PHILOSOPHY OF LANGUAGE

SYNTHESE LIBRARY

MONOGRAPHS ON EPISTEMOLOGY,

LOGIC, METHODOLOGY, PHILOSOPHY OF SCIENCE,

SOCIOLOGY OF SCIENCE AND OF KNOWLEDGE,

AND OF THE MATHEMATICAL METHODS OF

SOCIAL AND BEHAVIORAL SCIENCES

Managing Editor:

Ja a k k o H in tik k a , Academy o f Finland and Stanford University

Editors:

R o b e r t S. C o h e n , Boston University

D o n a l d D avidson, The Rockefeller University and Princeton University

G abriel N uchelmans, University o f Leyden

W esley C. S a l m o n , University o f Arizona

V O L U M E 71
 FR A N Z VON K U TSCH ERA

PHILOSOPHY OF
LANGUAGE

D. R E ID E L P U B L IS H IN G C O M P A N Y
DORDRECHT-HOLLAND / BOSTON-U.S. A.
 Library o f Congress Cataloging in Publication Data

Kutschera, Franz von.

Philosophy o f language.

(Synthese library; v. 71)

Translation of Sprachphilosophie.
Originally given as lectures at the Universitat Munchen
and the Universitat Regensburg.
Bibliography: p.
Includes indexes.
1. Languages—Philosophy. I. Title.
P106.K8514 401 75-12877
ISBN-13: 978-94-010-1822-7 e-ISBN-13: 978-94-010-1820-3
DOI: 10.1007/978-94-010-1820-3

SPRACHPHILOSOPHIE
Second edition published in 1975 by Wilhelm Fink Verlag, Munich
Translated from the German by Burnham Terrell

Published by D . Reidel Publishing Company,

P.O. Box 17, Dordrecht, Holland

Sold and distributed in the U.S.A., Canada and Mexico

by D . Reidel Publishing Company, Inc.
306 Dartmouth Street, Boston,
Mass. 02116, U.S.A.

All Rights Reserved

Copyright © 1975 by D . Reidel Publishing Company, Dordrecht, Holland
N o part o f this book may be reproduced in any form, by print, photoprint, microfilm,
or any other means, without written permission from the publisher
 TABLE OF CO N TEN TS

A U T H O R ’S PREFACE VII

IN TR O D U C TIO N 1

CHAPTER I / PRELIM INARY DISTINCTIONS 5

1. Language and Linguistic Utterances 5
2. Descriptive Statements 13
3. The Use and Mention of Signs 17

C H A P T E R II / T H E O R I E S O F M E A N I N G 19
1. Realistic Semantic Theories 19
2. Behavioristic Theories of Meaning 59
3. Quine’s Philosophy of Language 73
4. Wittgenstein’s Philosophy of Language in the Philosophical
Investigations 105

C H A P T E R III / T H E O R IE S OF G R A M M A R 165
1. Traditional Grammar 169
2. Logical Grammar 182
3. Generative Grammar 218

C H A P T E R IV / L A N G U A G E A N D R E A L I T Y 242
1. The Thesis of the Role Language Plays in Experience 242
2. The Role of Vocabulary 255
3. The Role of Grammar 261
4. The Epistemological Problematic of the Relativity Thesis 278

BIBLIO GRAPH Y 291

I N D E X OF S U B J E C T S 301
I N D E X OF L O G I C A L S Y M B O L S 305
 A U T H O R ’S P R E F A C E

This book has arisen out of lectures I gave in recent years at the Uni­
versities of Munich and Regensburg, and it is intended to serve as a
textbook for courses in the Philosophy of Language.
In my lectures I was able to presuppose that the students had taken an
introductory course in logic. Some knowledge of logic will also be helpful
in studying this book - as it is almost everywhere else in philosophy -,
especially in Section 3.2, but it is no prerequisite.
I would like to give my sincere thanks to Prof. Terrell for his excellent
translation of the book, which is based on the second, revised and en­
larged German edition.

Regensburg, May 1975 F ran z von K utschera

 IN TR O D U C TIO N

Language has become one of philosophy’s most important and pressing

themes during this century. This preoccupation with language has its ori­
gins in the most diverse areas of philosophical inquiry. It has come from
the theory of knowledge by way of a turn from the critique of reason to a
critique of language; from logic as a consequence of its concern with arti­
ficial languages and the logical analysis of natural languages; and from
anthropology by way of the emphasis on language as an accomplishment
essential to the definition of man and through the discovery of correlations
between linguistic form and man’s image of the world; from ethics be­
cause of its concern with the linguistic forms of ethical statements and the
demarcation of the boundary between them and descriptive sentences.
This philosophical turn in the direction of language can also be read
off in quite general terms from the change in the formulation of philosoph­
ical problems. Where a question used to be raised about ‘the nature of
causality’, for example, or ‘the content of the concept of causality’, now­
adays we feel it more appropriate to formulate the problem as a question
about the use of words, and to ask: “What is one saying when he says
that an event A brings about an event 2??” 1
During the course of this development philosophical attention has
sometimes concentrated so exclusively on language that it has simply been
identified with linguistic analysis. This is especially true of the two main
streams of analytical philosophy. It is true of the enterprise, beginning
with Bertrand Russell and continuing from Rudolf Carnap to Nelson
Goodman and Willard van Orman Quine, aimed at a logical analysis of
the language of science, especially of philosophical language, in which the
attempt is made to clarify and make precise the terms and propositions
of that language by using the instruments of modern logic. It is still
more true of the other principal direction of analytical philosophy,
originating in George Edward Moore, then stamped with the distinctive
mark of Ludwig Wi-genstein and developed further by the Ordinary-
Language-Philosophy of the Oxford School (Gilbert Ryle, John Lang-
2 I N T RO DUC TI ON

shaw Austin), in which the attempt is made to clarify philosophical lan­

guage by analyzing the ordinary usage of its terms.
In the Foreword to his Begrijfschrift, Gottlob Frege says: “If it is one
of philosophy’s tasks to liberate the human mind from the dominion of
the word, by revealing the illusions concerning conceptual relationships
engendered, often almost unavoidably, by linguistic usage, by freeing our
thought from the burden imposed on it by nothing but the structure of
our linguistic instruments of expression, then my Begrijfschrift (concep­
tual notation), further developed for these purposes, will be a useful tool
for philosophers.” 2
Frege’s ‘if’-sentence then turns into a categorical assertion and for
Russell himself the most important task for philosophy consists “in
criticizing and clarifying notions which are apt to be regarded as funda­
mental and accepted uncritically. As instances I might mention: mind,
matter, consciousness, knowledge, experience, causality, will, time.” 3
And Carnap writes in the Foreword to the 2nd Edition of his book, Der
logische Aufbau der Welt: “The new definitions [of the concepts] should
fit into a systematic structure of concepts. Such a clarification of concepts,
nowadays frequently called ‘explication’, still seems to me one of the
most important tasks of philosophy.” 4
For Wittgenstein even as early as the Tractatus, all philosophy is
language analysis5 and in the Philosophical Investigations philosophy is
nothing but a constant struggle against “the bewitchment of our intelli­
gence by means of language.” 6 Finally, for many adherents of Ordinary-
Language-Philosophy, the turn taken by philosophy to language, the
‘linguistic turn’, signifies the great revolution of modern philosophy.
Thus Austin says: “ ...it cannot be doubted that they [these methods of
linguistic analysis] are producing a revolution in philosophy. If anyone
wishes to call it the greatest and most salutary in its history, this is not, if
you come to think of it, a large claim.” 7
In view of the large number of enterprises and directions that are
grouped together under the heading of a philosophy o f language, it can be
no wonder that the title has no precise and well-defined content. We will
therefore sketch briefly the themes with which this work is supposed to
be concerned and the objectives that guide us in it.
Speaking quite generally to begin with, we will be investigating in what
follows the functional role of language, and the results accomplished by it.
 I N T RO DUC TI ON 3

We will pick out three problem areas from the wealth of questions which
can be put in this connection, areas that seem to us to be of a special and
fundamental significance: the problem of the grammatical articulation and
synthesis of linguistic expressions, the problem of their meaning and the
problem of the role of language in experience.
After some preliminary remarks we shall be concerned in Chapter II
with the question of what it is that makes up the meaning of linguistic
expressions, how it is established and how communication in language
functions. There then follows, in Chapter III, a discussion of the funda­
mental problem of grammar, how the combination in language of meaning­
ful expressions into new units of meaning functions. Finally, in the IVth
Chapter, the relationship between language and experience will be discus­
sed, the question as to the extent in which the forms of language express
particular schemata of interpretation for experience, to what extent the
forms of language leave their mark on the way in which reality is expe­
rienced.
These three problem areas are also basic problems of the linguistic
sciences, particularly of general linguistics as the fundamental linguistic
discipline. Consequently we find a great deal of overlap between philo­
sophical and scientific linguistic investigation. No sharp boundary can
be drawn between the two. For that reason, the philosophy of language
must inform itself of the results in linguistics that are relevant for its
questions and must take account of them, and the opportunity arises
to lend support to theses in the philosophy of language with arguments
drawn from linguistics. Philosophy is not operating in a realm independ­
ent of experience, then.
The reasons that lead us to interest ourselves here in the question con­
cerning the function and the accomplishments of language are above all
the following: Language is the most elementary, the most important
instrument for all of the sciences, one that we use constantly. If we assert
something (describe, classify, formulate hypotheses or theories), argue,
lay down rules (define, prescribe forms of measurement, etc.) we make
use of language.
Since language is such an important instrument of science, a basic
knowledge of this instrument is itself very useful for work in a particular
science. Natural languages, which we apply for the most part even in
the sciences, are not made primarily for scientific purposes and for their
4 I NT RO D UC T IO N

scientific use they must be supplemented and made precise in a great many
ways. The scientist, therefore, is not only a user of language but also to a
certain extent a maker. In the latter role, at least, he can not just commit
himself with naive trust to the conventions of language, but he requires
insights into the nature of linguistic means of expression. In this sense, the
function and effect of language are one of the themes of a propaedeutics o f
science.
The question of the effect of language on the formation and organiza­
tion of experience is especially of paramount interest to the theory o f
knowledge. Does speech consist only in the expression of contents of
thought or perception that are independent of language, changed in no
respect by the manner of their expression, or are these contents always
linguistically determined, so that the forms of our language are the forms
of our experience? But the problem of meaning, which, as we shall see,
is closely connected with this question, also has relevance to epistemology.
These two questions, from propaedeutics and from the theory of know­
ledge, guide the ensuing discussions of language and determine the philo­
sophical orientation of these investigations.

NOTES

1 P. Alston gives this example in [67], p. 388. -T h e numbers in square brackets identify
the author’s work listed in the index according the year in which it appeared.
2 Frege [79], p. Vlf.
3 Russell [56], p. 341
4 Carnap [28], p. X.
s Wittgenstein [22], 4.0031.
6 Wittgenstein [53], 109
7 Austin [62], p. 3f. - For analytical philosophy see the presentations by Savigny [70]
and Lorenz [70], Chapter I, for example. For Ordinary-Language-Philosophy spe­
cifically, see Savigny [69].
 CHAPTER I

P R E L IM IN A R Y DISTIN CTIO N S

1. L a n g u a g e and linguistic utterances

Before we take up the main themes of this book - grammar, meaning

and the role of language in experience - it is the intent of this chapter
first to introduce some preliminary distinctions. When we refer to ‘lan­
guage’ in what follows, this is to be understood in the sense of a general
expression covering all languages or all the languages taken into account
in that particular context. We understand what a language is by example
to begin with: In the discussions that follow we are referring for the most
part to natural languages, i.e. languages that have had a historical devel­
opment, such as German, English, Greek, Japanese, etc. Within these
languages we can distinguish both various stages of development and
various levels, such as the standard language, the level that has the status
of norm and ideal, and is used in literature, the schools, radio and press,
etc., and ordinary or everyday language, in which the norm represented by
the standard language is relaxed and sentence structure, choice of words
and expression handled more freely. Besides these we can further distin­
guish the dialect forms of the language and finally there are still further
countless special forms, which depend, for example, on their users’ social
status and often are differentiated only in small details.1 All of these
developmental forms, linguistic levels and variations must be specified
when assertions about a natural language are made, if they are to be exact.
We make a distinction between natural languages and artificial lan­
guages, which play a part in logic and mathematics especially and are defined
by explicit conventions. The most important examples of such artificial
languages are the languages of logical symbolism.
In a broader sense of the word ‘language’, a sense we will not be using
in what follows, one can also speak of a gesture language, a signal language
(consisting of particular flag or light or sound signals) or an animal
language, and understand a language generally speaking as a system of
signs.
6 CHAPTER I

We have before us a linguistic utterance when X says to Y, ‘Yesterday

I went to the movies’, when Y asks X , ‘How was the film?’, when X greets
Y with ‘Good morning’, when Y makes a request of X ‘Lend me your
umbrella!’, and so on. Such a linguistic utterance is an act of the speaker’s,
with which (as a rule) he addresses himself to a hearer. For this reason
utterances are also called speech acts or items of linguistic behavior. It is
also possible for a group of persons to play the part of speaker (a govern­
ment makes a declaration, for example, physicians publish a Bulletin,
etc.), and the hearer too is not always a single subject, but can be a more
or less strictly delimited group (a lecturer speaks to his auditors as a
group, a manifesto is addressed to the public, etc.).
We differentiate between the linguistic utterance as an act and the
product of that act, between the act of speaking and what is spoken. The
product of a linguistic utterance is also often referred to as a linguistic
utterance. Since in the following we shall be speaking of the products
of utterances above all, we will adhere to that usage. Wherever a distinc­
tion is necessary, however, we will speak of the utterance as performance
and the utterance as product.
A further distinction must be drawn between forms of behavior and
their particular realizations: between a particular waltz step (carried
out by a particular person at a particular time) and the form of the waltz
step, between the typical form of a curtsy and a particular curtsy, etc. In
the same way, we must also distinguish between forms of linguistic be­
havior and their concrete individual realizations, the utterance-perfor­
mances, distinguish the form of greeting ‘Good morning’, for example,
from 7 ’s greeting X with ‘Good morning’ on a particular occasion.
We also make the corresponding distinction within the realm of what
is produced [by an act] and so, for example, we make a distinction between
a particular occurrence of a written word and the graphical word-type to
which it belongs. Thus the word-type ‘and’ occurs in several places on this
page, i.e. there are several occurrences on this page of the word-type
‘and’. We shall also use type and occurrence to express the distinction in
question for utterances as performances. Type and occurrence are related
as class and member: an utterance as an act or as a result or product [of
an act] is always a member of a class of acts or products, which represents
the utterance’s type, its form.
We call the forms of utterances [linguistic] expressions. An expression,
 PRELIMINARY DISTINCTI ONS 7

then, is a form of utterance, whether taken as act or as product. Wherever

the distinction is important, we shall speak more exactly of expression as
performance and expression as product.
Now the distinctive feature of linguistic utterances lies in the fact that
by means of them we make ourselves understood to one another, commu­
nicate something, ask for something, etc., that they are, putting it gener­
ally, meaningful. Speech acts consequently have two aspects: the phonetic
or (graphic) aspect, that a sequence of spoken (or written) sounds is
produced, the sounds G-o-o-d-M-o-r-n-i-n-g are uttered. But at the same
time an act that goes beyond the production of sounds is consummated:
the act of greeting the person so addressed; that is the semantic aspect.
Every utterance, then, has a phonetic and a semantic aspect. These two
aspects are not included in two separate acts, but are two views of the
same act. The phonetic aspect disregards the utterance’s meaning, its
communicative function, and takes account of the distinctive features of
the sound pattern only. The semantic aspect abstracts from these phonetic
particulars and fixes exclusive attention on the utterance’s meaning. From
this point of view, ‘meaning’ is in the first instance nothing but the func­
tion of an utterance in the communicative process.
Just as phonetic and semantic aspects are differentiated for utterances as
performances, so are they for utterances as products, and this distinction
carries over in an analogous manner to statements. In what follows we
shall be interested in utterances and expressions as products only insofar
as they are products of the phonetic act parts, i.e. as sequences of sounds or
forms of sequences of sounds, or series of signs or patterns of series of
signs, and this is the way in which we shall understand statements about
phonetic or graphic objects. Now linguistic utterances do not have mean­
ing in and of themselves, just by virtue of their sound patterns, for
example (otherwise we could understand even foreign languages without
having to learn what expressions in them mean). They only take on
meaning in the context of a system of rules which say how they are applied,
singly or in combination, for specific purposes in specific situations.
Linguistic utterances are, then, acts governed by convention and for their
effectiveness they require rules of usage that are agreed upon. In the case
of natural languages, these rules are not explicitly formulated and the
conventions are not formal agreements. Rules and conventions are
customs and norms, rather, understood and followed intuitively; their
8 CHAPTER I

explicit formulation and definition is the responsibility of linguistic science

alone and something that is accomplished in grammars and dictionaries
that codify linguistic usage.
Linguistic conventions are general rules, and so they refer not to partic­
ular utterances, but to [linguistic] expressions. For example, a conven­
tion of this sort does not say that on the morning of January 1, 1973, in
Munich, under such and such particular circumstances, Hans can greet
Fritz by saying ‘Good Morning’. What it provides, instead, is that anyone
can greet anyone else in the morning by saying ‘Good Morning’. It is on
the basis of this general convention, then, that a specific utterance of the
expression ‘Good Morning’ acquires the function of a greeting. The
meaning of the utterance is consequently defined by way of the meaning
of the expression uttered - and by way of the pragmatic circumstances of
the utterance (who the speaker is, to whom it is addressed, when and
where the assertion is made, etc.) The meaning of an utterance like ‘I saw
you yesterday’, in which so-called indexical expressions (‘I’, ‘you’, ‘yester­
day’, ‘here’, etc.) occur, obviously depends very definitely on its pragmat­
ic circumstances. It is true or false depending on who is speaking to
whom and when and where he is speaking. The meaning of the expression,
‘I saw you here yesterday’, as an assertion-type, is consequently so tho­
roughly indeterminate that we can not attach any truth-value to it at all,
even if the relevant factual information is available to us.
Linguistic conventions have to do with expressions, then. They prima­
rily have to do with forms of behavior (e.g. the function of saying ‘Good
Morning’). But they can also be formulated as rules for expressions as
products, as rules which say how phonetic or graphic signs or sequences
of signs can be used. Although the question as to whether meaning per­
tains primarily to statements as forms of behavior or as objects is of great
interest in principle - we shall go into it in Chapter II.4.5 - it is never­
theless possible, without prejudice to that question, to take over the usual
manner of expression according to which the semantic conventions
govern the use of expressions as objects.
A linguistic expression, as bearer of a meaning, is not a mere object,
e.g. a sound sequence, but an object that is used according to definite rules.
And we can not understand a language to be a set of expressions as
objects but only as a set of expressions together with the rules for using
them.
 PRELIMINARY DISTINCTI ONS 9

Furthermore, a language in the narrower sense of the word is not only

a collection of individual expressions in isolation, but a system for the
construction of in principle infinitely many expressions (e.g. sentences)
from a finite number of basic expressions (e.g. words), so that the rules of
the language include not only rules of use for individual expressions, but
also rules for the combination of expressions and the use of such com­
pound expressions, according to which their meanings are defined in
terms of the meanings of their component expressions.
Ferdinand de Saussure, one of the founders of modern linguistic science,
illustrated the interpretation of a system of rules by comparing a language
with the game of chess.2 Chess is a system of rules concerning how the
pieces are to be set up and moved on the chess board and of rules con­
cerning the capture of pieces, winning and losing. Just as in this case it is
not just a matter of the pieces only, but of the rules governing their use,
so that it would be false to say that chess is defined by the set of pieces;
in the same way, a language has to do with the rules and not just the
expressions in the language. And as the function of a piece can not be
characterized all by itself, but only with reference to the functions of the
other pieces and accordingly with the entire game, so the functions of
linguistic expressions can be defined only with reference to the entire
language.
We ordinarily say that the sequence of sounds /tsu:k /Jpitso/ and the
sequence of written signs Z-u-g-s-p-i-t-z-e both represent the same word.
But how is that to be understood, when the two expressions, as types as
well as occurrences are entirely different?
When, following Lyons in [69], p. 56ff., we speak of different realiza­
tions of a word, e.g. phonetic and graphic, the following interpretation is
obvious: The realizations of words in spoken language are primary, not
only historically (because there was a spoken language before there was a
written language), but also to the extent that we would scarcely regard a
written language without phonetic interpretation as a language at all. The
written realization of linguistic expressions, however, appears to us to be a
Commodum which is of course very important in practice but expendable in
principle. (For the most part, the language we use is the spoken language.)3
To that extent we could identify the phonetic realization of the word with
the word itself and say that other realizations, e.g. graphic, represent the
word only to the extent that a phonetic interpretation for these realiza­
10 CHAPTER I

tions is stated, according to which they represent the word symbolically.

