sharp difference in the psy¬
chological strain imposed by communities above and below this point.
Thus, when we use a word that has the semantic characteristic of
vagueness, it may or may not be a liability. The failure to distinguish
this semantic characteristic from defects of discourse to which it may
give rise has led to an unfortunate transference of the negative evalu¬
ation of the latter to the former. Thinking about language has often
been dominated by the unformulated and unexamined assumption that
vagueness as a semantic characteristic is always undesirable and that an
“ideal” language would contain no vague words.
This kind of distinction is less likely to be missed in the case of
lack of specificity. It is tolerably obvious that to say that one word is
more general than another is not to say that any utterance employing
the former when the latter could have been used instead is going to
suffer from lack of specificity in a pejorative sense. It seems obvious that

sometimes we need a more rather than a less general term, because we
want to say something general. A physicist who says “Metals expand
when heated” rather than, for example, “Iron expands when heated”
is unlikely to be excoriated for lack of specificity.
Kinds of Thus far we have been concentrating on the kind of vagueness

vagueness: that involves the lack of a precise cutoff point along some dimen¬
degree and sion—age, number of inhabitants, or strength of opposition. This
combination of sort of vagueness furnishes the standard examples because it is the
conditions easiest to discern and to analyze, but it is by no means the only
sort. Another, more complex source of indeterminacy in application
can be found in the way in which a word can have a number of inde¬
pendent conditions of application. For a relatively simple example,
consider ‘fruit’ in its culinary rather than biological sense, the sense in
which it is contrasted with Vegetable.' Thus, peaches, apples, bananas,
and plums are clear cases of fruit, while lettuce, tomatoes, carrots, and
spinach are clear cases of vegetables. It would seem that being a fruit
in the biological sense, that is, being the part of the plant that bears
the seeds, is involved in being a fruit in the culinary sense but is
certainly not a sufficient condition, for many “vegetables” are fruits
biologically, for example, tomatoes, beans, squash. If we add the re¬
quirement that fruit in the culinary sense be typically served in a
sweetened form (in desserts, sweet drinks, etc.), then we can rule out
che vegetables that are fruits biologically. We then have at least two
fruit-making conditions, conditions each of which goes some way
toward constituting something a fruit: 1. being a fruit biologically, and
2. being typically served in a sweet form, either because it is naturally
sweet or because it has been sweetened. But just how are we to formu¬
late a sufficient condition for the application of the term? We cannot
say that anything satisfying either of the conditions is a fruit, for we
have seen that many vegetables satisfy the first. We might say that any¬
thing satisfying both conditions is unquestionably a fruit, although
it may be argued that it is not clear that a pumpkin, which satisfies both,
is a fruit. The chief source of indeterminacy comes when we ask whether
something is a fruit if it satisfies condition 2, but not condition 1. One
example of this is rhubarb, which is always served in sweetened form
but which is the stalk of the plant from which it comes. In fact, it is
not clear how rhubarb is to be classified; this indeterminacy became of
legal importance in Canada because of a law that levied different import
duties for fruits and vegetables.
This humble example illustrates the kind of vagueness that stems
from an indeterminacy as to just what combination of conditions is
sufficient or necessary for the application of a term. We often have this

kind of vagueness where there is a plurality of relevant conditions.
Where all the conditions are satisfied, we have an ideally clear case.
There will be some conditions, as in the preceding condition 1, the
satisfaction of which is clearly not sufficient for the application of the
term. But there will be some conditions and combinations of conditions
such that when they are satisfied we do not know what to say, or we
vacillate between one decision and the other, or mature native speakers
of the language disagree. An important and complex example of this is
the term 'religion.’ If we try to list characteristic features of religion,
features such that the possession of any of them does something to
make that which has it a religion, we might come up with the following.
1. Beliefs in supernatural beings (gods).

2. A distinction between sacred and profane objects.
3. Ritual acts focused around sacred objects.
4. A moral code believed to be sanctioned by the gods.
5. Characteristically religious feelings (awe, sense of mystery, sense of
guilt, adoration, etc.), which tend to be aroused in the presence of
sacred objects and during the practice of ritual, and which are as¬
sociated with the gods.
6. Prayer and other forms of communication with gods.
7. A world view, that is, a general picture of the world as a whole and
of the place of the individual in it, including a specification of its
over-all significance.
8. A more or less total organization of one’s life based on the world
view.
9. A social organization bound together by the preceding character¬
istics.
When a cultural entity exhibits all these features to a marked

degree, we have an ideally clear case of religion, as with Roman Ca¬
tholicism, Orthodox Judaism, Orphism.
We can say then that the conjunction of these features provides
a sufficient condition for the application of the term 'religion.’ But is it
also necessary? What happens if one or more of these features is absent,
or present only in an attenuated form? There are many cases of this
sort. Ritual, and the demarcation of certain objects as sacred, can be
sharply deemphasized, as in Protestantism and Islam. It can even drop
out altogether, as with the Quakers and other groups who are mainly
concerned with the cultivation of mystical experience. There are primi¬
tive societies where the moral code has no connection with the cultic
system or the theology; moral rules are thought of as handed down from
tribal ancestors rather than ordained by supernatural deities. Beliefs
in supernatural beings can be whittled away to nothing, while leaving
many of the other factors intact. Thus, in some Unitarian groups and
in Humanism, we have a religiously toned orientation around certain
ideals, such as social equality and a moral code based thereon, without
the ardor being directed toward a supernatural being and without any
cultus m which this ardor is expressed. One branch of Buddhism, the
Hinayana, ignores supernatural beings, at least officially. The emphasis
is on the cultivation of a moral and meditative discipline that will
enable one to attain a state in which all craving has ceased. Finally, the
social group can be reduced to one; that is, a person can develop his
own private “religion.” Spinoza, for example, worked out his own re-
ligion, which was based on a calm and joyful acceptance of everything
that happened as necessarily flowing from the impersonal nature of the
universe.
The important point is that with many combinations of these
features we get uncertainty about and/or disputes over application of
‘religion,’ even when all the “facts” are agreed on. If we have all the
features exemplified, we clearly have a religion; if none or almost none
are exemplified, as with baseball, it is clearly not a religion. Anyone who
disagreed with these judgments would thereby be showing that he did
not understand the word ‘religion.’ But in between there will be several
different sorts of cases in which the application of the terms is problem¬
atic. What are we to say about Humanism or Hinayana Buddhism or
Communism? None of these systems have anything to do with personal
deities. But Communism, for example, strongly exhibits other features;
there is an elaborate cultus, sacred objects (for example, the body of
Lenin, the works of Karl Marx), and a definite world view. It is just
not clear what we should say. And if we go into a primitive society of
the sort previously mentioned, are we going to say that the ritual system
is the religion of the society, despite its dissociation from the moral
code? And what if this ritual system involves no conception of personal
deities? Again, it is not clear what to say, even if we have all the relevant
facts about the society before us. A term like ‘religion’ gets its meaning
through being applied to certain “paradigm” cases like Roman Catholi¬
cism; it is then extended to other cases that do not differ from the para¬
digm in too many respects. But it is impossible to say exactly how many
respects are too many. (We should also note that it is not just a question
of how many conditions are satisfied, for they are unequally weighted.
As we ordinarily use the term ‘religion,’ the absence or near absence of
beliefs in supernatural beings is more of a reason for denying application
of the term than is the absence or near absence of ritual or the restriction
to a single person. Presumably, this usage is connected with the fact
that we live in a relatively nonritualistic culture.) Actually, ‘religion’
exhibits both kinds of vagueness. Even if we could say exactly which
or how many of the various religion-making characteristics a cultural

entity has to have in order to be a religion, we would be unable to say
with respect to a given characteristic, exactly what degree of it we must
have in order to apply the term. Many terms for cultural entities share
this double vagueness.1
is absolute Is every word vague to some extent? Since there are contexts in
precision which it is important to use language with as much precision as
possible? possible, this is a significant question. Quite often when one sets
out to make a term more precise, it will turn out that the terms
he employs to remove the vagueness in question will themselves be
vague, though perhaps to a lesser degree and/or in different respects.
Thus, if we try to remove the vagueness of the everyday word ‘city’ by
stipulating that a community is a city if and only if it has at least
50,000 inhabitants, this removes the vagueness consisting in an in¬
determinacy as to the minimum number of inhabitants required; how¬
ever, now the spotlight may be shifted to other areas of vagueness, for
example, the term ‘inhabitant.’ Under what conditions is a person to be
counted an inhabitant of a community? It is clear that a person who
resides and works within the boundaries of a community is an in¬
habitant; and it is clear that one who has never set foot within it is not
an inhabitant. But what if he owns a residence in the community that
he occupies only in the summer, renting it out and living elsewhere the
rest of the year? What if he attends college in the community, living
in a dormitory while the college is in session but living outside the
community while the college is not in session? What if he is living and
working in the community for a fixed two-year period, but owns a home
in another community which contains most of his belongings and to
which he plans to return after this assignment is completed? Is he an
inhabitant of the community during this two-year period? ‘Community’
is subject to a somewhat different sort of vagueness, which consists not
in an indeterminacy as to what is to count as a community but rather
as to what is to count as a single community. For political purposes,
these questions are settled by legislation. The boundaries of a given
community are established by law for matters of taxation, police au¬
thority, and eligibility to vote. But for other purposes—for example,
sociological research—these boundaries may be of no importance. Thus,
“a town” that straddles a state line may or may not be counted as a
single community, depending on the questions at issue. Similarly, de¬
pending on the kind of problems at issue, Staten Island may or may not
be counted as belonging to the same community as Manhattan Island.
1 For a discussion along these lines of ‘poem,’ see C. L. Stevenson, “On ‘What
Is a Poem’?” Philosophical Review, LXCI (July 1957).
This example is instructive in several ways. First, it would be a

serious mistake to suppose that we are making no progress but are
simply replacing one example of vagueness with another. We have re¬
moved one element of vagueness—the indeterminacy as to the number
of inhabitants required and we have not introduced any new vagueness;
rather by removing the first, we have made visible other vaguenesses
that were there all along. Whatever vagueness attaches to ‘inhabitant’
automatically attaches to ‘city,’ whether or not we have definitely estab¬
lished how many inhabitants are required before we have a city. It is
just that the indeterminacy over the number required is a more obvious
feature and until that is removed we don’t notice the problems as to
what makes someone an inhabitant. Second, it is worthy of note that
the problems raised about ‘inhabitant’ concerned the exact combination
of conditions required for the application of the term, rather than a
cutoff point along one or more dimensions, as with ‘city.’ Sometimes
the residual vagueness will be of the sort removed, sometimes of a
different sort, sometimes of both sorts.
The third point to be noted is that even if we could decide just
which combination of conditions was necessary and sufficient for the
application of the term ‘inhabitant,’ the terms in which these conditions
are stated are themselves more or less vague. For example, we made use
of the term works in the community.’ No doubt, there are many cases in
which the applicability or inapplicability of this term is unproblematic,
but there are problematic cases as well. What of a salesman, the home
office of whose company is in the community but who, by the nature of
his work, spends most of his working hours elsewhere? Or, conversely,
what of a man whose employer is elsewhere but who spends most of his
working hours in the community in question, as a consultant or lobbyist?
And what of a writer who happens to do most of his writing within the
boundaries of the community? Does he “work in the community”? The
term ‘occupies a home’ is also subject to vagueness. If a person owns
several houses, does not rent any of them to others, and spends part of
his time in each, does he occupy all of them or some one or more of
them? And so it goes.
Precision At this point we may feel that the removal of all vagueness from
through a given term is an unrealistic goal; the most we can hope to do is
quantification to approach it asymptotically. But before embracing that conclu¬
sion, we should consider the way scientists have tried to extricate
themselves from the bog by replacing qualitative with quantitative
terms. So long as we simply try to be more definite about the conditions
of application without stating these conditions in terms of exact quanti¬
tative limits, as with our discussion of ‘inhabitant,’ it seems clear that
we are going to wind up repeatedly with terms that are vague in one
way or another. But if we do something like replacing ‘hot’ and ‘cold’
with numerical degrees of temperature, we may be able to do away
with vagueness completely. We should not suppose, however, that the
introduction of numbers is itself a panacea. The above discussion of
‘city’ suffices to show that. The introduction of a quantitative limit did
not get rid of all vagueness, for the simple reason that we still had the
problem of identifying the units to be counted. This problem arises
for any activity of counting. In order to determine how many P’s there
are, we have to be able to tell 1. when we have P rather than Q, and 2.
when we have one P rather than more than one. Insofar as it is impos¬
sible to settle one or both of these questions, vagueness will attach to
the numerical statement that there are so many P’s. Difficulty over 1
has been illustrated with ‘inhabitant’ and difficulty over 2 with ‘commu¬
nity.’ Both kinds of difficulty attach to many attempts to attain precision
through specifying numbers of units required. Thus, we may try to
remove the vagueness of “mountainous region” by requiring the pres¬
ence of at least five mountains of 5,000 feet or over. But we run into
the first problem in trying to decide whether we should say there are
any mountains in a region that consists wholly of a plateau, the average
elevation of which is 7,000 feet and which contains twelve noticeable
elevations over the level of the plateau, ranging from 7,500 to 8,500 feet.
The second problem emerges when we take something that is clearly a
mountain range and tty to divide it up into constituent mountains. Are
two noticeable peaks divided by a saddle not very far below the height
of the lower of the two peaks two mountains or only one? 2
The introduction of measurements of positions along a continuum,
as with length, temperature, and weight, is significantly different from
the procedure of counting units specified with unreformed terms from
ordinary language. If we replace ‘large city lot’ with ‘city lot containing
at least 20,000 square feet,’ or replace ‘cold drink’ with ‘drink, the
temperature of which is 45 degrees F. or less,’ we do not run into any¬
thing like the problems we encountered in counting inhabitants. Of
course, we still have the problem of determining when we have a city
lot, and when we have one city lot rather than two city lots; but there
is no problem about identifying degrees of temperature or square feet,
or about determining that at a given moment we are dealing with one
2 This distinction between two problems cuts across the distinction already made
between two kinds of vagueness. That is, a vagueness as to what counts as one P,
as well as a vagueness as to when we have P rather than not-P, can be either of
the degree or the combination of conditions sort. Our initial illustrations of the two
kinds of vagueness were cases of vagueness as to when we have P rather than not-P.
But by now we have examples of both kinds of vagueness with respect to the other
sort of question. The point about 'community’ exemplifies combination of condi¬
tions vagueness, and the point about ‘mountain’ exemplifies degree vagueness.
square foot rather than two. This, however, does not mean that in-
determinacies of another sort might not emerge. Any measurement is
reported subject to a certain margin of error. That means that we can
never be absolutely certain that a lot boundary is exactly 100 feet—
neither a fraction more nor a fraction less. (Though, of course, this
measurement can be made in such a way that any uncertainty as may
remain is of no practical importance.) This could be put by saying
that 100 feet long’ is vague to a certain extent, because in no situation
are we ever absolutely certain that it is applicable. But this fact seems
to be significantly different from those we have been discussing. This
indeterminacy is due to inherent limitations on our powers of measure¬
ment, rather than to a feature of our language that might conceivably
be altered in some other language. In other words, this uncertainty as to
application is due to a certain insufficiency in the data (albeit an in¬
sufficiency that will never be remedied), rather than to a semantic
feature of the words used. Each particular case of vagueness that we
have considered can be removed by deciding to tighten up the criteria
of application in a certain respect; but the indeterminacy stemming from
the margin of error in measurements cannot be removed by any decision
that is in our power to make. It seems best, then, not to regard this as a
kind of vagueness, and to admit that in this one kind of case, at least,
we have terms that are completely free of vagueness. But it must be
remembered that it is only 200 square feet,’ not ‘city lot containing at
least 200 square feet,’ which has been declared vagueness-free. That is,
any application of the nonvague measurement terms is going to exhibit
whatever vagueness attaches to the terms we use to talk about whatever
is being measured. This consideration is particularly important in the
social sciences, where the use of precise measures is apt to mask the
vagueness of the terms we use to specify what is being measured. Thus,
we may get very precise looking results correlating degrees of prejudice
toward Jews with degrees of acceptance of oneself. But the fact that we
subject our data—for example, answers to questionnaires and clinical
psychologists’ ratings of responses to projective tests—to elaborate alge¬
braic manipulations should not lead us to forget that in taking our final
numerical results to give us degrees of acceptance of oneself, we are
subject to all the indeterminacy that attaches to questions of the form,
“Does Jones accept himself?”
Open texture Friedrich Waismann, in his well-known essay, “Verifiability,” 3