We will again follow de Saussure here,4 however, and adopt a more
abstract standpoint: it takes the various realizations of a word to be in
principle equally correct - even if they do have their own peculiar advan­
tages now and then5 - and the word itself to be an abstract object, which
we can identify with the class of its realizations. We have now seen that
words are not isolated objects, but function as meaningful expressions
only within the confines of the system of rules of a language. Consequent­
ly it is better to speak of various realizations of a language. A language,
then, is a system of rules for the use of certain abstract objects (e.g.
words) specified in the first instance only with respect to number and
variety. A realization of the language arises out of this system by virtue of
the identification of these abstract expressions with concrete objects
(sound sequences, sequences of graphic signs, etc.). Conversely, proceed­
ing from realizations of language, i.e. from systems <A , where A is a
set of concrete objects and 3t a set of rules for their use, one can define
an equivalence relation between two systems <A l9 and <A 2,
which holds between them when there is a mapping (j>of A ± on A 29 so that
what the rules from 31^ determine for objects a from A x corresponds to
what the rules from 3t2 determine for the expressions <j>a from A 2. A
language is then a class of such equivalent systems.6
This abstract interpretation of languages can also be illustrated by
comparison with the game of chess: What materials the pieces are made
of and how they look is not an essential feature of chess; what is impor­
tant is only the number of the pieces and the differences among them. In
the same way, how the expressions of a language are concretely made up,
what material they consist of (sounds or graphic signs) and what form
they have, is not a matter of importance for the language; what is impor­
tant is only the differences between them that are relevant with respect to
the rules. De Saussure also speaks in this connection of the abstract form
of a language and the substance of its realizations.
It emerges from these abstract ways of looking at languages that in the
following discussions we can leave the questions that relate to realizations
of language out of account. In particular, we shall not concern ourselves
with the theory of phonetic realizations, phonology.

The student of linguistic science sees his task to consist in describing

 PRELIMINARY DI STINCTI ONS 11

language as he finds it, in understanding the actual norms of linguistic

correctness, not in setting up new norms. His statements are thus intended
to be descriptive, not normative.
When the linguist sets forth a German grammar and explains the
meanings of German words in dictionaries, however, he is not simply
describing; he is also making them more precise; explicit grammatical
rules are more precise than the intuitive everyday customs, the demarca­
tion and explication of meanings and their differences define these
meanings more exactly. The work of the linguist can be linguistically
creative to a certain extent, if the scientific description of the language
reacts upon it and sets new linguistic norms. Standard German, for
example, has certainly been influenced as a norm by German grammars
and dictionaries. But this effect of feedback and normalization is rather
slight: Certainly it is not a matter of fixing new meanings but of explica­
tions that make things more precise. Besides, the linguist takes account of
a certain range of variation in grammatical rules for actual linguistic
usage; unlike the logician, he is not aiming at a system of rules that
governs every case unequivocally. Consequently, one can make the claim
that the statements of linguistic science are actually descriptive.
Now how can linguistic regularities be established? The following
problem arises here: The empirical material the linguist has to start with
consists of the concrete utterances of those who belong to the linguistic
community with which he is concerned. These linguistic utterances -
language as parole or as performance - as F. de Saussure or N. Chomsky
call it - or their products are not what the linguist is primarily interested
in, but rather language as a system of rules, as langue or competence.1
Now this linguistic norm must be inductively inferred from the available
utterances. That can not be carried through directly, however, for there
are many incorrect expressions included among the linguistic utterances,
arising from a speaker’s accidental or characteristic mistakes. But before
one knows the rules, one can not eliminate these incorrect expressions and
thus one can not use the actually available utterances directly in testing
one’s hypotheses about linguistic norms.
The linguist’s situation, however, is fundamentally in no way worse
than, for example, the physicist’s. In physics, too, one can put regularities
to observational test only if the results of measurement are not influenced
by disturbing factors. In this case, too, however, one can only say when
12 CHAPTER I

a disturbing factor is present if one already has knowledge of the regular­

ities. But there is no vicious circle here. Instead, by varying the conditions
of observation systematically one can discover the regularities that
remain constant, as against these variations, and the domain in which they
hold true.8
NOTES

1 See Grebe [661, P- 25ff.

2 See de Saussure [16], Introduction, Chap. V ; Part 1, Chap. Ill, §4; Part 2, Chap. III. -
The parallel between language and game is also fundamental to Wittgenstein’s philos­
ophy of language. See Chapter II. 4.2.
3 Of course note should be taken o f the fact that written and spoken language are, as is
the case with Japanese because of historical reasons, for example, different languages,
which only partially correspond, so that reproducing a written text orally often has the
character of a translation. On this point see note 6 also.
4 The statements on this point in [16] are not unambiguous, to be sure. Thus in the
Introduction, Chapter VI, § 2, de Saussure defends the interpretation of the relation
between phonetic and graphic realization sketched above. - On what follows see also
L. Hjelmslev’s expositions in [43], and Francis Bacon as well: “For the organ of tradi­
tion, it is either Speech or Writing: for Aristotle saith well, Words are the images o f
cogitations, and letters are the images o f words; but yet it is not of necessity that cogita­
tions are expressed by the medium of words. For whatsoever is capable o f sufficient
differences, and those perceptible by the sense, it is in nature competent to express cogita­
tions. And therefore we see in the commerce of barbarous people that understand not
one another’s language, and in the practice of divers that are dumb and deaf, that
mens’ minds are expressed in gestures, though not exactly, yet to serve the turn.”
(Proficience and Advancement o f Learning Divine and Human (1605), Bacon [57],
V. Ill, p. 399). And almost word for word to the same point: “Hoc igitur plane
statuendum est: quidquid scindi possit in differentias satis numerosas ad notationum
varietatem explicandam (modo differentiae illae sensui perceptibilis sint) fieri posse
vehiculum cogitationum de homine in hominem. Nam videmus nationes linguis
discrepantes commercia non male per gestus exercere.” {De dignitate et augmentis
scientiarum, Liber sextus, Caput I, in Bacon [57], V. I, p. 651). Brekle also cites these
passages in his introduction to Cordemoy [77], p. XXIX.
5 See Lyons [69], p. 62f.
6 This account involves a certain idealisation. Actually the correspondence between the
different realizations is not always one to one: The same graphic expression can be
phonetically ambiguous (e.g. ‘read’ in English - /red/or/ri :d/) and the same phonetic
expression can be graphically ambiguous (e.g. /k u :/ in French - ‘cou’ (neck) or ‘coup’
(blow)). - Lyons points out in [69], p. 39ff. that to the extent in which spoken and writ­
ten language differ not only in unessential particular points, but differ in their gram­
matical distinctions, for example, as well, they can take on the character of different
languages. According to Lyons, this divergence is explained by the diverse sorts o f
situations and purposes in and for which we make use of spoken and written languages.
7 On the distinction between performance and competence see Chomsky [65], p. 10, 25,
139, 187.
8 See also Chomsky [65], p. 3f.
 PRELIMINARY DISTINCTIONS 13

2. D escriptive statements

Linguistic utterances occur in the context of a great variety of situations

and kinds of behavior and serve the most diverse ends. Language has an
abundance of different functions, then. We are using language when we
ask, command, affirm, lecture, greet one another, prove, appraise, tell lies,
tell jokes, pleas or pray. Depending on the way in which language is
being used, on the type of discourse, as we choose to put it, the mode in
which linguistic expressions are meaningful will change, too, for ‘to be
meaningful’ means nothing other than ‘to have a specific function in a
context of linguistic usage’.
Within the science of linguistics it is not possible for us to give any of
these types of discourse or any mode of meaning an absolute status and
make it an exclusive object of consideration, if we do not wish to impose
very sharp limitations on the range of investigation from the very be­
ginning. But since our primary interest here, from the aspect of a propae­
deutics for science, is in the scientific type of discourse, in the way in which
language is used preeminently in the sciences, we shall be occupied
throughout long stretches of what follows with this type of discourse
and we shall not go into other types in greater detail except in Chapter
II.4.5.
What is important for the scientific use of language is in the first place
linguistic utterances of the type of declarative or assertive sentences. Sen­
tences of this type are used to make reports, to give information, to
describe something, to establish that something is the case. With such sen­
tences, we assert that the facts about something are thus and so. They
speak about something and they have, therefore, as we say, a descriptive
content or a descriptive meaning.
The following analysis will thus concentrate very heavily on declar­
ative sentences. The more detailed studies in scientific linguistics are
available for them, while the analysis of questions, commands, etc.,
has received the necessary attention only very recently. A certain (ex
post facto) legitimization of this focus of our interest will emerge in
Chapter 11.4.5-
In order to explain the distinctive character of declarative discourse and
descriptive meaning somewhat more precisely still, and to emphasize
from the outset that it is only one among a great many other ways in which
14 CHAPTER I

language is used, we will make a brief reference to a distinction drawn by

K. Biihler. Within the confines of these preliminary remarks, this reference
has a provisional character only; we shall return to it for more exact
consideration in Chapter II.4.5.
K. Biihler differentiates three components in connection with every
utterance: speaker, referent (what the utterance refers to) and hearer. He
coordinates a component of the utterance’s meaning to each of these
components: to the speaker, the expressive component or Kundgabe, to
the referent the descriptive component, and to the hearer the evocative
component or Appell. By Kundgabe is meant what the speaker expresses of
his own feelings, sensations, inclinations, of attitudes or goals in his utter-
rance. Thus, for example, his disgust when he says ‘Pfui’, his approval in
‘Bravo’, his intent in the sentence, T am going to Rome tomorrow’.
Kundgabe often involves the use of particular emotive or value-toned
words, as in the use of ‘nigger’ instead of ‘Negro’ or of ‘helper’ instead of
‘assistant’. Kundgabe does not mean, however, what conclusions can be
drawn about the speaker on the basis of his utterance - as, say, his excite­
ment can be inferred from his rapid, incoherent speech. Kundgabe is conse­
quently not what the utterance is a symptom or sign of, nor is it what the
speaker explicitly says about himself in his utterance, his feelings, goals,
etc., as in ‘I (personally) find that marvelous’ or T am planning on taking
a student tour to Africa’.
The Appell is supposed to be those components of an utterance’s mean­
ing by virtue of which it aims at a reaction of the hearer’s, seeks to call it
forth. This Appell becomes particularly clear in commands, which call
upon the hearer to do something, and in questions, which call upon him
for an answer. But an Appell lies also in the evocative meaning of value-
statements such as ‘That is good’ in the sense of calling on the hearer to
act accordingly or ‘That is bad’ in the sense of calling on him to refrain.1
And it may be a matter of the suggestive meaning of such statements as
‘Isnt’ that dreadful!’ or ‘It is quite obvious that he did that only to advance
himself’. Appell, however, does not consist in the actual or intended
effect of an expression on the hearer.2
These components of meaning enter into the various types of discourse
with varying strength. We will not, however, advance the opinion that the
various types of discourse can be adequately characterized solely in terms
of the relative strength with which these three components of meaning
 PRELIMINARY DISTINCTIONS 15

occur. That would be much too simple a picture of the functions of lan­
guage. Generally speaking, one will be able to characterize the various
types of discourse only by way of their functions in the total context of
action in which they are used. The modes of meaning will also be just as
diverse as these pragmatic contexts of language use. We shall go into this
more closely in Chapter II. Here it is sufficient that we can define descrip­
tive meaning more exactly by way of a contrast with evocative and ex­
pressive meaning.
Even in the sentences that are characterized as declarative sentences in
grammar (and contrasted with interrogative and imperative sentences),
expressive and evocative components have a role. In the sentence, ‘That
was an extraordinarily good performance’, the expressive component is
dominant. The speaker’s evaluation of the performance is very much in
the foreground, while nothing is said about the performance’s objective
characteristics, if no factual criteria of evaluation have been involved. The
sentence, ‘Presumably Fritz is sick’, includes, by virtue of the word
‘presumably’, an expressive component which expresses the speaker’s
attitude toward the sentence’s validity. And the sentence ‘You can’t
do th a t!’ contains a strong evocative component along with its expressive
component.
The descriptive component, on the other hand, is dominant in declara­
tive sentences as used in the sciences. The expressive component resides
only in the assertive character of these sentences and the evocative com­
ponent is almost entirely withdrawn.
We can cite as one last definition of descriptive statements that only
such statements are true or false, for they are the only ones to maintain
that something is the case and can thus be confronted with facts. To be
sure, the truth value of a declarative sentence, i.e. its truth or falsity, can
depend on the context of its utterance, if it includes index expressions the
reference of which is defined only in terms of the context of their utterance
and therefore can refer to different persons, things, places and points in
different contexts. Only when these index expressions are replaced by
expressions with meanings that do not depend on context will the state­
ment as such, independently of the circumstances of its utterance, be true
or false.
NOTES

1 The emotive and evocative components in ethical statements and value statements
16 CHAPTER I

have been elaborated by Stevenson in [44], e.g. His emotive theory of these statements
is based precisely on the denial of a descriptive content to them.
2 K. Buhler in [34], p. 28ff. distinguishes expression [Kundgabe] (more in the sense of
a sign of the speaker’s feelings, than in the sense of anything informational), representa­
tion [Darstellung] and Appell. Accordingly he speaks of linguistic expressions as
symptoms, symbols and signals.
 PRELIMINARY DISTINCTIONS 17

3. T h e u s e a n d m e n t i o n of s i g n s

The distinctive characteristic of signs, that of meaning something, re­

quires, especially when we are speaking about language, a precise distinc­
tion between sign and significatum, between the linguistic expression as a
sequence of sounds or written signs and its meaning. Simple as it is to
distinguish the name ‘Felix’ from the man, Felix, it designates; the ad­
jective ‘red’ from the property of being red; or the sentence ‘Felix has red
hair’ from the fact that Felix has red hair1, experience shows that a precise
distinction comes very hard when expressions are being spoken of.
When I say, ‘Felix has red hair’, I am using the word ‘Felix’ to say
something about Felix, the man. That sentence says nothing about the
word ‘Felix’. But when I speak of the sentence and say, for example,
that the word ‘Felix’ occurs in it, I am not using the word ‘Felix’. Instead,
I am mentioning it, referring to it. Likewise, in the sentence ‘Munich has
more than a million inhabitants’ I am using the word ‘Munich’ to say
something about the city of Munich; while in the sentence “ Munich’
has two syllables’ I am using the word ‘Munich’ to say something about
the word.2
A clear graphic distinction between the use and the mention of a word
is indicated. The usual convention is to place expressions that are being
mentioned, not used, within quotation marks, as we have already done
above. The expression together with these quotation marks then consti­
tutes a new expression, which we use in order to speak of the original
expression. According to this convention, then, an expression never
stands for itself: ‘Felix’ never stands for ‘Felix’, only for Felix. What
stands for ‘Felix’, on the other hand, is “ Felix” and once again,
‘“ Felix” ’ for “Felix”, and so on. Consequently we have to write
“ Munich’ has two syllables’ instead o f ‘Munich has two syllables’. The
expression ‘Munich’ does not occur in this sentence, then (we are not
speaking of Munich); it is rather the expression “Munich” that occurs,
since we are speaking o f‘Munich’.3
Related to the distinction between use and mention, there is also the
distinction between the language spoken of, as the object language, and
the language being used to speak of it, as the metalanguage. So when we
are speaking, in German, about English grammar, English is the object
language and German the metalanguage. This distinction always refers to
18 CHAPTER I

a relation involving the application of one language to another; one can

also speak in English about the German language, in which case English
is the metalanguage and German the object language. Further, object and
metalanguage can coincide, as in speaking in German about German
grammar. For systematic reasons, however, it is always to be recommend­
ed that we distinguish between object and metalanguage, and so separate
even in the latter case the metalinguistic use of German from its consider­
ation as object language. This is all the more important in that neglect of
this distinction can lead to the so-called semantic antinomies.4
To conclude these preliminary remarks, let us mention the important
distinction between syntax, semantics and pragmatics, first fixed as to
terminology by Charles Morris in [46]. It is oriented toward the distinc­
tion between the linguistic expression, its meaning and the context in which
it is used and it has to do with scientific studies about language. If an
inquiry is concerned only with linguistic expressions, their spoken or
written form, in abstraction from their meaning or context of use, it is
identified as syntactic. If the meaning of the expression is also being
considered, one speaks of a semantic inquiry and, finally, if the practical
context of its use is taken into account, one speaks of a pragmatic inquiry.5
Where the term ‘semantics’ is used in this sense, the ancient Stoic
designation ‘semiotic’ is often used for a theory that embraces all of these
aspects of linguistic expressions, or more broadly still: for a general
theory of signs.6
NOTES

1 These examples are not intended to anticipate any theory o f meaning, along the lines,
say, of generally presupposing that predicates refer to properties. The examples are
only supposed to illustrate the distinctions that are necessary in any study o f languages.
2 The terminology ‘use’ - ‘mention’ was introduced by Quine; see Quine [51b], 4. - The
distinction itself is quite old and was given extensive treatment in the scholastic theory
o f supposition, for example.
3 The usual procedure in literary quotation, which is also adopted in the text, is to
use a mixture o f single and double quotes. These may be understood as merely typo­
graphical variants of the same quotation device.
4 See Tarski [35] and [44], as well as the presentation in Kutschera [64].
5 According to some of the theories of meaning reported in Chapter 3, the meaning of a
linguistic expression is actually nothing other than its use, so that semantics can not
meaningfully be separated from pragmatics.
6 The label ‘semiotic’ is used in this sense by Locke, Peirce and Morris, for example.
 CHAPTER II

T H E O R I E S OF M E A N I N G

1. R e a l i s t i c se m an tic theories

In this chapter we will turn our attention to this work’s first main theme,
the question of the meaning of linguistic expressions. This question directs
itself in the first instance to all of the ways in which language is used, to
all types of discourse, such as assertions, questions, commands, etc. The
principal interest of the philosophy of language, particularly under its
aspect of scientific propaedeutics, is concentrated, however, as has already
been said in 1.2, on declarative discourse as the most important way in
which language is used in science. In this sense, we intend to concentrate
our attention in what follows preeminently on the investigation of prob­
lems of meaning connected with descriptive discourse.
When we speak o f ‘sentences’, what is meant by that is for the most part
descriptive sentences, and what we understand by ‘meaning’ is for the
most part descriptive meaning in the sense of 1.2.1
Since we intend to go into composition of meaning separately in
Chapter III, we shall be less concerned in this chapter with the question
as to how the meaning of compound expressions arises from the meaning
of their parts than with the question of how the meanings of simple words
or of simple linguistic structures can be characterized.
Now there are distinctly different theories about linguistic expressions
in connection with descriptive discourse. In the following, we will present
and discuss some especially important types of such theories.2
The first and the oldest group of theories of meaning that we will con­
sider in this chapter is constituted by the copy theories or realistic seman­
tic theories. The explanation of the latter designation is that these theories
presuppose a realism of concepts, as will become clear in what follows.
They interpret meaning as a conventional relationship between signs and
concrete or conceptual entities, which exist independently of the linguistic
signs. On this interpretation the meaning of a linguistic expression does
not depend on its use in concrete situations, but the use is determined by
20 CHAPTER II

the meaning, so that a sharp separation between semantics and prag­

matics is possible.

1.1. Naturalism and Conventionalism in Realistic Semantics

The oldest treatise on language that has survived to our times is Plato’s
‘Cratylus’ (ca. 388 B.C.). Plato concerns himself in it with a problem
posed at the beginning of any inquiry into linguistic meaning: By virtue
of what is it that an expression receives its meaning? Does it have this
meaning by nature or is it assigned to it by convention?
There are two opposed positions with respect to this question, which
have been designated in the history of semantics as (semantic) naturalism
and conventionalism.
Naturalism in the sense that one can infer the meaning of a word just
from its sound alone has never been seriously put forward and it would
be obviously absurd. For if that were the case we would not have to learn
foreign languages with much effort, but we would understand them imme­
diately. And the occurrence of homonyms, i.e. words with the same sound
and different meanings would be impossible to understand.3
Naturalism, as Plato describes it in the Cratylus, maintains only that
there is a natural affinity between the sound of a word and its meaning.
One possibility for such an affinity lies in the formation of words by way
of phonetic imitation of sounds in order to designate those sounds and
thereby indirectly to designate animals, objects or events for which those
sounds are typical, as in so-called onomatopoeia. Examples of onomato­
poeic words are ‘cuckoo’, ‘whisper’, ‘baa’.4 But such words are rare and
only the very fewest of the things we talk about are characterizable by
typical sounds. In his presentation of naturalism in the Cratylus, conse­
quently, Plato emphasizes the fact that the naturalistic thesis - words
characterize what they refer to by their sound - is not to be understood in
the sense that they imitate sounds, but in the sense that they present the
essence of the things.5
According to the naturalistic point of view, this presentation is based
upon an affinity sounds and sequences of sounds have with attributes,
events, actions, etc. Thus, for example r as a sound produced by vibration
is supposed to suggest movement, the occlusives d, t, b, p to suggest the
enduring, the binding, etc.6 By virtue of such affinities some words have
their meanings directly attached to them. Then other words are formed
 THEORIES OF MEANING 21

from them, words which also have a natural meaning on the basis of their
composition and etymological derivation.7 Finally there are metaphors,
i.e. transferences of words to things which manifest certain relationships
to the objects originally designated by the word (‘bottleneck’, ‘elbow-pipe’,
etc.) as well as a great many variations on the derived words, which can
obscure their origins.
Now Plato himself lays stress on the fact that the affinity between
sounds and attributes is very thin8 and even he looks on his etymological
analysis with skepticism.9 For this reason he considerably weakens the
naturalism he had advanced by emphasizing the fact that conventions
also play a part in determining the meanings of words.10 The criticism
directed against naturalism to the effect that it is unable to explain the
occurrence of synonyms, i.e. of expressions with the same meaning but
different sounds, nor the differences of one natural language from another,
Plato refutes by noting that just as there are different pictures of the same
object, all of which exhibit some relation, even if different ones, to it, so
there can also be different phonetic presentations of one and the same
fact.11
If Plato adhered to the essential features of the naturalistic position in
the Cratylus, despite the difficulties he recognized in it, it was on the basis
of the following argument:12 There are true and false sentences. Thus the
sentence ‘Cats meow’ is true and the sentence ‘Dogs meow’ is false. Now
if it were a matter of mere convention whether cats were designated as
‘cats’ one could just as well call them ‘dogs’ and say ‘Dogs meow’ instead
of ‘Cats meow’. But these sentences have different truth values and conse­
quently we can not simply say ‘dogs’ instead of ‘cats’, but cats must be
called ‘cats’; i.e. for all things there are designations that are just as
objectively correct and false as the sentences formed with them are ob­
jectively true and false.
But this argument is untenable because a sentence is not true or false
on the basis of its sound pattern alone, but only on the basis of a semantic
interpretation of that sound-pattern.13 I.e., the sentence ‘Dogs meow’ is not
false as such, but only on the usual interpretation of the words ‘dog’ and
‘meow’. If one interprets the word ‘dog’ as one normally interprets the
word ‘cat’, this sentence becomes true.
Along with this argument the most important rationale for a non-
conventionalistic interpretation of meanings collapses, however, and
22 CHAPTER II

conventionalism then becomes the most plausible position.14 This does

not require us to assume that linguistic signs were originally chosen quite
arbitrarily - in particular, for example, there may be affinities between the
sounds and expressive components of statements15 - but for languages as
we have them before us now a natural affinity between sound-structure
and meaning is certainly no longer demonstrable except in a few special
cases, as for example onomatopoetic formations. Therefore, we can
proceed on the practical assumption that the relation between sign and
significatum is arbitrary and depends on conventions.
Now this semantic conventionalism must, however, be restricted in
several respects. Thus Plato emphasizes in the Cratylus that linguistic
conventions are not explicit and formal agreements (ai)v0f|KT|), but
customs (zQoq) of linguistic usage handed down by tradition,16 and that
linguistic meanings can not be fixed by everyone to suit himself.17 Lan­
guage can fulfill its function of intelligible communication only if all who
participate use the words in the same way (i.e. in the same sense) and if
the norms of usage (vopoq) remain practically constant.18
While the previous statements on the controversy between naturalism
and conventionalism hold true for all theories of meaning, for a realistic
theory of meaning as Plato presents it,19 there arise still further restric­
tions on conventionalism, not required by other theories of meaning, or
not required in the same degree.20 According to Plato things in themselves
exhibit certain objective attributes, relations and differences. It is now the
task of language to give appropriate expression to these attributes. If,
for example, a language is unable to express an objectively given differ­
ence and has the same word for gold and iron, say, then it is not adequate.21
This requirement of Plato’s, that an adequate language must be able
to represent all real entities, one which essentially concerns the language’s
range of expression, is to be supplemented on one important point:
Language is definitely not just a set of isolated names for designating
entities (things, attributes and the like), but a system in which out of a
finite number of linguistic basic elements (e.g. words) we are able to con­
struct infinitely many sentences. These sentences have a certain grammat­
ical structure and this structure must correspond to their structure of
meaning and consequently, on the realistic interpretation, to the onto­
logical structure of the states of affairs of which they speak. That is to say,
one can not construct a language by arbitrarily introducing some signs
 THEORIES OF MEANING 23

for entities; these signs must form a system, in which they can be grammat­
ically combined in such a way that they reflect the ontological relation­
ships in which the entities designated stand. For example, if one wants to
express the fact that a thing a stands in the relation p to a thing /?, one
needs not only linguistic signs for a, /?, p (e.g., a, b, JR), but also a syntac­
tical relation between these signs with which the relation between a, /? and
p can be represented (e.g. as R(a, b)). Grammar must therefore be con­
structed in accord with ontology. Not all the systems of signs you might
choose come into question as languages, but only those that have a
grammar which permits the expression of ontological differences and
relationships.
Plato also compares language with a set of tools (opyavov), which must
be of such a kind that we are able to work on the things linguistically with
them, tools which must therefore be constructed properly for describing,
differentiating, comparing, etc., things. Just as one can not work on any
material you please with any tool you please, but only with such tools as
are suitable to the material and the work to be done on the occasion, one
can not describe, distinguish, etc., with any linguistic expressions you
please, but only with a linguistic organon that is properly constructed.
And conventionalism finds its limit in this requirement, that the linguistic
tools be true to their tasks.22
From these considerations the following interpretation of realistic
semantics emerges: language as an abstract system of signs in the sense
of 1.1 must correspond to the ontology of the range of objects to which
the language refers. In particular, the ontological structures must be pre­
sentable in terms of grammatical structures and every grammatical struc­
ture must correspond unambiguously to a definite ontological structure.
In the ideal case language and ontology would be isomorphic structures.
Language as an abstract system of signs is thus determined by the range
of meaning it is supposed to represent. Here conventions have no place.
The only conventional aspects are the various realizations of language
that are used.
Later on, Leibniz speaks to the same effect in his Dialogue:

Even if the marks are arbitrary, there is something about their use and combination
that is not arbitrary - a correspondence between marks and things and the relations
between different marks that mean the same thing. This relation or this reference is the
basis of truth.23
24 CHAPTER II

But taking all that into account, for a realistic semantics these limits of
conventionalism consist only in requiring that the conventions serve the
purpose of linguistic constructions, that of representing adequately ante­
cedently given ontological structures and distinctions, that they be suited
to their intended purpose from the realistic standpoint. But being suited
to an intended purpose is obviously a general requirement for convention­
alism as well.

1.2. Basic Ideas o f Realistic Semantics

What we will do now is first of all to present the basic ideas of the con­
ventionalist copy-theory and then discuss some important theories of
this type that have been proposed in the modern literature.
Realistic semantics takes on its simplest form when it is said that lin­
guistic expressions have only one semantic function, which consists in the
fact that (on the basis of convention) they designate certain entities.
Thus the meaning of a proper name consists in its designating an object:
I introduce a young man and say: ‘This is Frederick Schulze’; I point to a
town and say ‘That is Frieding’; I point to a mountain and say ‘That is the
Jungfrau’; I point to a constellation and say ‘That is Orion’. On each
occasion, then, I indicate an object and stipulate that a certain expression
is supposed to function as the name of this object. That constitutes the
entire semantic function of this expression.
One proceeds in a similar way with predicates, saying, for example: The
predicate ‘red’ stands for the quality you can observe on this object and
this object, but not on this one and that one; the predicate ‘larger than’
stands for the relation that holds between the members of this pair and
that pair of objects, but not between the members of such and such other
pairs. Predicates thus designate qualities and relations or, more generally,
attributes.
Finally, for sentences with the simple structure subject-predicate-
object(s), symbolically F(al9..., an), it is stipulated that they designate the
state of affairs that the attribute designated by F belongs to the H-tuple
designated by au ..., an. The sentence ‘The Jungfrau is a mountain’ thus
designates the state of affairs that the Jungfrau is a mountain; the sentence
‘Munich lies between Garmisch and Nuremberg’ designates the state of
affairs that Munich lies between Garmisch and Nuremberg, and so on.
What simple sentences designate is thus defined as a function of what the
 THEORIES OF MEANING 25

proper names and predicates that occur in them designate together with
the way in which the sentence is constructed out of those terms. What
determines the meaning of a sentence is the meanings of the words that
occur in it and their position in the sentence - ‘Fritz strikes Hans’ means
something different from ‘Hans strikes Fritz’, even though the same words
occur in both sentences, so that order also plays an important role.
Furthermore, a sentence is called true if and only if the state of affairs it
designates is a fact. The sentence ‘The Jungfrau is more than 6000 feet
high’ is true if and only if the Jungfrau is more than 6000 feet high.24
The meaning of complex sentences is then determined in a similar
fashion, with the aid of the semantic rules for grammatical combinations
and with reference to the meaning of the simple linguistic expressions.
This will be presented in III.2.1.
By virtue of stipulation, which is the gist of this theory, linguistic ex­
pressions are conventionally attached to certain entities (objects, attri­
butes, facts). This relationship is called the name-relation and the semantic
function of the expressions consists in this relationship entirely. Accord­
ing to this theory, that is all there is to the meaning of linguistic expres­
sions.
If we have previously spoken of attributes (qualities and relations) and
states of affairs, this requires some further elucidation. These terms are
often used in such a way that they are supposed to designate something
real, so that attributes and facts are something actually present in the
world. Attributes and states of affairs are then contrasted with concepts
and propositions, these latter as something in thought only, as in con­
ceptualism, or as entities that constitute a realm of reality of their own,
as in Platonism. While ‘red’, for example, is a quality found in the world,
and consequently an attribute, ‘prime number’ would be a concept; and
while ‘This rose is red’ is a fact which is actually encountered in the world,
‘17 is a prime number’ would be a proposition, which does not occur in
the world.
But this distinction between attributes and concepts on the one side
and states of affairs and propositions on the other would first have to be
made more precise: what ‘real’ or ‘actual’ means has not been generally
established. Indeed, the word “real” has quite different meanings in
different contexts, as the examples, ‘a real (i.e. true) friend’, ‘a real (i.e.
not an imaginary) illness’, ‘a real (i.e. effective) help’, demonstrate.
26 CHAPTER II

Take the case of clearly defined predicates. Shall we say that all of the
predicates that can logically be constructed out of them also denote attri­
butes, as they do (‘non-red’ along with ‘red’, ‘square circle’ along with
‘circle’ and ‘square’)? If yes, how is it that all logically possible conceptual
structures are also actually realized? If no, then there are sentences formed
with the aid of logical connectives out of sentences that denote the real
which do not themselves denote anything real, i.e. there are sentences that
change their meaning in certain contexts.
The following definitions for attributes and states of affairs would come
easily to mind: We could (1) interpret attributes (such as ‘red’, ‘unicorn’
but not ‘prime number’) as concepts that are defined only for empirical
objects (concrete things or animals, plants, human beings); or (2) as
concepts under which an empirical object falls (‘red’ but not ‘unicorn’).
And one could interpret states of affairs (3) as propositions that refer to
empirical objects (such as ‘Munich has less than a million inhabitants’ but
not ‘3 is a prime number’) or (4) as propositions which are true, i.e. which
say what is actually the case (such as ‘Munich has more than a million
inhabitants’ but not ‘Munich has less than a million inhabitants’). However,
according to definitions (2) and (4) the meaning of a predicate or a sentence
would depend on empirical circumstances: If marmots were to become
extinct, the word ‘marmot’ would change its meaning and the sentence
‘Munich has less than a million inhabitants’ would have changed its
meaning since 1950.25 On definitions (1) and (3), however, the term
‘empirical’ remains vague and, besides, the distinction between ‘em­
pirical’ - ‘nonempirical’ is scarcely any more important for semantics
than the distinction between ‘animate’ - ‘inanimate’.
In what follows we will make no such distinction, therefore, and use the
terms ‘attribute’ and ‘concept’ or ‘state of affair’ and ‘proposition’ as
synonyms. We will also not go into questions concerning the nature and
ontological status of concepts and propositions right away. These
questions will be cleared up in the course of further discussion.
Another version of realistic semantics arises if the conceptual entities -
possibly because their ontological status looks all too obscure - are iden­
tified with mental contents, for the most part called ‘ideas’, and in line
with this, linguistic expressions are said to denote ideas. On that view, the
word ‘this rose’ would not directly denote this rose. Instead, it would
immediately refer to an idea of the rose and only by way of the relation
 THEORIES OF MEANING 27

between the rose and the idea of it have any reference to the rose itself.
And the word ‘red’ would not denote the attribute ‘red’, but a “general
idea” of ‘red’. But the general nature of such ideas remains completely in
the dark - still more obscure, if possible, than that of concepts. Besides,
as a mental state an idea is in each case a state (or content of experience)
of a subject: two ideas belonging to different persons are different even
though they refer to the same thing: X can not have one of Y 9s ideas and
vice versa. But that would imply that the sentence ‘This rose is red’ would
mean quite different things for the speaker and the hearer. Besides that
surely we use the sentence ‘This rose is red’ to speak not of our idea or of
other people’s ideas of the rose but of the rose itself.26
After this brief excursion on attributes, concepts and ideas, back to
the simple form of realistic semantics!
This simple form of the copy-theory runs into some difficulties. There
is already a certain difficulty in the fact that we cannot point to attributes
and facts, as abstracta, in the same way as we can to concrete objects in
establishing the name-relation.27 If we want to say which attribute ‘red’
designates, we can not produce this attribute itself, not point to it as to a
mountain, a village, a constellation; we can only point out some instances
and counter-instances of the attribute: some red and some non-red things.
But it is clear that a finite number of examples and counter-examples
will not be sufficient to distinguish an attribute unambiguously. And so
one runs into the problem: Why does the learning of predicates by
examples work and to what extent does it work? But we will discuss this
question more exactly only somewhat later on.28 Right here we will
simply be satisfied with the fact that, as experience shows us, it does
work very well.
Furthermore, from a nominalistic standpoint, one could object that there
are only concrete objects, and no abstracta such as attributes, and that
consequently predicates could not be names because there is nothing they
could denote.29
According to the nominalistic interpretation, predicates, in contrast to
proper names and sentences, do not have meaning in and of themselves,
but are synsemantic or syncategorematic expressions, i.e. expressions
that have meaning only in the sentence context. The word ‘rose’ means
nothing alone, then, but only sentences such as ‘This rose is red’, ‘Fritz
has a red shirt’, etc., have meaning. The traditional formulation of this
28 CHAPTER II

point of view is universale est vox, that is, the objects that fall under a pred­
icate F have nothing in common except that they are called F. Red ob­
jects, then, do not have a common color or quality - what they have in
common is nothing but being called ‘red’. But this interpretation leads on
its part to the following fundamental difficulty: I f ‘red’ does not have any
meaning in and of itself, then the meaning of the sentences 6a is red’, ‘b is
red’, etc., is not determined by the meaning of the terms ‘a \ ‘b’ and ‘red’.
Rather, the meaning of all of these sentences is to be determined separate­
ly in each case, i.e. all of them are to be regarded as distinct sentence con­
stants. But then it is no longer possible to communicate new facts with
sentences of the form ‘c is red’, i.e. to make an assertion about the color
of a new object c, to which the predicate had not previously been applied,
because the meaning of ‘c is red’ must first be defined. A language in which
no new facts can be communicated is of no use, however, and therefore
we shall not be able to avoid assigning meanings to predicates also.
Since a nominalist does not admit propositions either, one can not
define the semantic function of predicates on the basis of propositions.
The meaning function of sentences is supposed to consist simply in the
fact that they are true or false. The stipulation that certain sentences ‘a is
an F \ 6b is an F \ ... are supposed to be true and other sentences ‘c is an
F \ ... false, says nothing about F - ‘F ’ is defined simply as a predicate that
applies to a, b, ..., but not to c, d, ... - and consequently the sentences ‘a
is an F \ ... etc., do not say anything about the objects a, b, c ,d ... either.
Nominalism offers no reasonable starting point for a theory of meaning,
then. We can therefore set it aside in what follows. The ontological ob­
jections against assuming abstract entities such as concepts and proposi­
tions will be solved in later discussions in a different way.
By way of contrast, the following objection to the simple form of realis­
tic semantics is of fundamental significance: We usually understand by
the meaning of a linguistic expression something based on the under­
standing of language alone, and independent of empirical data. In order
to know what an expression means, I have to master the language to which
it belongs; in doubtful cases I have to consult dictionaries and grammars,
but not any knowledge of facts. Therefore the question of the synonymy,
the identity of meaning of two proper names must be capable of being
decided on the basis of knowledge of language alone. But if we identify
the meaning of a proper name with the object it denotes, then the question,
 THEORIES OF MEANING 29