suggests that with respect to certain kinds of terms, particularly
nouns denoting physical objects, there is a virtually inexhaustible source
of vagueness, which will remain undisturbed after maneuvers of the
3 In Logic and Language, First Series, ed. A. Flew (Oxford: Basil Blackwell
1952).
preceding sorts. Waismann points out that apart from actual cases of
indeterminacy of application, one can think of an indefinite number of
possible cases in which one would not know what to say. He asks us to
envisage the possibility that something to which we had been applying
the word ‘cat’ with complete assurance should suddenly begin to speak,
or should grow to the height of twelve feet, or should vanish into thin
air and then reappear and vanish again from time to time. In such
cases, we would not know whether to apply the term. Waismann also
points out that a scientific term, like a word for a chemical element,
‘gold,’ which we would ordinarily think of as quite precisely defined, is
in fact defined in such a way that we are prepared to take any one of
several characteristics as conclusively showing that what we have is gold.
These include the specific gravity, the spectrograph of light emitted when
placed in a flame, the X-ray spectrum, and the ways it enters into chem¬
ical composition with other substances. Even if each of these criteria is
quite precisely conceived, we can easily imagine a situation in which
we would not know whether to apply the term ‘gold,’ namely, a situation
in which some of the tests indicated gold and the others didn’t. We
might say that we use ‘gold’ subject to a presupposition that positive
results on these tests will always go together, and that we use ‘cat’
subject to a presupposition that anything satisfying the ordinary criteria
for something being a cat is not going to suddenly grow to a height of
twelve feet, is not going to periodically vanish into thin air and reappear,
and so on. The difference between these two cases is that the features
Waismann mentions for ‘gold’ are quite explicitly involved in our ordi¬
nary use of the term, while the ones he brings out for ‘cat’ are not. We
would never dream of making sure that an object does not disappear
into thin air before saying that it is a cat. (This contrast holds for the
things that Waismann said about ‘gold’ and ‘cat,’ rather than for these
words themselves. We could undoubtedly parallel the point he made
about cat for gold’ and vice versa.) The ‘cat’ case is the more im¬
portant one, for no definite limit can be placed on the kind of condi¬
tions mentioned there. The number of wild situations we could envisage
in which we would not know whether to say that what was before us was
a cat is limited only by the extent of our ingenuity. As Waismann puts
it, when we form a concept, we only have certain kinds of situations in
mind; as a result, the concept is armed only against certain contingencies.
This feature of a term Waismann calls “open texture,” or “possibility of
vagueness. His claim is that this kind of indeterminacy can never be
completely eliminated; for although we can make a decision as to what
we would say in any given kind of case, for example, vanishing into thin
air and not reappearing, there will always be an indefinite number of
other conceivable cases with respect to which the concept is still not
delimited. It is not clear that it is, strictly speaking, impossible to decide

all such cases in advance, for it is not clear just how numerous they
are. But Waismann is surely justified in holding that, in fact, there will
always be this kind of penumbra of indeterminacy attaching to physical
object terms, and that in this they are to be distinguished from arithmet¬
ical terms, for example, with respect to which no such problems arise.
importance of Vagueness, like multivocality, can complicate specifications of

the notion meaning and identification of synonyms. For even if two expres-
of vagueness sions are perfectly synonymous on other grounds, they may differ
in that one is vague in a respect in which the other is not. A re¬
lationship of this sort is what we are looking for when we try to make a
term more precise; the hope is that, for example, ‘drink with a tempera¬
ture below 45 degrees F.’ has exactly the same meaning as ‘cold drink,’
except that there is an indeterminacy in the latter that is absent in
the former. But when we are trying to bring out the meaning of an
expression as it actually is rather than polish it up, what we need is
another expression that matches as exactly as possible the vagueness of
the former. Thus, in defining ‘adolescence’ as the period of life between
childhood and adulthood, we presumably have a good match. For the
indeterminacy of the boundaries of adolescence is just the same as that
which attaches to the upper boundary of childhood and the lower
boundary of adulthood. There are many other places where vagueness
should be, but often is not, taken account of. Thus, claims that a given
statement is analytic (true solely in virtue of the meanings of the con¬
stituent terms) are liable to founder on the fact that the terms, though
properly lined up in other respects, do not match with respect to vague¬
ness. Thus, it may be claimed that “If someone lives in a city, he lives
in a large community” is analytic. In adjudging this claim we should
try to determine whether the vagueness of ‘city’ matches that of ‘large
community.’ We should not suppose that it is possible to make any
precise determination of this. It is characteristic of vague terms that
there is no precise boundary between areas of clear application or non¬
application and areas of indeterminacy of application, any more than
there are sharp boundaries between application and nonapplication.
This is not surprising. It would be absurd to have the term ‘city’ sharp¬
ened to the point at which the area of indeterminacy of application is
bounded precisely at, for example, 25,000 inhabitants and 40,000 in¬
habitants, without going farther and making a sharp boundary between
application and nonapplication, that is, removing the area of indeter¬
minacy altogether.
Vagueness should also be recognized in trying to give criteria of
meaningfulness, like the verifiability criterion. Insofar as a term is vague,
the question as to what evidence would confirm or disconfirm a state¬

ment asserting it of something does not have any precise answer. Thus,
to recur to an earlier example, we can give no clear-cut answer to the
question: “What evidence would confirm or disconfirm the thesis that
there is no society without a religion?” For as we saw, in the face of
certain collections of evidence, for example, a ritual system, a set of
beliefs about spirits, and a moral code—none of which have any close
connection with the others—it is, in principle, unclear whether we
should say that the society has a religion.
Finally, the fact of vagueness forces us to make some sort of
qualification in the supposedly self-evident “Law of Excluded Middle,”
that every statement is either true or false. For, as we have seen, where
we have a borderline application of a vague term, it is, in principle,
impossible to pronounce the statement either true or false. This will
have to be handled either by qualifying the “Law,” by denying that in
these cases we really have a statement, or by saying that the Law holds
only for a language in which all the terms are absolutely precise (and
then what are we to say about languages that actually exist?).
Metaphorical Another thing that complicates the job of the semantic theorist
and other is the fact of metaphorical uses of terms. After one recognizes the
figurative complications introduced by multivocality, he is still likely to think
uses of that one can give a complete account of the semantics of a lan¬
expressions guage by specifying each of the senses of each of the words (or
whatever is taken as minimum meaningful units), together with a
set of recipes for deriving the meanings of larger units from the meanings
of their elementary components plus the mode of combination. But even
a complete system of this sort will not cover what is going on when
e.e. cummings says:
the sweet small clumsy feet of april came into the ragged meadow of
my soul
By any recognized method of distinguishing senses of terms, there is

no sense of the word ‘feet’ in which we can speak of the feet of a
month; nor is there any sense of ‘meadow’ in which a soul can have a
meadow. Nor is cummings introducing new senses for them; what he
says is intelligible to a reader only if that reader knows certain estab¬
lished senses of these words. The uses they have here are somehow
parasitic on those established uses; they constitute a certain kind of
extension of them. It is an extremely important fact about language
that it is possible to use a word intelligibly without using it in any
of its senses. There are other places in the philosophy of language where
this fact has the effect of blurring sharp distinctions. ‘Green ideas sleep
furiously’ is a common example of a meaningless sentence, one that is

meaningless by reason of a “category mistake.” (Ideas belong to the
wrong category for being qualified by color terms.) But it is not at all
difficult to imagine a poetic context in which this sentence would be
quite appropriate, in which it could be used by the poet to communicate
what he wanted to communicate. And so if we want the term meaning¬
less’ to preserve contact with ‘uselessness for communication,’ we must
not say, unqualifiedly, that the sentence is meaningless, but only that it
is meaningless in literal or prosaic discourse. And to the extent that such
qualifications are necessary, it will be necessary to clarify the contrast
between literal and figurative discourse.
Let us use the term figurative’ in the following way. WTerever
an expression is used so that, even though it is used in none of its
established senses, nevertheless, what is said is intelligible to a fairly
sensitive person with a command of the language, the expression will
be said to be used figuratively. It is obvious that this sort of thing is pos¬
sible only if these uses are somehow derivative from uses in established
senses. Otherwise a knowledge of the standard senses would do nothing
to enable a listener, however sensitive, to see what is being said. We
can distinguish different kinds of figurative use in terms of the basis of
derivation. Where the derivation is on the basis of a part-whole relation¬
ship, as when we say “The first ship opened fire” (it is a part of the ship
that literally opened fire) or a genus-species relationship, as “I have not
spoken to a single creature for a week” (that is, I have spoken to no
men, a species of creature), one traditionally speaks of synecdoche. The
term ‘metonymy’ has been used to cover cases in which the transfer is
made on the basis of any one of a number of relationships, such as
cause-effect, as when we say of a performer “He got a good hand” (that
is, he got a lot of something produced by hands) or container-contained,
as in “The White House had no comment.” (None of these are figura¬
tive uses, as I have just defined that notion, because in all these cases
the expressions in question are being used in established senses. These
are examples of figurative senses. But they can serve as examples of
kinds of basis of derivation.) Metaphor is that sort of figurative use in
which the extension is on the basis of similarity. Consider the following
passage from Shakespeare’s Macbeth:
macbeth: Methought I heard a voice cry “Sleep no more!

Macbeth doth murder sleep,” the innocent sleep,
Sleep that knits up the ravell’d sleeve of care,
The death of each day’s life, sore labour’s bath,
Balm of hurt minds, great nature’s second course,
Chief nourisher in life’s feast, . . .
Act II, scene ii.
It is because of some kind of similarity between what Shakespeare is

attributing to sleep in the third line of the above quotation and what
happens when one knits up a ravelled sleeve that one can understand
what is being said, even though the words ‘knit’ and ‘sleeve’ are not
being used in any established sense. Since metaphor is the most wide¬
spread and theoretically the most interesting of the figures of speech,
we shall concentrate our attention on it. This treatment can, to a large
extent, be transferred to the other figures of speech.
The nature In his essay, “Metaphor,” Paul Henle has given an illuminating
of metaphor analysis, which employs Peirce’s notion of an icon,4 (See Chapter
3.) As pointed out previously, it is clear that in a metaphor one
of the established senses of the expression is involved. This is clear, for
unless one understands the relevant established sense, he will not be able
to understand the metaphor. Unless one can understand sentences like
“I knitted up the ravelled sleeve of that sweater,” he will have no chance
of understanding the third line of the above quotation. But one is not
actually using the term in this sense; or, at least, one is not simply using
it in this sense. One is somehow using the term to say something dif¬
ferent, though related, and working through the established sense in
order to do this. Although so much is abundantly clear, it is no easy
task to give a precise specification of the mechanics of the operation.
Henle suggests that we should think of the expression functioning in
one of its established senses to specify a kind of object or situation that
we are directed to use as an icon of what we are speaking about
metaphorically.
Metaphor, then, is analyzable into a double sort of semantic relation¬

ship. First, using symbols in Peirce’s sense, directions are given for
finding an object or situation. This use of language is quite ordinary.
Second, it is implied that any object or situation fitting the direction
may serve as an icon of what one wishes to describe. The icon is
never actually presented; rather, through the rule, one understands
what it must be and, through this understanding, what it signifies.5
On this analysis, to speak of sleep as knitting up the ravelled sleeve

of care would be to say something like: “Consider a woman knitting
up the ravelled sleeve of a sweater, and you will have an icon of the
action of sleep on a careworn person.” Instead of saying directly and
explicitly what sort of effect sleep has on a careworn person, Shakespeare
has presented us with another situation in which an agent is altering
something in a certain specified way and is, in effect, suggesting that
4 In P- Henle, ed.. Language, Thought, and Culture (Ann Arbor, Mich.: Uni¬
versity of Michigan Press, 1958).
5 Op. cit., p. 178.
by considering this situation we can realize something of the effect

of sleep on a careworn person. But this is only possible if there is some
important and readily noticeable similarity between the two situa¬
tions; such similarity is a necessary condition of successful metaphor.
If sleep had been apostrophized as that which “hammers nails into
arrogance” we would not know what to make of it. This is not to say
that metaphor is the same as simile, an explicit assertion of a similarity.
In “The action of sleep on a careworn pi son is similar to the action
of a knitter on a ravelled sleeve,” no expression is used metaphorically.
Nevertheless, it remains true that the existence of such a similarity is
presupposed by the metaphor. Thus, the difference between metaphor
and simile is somewhat analogous to the difference between 'My son
plays baseball’ and 'I have a son and he plays baseball,’ where what is
presupposed but not explicitly asserted in the first is explicitly asserted
in the second.
We begin to see something of the pervasive importance of the
capacity of language for figurative uses when we consider the phe¬
nomenon of “dead metaphor.” The language is full of senses of terms
that we can plausibly suppose, and in some cases even show by historical
research, to have developed out of metaphorical uses of words. Consider
such phrases as ‘fork in the road,' ‘leg of a table,’ ‘leaf of a book,’ ‘stem
of a glass,’ and ‘eye lids.’ In the present state of the language, the word
‘fork’ has as established a sense in this phrase as it does in the phrase
‘knife and fork.’ But we can well imagine that at an earlier time when
the word was regularly applied only to the eating and cooking imple¬
ment, people would use the word metaphorically in speaking of a place
in which a road divided into two parts, each of which continues in
roughly the same direction but making an acute angle with the line of
direction of the original road. This use then “caught on,” and because
new generations could learn to apply the term directly to situations of
this sort without needing to go through the older use, the sense in
which the word is applied to roads came to be one of the established
senses of the term. This example illustrates the very important role
of metaphor in initiating uses of words that can eventually grow into
new senses.
Basis of the We have been putting a great deal of weight on the notion of
literal- an established sense of a term. The problem was presented as
metaphorical arising from the fact that words can be used intelligibly without
distinction being used in any established sense; and the account of metaphor
given is such as to separate metaphorical uses of words from those
uses in which the word is being used in a sense it actually has in the
language. But how do we know that in ‘He knit his brow,’ ‘knit’ is being
used in one of its established senses, whereas in 'Sleep knits up the