for example, of whether the expressions ‘morning star’ and ‘evening

star’ are synonymous becomes an empirical question: the synonymy of
these two expressions can only be determined by astronomical observa­
tions.30
Furthermore, the sentences ‘The morning star is identical with the
morning star’ and ‘The morning star is identical with the evening star’
have different meanings as we usually understand them. The first is a
logical tautology, the second an empirical statement. Now since the
meaning of a sentence can depend on nothing other than the meanings of
the words that occur in it and their position in the sentence, and since the
two sentences differ only in the words ‘morning star’ and ‘evening star’,
the meaning of these two words must be different even though they name
the same object. And so we can not identify the meaning of a proper name
with the object it names.
There are also meaningful proper names that denote no (real) object, as
for example, ‘Odysseus’, ‘the smallest real number greater than O’, ‘the
present king of France’. These names are not completely meaningless,
however, as are expressions like ‘Ripht’, ‘Kaluphem’ etc., but they are
different from names like ‘morning star’, ‘the smallest natural number
greater than O’, ‘the King of France in 1182’ only by virtue of the fact that,
because of the constitution of the world or the real number axioms, there
is nothing they name. Thus Quine says that “A singular term need not
name to be significant.” 31 And in [53], 40 Wittgenstein pointed out that a
proper name does not lose its meaning by virtue of the fact that the object
it names ceases to exist. The name ‘Socrates’ is also meaningful today, when
there is no longer any Socrates. Nor can we say that the meaning of the word
‘Zugspitze’ is 3000 meters high. It follows from this, too, that the mean­
ing of a proper name can not be identified with the thing it names. We
must, then, distinguish the meaning of a proper name from its reference,
i.e. from the object it names - if there is one.32 And we must distinguish
between proper names that are meaningless and so have no reference
either (like ‘Kaluphem’, ‘Ripht’), proper names that have meaning but no
reference (like ‘the present King of France’) and proper names that have
both meaning and reference (like ‘morning star’, ‘the smallest natural
number greater than O’). In order to emphasize terminologically the
difference in the relations of the proper name to its meaning and to its
reference, it is often said that a proper name expresses its meaning and
30 CHAPTER II

designates or names the object for which it is a name. In German Bedeu-

tung (meaning) is distinguished from Bezug (reference).33
What, now, does constitute the meaning of a proper name if it is not
to be identified with its reference? We will defer answering this question
and return once again to the meaning of predicates and sentences.
For a predicate we can continue to identify its meaning with the concept
it stands for. There are no two predicates that stand for the same concept
of which we would claim that they are different in meaning (in the usual
sense). And there is no meaningful predicate of which we could not say
that it stands for a concept.34 Concepts, as we saw above, are not present in
the world like objects: it is not an empirical but a logical question whether
a specified concept exists. So concepts are meanings of predicates and we
say now, following the terminological distinction between ‘name’ or
‘designate’ and ‘express’, that predicates express (not designate or name)
concepts.35
Predicates lack direct reference and it is not necessary to construct such
a reference. For reasons of symmetry, however, we can say that a predicate
refers to (designates) the class of things that fall under the concept the
predicate expresses. What this class looks like, i.e. what elements it includes,
is then once more an empirical question and it is also an empirical question
whether two predicates refer to the same class, whether for example the
predicates ‘animal with kidneys’ and ‘animal with a heart’ have the same
reference. This construction of a reference for predicates will prove to be
very useful in what follows.
We can also, in the case of a sentence, identify its meaning with the
proposition it stands for. Again, there are no two sentences which stand
for the same proposition and which we would claim to be different in
meaning (in the usual sense), and there is no meaningful sentence of which
we could not say that it stands for a proposition.36 Propositions are no more
empirically given things than concepts are: it is not an empirical question
whether a proposition exists or not or whether two propositions are iden­
tical. So we say that a sentence expresses (not designates) a proposition.
There is also a natural way in which one can assign a reference to prop­
ositions: their truth-value. Whether a sentence is true or false is a question
of fact, which generally speaking cannot be answered if we know only
the sentence’s meaning. We know what the sentence ‘It is now raining in
New York’ means; but in order to determine whether it is true, one must
 THEORIES OF MEANING 31

make empirical observations. But since ‘true’ and ‘false’ are predicates,
we do not say - as Frege does - that a sentence designates its truth value,
that it is a name for it.
Since a simple (atomic) sentence is constructed out of proper names
and a predicate, its meaning depends entirely on the meanings of the words
that occur in it (the proper names and the predicate) and the way in which
these words are connected together in the sentence. The reference of a
sentence will also depend on the reference of the proper names that occur
in it: the truth-value of the sentence ‘Fritz is blond’ depends on which
individual is designated by the name ‘Fritz’. And if a proper name that
designates no object occurs in a sentence, such as ‘Odysseus’ or ‘the
present King of France’, then reference can also be denied of it.37
Whether a sentence has a reference, then, is also an empirical question.
Now if one were not to assign any reference to predicates, then the refer­
ence of a sentence would not only depend on the words that occur in it
and their connection, but on their meanings as well. But if we assign a
reference to predicates in the way described, then the reference of the
sentence can be interpreted as a function of the references of words oc­
curring in it and their connection in the sentence.38
Now proper names and sentences can also be constructed from a predi­
cate F without the use of proper names. Examples such as the expressions
‘XxF{xy (‘the class of objects with the property F ’),‘ixF(x)’ (‘the thing that
has the property i 7’), ‘A xF (xy (‘All things have the property F ’), ‘ VxF(x)’
(‘Some things have the property F ’) show that this is so. The reference of
these expressions depends on the range of objects being spoken of in the
specific context and on the meaning of the predicate. The truth-value of
‘All numbers are divisible by two’ depends on whether I am considering
only even numbers, or any natural number you please, and the truth-
value of the sentence ‘There is a square root of —1’ depends on whether
I am considering real or complex numbers. The reference to a domain of
objects in the quantifiers (over which the variables within these quantifiers
range) thus determines the reference of the sentence in these cases.
Now let us return to the meaning of proper names! In the first place
there is the group of proper names that take the form of a definite descrip­
tion, such as ‘the author of Kabale und Liebe\ ‘the present President of the
United States’, ‘the first place you come to when you go south on B2
from Murnau’, etc. They are constructed with a predicate F and we write
32 CHAPTER II

them symbolically in the form ixF(x). If the description is supposed to be

successful, i.e. to designate a well-defined object, then exactly one object
must fall under the predicate.39 Whether the description is a name and,
if it is, for what object is a question of fact and it is equally a question of
fact whether two descriptions such as ‘the morning star’ and ‘the evening
star’ refer to the same object. If there is no object that falls under F, then
ixF(x) is not a name (as, for example, ‘the present King of France’, ‘the
smallest real number greater than O’). Likewise, in the case in which
several objects fall under F, one will also say that ixF(x) is not a name
(does not denote anything), as for example ‘Heidegger’s book’, ‘the son
of Jacob’. In these cases, then, definite descriptions have no reference.
We can now determine the meaning of a definite description by way of
the meaning of the descriptive predicate. So for this case we can reduce the
question of the meaning of proper names to the question of the meaning
of predicates. The like also holds for other proper names constructed
from predicates, as for example the class-names ‘AxF(x)’ (‘the class of
things that have the property F ’), or for function-expressions ‘/ (a )’ (‘the
value of the function / for argument a’) since functions can be interpreted
as many-one relations.40 In these cases, too, we can define the meaning of
these names in terms of the meaning of the predicates from which they
are constructed. The meaning of these proper names, then, is the way in
which the objects they designate - if there are such - are conceptually
defined by the names.
How does it stand with the remaining proper names now, those that do
not have the form of a definite description, a class term or a function
term? We will call them ostensive proper names. ‘Socrates’, ‘Munich’, ‘the
Zugspitze’, for example, are proper names of this sort. We can say of
these proper names that their primary semantic function consists in their
reference. There are, however, meaningful proper names with no reference,
such as ‘Odysseus’, so that we can not disregard meaning entirely in con­
nection with these names. What one could now say is that ostensive
proper names are more or less exactly defined in their meaning by a more
or less sharply delimited set of assertions in which they occur and which
have a quasi-analytic character. On this view, the meaning of ‘Aristotle’,
for example, would be defined by assertions such as ‘Aristotle was a pupil
of Plato’s’, ‘Aristotle was the teacher of Alexander the Great’, ‘Aristotle
wrote the Organon’, etc., thus by assertions which for us are always bound
 THEORIES OF MEANING 33

up with the name ‘Aristotle’. There are such assertions even for a name
introduced by indicating the object for which it stands, for an object is
never given to us as a bare individual but always under certain conceptual
specifications. It is clear, however, that - apart from special cases, like the
introduction of a proper name in a system of implicit definitions (such as
the term ‘0’ in the Peano axioms, for example) or names for fictitious per­
sons of whom assertions are made in a story (‘Odysseus’ as this name is
used in the Iliad and the Odyssey) - for ostensive names no group of
such identifying descriptive assertions with strictly fixed boundaries can
be given. Different people connect quite different identifying assertions
with the same name and even the single individual can not give a sharp
boundary between assertions which serve as identificatory for him and
other assertions.41 One can not regard all of the assertions that hold true
of a proper name as identificatory for it in the sense of an implicit defini­
tion either, for otherwise all of these assertions would be analytic.
Even if one defines the meaning of ostensive proper names as the mean­
ing of the predicate that arises from the conjunction of the identificatory
assertions by replacing the proper name in question by a variable, it is
clear that it follows from this that the meaning of such proper names is
generally very indefinite.42 This indefiniteness of meaning is pointed out
in many accounts of semantics where meaning is not interpreted in quite
another sense and identified, for example, with reference.43 The meaning
of ostensive names may, however, also be defined so that it is determined
solely by their reference. This interpretation as standard names is discussed
in Chapter II. 1.5.

1.3. Wittgensteins Picture Theory o f Language in the Tractatus 44

We shall now present and discuss some realistic semantic theories. In
doing so, however, we will not make any great, wide-ranging excursion
into history, but limit ourselves to a few particularly important modern
theories. This section is intended to deal with the basic features of Witt­
genstein’s semantics in the Tractatus ([22]). We shall not take account of
the many complex details, which are not necessary for the theory, nor of
the other themes the Tractatus deals with. What we are about in the follow­
ing is not an interpretation of the entire Tractatus but only its basic se­
mantic ideas. For the interpretation of the difficult text we are making use
of the detailed and thorough work by E. Stenius [60].45
34 CHAPTER II

Wittgenstein starts from ontological presuppositions, entirely in the

realistic manner.46 According to his presuppositions, entities are to be
categorically subdivided into states o f affairs, objects and attributes. Facts
are existing (realized) states of affairs ([22], 2). The world is presented as
the fact that such and such states of affairs, such and such relationships
among things exist ([22], 2.01). States of affairs can be simple or complex.
A complex state of affairs can be analyzed into simpler states of affairs.
According to Wittgenstein the world is a complex fact which can be anal­
yzed without ambiguity (herein lies the absolutism of the Tractatus) into
atomic facts of the utmost simplicity (in the postulation of such ultimately
simple facts lies the atomism of the Tractatus). Wittgenstein calls the
(simplest) objects and attributes that are present in these atomic facts the
things, which together constitute the substance of the world ([22], 2.021).
How these atomic states of affairs and things are supposed to look re­
mains unclear. In any case, it becomes apparent from Wittgenstein’s
further remarks that things can not be identified with the concrete ob­
jects and their attributes that we usually have in front of us, whether in
everyday life or in the sciences. Their existence is merely postulated; they
are not described more precisely or exemplified.
Now according to Wittgenstein an atomic sentence is a picture of an
atomic fact ([22], 3 and 4), while complex sentences can be interpreted as
pictures of facts only indirectly, namely insofar as they can be transformed
into compounds of atomic sentences which can be represented as pictures
of factual complexes. Since these transformation rules are not given more
exactly, however, and the treatment of complex sentences does not have
to do with the basic semantic ideas in the Tractatus, we will restrict our­
selves here to the semantic function of atomic sentences.
Obviously, by the claim that sentences are pictures of facts it can not be
intended that a sentence is a picture in the naturalistic sense, as a photo­
graph or a portrait is a picture of a person. For sentences have no super­
ficial similarity with facts. To define his picture concept Wittgenstein
proceeds rather from the concept of isomorphic mapping. One says that
there is an isomorphy between a structure A = <A , ..., R™m}, composed
of a (non-empty) set of objects A and m nr place attributes R lnt (i = 1,..., m),
and a structure B = (B , S £l5..., S ”m> of the same type, composed of a
(non-empty) set B and m «r place attributes S lm, if there is a one-to-one
mapping $ of A on B, so that the following holds:
 THEORIES OF MEANING 35

(I) ^ ni{au ... ,ant) = ((£ (aO. ••• ><£(a„,)) for all i and all «r tuples
al9...9ant from A. (j) is then called an isomorphy-correlator.

For example, if A = ({a, b, c}9 i?}, R \y and5=<{//, e9f } 9 S {9 5*>and

if it is the case that R{a, ~iR\b9 R [ c9 R 2ab9R\cb and ~\R \xy for all other
pairs x, y from {<a9b, c}9 and also that S{e9 S [ f 9"i S \d 9S\ed9 S \ f d and
-i S \xy for all the remaining pairs x, y from {d9e , f }, then the mapping <£,
defined by $ (a) —e9^>(b) = d,(j) (c) = f 9is an isomorphic mapping oi-A on B.
Both structures can be illustrated graphically by the following figures,
in which R \ a is represented by@and R\ab by a -» b9 S \a by[£]and S\ab
by a=>b.

A simple example may make this picture concept clear. Five tokens of
two colors, viz. green or red and blue or yellow, and with different sizes
can represent the relationships in a family of five persons, if a token is
assigned to each person according to the rules that larger tokens corre­
spond to older persons, red tokens to the men in the family, green to the
women, blue to parents and yellow to children.
Now in order for Wittgenstein’s idea of the sentence as a picture to be
carried out, he interprets the sentence not as an object but as a fact: The
proper names in the sentence - e.g. 6a’ and *b9 in ‘aSb’ - are objects to
which objects a and b in the fact to be pictorially represented are assigned,
by virtue of a rule of interpretation (an isomorphic mapping of the (atomic)
objects in the world onto the proper names). The predicate sign ‘5 ’ is to
be distinguished from the syntactic 5-relation, in which the objects V and
‘6’ stand in the sentence 6aSb\ namely from the relation of standing to the
left or to the right of the sign ‘5 ’. It is this 5-relation, not the sign ‘5 ’,
which corresponds to a (here two-place) attribute 5 in the fact to be
pictured.47
According to Wittgenstein, then, it is not the predicates as objects, i.e. as
expressions, that are names of attributes, but the syntactic relations. And
the pictures (names) of facts are not sentences in the usual sense (as objects,
sound sequences) but syntactic facts. So names, according to Wittgen­
stein, are always of the same ontological category as what they stand for.
In order for a sentence to be a picture of a fact, on Wittgenstein’s
36 CHAPTER II

theory, it is sufficient that the sentence and the fact are of the same struc­
tural type, i.e. that the same number of objects, or proper names, occur
in them and the same number of attributes with a like number of places
and that a rule of interpretation <j>is available. He does not require that
the isomorphy conditions (I) are fulfilled, for example that aSb holds
when the sentence reads 6aSb\ but bSa, for example, can also be the case.
A sentence is a correct picture of a fact, i.e. true, if and only if the iso­
morphy conditions are fulfilled.48
The sentence represents its sense as a fact, i.e. it represents ‘how things
are, if it is true’ ([22], 4.022). “To understand a sentence is to know what
is the case when it is true” ([22], 4.024). “The meanings of the simple
signs must be explained for us to understand them. But we use sentences
to explain and make ourselves understood” ([22], 4.026).
In Wittgenstein’s system, then, the rule of interpretation (j>corresponds
to the name-relation for proper names. Combined with it there must be a
coordination of syntactic relations and attributes.49 On the basis of these
two correlations the sentence as a specific relation between names then
represents a fact as a relation between things, just as the model using
tokens, in the example above, represents which family relationships are
before us.
In view of all this, the semantics of theTractatus essentially differs from
the basic form of realistic semantics delineated in II.l. in that predicates
and sentences are not interpreted as objects, but as relations and facts.
Consequently, the semantic function of language does not consist simply
in naming or expressing, but also in a correspondence of the categorical
structure of the expressions to the categorical structure of reality. The
predicative nature of attributes is communicated by the predicative na­
ture of the syntactic predicate relationships and the inner structure of the
fact is communicated by the inner structure of sentences as syntactic
facts.50
If Wittgenstein goes beyond the realistic account in this point, as we
have seen above51, there is in his system, on the other hand, no distinc­
tion between meaning and reference. For him proper names mean objects,
predicates, classes. That may be connected with the fact that in making
his observations he has in mind an ideal language that is regarded as being
purely extensional as the artificial languages of modern logic, which like­
wise have expressions for which only a reference is stipulated, but not a
 THEORIES OF MEANING 37

meaning. For a general semantics, however, the distinction between refer­

ence and meaning is of fundamental importance. Wittgenstein first
reinstated this distinction in [53].
Many more critical remarks can still be made concerning Wittgenstein’s
semantics, particularly in connection with his absolutism, his atomism
and his theory of the semantic function of compound sentences. Since we
have not presented these theories here and since the basic semantic ideas
of the Tractatus are independent of absolutism and atomism, we will not
go into them any further here. What is important in our context is only
that Wittgenstein’s semantics in the Tractatus is a good example of the
simple foim of realistic semantics, since it develops the ontological pre­
suppositions and the correlation theses of this semantics in a completely
explicit manner.

1.4. Frege's Semantics

Frege formulated his ideas on semantics, which had the greatest influence
on the development of modern theories of meaning, primarily in [92] and
in the fragment ‘Ausfuhrungen uber Sinn und Bedeutung’ (ca. 1895).52
He proceeds from the idea that proper names mean objects. Starting from
here, however, as we have already shown in terms of a Fregean example
in II.1.2, one falls into the difficulty that the meaning of a sentence can be
changed by replacing a proper name by a proper name with the same
meaning. Frege concluded from this that proper names have two semantic
functions: in the Fregean terminology they mean (bedeuten) an object and
express (driicken aus) a sense (Sinn). Frege thus distinguishes between the
sense and the meaning (Bedeutung) of proper names: every proper name
that is meaningful at all (in our earlier sense, that it stands for something,
and is not, like the expressions ‘Kaluphem’ or ‘Ripht’, for example, a
mere series of letters without a semantic function) has a sense. Proper
names with the same sense also have the same meaning in Frege’s
terminology, but proper names that have the same meaning in Frege’s
terminology do not always have the same sense.
It is an unfavorable feature of Frege’s terminology that he designates
the reference as ‘meaning’, for meaning as usually understood is something
given as immanent within language, the existence and character of which
does not depend upon extralinguistic facts. For this reason we have not
adopted the Fregean terminology. If one makes use of the Fregean expres­
38 CHAPTER II

sion ‘meaning’, then, one must always keep it in mind that this word as
used by Frege is not used in its normal way. That will avoid the appearance
of inadequacy that some of Frege’s statements have. In order to draw a
clear distinction between ‘meaning’ in our sense, described in II. 1.2, and
‘meaning’ in the Fregean sense, in the following we will identify it, and
related terms, by attaching the subscript tF \
Now if one assumes along with Frege that the assertive content of a
sentence depends on the sense of the proper names that occur in it, then
the different content of the sentences ‘The morning star is identical with
the morning star’ and ‘the morning star is identical with the evening star’
can be explicated by way of a difference in sense between the two (identical
in meaningF) proper names ‘morning star’ and ‘evening star’.
Now what, according to Frege, constitutes the sense of a proper name?
He interprets the sense of a proper name as the way in which the designat­
ed object is given by means of the name. Thus by way of the proper name
‘morning star’ the planet Venus is given as the star that shines brightest
in the morning, by way of the name ‘evening star’ as the star that shines
brightest in the evening. And the point S in the figure below is character­
ized in different ways as ‘the intersection of the medians on a and V and
as ‘the intersection of the medians on b and c \

For Frege does every proper name have a determinate sense? We have
seen that this is questionable for ostensive proper names like ‘Aristotle’.
Frege, too, concedes that not all proper names have a determinate sense,
nevertheless he wants to exclude such cases for precise scientific lan­
guages.53 Frege also acknowledges proper names that have a sense but
are without meaningsF, such as ‘Odysseus’, ‘Pegasus’, ‘the smallest real
number greater than 1.’ Proper names of this sort are to be excluded from
assertive discourse, however, especially in scientific contexts, because
sentences that include such proper names are neither true nor false.
 THEORIES OF MEANING 39

The sense of a proper name for Frege is a conceptual, platonistic entity,

i.e. something intersubjective, objective and to be strictly distinguished
from the subjective ideas and associations which a given individual idio-
syncratically attaches to a proper name. For him the sense of a proper name
is something that various people attach to this proper name in the same
way.54 Frege gives no criterion for identity of sense for proper names. But,
as the following shows, one will have to interpret the concept of sense very
narrowly, so that, apart from trivial cases of identity of sense such as
arise from explicit definition, for example, different proper names as a rule
also have different sense. Although Frege never uses identifying descript­
ive concepts to define the sense of a proper name in the form of a descrip­
tion, such an interpretation, as we have seen in II. 1.2, would nevertheless
be very compatible with Frege’s assertions.
Frege now ascribes to predicates and sentences as well a double seman­
tic function in analogy with proper names. Sentences too have sense and
meaningF. Frege assumes that the meaningF of a sentence depends only on
the meaningF of the proper names that occur in it, not on their sense.
This implies that the sentence content, the proposition, can not be the
meaningF of the sentence, as the above examples show. So Frege holds
that the proposition or the thought (Gedanke), as he says, is the sense of a
sentence. Now if one takes account of the fact that what Frege under­
stands by ‘meaning’ is the reference, there is nothing artificial about
Frege’s regarding the truth-value of a sentence as its meaningF, so that
there emerges here an exact analogy with the meaningF of proper names.55
Corresponding to his acceptance of meaninglessF proper names Frege
also admits sentences that have sense well enough, but are not meaning-
fulF, i.e. are neither true nor false. This is true in particular of sentences
like ‘Odysseus landed in Ithaca’ which contain meaninglessF proper names.
But as we have said, such sentences are to be excluded in scientific con­
texts. Nevertheless, every sentence whatsoever with a semantic character
has a sense.56
Frege’s account of the sense and meaning of sentences and the general
principle of substitution, that in all sentences expressions with the same
meaningF can be exchanged for each other without altering the truth-
values of the sentences (because the meaningF of a sentence depends on
the meaningp of the expressions that occur in them, not on their sense)
can only be sustained if it is assumed that the expressions do not have their
40 CHAPTER II

usual meaningF in all contexts. For there are contexts in which the ex­
change of expressions with the same meaningF alters not only the sense,
but also the truth-value, the meaningF of a sentence. Along with Frege,
Russell in [05] and Quine in [64b] have given examples of such cases.
Frege brings in the example: ‘A lied in saying that he had seen B \ 57 If this
sentence is true then 6A saw B ’ is false. Now le t6A saw C ’ also be false;
despite that, ‘A lied in saying that he had seen C ’ may be false. Russell
offers the example of the true sentence: ‘George IY wished to know
whether Scott was the author of Waverly’. From it, by replacing the proper
name ‘the author of Waverly’ with the proper name ‘Scott’, which has the
same meaningF, there arises the false sentence: ‘George IV wished to
know whether Scott was Scott’. Quine, finally, introduced the example of
the true sentence: ‘It is a truth of mathematics that 9 = 9’, from which by
replacing the proper name ‘9’ with the proper name ‘the number of pla­
nets’, which has the same meaningF, there arises the false sentence ‘It is a
truth of mathematics that the number of planets= 9 ’. Frege now assumes
that in all contexts in which such an exchange of expressions which have
the same meaningF is not possible salva veritate - he calls them indirect or
oblique contexts - the meaningF of an expression is its sense. This assump­
tion is natural enough, since in indirect discourse, such as ‘Fritz said that
he had read the book’, for example, we are speaking only about the con­
tent of a statement of Fritz’s without making that statement ourselves.
Thus we are not speaking, as in the sentence ‘Fritz has read the book’
about the book and Fritz’s reading of it, but about the sense of an asser­
tion which is an assertion about the book only as a direct statement.
Frege shows in [92] in a detailed analysis of various types of subordi­
nate clauses - indirect contexts are mostly subordinate clauses - that this
account can be carried through.
Frege said nothing about the sense and the meaningF of predicates in
[92]. Some intimations about them are to be found, however, in the frag­
ment cited above Vber Sinn und Bedeutung. In it Frege introduces as the
sense of a predicate the concept it expresses, and as its meaning something
that could be called an extensional concept. Extensional concepts are
supposed to be predicative, i.e. they are not objects (and so, particularly,
they are not classes). Furthermore, they are supposed to be identical when
they apply to the same objects, i.e. when they determine the same classes.
According to this stipulation, the meaningF of a sentence then depends
 THEORIES OF MEANING 41

only on the meaningF of the predicates that occur in it, but not on their
sense. In this case Frege’s construction is somewhat artificial. The only
reason for not introducing classes as predicate meaningsF is that they
have no predicative character and so their application to objects does not
produce any truth-values as objects.
A paradox has been stated in connection with concepts similar to the
paradox of identity illustrated by the sentence pair ‘The morning star is
identical with the evening star’ - ‘The morning star is identical with the
morning star’. It originates with G. E. Moore and is called the paradox o f
analysis.58 According to it, sentence (a) ‘The concept “brother” is iden­
tical with the concept “male sibling” ’, has a different sense from sentence
(b) ‘The concept “brother” is identical with the concept “brother” ’,
because the sense of the former is not trivial and is relevant in connection
with the analysis of concepts. But if these sentences are different in sense,
then the expressions ‘the concept “brother” ’ and ‘the concept “male
sibling” ’ must be different in sense and therefore the two concepts must
be different. Analysis of concepts, so it seems, must be either trivial, like
(b), when the concept analyzed is identical with the concept given as anal­
yzing it, or false, like (a), when these concepts are different.
In [46] A. Church has recommended resolving this paradox analogously
with the Fregean paradox of identity of objects, by interpreting the expres­
sions ‘the concept “brother” ’ and ‘the concept “male sibling” ’ as identi­
cal in meaningF but different in sense. But what is the sense of a concept
name supposed to be in contrast with its meaningF in this case? Frege’s
distinction with regard to this point can not be used here, since as analy­
ses of concepts (a) and (b) do not concern classes as ranges of the concepts
(or extensional concepts as Frege understands them), but the concepts
themselves. Otherwise we would have before us not any analysis of a
concept but an empirical claim. The sense element, the meaning, lies
precisely in the conceptual, and what - in Frege’s way of speaking - is
supposed to be the way in which a concept is given by a name? Concepts
are not defined by (other) concepts nor identified like objects, but are the
instrument of definition.
Statement (a) will be more correctly formulated in this way: (c) ‘The
predicate “brother” is identical in meaning (synonymous) with the predi­
cate “male sibling” ’. Then the paradox does not arise, for the terms
‘ “brother” ’ and ‘ “male sibling” ’, as they occur in (c), are not only differ­
42 CHAPTER II

ent in sense, but also different in meaningF - they do in fact designate

different expressions - and so the substitution of identical items is out of
the question from the very beginning. Formulation (c) is also more ade­
quate than (a) because so-called analyses of concepts do not have to do
with establishing the identity of concepts but with assertions about the
meanings of predicates.
If one wishes to speak of concepts, however, then either the concepts
‘brother’ and ‘male sibling’ are identical - in that case this identity is not
an empirical or a linguistic, but a logical fact and just as trivial as the
identity of ‘brother’ and ‘brother’, or they are not identical and then
statement (a) is false. That (a) has a non-trivial look to it is probably
attributable to nothing but the fact that we understand it in the sense of
(c).«9
Frege’s semantics has been criticized from many sides. The only sort
of criticism we will go into here is that which stands on the ground of a
realistic semantics itself, for we shall concern ourselves extensively later
on with the more radical criticism of the realistic account as such. Looked
on from the standpoint of the semantic scheme given in II. 1.2, one could
especially object to the Fregean construction of predicate meaningsF, but
still, taken as a whole, the differences are not very troubling.
Carnap has offered a more fundamental criticism of Frege’s semantics
in [56]. Above all, Carnap makes the following objections to Frege’s
account:
(1) According to Frege, the same expressions can have different mean-
ingsF in different contexts. - That is correct, but not a decisive objection,
especially when one reflects that the meaningF is in fact the reference and
that naturally in indirect contexts the usual reference of an expression is
replaced by its sense.
(2) According to Carnap it can even come about that one and the same
occurrence of an expression has different meaningsF, namely its usual
meaningF and its usual sense, for instance, in the example Frege gave in
[92], ‘Bebel has the illusion th a t...’. In this case ‘has the illusion th a t...’
stands for ‘believes th at..., and n o t...’. In the latter expression ‘...’
stands at one time in an indirect context and at another time in a direct,
and so in its abbreviated form ‘has the illusion t hat . . must have both
its usual meaningF and its usual sense at one and the same time. - But it is
sufficient for a correct response to all questions of substitution and syn­
 THEORIES OF MEANING 43

onymity concerning this sentence to assume that ‘...’ stands in an in­

direct context here and meansF its usual sense.
(3) A name can be introduced for every entity, including sense-contents.
But what is the sense of such names of senses, the sense of names for the
latter sense contents and so on? An infinite hierarchy of sense contents
would have to be assumed. - A. Church has attempted to give such a hier­
archy of sense contents and so make the ontology of the Fregean seman­
tics more precise.60 But this system is of only formal interest and of little
relevance to semantics, for the sense contents of higher levels are not
precisely characterized and described in intuitive terms. On the other hand,
however, it is not at all necessary to introduce names for sense contents
and then ascribe to them a sense. Names for sense contents are needed in
Frege’s system no more than they are in Carnap’s. In order to speak of
sense contents, not new names, but indirect contexts are used.
To be sure, Carnap says that sense contents of higher order would
actually be needed in a variety of oblique contexts, as for example ‘It is
not necessary that Hans believes that it is possible that p \ But here too it
is sufficient for all pertinent semantic considerations to assume that p
simply stands in indirect context and meansF its usual sense.
Actually Frege’s semantic account is not so entirely different from Car­
nap’s method of extensions and intensions in [56], which we will consider
in the next section. While Frege holds to the general substitution principle
and therefore has to change the meaningF of expressions in indirect con­
texts, Carnap restricts the substitution principle and can therefore require
that expressions have the same meaningp in all contexts.
Finally, Quine too has criticized Frege’s (as he has Church’s) notion
that expressions in indirect contexts meanF their usual sense. According to
him this account is insufficient: For example if p is a factually true sen­
tence, then it is true that f = i g ( p A g = f ) , i.e. the concept / i s identical
with that concept g of which this holds: p, and g is identical with /. The
predicates / (x)’ and 6ig(p A g = f ) (x)’ have the same meaning, but despite
that it is not true that along with ‘N ( f (a))’ ( / (a) is logically necessary’),
*N(ig (p a g = / ) (a))’ is also true, for from the latter sentence there follows
the false proposition ‘N(p)\ There are cases then, in which the exchange
of expressions with the same sense can change even the meaning of the
context.61 The way out of the difficulty by saying that the expressions / ’
and ‘ig {p a g = / ) ’ have the same meaningF, to be sure, but differ in sense,
44 CHAPTER II

is not available here, for it is not these expressions (as proper names for
concepts) that are exchanged in the context in question but the predicates
‘/ ( x ) ’ and ‘ig(pA g = / ) ( * ) ’, which stand for the same concepts and so
are alike in sense. A possible difference in the sense of the concept names
is thus irrelevant to the problem.
This objection of Quine’s again raises the question we have already
discussed above, whether there are predicates that differ in meaning but
which stand for the same concept. We have answered this question in the
negative. And Quine’s criticism is no reason to give this position up now.
The expression ‘ig(p Ag=f ) ( x) ' does not make sense in this form, for in
it a proper name ^ig{p a # = / ) ’) is placed in the position of a predicate
sign in front of the argument ‘(x)’. The correct way of writing it must read:
‘xei g( p Ag = / ) ’ (‘the object x falls under the concept i g ( p Ag = / ) ’) -
but in that case the proper name ig(p a # = / ) ’ occurs in 6N(aeig (p a g =
= / ) ) ’ for which one can assume that it has a different sense from ‘/ ’, so
that now a difficulty for the Fregean semantics no longer arises. Alterna­
tively, one must replace the proper name by a predicate, namely by the
expressionp a f ( x) (It is surely supposed to be true that f\x(ig (p a g = / )
(x)=p a f (x))) - but in that case it is obvious that the concepts / (x) and
p a f ( x) are not identical and so the predicates ‘/ (x)’ and ‘p a f (x)’ do not
have the same sense, so that the difficulty is once again avoided.
In the case of Frege himself a formula like ig(jp A g = f ) is also com­
pletely impossible because identity is defined only for objects and only
objects can be identified by description, not predicates which according to
their basic definition are always predicatively applied.

1.5. Carnap's Method o f Extensions and Intensions

To develop his basic semantic ideas in [56], Carnap proceeds from the
distinction between intension and extension. This distinction is especially
natural for predicates: We construe the intension of a predicate as its
meaning, thus for example a one-place concept as the intension of a one-
place predicate62; we construe its extension, on the other hand, as the
class of objects that fall under the concept. Thus on this view, for example,
the extension of a two-place predicate is a class of ordered pairs that stand
in the relation the predicate signifies. Two «-place predicates have exactly
the same extension, therefore, if they apply to the same ^-tuples of objects.
While the question of the identity of intension of predicates is a linguistic
 THEORIES OF MEANING 45

question, the question of the identity of their extensions is (for empirical

predicates) an empirical question. Predicates with the same intension al­
ways have the same extension, but the converse does not hold, as is shown
by the example of the predicates ‘creature with a heart’ and ‘creature with
a kidney’, if we may assume that precisely those creatures that have a
kidney also have a heart.
The distinction intension - extension is now carried over by Carnap to
sentences and proper names as well. The proposition is regarded as the
intension of a sentence, its truth-value as its extension. What is specified
as the intension of a name is the individual concept - this term remains
without any more exact explication, but perhaps we can regard the individ­
ual concept as an identifying concept, however chosen. The object it
designates is specified as the extension of a name. It is plain that
this distinction corresponds to the Fregean distinction between sense
and meaningF. There are two differences between Carnap and Frege,
however:
(1) In the first place, Carnap misses in Frege a more precise criterion
for identity of sense and he himself offers two explications of the concept.
The first concept is that of identity o f intension. In this connection inten­
sion is understood in an essentially broader way than Frege’s sense.
Namely, it is the case for Carnap that two predicates have the same in­
tension when it is logically provable that they apply to precisely the same
arguments, that two names have the same intension when the identity of
the objects they designate is logically provable and that two sentences have
the same intension when their equivalence is logically provable. For the
sake of brevity, we will, like Carnap, speak of the logical equivalence (in
short, L-equi valence) of expressions in all three of these cases. Carnap
thus defines identity of intension by way of L-equivalence.63
A sharper concept of identity of sense than that of identity of intension
is that of intensional isomorphism, which we will discuss below.
(2) In order not to have to admit as Frege does that the same expres­
sions have different meanings (or intensions and extensions) in different
contexts, Carnap furthermore abandons the general principle of substitu­
tion, according to which substitution of expressions with the same mean­
ing in any context you please is possible salva veritate. Instead Carnap
defines extensional and intensional contexts as follows:
He calls an occurrence of an expression in a sentence extensional (inten-
46 CHAPTER II

sional) (or says that this expression stands in an extensional (intensional)

context) if substituting for it an occurrence of any other arbitrarily
chosen expression with the same extension (intension) does not change
the extension (intension) of the sentence (and if it is not extensional).64
So far as its content is concerned, this procedure comes out the same
as Frege’s: There are contexts in which reference is made essentially
to the sense of linguistic expressions (in which what is being spoken of is
not their meaningF but their sense), and so in these contexts expressions
with the same meaningF can not be exchanged salva veritate.
The substitutability of expressions with the same extension salva veri­
tate is restricted to extensional contexts, then, and the substitutability of
expressions with the same intension salva intensione is restricted to inten­
sional contexts.
Now there are, however, also non-intensional contexts, i.e. contexts in
which the unregulated substitution of intensional expressions can alter
the context’s intension, indeed not only the intension but the truth-value
as well. For example, belief sentences like ‘Hans believes that p’ are of this
kind. This sentence can be true and a sentence ‘Hans believes that q’ can
be false, although p and q are L-equivalent.
No matter in which particular way one may interpret belief sentences65,
whether in the sense of ‘Hans has the disposition to affirm the sentence
‘p ’ (or a synonymous sentence)’ or in the sense of believing a proposition-
in any case of a rational and adequate interpretation of the concept of be­
lief it can be shown that belief contexts are not intensional. One must have
a narrower concept of sense than that of intension in order to find a crite­
rion of substitution in these cases.
Carnap introduces the concept of intensional isomorphism for that
purpose66 in the following way: two sentences are intensionally isomor­
phic precisely when they are built in the same way (with the same logical
operators in the same order) out of descriptive constants of the same type
that have the same intensions. For example, the two sentences ‘F(tf) a
a “ i G(by and ‘H(c ) a ~"i/(e)’ are intensionally isomorphic if the sentences
‘ f\x(F(x)==H(x))\ ‘ A x(G (x)= I(x))\ ‘a= c’ and ‘e —V are logically true.
On the other hand it does not hold true that sentences that are produced
from each other by logical transformations are intensionally isomorphic or
that, for example, the two sentences F(a, b) and F(b, a) are intensionally
isomorphic, even if the sentence A xy(F (x9y)= F (y, x)) is logically true.68
 THEORIES OF MEANING 47

Carnap applies this concept of intensional isomorphism for an analysis

of belief sentences and the paradox of analysis.69 We do not wish to go
into it any more deeply here, however, and will only remark that as an
explication of the concept of identity of sense according to Frege the
concept of intensional isomorphism is on the one hand too narrow - we
saw above that Frege regards the sentences A a B and B a A 9 for example,
which are not intensionally isomorphic, as having the same sense; and on
the other hand, it depends on the choice of basic constants whether this
concept is not too broad for the criterion of substitution in belief sentences.
For if F(x) and G(x) are two concepts in number theory, for which
f\x(F (x) = G{x)) holds true as a non-trivial mathematical principle, then
the sentences F(a) and G(a) would be intensionally isomorphic: it could
be, however, that Hans believes that F(a) but not that G(a) is true.70
With an adroit choice of basic constants, however, one will always be
able to bring it about that the concept of intensional isomorphism is not
in any case too broad.