ravelled sleeve of care/ it is not? One might object that the fact that the
second sentence is so readily intelligible shows that the word is being
used in a sense that a native speaker of the language would recognize
it as having. On the other side, it can be pointed out that there are so
many contexts of this sort that the project of listing a separate sense for
each context would not recommend itself to the dictionary-maker as a
feasible one. (And it may not be possible, in principle, to list all the
contexts in which a word could be used intelligibly without being used
in any of its commonly recognized senses.) But it may be said that al¬
though this is an important practical consideration for the practicing
lexicographer, it does not decisively settle the theoretical question as to
how many senses a word, in fact, has. To this it may be replied that it
is a great mistake to suppose that there is any such thing as the number
of different senses a word really has, independent of any considerations
as to what is the most perspicuous way of representing its use in the
language. However, it would be desirable to find some basis for refusing
to distinguish separate senses for each metaphorical context, other than
the difficulty of managing that many different senses. Perhaps we can
find such a basis by considering the different ways in which what is
being said on a given occasion can be explained to someone.
If someone does not understand what was said, the simplest way
of bringing him to understand is to paraphrase the remark. The whole-
sentence can be paraphrased: when I said “Is that all right?” what I
meant was “Will I be penalized for that?” Or the restatement may be
confined to some component of the sentence, if it is clear that the
failure to understand is concentrated at that point: “What do you
mean, you've got a new case?” “By 'case' I mean example.” This tech¬
nique can be used with metaphorical utterances. Thus, I can explain
‘Sleep knits up the ravelled sleeve of care’ by saying, “That means
that after a good night’s sleep your cares and worries will not seem
as pressing as they did before.” But with utterances ordinarily regarded
as metaphorical, this method is not as adequate as it is elsewhere. This
kind of paraphrase fails to bring out the way in which what we are
saying about sleep is based on the notion of someone repairing the
sleeve of a garment by knitting. In bypassing the metaphorical exten¬
sion, it thereby fails to bring out the richness of what had been said.
Therefore, an explanation that explicitly brings out the comparison will
be more adequate. Just as in knitting up a ravelled sleeve one makes
it whole again, restores it to its proper use, so when a careworn person
gets a good night’s sleep he is thereby restored to a condition in which
he can function with normal effectiveness.” (I am not suggesting that
this is an ideally adequate or complete example of this type of expla-
nation. The richer and more suggestive a metaphor is, the more im¬
possible it is to spell out explicitly all the similarities that underlie it.)
This will be more adequate because it not only makes clear what fact is
being asserted concerning sleep but also makes explicit the way in which
this fact is being asserted. Thus, we can use the necessity of this kind of
explanation as a criterion of metaphorical use as opposed to use in an
established sense. This fits in with the account given of metaphor. It
is because the metaphorical use of an expression involves a double
operation, in which we operate on the basis of an established sense
but go beyond it, that the more elaborate explanation is needed.
What this criterion actually gives us is not a black and white
distinction of kinds but a continuum of degrees. At one end, we have
the clear cases of literal uses of terms, like “She is knitting a sweater.”
At the other end, we have clear cases of metaphors like “The sweet
small clumsy feet of April came into the ragged meadow of my soul.”
Near the latter end, we have relatively standard metaphors like ‘He blew
his top,’ ‘Russia has dropped an iron curtain across Europe,’ ‘Religion
has been corroded by the acids of modernity,’ or ‘He got into a stew,’
where the frequency of sentences like this may lead one to distinguish a
special sense of the word for such contexts. Thus, Webster’s New
Collegiate Dictionary, 1959 edition, lists as one of the senses of ‘curtain’:
“anything that acts as a barrier or obstacle by protecting, hiding, or
separating; as, a security curtain” (p. 204). And one of the senses listed
for ‘stew’ is: Colloq. A state of agitating worry” (p. 831). Neverthe¬
less, in each of these cases a more standard sense of the term is so clearly
in the background that we are perhaps justified in feeling that one has
not brought out the full force of what is being said unless one has made
explicit the underlying comparison between, for example, the violent
random motion of the meat and vegetables in a stew and the typical
activity of a person who is “in a stew.” Further down the scale toward
literal meanings are the senses sometimes termed “figurative,” such as
the sense of ‘cold’ in ‘He’s a very cold person,’ the sense of ‘dead’ in
‘The socialist movement in this country is dead,’ and the sense of ‘hard’
in ‘hard liquor.’ These are established senses, but very little reflection
is needed to realize that they are derivative from more basic senses in
the same sort of way as that in which a figurative use is derivative from
use in an established sense. These senses can be separately specified;
‘cold’ in ‘He’s a cold person’ means ‘lacking in emotional expression.’
Still, it seems that we would have some tendency to feel that a person
who learned this just as a separate sense without seeing that to be cold
in this sense is importantly like being at a relatively low temperature
would be missing something. Still further down the scale will be what
we earlier called “dead metaphors.” Here there is little tendency to
insist that one has not fully understood what one is saying when he
speaks of a fork in the road unless he sees the similarity between this
and a kitchen fork. The later sense has become almost completely auton¬
omous. Nevertheless, the relation of derivation can be recovered on
reflection; thus, we still have something that is distinguishable from
senses so related, or unrelated, that we cannot discover even an archaic
figurative derivation.
Until the last paragraph, I have succeeded in avoiding the term
‘literal,’ the usual contrast with ‘metaphorical.’ I have tried to avoid it
because in the hands of the advocates of the verifiability theory, and
other partisans of sweeping dichotomies, it has all but lost its useful¬
ness. Originally, a literal use was one “according to the letter,” which
meant pretty much what I have been trying to convey by the term
‘established sense.’ But logical positivists began to use the term ‘literal
meaning’ for what was countenanced by their criterion, and along with
this they tended to use ‘metaphorical,’ ‘emotive,’ and ‘poetic’ indiscrimi¬
nately for what was rejected by their criterion. Such usages only spawn
confusion. There is even less justification for the phrase ‘literal meaning’
than for the phrase ‘cognitive meaning’ or ‘factual meaning.’ A term
can be said to be used literally when it is used in such a way that the
meaning of the sentences in which it occurs is a determinate function
of one of its senses. It is a mistake, however, to think that ‘literal’ de¬
notes a kind of meaning. As should be clear from the above account,
whenever we use an expression with an assignable meaning we are using
it literally. Thus, all meanings would be “literal”; the term as so used
has no distinguishing power. The confusion thickens when it is supposed
that “literal meaning” is empirically respectable meaning, as opposed
to what is indifferently labelled “emotive meaning” or “metaphorical
meaning.” Quite apart from the confusion involved in using these terms
to mark “kinds of meaning,” there is no justification for regarding a
metaphorical statement as ipso facto unverifiable. It is true, of course,
that when one is speaking metaphorically, it is generally more difficult
to be sure of exactly what he is saying than when he is speaking literally.
In effect, he has given us something as a model for something else
without making explicit in exactly what way it is supposed to be a
model; we have lost the controls that come from using words in estab¬
lished senses. But this simply means that what he is saying is, to a
certain extent, indeterminate, just as it is if the expressions he uses
are vague. It does nothing to show that what he is saying, insofar as one
can be sure of what it is, cannot be put to an empirical test; and still
less does it show that what he is saying is distinctively “emotive.” When
I say “Sleep knits up the ravelled sleeve of care,” to a large extent I am
saying the same thing one would normally say in uttering the sentence,
As a result of getting a good night’s sleep a careworn person will be

less disturbed and distracted by his cares.’ And to the extent that it can
be determined that I am saying the same thing, there is no difficulty in
seeing what sorts of observations would confirm or disconfirm what I
am saying. There are, however, many metaphorical statements for which
it is difficult to imagine any empirical test, for example, “Life’s ... a
poor player that struts and frets his hour upon the stage and then is
heard no more.” But this is not because the statement is metaphorical.
If we take an approximate literal paraphrase, for example, “Life is
futile,” we will have the same kind of difficulty in imagining an empir¬
ical test for it. Although it may be the case that empirically untestable
statements often assume a metaphorical form, it is not the fact that
they are expressed metaphorically that makes them untestable.
Irreducible We have briefly considered the topic of dead metaphors. A class

metaphors: of metaphors that is of special interest philosophically is made up
God and of those that cannot die. They are importantly involved in theology
inner feelings and in descriptions of feelings. To see that this is so, we shall have
to go more deeply into the distinction between the words ‘literal’
and ‘metaphorical’ than we did previously.
Thus far we have rested the literal-metaphorical distinction on
the notion of established senses of an expression, without looking very
far into what is involved in a sense becoming established. If we accept
uncritically the usual lexicographical procedures, there is no doubt that,
on the account of the literal-metaphorical distinctions that we have
been giving, the following statements involve literal rather than meta¬
phorical uses of words.
1. God made the heavens and the earth.
2. God spoke to the prophets in days of old.
3. God has punished me for my sins.
4. I felt a stabbing pain.
5. When I heard the news, I was transported.
6. I felt constricted.
7. When I hear that music, I get the feeling of marching along in a
triumphal procession.
With respect to the feeling terms, this is so because we have used ‘stab¬
bing,’ for example, so often in application to pains, that by the usual
lexicographical criteria one is justified in distinguishing a separate sense.
As for the theological contexts, it is not that special senses are involved;
rather, senses are specified in such a way as to cover application both to
man and God. Thus, one of the senses of ‘make’ in Webster’s New Col¬
legiate Dictionary is “To cause to exist, appear, or occur” (p. 507), and
one of the listings for ‘punish’ is “to afflict with pain, loss, or suffering
for a crime or fault” (p. 685). Nevertheless, it is a very important fact