Carnap’s most striking contribution to realistic semantics in [56] lies

in his formulation of criteria for identity of intensions which show how
intensions may be defined.
According to Carnap two expressions, e.g. two sentences A and B have
the same intensions if they are L-equivalent, i.e. if the sentence A = B
(A iff B) is logically true. A = B is logically true, if it is true in all interpre­
tations of the individual and predicate constants occurring in A and B.
This can also be expressed by saying: A^=B '\s logically true, if it is true in
all logically possible worlds. Then its truth value is independent of contin­
gent facts. This means: A and B have the same intensions, if they have the
same extensions, the same truth-value, in all possible worlds. And this
may be generalized: Two expressions have the same intensions, if they
have the same extensions in all possible worlds. But this means that we
can determine the intension of an expression A as that function which
assigns to A in every possible world the extension A has in that world.
If we know the meaning of A , and therefore its intension - synonymous
expressions always have the same intensions - we can determine the ex­
tension of A in a world i in principle (i.e. if we are provided with the neces­
sary factual information on i): the object which a proper name A desig­
nates in j, the class of objects having the property expressed by a predicate
48 CHAPTER II

A , and the truth-value of a sentence A. We can, therefore, coordinate

to A a function, assigning the extension of A in i to / for all possible
worlds i.
The inversion of this principle, that this function also uniquely deter­
mines the intension of A , defines the notion of intension in contradistinc­
tion to that of meaning. As S. Kripke, R. Montague and others have
pointed out, this Carnapian idea for the definition of intensions may also
be generalized into a definition of meanings. As we have seen, the inten-
sional identity of two expressions A and B does not imply their synonymity.
Only if A and B may be substituted salva veritate for each other in all
contexts, can we say that they are synonymous. Belief-sentences, for
instance, show, however, that such a general substitutivity does not hold
for intensionally identical expressions. But we can narrow the concept of
intension into a concept of meaning by admitting not only logically
possible worlds but also logically impossible ones. Not every logical
inconsistency is obvious; so there may be worlds that some person con­
siders possible although they are inconsistent. If a function / assigns ex­
tensions of A also to those worlds that are possible only in a weaker sense,
then we may define the meaning of A by /. A and B are then synonymous
if they have the same extensions in all weakly possible worlds.
We shall not elaborate this point. It was only mentioned to show how
meanings may be brought into intensional semantics.
The effect of Carnap’s definition of intensions may be illustrated by the
problem of ostensive proper names that was pointed out in Chapter II. 1.2.
We can now determine the sense of a proper name A , its individual con­
cept, as that function g, which assigns each world i the object designated
by A in it. If A has the form of a definite description ixF(x) its extension
g (i) in i depends on the extension of the defining predicate F in i, i.e. on
the function / which determines the extension of F in every world. From
/ we obtain g by postulating that g (i) be that object a such that g (/) is the
unit class of a in case such an object exists; otherwise g(i) is to be an
arbitrary object b (the same for all /). (If F does not fulfill the normal
condition for a description that it applies to only one object then it is
usual to interpret ixF(x) as a name foi such an object b.) The intension g
of ixF(x) is therefore determined by the intension/ of F, as we maintained
in II. 1.2. The reference of ixF(x) will generally be different in different
worlds. For ostensive proper names, however, it seems adequate to inter­
 THEORIES OF MEANING 49

pret them as standard names, i.e. to assign them the same reference in all
worlds. For with such names we designate objects as individuals, inde­
pendently of their contingent properties. There is no reason, then, to
designate Chicago by ‘Munich’ in another world. This procedure implies
that the function / which represents the intension of an ostensive proper
name A has the same value in all worlds; the reference of A , therefore,
also determines the intension of A. That is the reason to say that such
names are semantically completely characterized by their reference and
have no meaning over and independent from that. For a detailed discus­
sion of these questions see Kripke [72].

1.6. The Concept o f Truth in Realistic Semantics

We will not go into the criticism of the fundamental principles of realistic
semantics until the sections following. In concluding the presentation of
this semantics, however, reference should be made to one further point
that has on occasion been the basis of an unjust criticism of semantic
realism. The criticism in question has to do with its concept of truth.71
In semantics nowadays the predicate ‘true’ is applied for the most part
to sentences, not to propositions.72 In ordinary usage, on the contrary,
‘true’ is certainly more frequently applied to propositions. Thus for
example, we say ‘It is true that Kuno is 40 years old’ rather than ‘ “Kuno
is 40 years old” is true’. But since one can move from a concept of truth
for propositions to a concept of truth for sentences and vice versa - the
following principle does hold: A sentence is true if and only if the pro­
position is true that it expresses - this distinction does not play any im­
portant role.
The correspondence theory describes the concept of truth in realistic
semantics. This concept has been formulated for modern logic by Alfred
Tarski in [35], by the following truth convention: (K) A sentence ‘. ...’ is
true if and only i f .... On this account, then, the sentence ‘It was raining
in New York at 10:15 local time on February 4, I960’ is true if and
only if it was raining in New York at 10:15 local time on February 4,1960.
This concept of truth is derived from Plato and Aristotle and in the
philosophical tradition it was usually formulated in the following way, or
something similar: ‘A sentence is true if and only if the proposition it
expresses agrees with reality’. The abbreviated formula reads: veritas est
adequatio intellectus ad rem. This formulation is supposed to give an
50 CHAPTER II

answer to the question of the content of the concept of truth. In contrast

to the meaning relationship, one cannot stipulate the concept of truth in
a purely conventional way: If the meanings of the terms of a sentence are
fixed, then the question of its truth is a question of fact, not any matter of
stipulation.
The traditional definition of truth has been legitimately criticized. For
what does ‘agreement’ mean in this case? When does an abstract, con­
ceptual structure like a proposition agree with concrete reality, or a sen­
tence like ‘Emil is happy’, as a sequence of written signs, with Emil’s
happiness? To explain the concept of truth by the concept of agreement is
to explain one obscure concept by another one that is still more obscure.73
On the basis of this difficulty and other inadequacies in the correspon­
dence concept,74 not always understood quite correctly, to be sure, at­
tempts have been made to replace this concept with others. For example,
Franz Brentano wishes to replace the correspondence concept with a
concept o f evidence, by stipulating:
A judgment (sentence) is true if it is made with evidence or if anyone
who made an evident judgment about the same object would accept that
judgment.75
But apart from the fact that the truth-conditions for the subjunctive
conditional occurring (necessarily) in the second part of the definition are
not determined,76 there is still no absolute evidence from which the truth
of a sentence would follow with rigor. And so Brentano’s definition gives
an answer rather to the question of when we are justified in making a truth
claim - namely when we have good reasons, i.e. ‘evidence’ - than to the
question of truth.
Further, an attempt has been made to replace the correspondence
theory with a coherence theory of truth, according to which the truth of
sentences is not defined in terms of their agreement with reality, but in
terms of their (logical) compatibility with each other. But what that
amounts to is the acceptance after the manner of the convention theory of a
contradiction-free set of sentences as true by convention and then
defining the truth of other sentences in terms of their compatibility with
the sentences already accepted as true. Such a theory was formerly ad­
vanced by Carnap and the Vienna Circle, for example.77 But we do not
call a sentence true because we accept it by convention, but we accept it
because, on the basis of observation, for example, we hold it to be true.
 THEORIES OF MEANING 51

The convention theory is thus hard to reconcile with the empirical proce­
dure of the sciences.
Another objection to the correspondence theory, which also concerns
formula (K), has been advanced by P. F. Strawson in [49] and [50].
According to Strawson the predicate ‘true’ is not a descriptive predicate,
but serves an expressive purpose in ordinary discourse, namely the con­
firmation of one’s own or someone else’s statements, as in ‘What I said is
true’ or ‘What Fritz said is true’, for example. In such sentences I am not
talking about sentences but confirming them. In its descriptive use, on the
other hand, the sentence ‘The sentence A is true’ says nothing more than
the sentence A itself, i.e. adding the word ‘true’ says nothing, the assertive
power of the sentence lies not in the addition of the word ‘true’, but, as
Frege says, in the form of the sentence itself.
Even if there are other non-descriptive uses of the word ‘true’ besides
its descriptive use, it does also have an important descriptive function, so
that by no means do we lack an explanandum for convention (K) in ordin­
ary language, as Strawson says. For we divide sentences into true and
false and we say, for example, ‘If A and B are true, then A a B is also true’,
‘All logical consequences of true sentences are true’ and so on. In the
semantic characterization of sentences the word ‘true’ actually plays an
important role, then, and in this role the word is purely descriptive.78
The difficulties in the traditional formulation of the correspondence
theory that have been cited are best evaded by asking about the meaning,
or the use of the predicate ‘true’ instead of about the nature of truth - as
if truth were a pre-existing concept we could describe by virtue of a pla-
tonistic talent for observation. For just as the question of the essence
of redness leads to difficulties, although we are familiar with the sense of
the word ‘red’ and know how to use it correctly, the question of the essence
of truth is only a misleading formulation of the question of the use of the
predicate ‘true’. If we ask about use, we can say along the lines of the
correspondence theory: A sentence ‘^4’ is called true if and only if reality
is just as the proposition represents it. But reality is just as the proposition
A represents it if and only if A . That is, we can say ‘A sentence 6A ’ is true
if and only if A \ The use of the word ‘true’ is so specified and this conven­
tion is simply what remains of the correspondence theory of truth as it
was formulated by Tarski.79
The semantic realist may now feel his real intentions betrayed by these
52 CHAPTER II

reductions, as he wanted to understand truth as a relation between

sentences and reality. But this relation is not some third thing that ties
reality and the sentence together, rather declarative sentences are
used to make true statements about reality. The reference to reality in
question lies in the sentences themselves, then. To that extent the sentence
‘A is true’ says no more than the sentence A itself, and the concept of
truth is reduced to convention (K), to characterizing the truth of A by A
itself. There is no general, independent meaning of the predicate ‘true’.
What it means that A is true can only be explained in terms of A or in
terms of conditions equivalent to A that vary from one sentence to another.
Likewise there is no general criterion of truth, but the testing of sentences
looks quite different from case to case. In order to determine whether it is
raining, one must look out the window, to determine whether a material
contains iron, one must make a chemical analysis, and in order to deter­
mine whether there is an infinite number of prime numbers, one must
carry out a proof. The criteria of truth for sentence A are nothing but the
criteria for the use of A .80
Therefore the general problem of truth that may originally have been
intended does not even exist.
The situation is similar with the relation of concept and object. That a
concept applies to an object does not consist in the fact that a relation
holds between the two, such that the one satisfies the other, a relation that
could be defined generally in and of itself. On the contrary, the concept is
itself predicative, the concept itself applies to an object or does not apply.
Therefore the question of the nature of the satisfaction relation also
reduces to the convention: Object a satisfies concept F if and only if
F(a) 81
NOTES

1 The boundary between descriptive meaning and other modes o f meaning will be
stated more precisely in II.4.5.
2 Along with many other theories o f meaning that could not arouse any general
interest, we shall not take up in what follows the so-called verification theory, which is
expressed in Moritz Schlick’s formula: “The meaning of a proposition is the method o f
its verification” ([36], p. 148), although it has played a certain part in discussions falling
within the philosophy of language. The reason is that this theory belongs in the context
o f a discussion o f empirical criteria o f meaning, and that belongs to the methodology o f
the empirical sciences. See Kutschera [72], 3.4.
3 See also the presentation of naturalism in Lyons [69], p. 4ff.
4 See also Grebe [66], p. 419ff.
 THEORIES OF MEANING 53

5 Cratylus 423b-e. (Citations follow the edition by H. Stephanus, Paris 1578.)

6 426c-427b.
7 See 422d-e. - In 393b-421c Plato gives a series of etymological analyses that are
almost without exception fully untenable.
8 See 435c.
9 See e.g. 399a.
10 435a-b.
11 432d. Thus the German ‘kikeriki’, the English ‘cock-a-doodle-doo’, and the French
‘coquerico’ are different onomatopoetic formations with the same representational
content. See Grebe [66], p. 419.
12 See 385b-d.
13 For this reason the predicate ‘true’ is often applied not to sentences but to the mean­
ings of sentences, propositions, as the Stoics did, for example.
14 The formulation of conventionalism in the Cratylus is given by Hermogenes in
384c-d: “ . . . I can not convince myself that there is any principle of correctness in
names other than convention and agreement. Any name which you give, in my opinion,
is the right one, and if you change that and give another, the new name is as correct as
the old - we frequently change the names of our slaves, and the newly imposed name is
as good as the old. For there is no name given to anything by nature (cpuosi); all is
convention (vopog) and habit (e0og) of the users.” [Benjamin Jowett’s translation.]
15 Indeed it is generally the case that not all non-linguistic signs are equally suited for
all purposes. Thus the sign because it suggests a direction, is better suited to serve
as a signpost than the sign ‘0’ is, etc.
16 See 435a-b.
17 390d-e.
18 See 388d.
19 See 368d-e.
20 See Chapter II.4 and Chapter 4.
21 Since the philosopher (SiaX-EKtiKog) is the one who correctly comprehends the
essence of things, it is also his task to lay down the correct linguistic norms, i.e. he must
function as linguistic lawgiver (vopoOstrig, ovopatoupyog) (390c-d).
22 Thus this comparison of words with tools (opyavov a pa xi s o n Kai ovopa 388a),
which later became fundamental in Wittgenscein’s philosophy of language, appears as
early as the Cratylus.
23 Dialogus, in : Die philosophischen Schriften von G. W. Leibniz, edited by C. J.
Gerhardt, VII, Berlin 1890, p. 192. The passage cited reads: “Nam etsi characteres
sint arbitrarii, eorum tamen usus et connexio habet quiddam quod non est arbitrarium,
scilicet proportionem quandam inter characteres et res, et diversorum characterum
easdem res exprimentium relationes inter se. Et haec proportio sive relatio est funda-
mentum veritatis.”
24 On realistic semantics’ concept of truth see also Chapter II. 1.6.
25 The meaning of this sentence depends on the time at which it is uttered. But accord­
ing to (4) - and this is our argument - it would not depend only on that.
26 G. Frege and E. Husserl especially reacted strongly against psychologism in seman­
tics (and in logic), as it dominated the last half of the 19th century. See Frege [18],
Part I, and Husserl [00], as well as the correspondence between Frege and Husserl.
27 Naturally that is also true of proper names for abstract objects such as numbers,
classes and the like.
28 See Chapter 11.4.4.
54 CHAPTER II

29 On the problem of universals see Stegmiiller [56], for example.

30 The example comes from Frege, see [92].
31 Quine [48], p. 9.
32 The distinction between meaning and reference is just as old as semantics itself.
According to Aristotle (De interpretatione) linguistic expressions mean (in a conven­
tional manner) ideas or concepts and not just objects.
And Sextus Empiricus says in Adversus mathematicos VIII, 11; “The adherents to the
Stoa say that the following three things belong together: What is meant, what is sup­
posed to mean, and the thing. That which has the function of meaning is supposed to be
the sound (in language) itself, e.g. ‘D ion’. What is meant is the proposition, made
intelligible through the words and which we comprehend because it is in our mind, but
which the barbarians do not understand even though they do hear the spoken sound.
The thing itself is what exists outside (of our consciousness), e.g. Dion himself. Among
the items named here, there are supposed to be two of a corporeal nature, namely the
sound and the thing, and one of an incorporeal, namely what is meant, the lekton,
which also possesses the property o f truth or falsity.”
33 In the German edition, the reference is to the way in which the distinction is drawn
in English, and Quine [64c] is cited.
34 This view is not shared by a large number of authors, such as e.g. A. Church [51a]
and W. V. Quine [64b]. We shall come back to this point in the context of Chapter
11.1.4.
35 Frege made this terminological distinction in [92].
36 What was said in note 34 is correspondingly true here.
37 How one chooses to proceed here is also a matter o f expediency. - In logical atomism
(see e.g. Russell [18]), in order to assure a reference for all proper names and a truth-
value to all sentences, an ideal language was postulated in which proper names stand
only for unanalyzable atomic objects. See also Chapter III.2.1.6.
38 That holds true only for so-called direct or extensional contexts. Generally speaking,
the statement is true only if supplementary assumptions as to the reference of words in
oblique contexts are made, as Frege did. See Chapter II. 1.4.
39 On the theory of definite descriptions see e.g. Carnap [56], §§ 7, 8, Linsky [67],
Searle [58] as well as Kutschera [71], 13.2.
40 See e.g. Kutschera [67], 5.3.
41 See also what Wittgenstein says in [53], 79.
42 Quine, taking this line in [48], recommended that all proper names be replaced by
definite descriptions. This procedure - if agreement as to the descriptive predicates
could be reached in translating proper names into definite descriptions - would be
logically correct. But it is another question whether ostensive proper names are not
practically indispensable nevertheless. For proper names give sentences a direct and
simple reference, while the reference of sentences without proper names is significantly
more complex. That comes out in the fact, for example, that a sentence like ‘Hans is
blond’ can be decided by direct observation, by checking only one object, whereas the
sentence ‘There is one and only one thing with such and such a nature and that thing is
blond’ would require examination o f the whole universe in order to determine its truth-
value, and as a combined universal and particular proposition would not be definitively
decidable. Furthermore, it is extraordinarily difficult to cite identificatory predicates
that do not contain ostensive proper names (e.g. names o f spatial and temporal loca­
tions) themselves. - Ayer takes this line in [63], p. 150.
43 G. Ryle in [66] calls attention to the fact that ostensive proper names are not entered
 THEORIES OF MEANING 55

in dictionaries, in which meanings of words are given, and that there are no transla­
tions for such names. That too shows: their primary function lies in their reference, not
in their meaning. One does not ask ‘What does “Salvador Dali” mean?’, but ‘Who is
Salvador Dali?’ On the meaning of proper names see also Strawson [50a] and Searle
158].
44 Wittgenstein [22].
45 See also the clear and brief presentation in Stegmiiller [65], pp. 525-561, which
follows Stenius.
46 There is no lack, among the many interpretations of the Tractatus, of such as
interpret the semantics of the work, despite the large number of Wittgenstein’s un­
ambiguously realistic formulations (which are then dismissed as ‘metaphorical’ or
‘non-essential’), as non-realistic. See e.g. Schwyzer [62], Shwayder [60] and [63], and
Lorenz [70], Chapter I, 3. So far, however, these interpretations have not been sup­
ported by sufficiently convincing arguments.
47 In his remarks Wittgenstein always presupposes an ideal language along the lines
of the symbolic language of modern logic; correspondingly simple assignments cannot
be given for ordinary languages.
48 Instead of this Wittgenstein could also have said: A sentence is a picture of a state
of affairs [Sachverhalt] if the isomorphy condition is satisfied, and it is true if the state
o f affairs of which it is a picture is a fact. It may be that Wittgenstein did not take this
course because he wanted to admit only pictures o f ‘existing’ entities (facts, not states of
affairs), and only facts as pictures of facts. - Wittgenstein also speaks of a sentence
presenting a state of affairs but picturing a fact.
49 For structures this coordination is expressed in the fact that they are taken to be
ordered ^-tuples, in which the zth relation in the first structure corresponds to the /th
relation in the second.
50 From the differentiation between the function of language as representing and as
communicating or showing, which is based upon naming, Wittgenstein derives some
quite drastic consequences for epistemology. We shall come back to them in Chapter
IV.4. This epistemological thematic is the central concern o f the Tractatus', the semantic
theory on the other hand serves only to provide a foundation for the epistemological
theses. Thus Wittgenstein writes to Bertrand Russell on August 19, 1919: “N ow I’m
afraid you haven’t really got hold of my main contention to which the whole business
of logical propositions is only corollary. The main point is the theory of what can be
expressed by propositions, i.e. by language (and, which comes to the same thing, what
can be thought) and what cannot be expressed by propositions, but only shown: which
I believe is the cardinal problem of philosophy.” (Cited in Griffin [64], p. 18.)
51 In this point Wittgenstein is perhaps already preparing his later notion, which we
shall go into in II.4.5, that sentences and predicates are meaningful not as objects, but as
speech acts.
52 Reprinted in Frege [69]. - On what follows see also Ch. Thiel’s presentation in
[65].
53 See Frege [92], p. 27, Note 2.
54 See Frege [92], p. 29ff.
55 Frege knew only the categorial distinction, concept (or more generally: function)
and object. For him propositions (thoughts) are objects. But now the property of being
true is not a genuine property, for the sentences ‘It is raining’ and ‘ “It is raining” is
true’ have the same content. The declarative character of sentences does not lie in the
addition of ‘is true’, but in the form of declarative sentences (see Frege [92], p. 34). So
56 CHAPTER II

all that remains is to regard truth values as objects. - What lies behind this rather
inadequate solution is just Frege’s antecedent decision to interpret concepts as functions
with the values ‘true’ and ‘false’, which is possible only if truth values are regarded as
objects. This artificial simplification permits him to unify his symbolism in an important
respect even if it does not contribute to its perspicuity.
56 See Frege [92], p. 32f.
57 Frege [92], p. 37.
58 See Schilpp [42], p. 660-667
59 For discussion of Moore’s paradox, see also Carnap [56], p. 63, and Langford [42].
69 See Church [43a], [43b], [51a], [51b].
61 See Quine [64b].
62 Carnap is more inclined to the locution which has it that the intension of a predicate
is an attribute, see [56], p. 21, but this way of putting it surely has to do above all with
the drawing of a line between concepts and ideas in Frege’s sense, i.e. subjective mental
data.
63 Frege formulates a corresponding criterion for predicates of which the substitution
instances are not (not all, or all not - that remains open) logically determined in the
fragment cited, ‘Ausfiihrungen uber Sinn und Bedeutung\ Such predicates are supposed
to have precisely the same meaning if and only if they have identity of intension in
Carnap’s sense. In [18], Part III (pp. 39,44,48) Frege gives many L-equivalent proposi­
tions an identical meaning, e.g. the propositions A AB and B ^ A , A and ~i ~i A, A^>B
and ~iA, and so on. But it is very much in question whether this criterion is not
too broad for Frege’s purposes (validity of the general substitution principle): For
belief sentences a more narrowly defined identity of meaning for predicates and
sentences perhaps has to be adopted. - In bis article Satz und Tatsache, reprinted in
Patzig [70], G. Patzig discusses whether facts can be conceived of as Carnapian inten­
sions of true sentences.
64 The case of non-extensional contexts, which on substitution of expressions with the
same intension do change their intension but not their extension, is not expressly
characterized. The following survey may clarify the terminology.
(1) If substitution of expressions with the same extension leads to contexts with the
same extension, then we have an extensional occurrence.
(2) If substitution of expressions with the same extension does not lead to contexts
with the same extension, then their occurrence is non-extensional.
These non-extensional contexts subdivide as follows:

(2a) If L-equivalent substitution leads to L-equivalent contexts, then that is

an intensional occurrence
(2b) If an L-equivalent substitution does not lead to L-equivalent contexts,
then it is a non-intensional occurrence.

These non-intensional occurrences can be further subdivided - as noted, Carnap does

not do this -

(2ba) An L-equivalent substitution leads to contexts with the same extension.

(2bb) An L-equivalent substitution leads to contexts that do not have the same
extension.

We can illustrate this with the following figure:

 THEORIES OF MEANING 57

(2b)
(2a) non - intensional occurrence

intensional occurrence 2ba 2 bb

extensional occurrence non-extensional occurrence

( D (2)

65 See Carnap in [56], pp. 53-55.

66 See Carnap [56], p. 56ff.
67 This definition presupposes an artificial language constructed according to strict
formation rules.
68 Cp. C. I. Lewis’s similar concept of synonymity in [43], discussed in Carnap [56],
p. 60ff.
69 See Carnap [56], 15, especially p. 61f. Church’s critique of these analyses is to be
found in [50], [54], Carnap’s reply to them in [56], p. 230fF.
70 See Mates [50], Scheffler [55], Putnam [54], Pap [55] and [57], Church [54],
Davidson [63], Linsky [49].
71 On what follows see Stegmiiller [57], Chapter XII also.
72 At any rate, in the case of states of affairs (Sachverhalten) we do not speak of ‘true’
and ‘false’, but of ‘subsistent’ and ‘non-subsistent’ states of affairs.
73 This difficulty is also illustrated by a notation in Wittgenstein’s journal for Oct. 27,
1914: “The difficulty faced by my theory o f logical picturing was that o f finding a connec­
tion between the marks on the paper and a state o f affairs out there in the world. I always
said truth is a relationship between the sentence (Satz) and the state of affairs
(Sachverhalt), but was never able to make out such a relationship.” ([61], p. 19f.)
74 Franz Brentano criticizes it, contending that:
(a) mathematical or logical propositions (Satze), that do not say anything about
real, but only about ideal structures, would not be true according to this definition (but
that holds only if one is not a Platonist and takes the concept ‘reality’ too narrowly in
the sense of ‘physical reality’),
(b) true sentences to the effect that a certain object a does not exist could agree only
with non-existent things, and consequently there could be no talk of agreeing or corre­
sponding with reality in this case (but propositions are not supposed to correspond to
things, but rather to states of affairs, and it is a real state of affairs that there is no a),
(c) on this definition, establishing the truth of a proposition would lead to an
infinite regress, since in order to establish that a proposition A is in accord with a state
o f affairs B, one would have to have already formulated a true judgment about B (but
the correspondence theory is supposed to characterize the concept of truth, not to
provide a criterion of the truth of propositions).
75 See Brentano [30].
76 Cf. however D . Lewis [73].
77 See Carnap [32], Hempel [34] and Neurath [32]. - We will come back again to the
role of conventions in defining true sentences in another context in Chapter II.4.4.
78 Stegmuller also expresses himself along these lines in [57], p. 225ff.
79 In the formulation of the truth convention as ‘“/4” is true if and only if A \ ‘A ’ is to
58 CHAPTER II

be read as a quasi-quotation, i.e. much in the same sense as ‘the mention of A \ (See
Kutschera [67], I.3.I.I.)
80 Kant, too, emphasizes that there is no general criterion o f truth in the Critique o f
Pure Reason, B 83, where he says:
“If truth consists in the agreement of knowledge with its object, that object must
thereby be distinguished from other objects; for knowledge is false, if it does not agree
with the object to which it is related, even although it contains something which may be
valid of other objects. Now a general criterion of truth must be such as would be valid
in each and every instance of knowledge, however their objects may vary. It is obvious
however that such a criterion [being general] cannot take account of the [varying]
content of knowledge (relation to its [specific] object). But since truth concerns just this
very content, it is quite impossible, and indeed absurd, to ask for a general test of the
truth o f such content. A sufficient and at the same time general criterion of truth
cannot possibly be given. Since we have already entitled the content of knowledge its
matter, we must be prepared to recognize that of the truth of knowledge, so far as its
matter is concerned, no general criterion can be demanded. Such a criterion would by
its very nature be self-contradictory.” [Norman Kemp Smith’s translation.]
81 The satisfaction relation was introduced by Tarski in [35]. On the predicative nature
o f concepts see also Frege [92b].
 THEORIES OF MEANING 59

2. B e h a v i o r i s t i c t h e o r i e s of m e a n i n g

We discussed the realistic semiotic first above because historically it

represents the oldest type of theory of meaning and is at the same time the
reference point from which all other theories set out. They have all been
built up from a criticism of the realistic account, which at first glance
seems so natural and convincing.
There are two points in particular in which all of the semiotic theories
discussed in what follows basically differ from realism. Realistic semantics
starts out, as we have seen above, from the idea that the meaning of lin­
guistic expressions consists in a representational relationship to things,
established by convention and capable of being considered in abstraction
both from the expression’s relation to speaker and hearer and its use on
particular concrete occasions. This isolation of language from its context
of use and this rendering absolute of its representational function is cut to
fit the descriptive meaning of declarative discourse above all, which is the
almost exclusive subject of inquiry in realistic semantics.
The semiotic theories referred to in the following - if one wanted to have
a single general name for these very diversified theories, one could contrast
them to realistic theories as pragmatic theories, with reference to the
philosophical direction of pragmatism, which was founded by Charles
Peirce, William James and John Dewey and was understood by Peirce to
be a semiotic theory to begin with1 - start out, as against realism, from
the following two ideas:
(1) Speech is a form of human behavior that has its place in the general
course of human life. Therefore its function is always to be analyzed with
this context as a background. The realistic account, according to which
the semantic function, the meaning, of a linguistic expression can be
defined independently of the context of its use, is therefore false. Language
is used in the course of various activities, for various purposes, in various
situations. Therefore one can not ascribe to language just one semantic
function, but there are as many semantic functions as there are contexts
of activity in which language is used. Semantics can consequently
not be pursued independently of pragmatics. As against the realistic
account, which is directed from the beginning at declarative discourse
alone, pragmatic semiotic thus reveals anew the variety of types of
discourse and wishes to undertake a more comprehensive analysis of lin­
60 CHAPTER II

guistic behavior, one which brings in the particulars of the context of

utterance.
(2) Abstract entities, such as concepts and propositions and the mean­
ings of linguistic expressions in the realistic account generally, are no more
available to scientific, intersubjective, empirical observation than ideas
and intentions in the mind, which represent meanings in psychological
semiotic theories. The only thing available to such observation and hence
the only thing a scientific study of language has to go on is language use.
Meaning has to be determined by it. On this view, whenever two expres­
sions systematically differ in their use, one will also have to make out a
difference in meaning between them, just as conversely differences in
meaning can only be made out where there are differences in use. There is,
then, a one-to-one relationship between meaning and language use and
it requires but a small stroke of Occam’s razor to give up the assumption
of abstract meanings as entities entirely and to identify the meaning of a
linguistic expression with its use. Pointing in this direction, Peirce, the foun­
der of pragmatism, had already said “ ... there is no distinction of meaning
so fine as to consist in anyting but a possible difference of practice.” 2
Behavioristic theories of language constitute an initial group within
what we have called pragmatic theories of meaning. In them the identifi­
cation of language use and meaning and the methodological viewpoint
that language theory has to start from intersubjectively observable lin­
guistic phenomena is fulfilled in a very radical way. The phenomenon of
speech is understood as overt conduct in the behavioristic sense, and
language use as linguistic behavior, which is then described in terms of
the classical behavioral concepts, stimulus, response, etc.
In the following we shall present behaviorism in terms of the ideas of
two of its principal advocates, Charles Morris and Burrhus F. Skinner.

2.1. Charles Morris

We will first go into semiotic as Charles Morris developed it in his book,
Signs, Language and Behavior (1946). Morris bases his theory on a behav­
ioristic conceptual apparatus in which behavior is described in the stimu­
lus-response scheme.

(a) Signs. Morris starts out from a general concept of sign, which is not
restricted to linguistic signs, but according to which the sound of a bell
 THEORIES OF MEANING 61

that regularly precedes a dog’s feeding is for the dog a sign of food. He
refers to older behavioristic definitions according to which a sign for a is
a substitute stimulus that evokes the same reaction a would evoke if a
were present. The sound of the bell is a standard example of such a sub­
stitute stimulus for the food, which evokes the same response as food
itself does (Pavlovian response).
This definition, however, is not usable, since linguistic expressions or
signs do not have to evoke immediate reactions. (A sign that an Alpine
pass I do not wish to travel over is closed does not interest me and so
evokes no response in me, but if I do want to travel by it, the response,
instead of being immediate, does not come into play until the next cross­
roads, where I can turn off in the direction of another pass.) Further­
more, a word like “boss” does not evoke the same response as the boss
himself in the flesh, but may evoke quite different responses (one might
curse him out, something one would refrain from doing in his presence).
In view of these flaws in the definition, Morris developed a more com­
plicated account along the lines of a definition which reads somewhat as
follows:
(I) If Z is a preparatory stimulus which produces in an organism
a disposition to respond under certain circumstances with a
(goal-directed) form of behavior of type T and if a is a stimu­
lus object that evokes (goal-directed) behavior of type T, then
Z is a sign for a.3
A preparatory stimulus in this connection is a stimulus that affects
responses to other later stimuli.
Despite these modifications (preparatory stimulus, which does not
have to be sufficient to release a response all by itself, evocation of dispo­
sitions to respond instead of responses, similarity instead of identity
between the responses to the sign and what it signifies, and response not
to the sign itself but the inclusion of further stimuli from the environ­
ment) and qualifications (Morris understands condition (I) to be only a
sufficient, not a necessary condition for signs4) of the original definition
this account is not yet sufficient. For in the first place, this stipulation
does not cover a broad spectrum of linguistic signs (for example, reports
of past events, such as ‘Thutmosis I died in 1510 B.C.’, generally do not
evoke any disposition to respond and sentences the hearer does not hold
62 CHAPTER II

to be true do not evoke any disposition to respond either5). Furthermore

(I) is also too broad, as a sufficient condition, (a drug that increases
excitability and so produces a disposition to respond violently to a dero­
gatory remark (‘certain conditions’) - a similar response, then, as might
otherwise be provoked by a physical assault - we will scarcely want to
claim is a sign of a physical assault). And finally the definition’s condi­
tions are formulated very vaguely. (How is the talk of similarity of re­
sponses to be understood, for example? What are the ‘certain conditions’?
There is also no reference to the subjects in whom responses or disposi­
tions are evoked: Is it all human beings or some of them or most of them?6)
Furthermore, ‘definition’ (I) takes no account of the fact that linguistic
signs at least have meaning in a conventional way, i.e. they do not evoke
certain dispositions by virtue of their syntactic nature alone, but by virtue
of their use as established by means of conventions. Signs have to be
understood in order to produce a behavioral disposition. If someone does
not understand a linguistic sign, it will not evoke any disposition in him.
I.e., only those persons who do understand the sign in question can be
admitted as organisms in which a disposition is evoked according to (I).
A sign does not count as such without qualification but only for certain per-
sons.The concept of understanding, however, is not a behavioristic concept.
Besides that, it should be noted that under the same circumstances
different hearers can respond quite differently to the same linguistic signs.
Thus for example, X t is very happy at the news that Mrs. N. N. is dead
(he owed her money), X 2 is sad (she owed him money), the news does not
affect X 3 at all, since he did not know Mrs. N. N. If need be - although it
would be very artificial - these differences can be blamed (‘certain condi­
tions’) on existing dispositions in X u X 2, and Z 3, but Morris’ account
makes no provision for that, since he understands ‘certain conditions’ in
the sense of environmental conditions.
The example of Morris’ attempt at a definition already shows how extra­
ordinarily difficult it will be to define the concept of sign behavioristically,
since we have to do with a very general, abstract concept that does not
relate to any homogeneous type of responses or dispositions to respond.
Going on from this concept of sign Morris then defines the interpretant
of a sign as the behavioral disposition the sign evokes in the hearer, its
referent (denotatum) as an object toward which the action the sign disposes
the hearer to perform is directed, its meaning (significatum) as the condi­
 THEORIES OF MEANING 63

tions of which it is true that everything which fulfils them is a referent of

the sign. In the example of the bell as signal of the dog’s food, the sound of
the bell is the sign, the dog’s disposition to look for food in a particular
place the interpretant, the food the referent and the possibility of being
eaten in a certain place the meaning of the sign.7
Linguistic signs are then defined as signs that come into play in the
context of social behavior and have the same meaning for speaker and
hearer,8 and are limited as to the possibilities of their being combined
by syntactic rules of language, for example. A language is then a set of
such linguistic signs.9
That a linguistic sign has the same meaning for X and Y presupposes
that it evokes the same dispositions to goal-directed actions in X and Y,
for same meaning presupposes same referent (the same class of reference
objects) and same referent once again presupposes dispositions to actions
directed toward the same objects. But if X says to Y, ‘There is a hornet
on your head’, there can be no talk of like dispositions, i.e. the condition
is much too narrow for linguistic expressions. Furthermore, single words
like ‘red’ or ‘Zugspitze’ would not be linguistic signs on this view, be­
cause - outside the context of a sentence - they do not evoke any definite
behavioral dispositions to respond in and of themselves. Sentences on the
other hand do not evoke behavioral dispositions similar to those evoked
by stimulus objects, but at most like those evoked by stimulus situations.

(b) Types o f Expression. Morris next attempts to define behavioristically

certain basic types of expressions, namely identifying expressions such as
‘here’, ‘in this or that place’, ‘at this or that time’, for example, descriptive,
such as ‘blue’, ‘warm’, ‘hard’, evaluative (appraisive) such as ‘good’, ‘bad’,
and prescriptive expressions such as ‘should’. His project is already sub­
ject to the objection that words are not used either descriptively or evalua-
tively, etc., but can have different characters in different contexts and can
be both descriptive and evaluative in the same context. Now this classi­
fication is supposed to be carried out behavioristically on the basis of the
way in which these expressions signify, i.e. what sorts of disposition they
evoke in the hearer. If the disposition evoked is inclined to concentrate
responses on certain space-time locations, then we have an identificator
before us, if it is inclined to direct behavior toward things with certain
characteristics, then we have a descriptor, if the disposition is directed
64 CHAPTER II

toward preferring certain objects, we have an evaluator (appraisor), and

if the disposition is directed toward preferring certain forms of behavior,
we have a prescriptor.10
Now these definitions are obviously much too primitive, for the ex­
pression ‘on November 4, 1940’ evokes no disposition to direct responses
to November 4, 1940, in me - how would I even do that?! And if a low­
brow philistine says, ‘This picture is beautiful’ - I don’t prefer it to
another on that account. And likewise a beggar’s appeal, ‘Lend me a
hundred dollars, please’, does not yet bring it about that I carry out that
action or even have a disposition to do so.

(c) Types o f Discourse. Morris also attempted to describe the differences

among the various types of discourse in terms of his behavioristic concepts
and to give a classification of types of discourse. He chooses his arrange­
ment of classes according to modes o f signifying, i.e. according to the main
type of expression in the statement and according to its purpose (from the
speaker’s point of view) and arrives at Table II.11
In this scheme, formators are formal expressions such as logical opera­
tors, punctuation marks, etc., for example,12 and systematic use is a use
for the systematization of a body of knowledge or a set of statements.
Morris says: The systematic use of signs is a use with the intention of
systematizing the behavior evoked by other signs.13 Since these two con­
cepts both remain very vague, we will pay no attention to the last column
on the table. The types of discourse entered in the individual compart-

TABLE n

Mode of Informative Evaluative Evocative Systematic

Signifying (incitive)

Designative Scientific Story-telling Juridical Cosmological

(predominantly Language (Active)
descriptors)

Evaluative Mystical Poetic Moral Critical

(Evaluators)

Prescriptive Technological Political Religious Propagandists

(Prescriptors)

Formative Logico- Rhetorical Grammatical Metaphysical

(Formators) mathematical
 THEORIES OF MEANING 65

ments are not supposed to be defined by this classification, but merely

constitute illustrations of the classification types. This looks somewhat as
follows:
A story (fiction) is written in descriptive language, but it does not serve
the purpose of informing about facts, but of evaluation, as in a novel of
social protest, for example.
A poem is written in evaluative language and serves valuational purposes
as in Schiller’s ‘Ode to Joy’, for example.
In myth a story is told in valuational language for the purpose of infor­
mation, as in the tale of Buddha having been a hare in an earlier incarna­
tion, who offered himself to a starving monk as food and before he leaped
into the fire shook himself so that no insect in his fur would be killed.14
It is obvious that the characterization of concrete types of discourse by
way of this classification is much too rough. The classification can only be
understood as an indication of how types of discourse could be defined
from the behavioristic point of view.