that some of these established senses (or, alternatively, some of the uses
of these established senses) are related to others in a way similar to that
in which metaphorical uses are related to literal uses. This begins to
appear if we ask how we would explain to someone what is meant by
“God has punished me” and by “My father has punished me.” It is a
striking fact that although the latter can, in principle, be explained by
directing the learner’s attention to certain observable transactions, the
former can be explained only by directing the learner to take cases
of one man punishing another as models (or icons) of what is meant,
much as a clearly metaphorical statement directs the listener to take a
certain kind of situation as a model of something else. This priority is
necessary and irreversible. There is no way of explaining what it is for
God to punish someone other than by saying, in effect, that it is some¬
thing like one man punishing another, for it is impossible to use any¬
thing like the “ostensive” teaching device usable for ‘My father punished
me.’ Whether or not a person can be aware of God punishing him
need not be decided in order for us to make our point. Ostensive
teaching is impossible here because it is impossible for the teacher to
tell when the learner is aware of God punishing him (unless the learner
can tell him that he is so aware; but in that case, the “learner” has
already learned his lesson in some other way). For divine punishment,
there is nothing analogous to the publicly observable direction of visual
attention that shows the teacher when the learner is paying attention
to an act of human punishment. Furthermore, it is impossible to define
God punished me in any terms that are not analogically derivative from
talk of human activity. I do not have the space to justify this last claim,
for that would involve going through all the possible sorts of definition
that seem initially plausible; the reader is invited to try any verbal ex¬
planations of the sentence he can think of and see what happens. Let
us take a brief look at one possibility. It may be said that ‘God punished
me’ means ‘God caused me to suffer as a result of my doing something
wrong.’ This latter sentence may be acceptable as an equivalent, but the
fact remains that ‘cause’ in this context raises exactly the same problems
as ‘punish.’
We do not have to rely on so controversial an area of discourse
as theology for examples of this phenomenon. The point can be amply
illustrated from our talk about feelings and sensations. There are some
feeling and sensation terms that can be intersubjectively stabilized by
reference to more or less constant situational or behavioral concomi¬
tants. These include ‘pain,’ ‘depressed,’ ‘excited,’ and ‘sleepy.’ But our
talk about feelings and sensations is by no means confined to these
terms. If we consider the terms we use to distinguish one kind of pain
from another, we shall find it impossible to tie them down in this

way. If we want to explain to someone what a stabbing pain is, as
opposed to a dull ache or a burning pain, we shall be forced back to
analogy. A stabbing pain is one that is like the pain one would receive
if he were stabbed. There is no other way to make explicit the meaning
of the expression. And to express our feelings with any fineness of
detail, we must use many terms that depend on analogies for their
significance. There is no publicly observable situational feature, facial
expression, or behavior that is specially tied to feeling constricted, as
contrasted with feeling embarrassed, apprehensive, or out of place.
Again, it seems that one can explain the term only by saying that to
feel constricted is to feel something like the way one feels when he is
physically constricted. Let us use the term 'quasi-metaphorical’ for
senses or uses of words that can, in the last analysis, be explained only
by analogy with what is talked about when these or other words are
used in senses that can be explained more directly.
Quasi-metaphors are, in the epistemologically crucial respects, in
the position of metaphors that cannot die. They have the character¬
istic indeterminacy of metaphors, but they lack the means, available
for many metaphors, for removing this indeterminacy. As we have seen,
a metaphor in the raw simply consists of specifying a model or icon for
something without specifying the respects in which it is an icon. Never¬
theless, in many cases there are devices for removing, or at least sharply
reducing, this indeterminacy. Sometimes we can point out examples of
what we are talking about and leave the hearer to see for himself what
the relevant similarity is. In other cases where, owing to the abstractness
or pervasiveness of the subject matter, this cannot be done, we can give
at least an approximate paraphrase in literal terms, as in the preceding
treatment of 'Sleep knits up the ravelled sleeve of care.’ But in a quasi¬
metaphor, as we have defined that term, neither of these devices is
available. We cannot point to cases of God punishing someone or to
cases of someone feeling constricted; even if we could point to such
cases, doing so would do nothing toward helping someone realize what
features of human punishing and physical constriction are being carried
over into these concepts. And we cannot give paraphrases of 'God has
punished me’ or ‘I feel constricted’ in terms that do not themselves
involve quasi-metaphors. This means that the indeterminacy character¬
istic of metaphors is indelibly stamped on these areas of our talk. It is
not clear either exactly what we are saying or what would be required
to confirm or disconfirm it. Concentration on statements of this sort
is perhaps partly responsible for the mistaken view that metaphorical
statements are inherently untestable. (There may well be other reasons
why theological statements, in particular, are refractory to empirical
4
test; the present point is that, apart from other reasons, empirical testing
is made difficult because of this kind of ineradicable indeterminacy.)
These considerations have led logistically and positivistically minded
philosophers to become extremely impatient with these modes of dis¬
course. They have typically favored complete annihilation for theology
and replacement of our ordinary talk about feelings with “physicalistic”
talk about states of the nervous system. Without going this far, we can
at least point out that philosophers of religion and philosophers of mind
should give more attention to the semantic status of the quasi-meta-
phorical terms in their domains than they have done to date.6
6 For a detailed discussion along these lines of the status of theological terms,
see W. P. Alston, “The Elucidation of Religious Statements,” in The Hartshorne
Festschrift: Process and Divinity, ed. William L. Reese and Eugene Freeman (La
Salle, Illinois: Open Court Publishing Co., 1964), pp. 429-443.
For
further
reading
Because of limitations of space, I have omitted from this bibliography any

works referred to in the course of the book. Thus, the reader can skim the
footnotes for additional suggestions.
chapter 1
The more sophisticated versions of the reference theory of meaning are to

be found in the writings of logicians. These range from the nineteenth
century logicians, J. S. Mill, A System of Logic, Bk. I (London: Longmans,
Green & Company, Ltd., 1906) and Gottlob Frege, “On Sense and Refer¬
ence,” in Philosophical Writings, ed. Peter Geach and Max Black (Oxford:
Basil Blackwell, 1952), to the more contemporary treatments of Alonzo
Church in “The Need for Abstract Entities in Semantic Analysis,” Pro¬
ceedings of the American Academy of Arts and Sciences, 80 (1951) and
C. I. Lewis, “The Modes of Meaning,” in Leonard Linski, ed., Semantics
and the Philosophy of Language (Urbana, Illinois: University of Illinois
Press, 1952). Rudolf Carnap has pursued this kind of approach by con¬
structing elaborate logical systems in Introduction to Semantics (Cambridge:
Harvard University Press, 1942) and in Meaning and Necessity (Chicago:
University of Chicago Press, 1947). Modern refinements of the ideational
theory of meaning include C. L. Stevenson’s formulation in terms of dis¬
positions of linguistic expressions to produce psychological effects in hearers,
in Chapter 3 of Ethics and Language (New Haven, Conn.: Yale University
Press, 1944) and a formulation in terms of intentions of speakers to produce
psychological effects in hearers, in Unit 14 of Henry Leonard’s An Intro¬
duction to Principles of Right Reason (New York: Holt, Rinehart &
Winston, Inc., 1957) and in H. P. Grice’s “Meaning,” Philosophical Re¬
view, 66 (1957). An historically important source for the stimulus-response
theory is Chapters 3 and 9 of C. K. Ogden and I. A. Richards, The Meaning
of Meaning, 5th ed. (New York: Harcourt, Brace & World, Inc., 1938).
107
108 For Further Reading
Bloomfield s theory undergoes some refinement and elaboration in C. C.

Fries, “Meaning and Linguistic Analysis,” Language, 30 (1951). Paul Ziff,
Semantic Analysis (Ithaca, N. Y.: Cornell University Press, 1960) and
W. V. Quine, Word and Object (New York: John Wiley & Sons, Inc.,
1960) represent divergent but equally sophisticated developments of the
notion that the meaning of an expression is a function of the conditions
under which it is uttered. B. F. Skinner’s Verbal Behavior (New York:
Appleton-Century-Crofts, Inc., 1957), represents the most determined at¬
tempt to apply stimulus-response concepts to language; and even though he
explicitly eschews use of semantic terms, his pragramme has clear impli¬
cations for their analysis. Chapter 3 of Roger Brown’s Words and Things
(New York: Free Press of Glencoe, Inc., 1958) provides a useful review
of stimulus-response theories, and Chapter 7 of Max Black’s Language and
Philosophy (Ithaca, N. Y.: Cornell University Press, 1949) is a penetrating
discussion of Morris’ account of signs. '
CHAPTER 2
Paul Feyerabend’s “Wittgenstein’s Philosophical Investigations,” Philo¬

sophical Review, 64 (1955) presents one interpretation of Wittgenstein’s
philosophy of language. All of the essays collected in Ordinary Language
fd- V'i S'ChaPPe11 (Englewood Cliffs, N. Prentice-Hall, Inc., 1964)
throw hght on the sense in which contemporary philosophers are concerned
with the ordinary use of linguistic expressions. Good examples of standard
treatments of uses of language” can be found in Chapter 2 of I M CoDi’s
Introductwn to Logic, 2nd ed. (New York: The Macmillan' Company,
JSS1)’ .f.nd, in, ia™ Frankena, “Some Aspects of Language’’ and
Cognitive and Non-Cognitive, m Paul Henle, ed.. Language, Thought
and Culture (Ann Arbor, Michigan: University of Michigan Press 1958)'
^np5°Jve discussi0n °f the nature and varieties of definition is to be found
n Part IV of Henry Leonard s An Introduction to Principles of Rinht
Reason (New York: Holt, Rinehart & Winston, Inc., 1957) For stimulating
d,scuss,0„s Of problems connected with the notion if synonym” seeNelZ
oodman. On Likeness of Meaning” and Benson Mates, “Svnonvmy ”
both reprinted in Leonard Linsky, ed., Semantics and the Philosobhv of
Language (Urbana, Illinois: University of Illinois Press, 1952) ^
chapter 3
Chapters 4-6 of H H. Price, Thinking and Experience (London- Hutchin

sons University Library, 1953) are very good on the distinction between
signs and symbols. Interesting discussions of the varieties of svmbohsm fn
elude Suzanne Langer, Philosophy in a New Kev (New York-?? a
,gSxng Fountfn (Bloomington, Indiana: University of Indiana Press

1954). For useful discussions of the nature of lanrmape ri ?
dward Sapir Language (New York: Harcourt, Brace & World In? T92I)
Chapter 2 of J. B. Carroll, The Study of Language fCambrWM O’
setts. Harvard University Press, 1955), and Chapter l o7A
ductwn to Linguistic Structures (New York: Harcourt, Brace & Wolld,
For Further Reading 109
CHAPTER 4
The spirit of the traditional empiricism of Locke and Hume lives on in C. I.

Lewis and H. H. Price. For Lewis, see Book I of his Analysis of Knowledge
and Valuation (LaSalle, Illinois: Open Court Publishing Co., 1946). For
Price, see Thinking and Experience (London: Hutchinson’s University Li¬
brary, 1953). The Bible of logical atomism was Ludwig Wittgenstein’s
Tractatus Logico-Philosophicus (New York: Harcourt, Brace & World, Inc.,
1922). For an excellent critical review of the movement see J. O. Urmson,
Philosophical Analysis, Its Development Between the Two World Wars
(Oxford: The Clarendon Press, 1956). Many of the classic papers of the
logical positivist movement are collected in A. J. Ayer, ed., Logical Positiv¬
ism (New York: Free Press of Glencoe, Inc., 1959). For readily intelligible
expositions of the verifiability criterion plus many applications to philo¬
sophical problems see Rudolf Carnap, Philosophy and Logical Syntax (Lon¬
don: Psyche Miniatures, 1955) and A. J. Ayer, Language, Truth, and Logic,
2nd ed. (London: Victor Gollancz, Ltd., 1946). For recent restatements of
the verifiability criterion see C. G. Hempel, “The Concept of Cognitive
Significance: A Reconsideration,” Proceedings of the American Academy
of Arts and Sciences, 80 (1951) and, with special reference to the applica¬
tion of the criterion to scientific theories, Rudolf Carnap’s essays, “Testabil¬
ity and Meaning,” reprinted in Herbert Feigl and May Brodbeck, eds.,
Readings in the Philosophy of Science (New York: Appleton-Century-
Crofts, Inc., 1953) and “The Methodological Character of Theoretical
Concepts,” Minnesota Studies in the Philosophy of Science, Vol. I (Minne¬
apolis: University of Minnesota Press, 1956). Important criticisms of the
verifiability criterion are to be found in A. C. Ewing, “Meaninglessness,”
Mind, 46 (1937) and in G. J. Warnock, “Verification and the Use of Lan¬
guage,” Revue Internationale de Philosophie (1951).
chapter 5
Important treatments of vagueness can be found in Max Black, “Vagueness:

An Exercise in Logical Analysis,” in Language and Philosophy (Ithaca,
N. Y.: Cornell University Press, 1949) and in Chapter 4 of W. V. Quine,
Word and Object (New York: John Wiley & Sons, Inc., 1960). Stimulating
discussions of metaphor include Chapters 5 and 6 of I. A. Richards’ The
Philosophy of Rhetoric (London: Oxford University Press, 1936), Chapter
6 of Philip Wheelwright, The Burning Fountain (Bloomington, Indiana:
Indiana University Press, 1954), and Max Black, “Metaphor,” in Models
and Metaphors (Ithaca, N. Y.: Cornell University Press, 1962). Monroe
Beardsley’s “The Metaphorical Twist,” Philosophy and Phenomenological
Research, 22 (1962) contains an acute criticism of the view presented in
this book, as well as the development of an alternative position.
INDEX
A
“A priori” knowledge, 4 Denotation, 16ff, 17n
Analytical philosophy, 7 Dialogues; see Plato
Aristotle, 6, 14n
Metaphysics, 1-2
Assertions, 28, 48, 73, 74, 75, 77, 80
Association, 45, 48, 52, 63, 67-68
“Emotive force,” 45, 48
ideational, 52, 59
Emotivists, 49
Austin, John, 35
Empiricists, 63ff, 72
Epistemology, 4, 48, 63, 66
Ethics, 49, 63
B
Behavioral theory; see Linguistic mean¬
ing F
Bergson, H., 5 Falsifiability, 7Iff
Berkeley, Bishop, 63ff
Bloomfield, Leonard, 26, 32
vjr
Gardener parable, 80
€ General sign theorists, 51
General term, 1, 85, 86-87
Carnap, R., 6
Communication, 22-23, 25ff, 56-57, 97
Conceptual analysis, 8
Conjunctive function, 18
Connotation, 16ff, 17n Henle, Paul, 98, 98n
Convention, 56ff Hospers, J., 57n
Hume, David, 63ff
D I
Definition Ideational association; see Association
ostensive, 65, 66-67 Ideational theory; see Linguistic mean¬
true by, 4 ing
111
112 Index
Ideas, 23, 24-25,63-64,97 Multivocality, 84, 96

Illocutionary-act potential, 37ff, 73; see
also Linguistic action
Imperatives, 27-28
Indeterminacy; see Vagueness N
Inference, 3-4 Naturalists, 49
T
JLt O
“Law of Excluded Middle,” 96 Osgood, Charles, 28
Leibniz, G. W., 6 Ostensive teaching, 104
Linguistic action, 34ff
illocutionary, 35ff, 48-49, 74-75, 78,
79
locutionary, 35-36 P
perlocutionary, 35-36, 47ff Paradigm, 89
Linguistic meaning Peirce, C. S, 21, 51, 51n, 53, 55ff, 69n,
behavioral, 11-12, 19, 22, 25ff, 31 98
ideational, 11-12, 19-20, 22ff, 26, 52, Phenomenal statements; see Statements
64, 66n Plato, 1, 2
referential, 11, 12ff, 18, 19, 22 Dialogues, 6
See also Meaning Republic, 1
Locke, John, 5,22, 23, 25, 63ff Platonist, 80, 82
Logic, 3, 48, 70n Plotinus, 5
Logical atomism, 2, 69, 71
Logical positivists, 69ff, 102, 106
"D
JL lb
Referential theory; see Linguistic mean¬

ing
Mathematics, 4, 70n
Republic; see Plato
Meaning, lOff
Ruby, L., 57n
cognitive, 74-75, 102
Russell, Bertrand, 2, 6, 12n, 13, 14, 45
emotive, 74-75, 102
57n,69
expressive, 74-75, 102
factual, 74-75, 102
literal, 74-75, 102
metaphorical, 102
See also Linguistic meaning, Verifiabil¬
S
ity, Word meaning Schlick, Moritz, 69
Meaningfulness, 8, 12, 18, 31, 36, 63 Semantics, 8, 15, 16, 17, 26, 28, 33, 43,
65, 67-68, 72ff, 78, 79, 81ff, 95- 46, 57, 66, 67, 84, 85-86, 93,
96 96, 98, 106
Meinong, A., 2 Sentence, 10-11, 26-27, 32ff, 36-37, 43-
Mental image, 25 44, 50, 60, 82
Metaphor, 8, 85, 96, 97ff intelligible, 62-63
“dead,” 99, 101-102, 103 meaningful, 36ff, 73, 78
quasi-, 105ff meaningless, 62-63, 97
nonobservation, 82
Metaphorical-literal distinctions, lOlff
observation, 77
Metaphysics, metaphysical, 1, 2-3, 63
77, 80, 81, 82 See also Linguistic meaning, Meaning¬
Metaphysics, see Aristotle fulness, Statement
“Sign-functioning,” 50, 51-52, 59
Metonymy, 97
Signs, 28, 30, 50ff
Morris, Charles, 28ff, 53
icon, 55, 57, 58-59, 104, 105
Index 113
index, 55-56 Synonym, synonymy, 20, 22, 37, 39,

symbol, 55, 59-60 44ff
Simile, 99
Social environment, 45ff
Socrates, 6
Statement, 11, 13, 48, 55, 96
¥
classification of, 4 Vagueness, 84ff
hypothetical, 79, 81 Verifiability, 6, 70ff, 81-82, 102
objective, 76
observation, 77, 78
phenomenal, 76-77
subjective, 76
w
Waismann, Friedrich, 93ff
See also Assertion, Sentence Wisdom, John, 79, 80, 80n
Stimulus-response connection, 31 Wittgenstein, Ludwig, 2, 5, 6, 32n, 53n
Substances, 2 Word meaning, 21ff, 33, 36ff, 44, 45;
Symbols, symbolism, 2, 12, 51, 52, 98; see also Linguistic meaning,
see also Signs Vagueness
William P. Alston
William P. Alston, Professor of Philosophy at the

University of Michigan, has also held visiting
appointments at the University of California at
Los Angeles and at Harvard University.
He has edited two books of readings,
Readings in Twentieth Century Philosophy
(with George Nakhnikian) and Religious
Belief and Philosophical Thought.
it;
PRENTICE-HALL FOUNDATIONS OF PHILOSOPHY SERIES

Elizabeth and Monroe Beardsley, editors
Virgil Aldrich Philosophy of Art
William Alston Philosophy of Language
Stephen Barker Philosophy of Mathematics
Roderick Chisholm Theory of Knowledge
William Dray Philosophy of History
William Frankena Ethics
Carl Hempel Philosophy of Natural Science
John Hick Philosophy of Religion
Sidney Hook Political Philosophy
John Lenz Philosophy of Education
Richard Rudner Philosophy of Social Science
Wesley Salmon Logic
Richard Taylor Metaphysics

Philosophy of Language (Alston)

Uploaded by

Copyright:

Available Formats

Philosophy of Language (Alston)

Uploaded by

Document Information

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

Philosophy of Language (Alston)

Uploaded by

Copyright:

Available Formats

What is the book about based on its title and table of contents?

What is the book about based on its title and table of contents?

What subjects are covered in the Foundations of Philosophy series?

What subjects are covered in the Foundations of Philosophy series?

WILLIAM P.

UNOATIONS OF PHILOSOPHY SERIES

Virgil Aldrich PHILOSOPHY Of ART

William Alston PHILOSOPHY OF LANGUAGE

Sl«ph«» Barker PHILOSOPHY OF MATHEMATICS

Roderick Chisholm THEORY OF KNOWLEDGE

William Dray PHILOSOPHY OF HISTORY

William Pironkona ETHICS

Carl Hempel PHILOSOPHY OF NATURAL SCIENCE

John Hick PHILOSOPHY OF RELIGION

Sidney Hook POLITICAL PHILOSOPHY

John ten* PHILOSOPHY OF EDUCATION

Richard Rudner PHILOSOPHY Of SOCIAL SCIENCE

Wesley Salmon LOGIC

Richard Taylor METAPHYSICS

Elizabeth and Monroe Beardsley, editors

: * ' O - , 1 , . * > W ' * . ' i *s ' K t $ E i 1 E S

PHILOSOPHY OF LANGUAGE, Alston

FOUNDATIONS Of PHILOSOPHY SERIES

PRENTICE-HALL INTERNATIONAL, INC, London

PRENTICE-HALL OF AUSTRALIA, PTY., LTD., Sydney

PRENTICE-HALL OF CANADA, LTD., Toronto

PRENTICE-HALL OF INDIA (PRIVATE! LTD., New Delhi

PRENTICE-HALL OF JAPAN, INC., Tokyo

PRENTICE-HALL DE MEXICO, S.A., Mexico City

without whom not

Many of the problems of philosophy are of such broad relevance

ELIZABETH BEARDSLEY MONROE BEARDSLEY

Though the philosophy of language might reasonably be thought

I am extremely indebted to the editors of this series, Monroe and

Introduction, 1 Sources of the philosopher’s concern with language: metaphysics, 1.

Empiricist Criteria ef Meaningless sentences, 62. Traditional form of an empiricist criterion,

The philosophy of language is even less well-defined and less in

individuals partakes. If this were not the case, it would be impossible

. . . m a logically correct symbolism there vail always be a certain

Note that this identity of structure is postulated to hold not between

Logic Another branch of philosophy in which concern with language

b. 'Someone sells insurance in our town.’ There are other indications of

3. There is someone who sells insurance in our town.

It is clear from such examples that an important part of logic consists

Epistemology The branch of philosophy known as Epistemology or Theory of

into questions concerning what it is for a term to have a certain

clarity and explicitness with respect to the basic concepts in terms of

expression to have a certain meaning and of the conditions under which

types of expressions. But if these problems belong in principle to the

1. That is no mean accomplishment, (insignificant)

rathe/ftanTw”6 ab°Ut pe°ple’ actions’ events’ or situations

1 this is the canonical form. I shall

8. Keep off the grass. This means you.

Here it seems plausible to regard 'this’ as referring to the sentence,

9. Lucky Strike means fine tobacco.

Here we are not talking about a linguistic expression, although it may

That look on his face means trouble.

of The literature on this subject contains a bewildering diversity of

Press61903n)dpR47Se11’ FHncipleS of Mathem^s (London: Cambridge University

di*cult \° find a full-blown presentation of this version in the works of