In estimating the value of Morris’ book, one must take note that Morris
himself was very well aware that what he had provided was more a pro­
gram than a scientific theory. He writes: “Our account bristles with prob­
lems, it sketches a program more than it records an achievement.” 15 So
far as we have not already done so above, here too we will make no criti­
cism of the details - for the details can be modified - nor will we make any
criticism of the fact that Morris has used concepts in his analysis that do
not belong to the behavioristic vocabulary, such as ‘valuational’, ‘purpose’,
‘command’, etc.16 Instead we will go into nothing but the fundamental
features.
If Morris’ definitions were supposed to lay down the foundations of his
semiotic and to show that the basic semantic concepts can be satisfacto­
rily defined on a behavioristic basis,17 then this plan can not be described
as successful. The definitions can rather serve as illustrations of the extent
of the chasm that yawns between program and execution in that theory.
As in the case of the behavioristic definitions of psychological concepts
we are left at the beginning with general declarations of intention. Pre­
cisely by their total failure to satisfy, the definitions give an impression of
the dreadful difficulties any execution of this program encounters. Cer­
tainly a sign is for the hearer a stimulus - the sound or the mark is an
66 CHAPTER II

acoustic or optical stimulus - otherwise the sign would not be perceived.

And certainly the understanding of signs can be described as a behavioral
disposition, if it be only a disposition to give a suitable answer to the
corresponding question. But the nature, the manner of the connection
between stimulus and disposition is so complex - and the dispositions take
so many different forms - that a description of the semantic function of
linguistic expressions in terms of these concepts is, at least at this time, an
extremely - one might almost say a hopelessly - complicated enterprise. We
lack any terms in which to characterize semantic categories (proper name,
predicate, sentence), to analyze phenomena such as indirect discourse,
etc. To that extent is the behavioristic conceptual apparatus wholly in­
capable of being used to analyze concrete, even simple and fundamental
linguistic facts.
Further, it is certainly possible to study speech as concrete behavior,
to ask: How does a hearer react to being stimulated by certain speech
sounds, how does he put them to use, and so on. And it is entirely possible
that these investigations will bring interesting results to light. But the
question is whether such investigations are relevant to the problems in the
philosophy of language that concern us here.
Let us take an example: The mathematician also uses signs and one can
inquire about his behaviour in using these signs, can study which mathe­
matical signs he responds with upon which stimuli, and so on. But all of
that is wholly irrelevant to mathematics itself. For arithmetic, the asser­
tion that one often responds to the sign ‘2 + 2 ’ with the sign ‘4’ or the like
is totally insignificant. The mathematician is not interested in the behavior
of mathematicians but in mathematical concepts and operations. Behav­
ioristic analyses provide no information about them, however. The behav­
ioristic approach to language resembles the attempt to understand mathe­
matics by observing the behavior of mathematicians.
Furthermore, the behavioristic theory of language is also mistaken be­
cause it attempts to grasp the phenomenon of language as a naturalistic
phenomenon. When a behaviorist speaks of language neither he himself
nor his audience understands it as a natural process of the sort he describes:
as responses to stimuli or as the production of substitute stimuli. Under­
stood in this way, his utterances remain totally irrelevant scientifically
speaking, as much so as his yawning or clearing his throat or his imposing
mannerisms.18 In fact, the behaviorist’s speech is of quite a different sort
 THEORIES OF MEANING 67

from the language he is speaking of; but it remains beyond the horizon
of his inquiries. In terms of the philosophy of language, however, the
question as to how the language behavior of others can be described scien­
tifically is not relevant. The relevant question is: What are we doing when
we are talking about something, when we make statements with the claim
that they are true, for example?19

2.2. B. F. Skinner
In order to round out the picture of behavioristic semantics, we will
briefly comment on B. F. Skinner’s book Verbal Behavior [57], which
plays an important role in recent discussions of the philosophy of language.
Skinner, who has worked predominantly in the area of research in animal
behavior, attempts in this book to apply the conceptual apparatus of
experimental behavioral research to the phenomenon of language. He
believes that recent progress in the area of behavioral research makes the
prospects for such attempt appear to be very good.20
Skinner’s work has been extensively discussed by N. Chomsky in [59b]
and we shall refer to his excellent critique often in what follows.21
Skinner starts from the model of simple behavior experiments. One
experiment has already been considered above, that a dog learns to look
for food in a certain place (response) when he hears the sound of a bell
(stimulus). What we have to do with in this case are conditioned responses
evoked by a specific stimulus, and Morris has responses of this type in
mind when he wishes to analyze linguistic activity behavioristically. An
experiment of another type consists in placing a rat in a cage in which
there is a lever the rat can work. Now if a pellet of food falls into the cage
when the lever is pressed, the rat learns to get food by working the lever
when it is hungry. In this case it is not a matter of behavior evoked by a
stimulus, but of instrumental behavior, and Skinner as opposed to Morris
interprets language behavior as instrumental behavior of this sort. One
can go on to put a light in the cage as well and arrange things so that a
pellet falls into the cage only if the light is on. Then the rat learns to press
the lever only in case the light is on.
With experimental designs of such a simple kind the rat’s behavior
can be controlled by experimentally specifiable parameters and so can be
described as a function of those parameters. Skinner’s intention is to
analyze language behavior in a similar way as a function of observable
68 CHAPTER II

parameters. Now since language behavior is not behavior that is control­

led through simple experimental conditions, however, the question arises
as to what behavioral parameters to take into account in this area, what
counts as stimulus and what as responsive or instrumental behavior.
In general, a stimulus is supposed to be a physical object or event that
regularly produces certain responses. A response, on the other hand, is an
item of behavior that regularly follows upon certain stimuli.
Right here we encounter the first fundamental difficulty for the behav­
ioristic conceptual apparatus. If we define stimulus and response inde­
pendently of each other, and so admit as stimuli everything that can
release responses - thus practically all of the objects and events the orga­
nism can perceive - and as response every form of behavior, then it is not
possible in general, on the basis of the experimental results, to give any
functional analysis of behavior within the stimulus-response scheme,
in which behavior is determined as a function of external conditions ac­
cording to laws. The connection between the behavior and the external
conditions is in general much too complex and so far it has evaded ade­
quate description. On the other hand, if we define stimulus and response,
as above, by way of their law-governed connection, then if we explain
responses in terms of stimuli, we are in effect explaining responses by
themselves, for a stimulus is only a stimulus if the response follows it. To
put it another way: A specific, non-trivial definition of stimulus and
response can be given only for the simplest cases, in which a lawful con­
nection between stimulus and response phenomenon has already been
demonstrated or is suspected. Thus for example, in the simple experimen­
tal set-up described turning on the light can be defined as stimulus and
pressing the lever as response. But that holds only for this experiment; in
general not every case of turning on a light is a stimulus and not every case
of pressing a lever a response.
Along with the basic concepts of stimulus and response, Skinner also
uses other behavioral concepts in his book, in particular so-called operants.
These are forms of behavior for which no stimulus directly releasing them
can be observed. An example of an operant is pressing the lever when the
light is not on. If pressing the lever results in food being delivered, the
operant is thereby re-inforced. The strength of an operant can be defined in
terms of the number of actions (lever-pressings) during the period of
extinction, i.e. from the last re-inforcement to the time action stops. If
 THEORIES OF MEANING 69

pressing the lever when the light in the cage lights up always produces
food, and the rat thus learns to press the lever when the light is on, light­
ing the light then serves as a secondary re-inforcement of the operant,
pressing the lever.
Skinner now wishes to apply these and similar behavioral concepts to
language behavior. Once again the difficulty arises that these concepts
are defined only for simple experimental set-ups. What the strength of an
operant is in general, for example, remains completely open. If one wishes
to describe speech-activities, say, as operants, what then is the strength
of these operants? What is the period of extinction in this case? Skinner
generalizes the use of these terms without any more detailed explanation.
The terms still have a scientific ring to them, to be sure, but they are no
longer precise and well-defined scientific terms. Foi Chomsky has shown
in [59b] that on the basis of the narrow, well-defined use of these terms
derived from behavioral research Skinner’s analyses seize on almost no
relevant aspect of linguistic behavior, but that upon a metaphorically
generalized use they are unclear and their use signifies no scientific pro­
gress in any case.22
Going beyond that, many analyses and definitions on Skinner’s part
are simply inadequate. When, to take a Skinnerian example, a sonata is
played and a listener says ‘Mozart’, the music is supposed to be the stimu­
lus and the word ‘Mozart’ the linguistic response aroused by this stimulus.
The listener might equally well say ‘awful’, ‘beautiful’, ‘My grandmother
always liked to hear that’, and so on. In order to explain a specific response
Skinner has to assume a specific stimulus, which is different in all the
cases of responses cited. But what objective stimulus corresponds to the
last answer? It is obvious that subjective factors define the stimulus in this
case, i.e. stimuli lose their external, physical character. Besides, we can
only identify the stimulus by way of the response, and so we can not
explain or predict responses on the basis of stimuli. With that this analysis
becomes worthless.
Further, according to Skinner a proper name, for example, is a response
to a specific person or a specific thing,23 but proper names are also used
in the absence of the object they denote, indeed even when the speaker has
never seen that object. Where then is the stimulus in this case? And if
it is said that for the speaker the referent of an expression (which Skinner
identifies with its meaning) is nothing but the probability that the speaker
70 CHAPTER II

will utter the name in the presence of the relevant object,24 that too is
inadequate. The probability that a German will use the word ‘Munich’
while he is in Munich is rather less than when he is in America, for example.
Going on beyond general behavioral terms, Skinner also uses special
terms for linguistic behavior. According to him25 ‘language behavior’ is
behavior that is re-inforced through the intervention of other people. But
this definition is obviously much too broad, since otherwise even training
an ape, for example, would be language behavior.26 Besides that, the
definition is also much too vague for anything to get started with it.
Going on beyond this, there are still further specific forms of linguistic
behavior to be found in Skinner, such as ‘mands’, ‘tacts’, 27 ‘echoic’ and
‘autoclitic operants’ and the like, none of which are well-defined and which
are used again and again - so far as one can determine - in ways that
deviate from their definitions. Chomsky is right, then, in saying that Skin­
ner’s procedure is “just a kind of play-acting at science” . 28 All that is
presented is a bombastic pseudo-scientific terminology which offers
nothing for any relevant distinctions, descriptions or explanations.
Finally, K. S. Lashley has emphasized in [51] that the behavioristic
accounts of linguistic theory in no way do justice to grammatical pheno­
mena. Sentences are not sequences of words with which some stimuli
happen to be coordinated. So far, however, we lack any usable account
that interprets grammatical structures behavioristically.29 In the final
analysis, then, not a single relevant linguistic distinction can be adequately
expressed in terms of behavioristic concepts. And so in what follows we
can neglect the behaviorists’ attempts at philosophy of language with a
good conscience.
NOTES

1 ‘Pragmatism’, too, is an over-all title for very different sorts o f philosophical enter­
prises, so that we are not running too great a risk that our designation of semiotic
theories as ‘pragmatic’ will cause their authors to be regarded as pragmatists in some
special sense.
2 Ch. S. Peirce, Collected Papers V, ed. Ch. Hartshome and P. Weiss, Cambridge,
Mass. 1965, p. 257.
3 See Morris [46], p. 10.
4 See Morris [46], p. 12.
5 On this point see Stenius [67], p. 261 also.
6 The ‘definition’ (I) has the following form: *R(Z) AD(Z, X, T, B)AO(Y)A
AB(Y, T, X)^S (Z, Y)\ where \K (Z )’ stands for ‘Z is a preparatory stimulus’,
‘Z>(Z, X, T, B)9for ‘Zproduces in l a disposition to behave along the lines of T under
 THEORIES OF MEANING 71

condition B \ ‘0(F )* for ‘F is a stimulus object’, ‘B ( Y , T9X )* for ‘F produces behavior

o f type Tin X \ and ‘*S(Z, F)* for ‘Z is a sign for F*. Either that is to be understood in
such a way that there are still quantifiers to take their places in the antecedent (other­
wise free variables occur in the ‘definiens* that do not occur in the ‘definiendum’) - but
in that case the question is open as to where to put the quantifiers: different groupings
o f quantifiers lead to quite different stipulations - or these variables are actually sup­
posed to be free - then one can also write (I) as *jR(Z) A v T B X ( D ( Z , X, T, B ) ^ O ( Y ) ^
AB(Y, T9 X ) ) ^ S (Z, YY thus arriving at the absurd conclusion that any preparatory
stimulus Z is a sign for any stimulus object Y that evokes a goal-directed behavior
pattern T9 for there is for every Y a. T and a B - namely the condition that Y itself is
present - such that D ( Z 9 X, T, B ) ^ 0 ( Y ) ^ B ( Y 9 T9 X ) is true. Presumably it should
read: There is a T and a B (dependent on Y and T) such that for all X .... We base what
follows on this interpretation.
7 See Morris [46], p. 17.
8 Morris calls such signs consigns, see Morris [46], p. 33.
9 See Morris [46], p. 36.
10 See Morris [46], p. 66.
11 See Morris [46], p. 125.
12 See Morris [46], p. 87.
13 See Morris [46], p. 104.
14 Morris gives this example in [46], p. 134f.
15 Morris [46], p. 246.
16 It is quite dubious and in any case an entirely unproved thesis, up to the present, that
there will be any success in translating these terms into behavioral terms.
17 See Morris [46], p. 60.
18 See Russell [40], p. 4 also.
19 In the philosophy of language we are not interested in describing linguistic behavior
as a natural process but in understanding the meaning and purpose of speech acts as
actions. This makes for a vast difference in the conceptual apparatus involved.
20 Skinner writes ([57], p. 3): “It would be foolish to underestimate the difficulty of the
subject matter, but recent advances in the analysis o f behavior permit us to approach it
with a certain optimism. New experimental techniques and fresh formulations have
revealed a new level of order and precision. The basic processes and relations which
give verbal behavior its special characteristics are now fairly well understood. Much of
the experimental work responsible for this advance has been carried out on other
species, but the results have proved to be surprisingly free of species restrictions. Recent
works have shown that the methods can be extended to human behavior without serious
modification.” Other behavioral researchers do not share this optimism. N. Tinbergen,
for example, says: “We may now draw the conclusion, that the causation of behavior
is immensely more complex than was assumed in the generalizations of the p ast....
Second, it will be obvious that the facts at our disposal are very fragmentary indeed.”
([51], p. 74, quoted in Chomsky [59b]).
21 For criticism of Skinner see also Hormann [67], p. 214ff.
22 See Chomsky [59b], p. 31.
23 See Skinner [57], p. 113.
24 See Skinner [57], p. 115.
25 See Skinner [57], p. 2.
26 See Skinner [57], p. 108, note 11, on this point. (‘The animal and the experimenter
comprise a small but genuine verbal community*.)
72 CHAPTER II

27 Skinner gives this definition: “A ‘mand’, then, may be defined as a verbal operant in
which response is reinforced by a characteristic consequence and is therefore under the
functional control of relevant conditions of deprivation or aversive stimulations”
([57], p. 35f.). And: “A tact may be defined as a verbal operant in which a response of
given form is evoked (or at least strengthened) by a particular object or event or
property of an object or event.” ([57], p. 8If.). Mands are thus supposed to be some­
thing like imperative (evocative) expressions, tacts something like descriptive expres­
sions, but it is not possible to extract that from the ‘definitions’. They are so vague that
one can not assess them as anything but helpless appeals to the reader’s good will.
28 Chomsky [59b], p. 39.
29 Chapters 12 and 13 in Skinner [57] also illustrate the total inadequacy of the behav­
ioristic conceptual apparatus for characterizing grammatical structures.
 THEORIES OF MEANING 73

3. Q u i n e ’S p h i l o s o p h y of l a n g u a g e

Quine’s most important ideas on semiotic are contained in his books,

From a Logical Point o f View [64a], The Ways o f Paradox [66], and Onto­
logical Relativity [69a].1
Quine’s pragmatic approach is already distinct in the introduction to
his principal work in the philosophy of language [60], when he writes:
“Language is a social art. In acquiring it we have to depend entirely on
intersubjectively available cues as to what to say and when. Hence there
is no justification for collating linguistic meanings, unless in terms of
men’s dispositions to respond overtly to socially observable situations.” 2
These statements even bring Quine into the neighborhood of linguistic be­
haviorism. It may not be ignored, however, that Quine contrasts favor­
ably with the proponents of behaviorism we discussed in II.2. His ideas are
essentially more detailed, more precise and better founded. That becomes
possible because he perceptibly goes beyond the, as we saw, much too nar­
row compass of behavioristic analysis of language. Actually the behavior­
istic approach expressed in the above quotation remains with Quine only a
program which is never put into practice. Quine almost never attempts to pre­
sent semiotic distinctions in terms of purely behavioristic concepts. Indeed,
the ‘hence’ in the above quotation is a non sequitur, for the acceptance of non-
behavioristically defined meanings is justified if the function of language
and linguistic behavior can be better understood and analyzed with them.
Referring to Quine’s formulation, N. Chomsky in [68] has shown in
particular that linguistic response dispositions are very different from
individual to individual (they depend on mood, information, education,
capacity for perception, etc.), so that they are not eligible as (intersub-
jective) meanings. On account of the wealth of possible ways of formulat­
ing something in language, furthermore, the probabilities of a specific
linguistic response that define such dispositions lie close to 0 for all
expressions.3 We can therefore leave Quine’s behavioristic-programmatic
statements out of account.
In the following we will select from what Quine has said about the
philosophy of language three topics that have special systematic interest.4

3.1. Quine's Elimination o f Meanings

Quine differentiates two theories in semiotic: a theory o f reference of lin­
74 CHAPTER II

guistic expressions and a theory of meaning.5 Now while according to

Quine the theory of reference, thanks to the work of Tarski,6 is an essen­
tially sound scientific discipline, the theory of meaning suffers from the
fact that its basic concepts are unclear and its basic assumptions doubtful.
Without yet presupposing the pragmatic identification of meaning and
use, Quine wishes to show first of all that assuming meanings along the
lines of realistic semantics is not justified.
In discussing this thesis, we will restrict ourselves to the realistic
meanings of sentences, propositions, since the arguments for concepts
sound much the same.
The starting point of Quine’s claim is that (1) it is superfluous to accept
propositions, that (2) besides, an unobjectionable definition of proposi­
tions is not possible and that (3) the acceptance of propositions would
imply making meanings something absolute, for which there is no empi­
rical basis.
We will now go into these three claims in greater detail:

(1) According to Quine the argument most commonly offered in favor of

assuming sentence meanings reads somewhat as follows: We distinguish
between meaningful and meaningless sentences, and so the predicate
‘meaningful’ obviously has a well-defined content. But a sentence can be
meaningful only if there is something it means, for ‘to mean’ is a transitive
verb. If someone claims that a sentence A is meaningful he must be able
to say what A means. If there were no meanings of sentences, then there
would be no meaningful sentences either.7
Quine argues against this that we in no way have to interpret the predi­
cate ‘x is meaningful’ - symbolically S (x) - in the sense of V yB (.x, y) where
B{x, y) stands for ‘jc means y \ Instead, we can regard this predicate as a
basic predicate. Indeed there are many linguistic expressions whose
grammatical usage does not correspond to their logico-semantic func­
tion.8 Besides that, we do not answer the question of the meaning of a
sentence by indicating a proposition, but by giving a synonymous sentence
(which is formulated in another language or in simpler or more precise
phrasing, etc.) And so we do not need to assume any meanings of sen­
tences; it is sufficient to make use of the two concepts ‘meaningful’ and
‘synonymous’ - in symbols G(x, y).9
The significance of Quine’s proposal lies in the fact that in this case the
 THEORIES OF MEANING 75

realm of dubious propositions is eliminated. The range of application of

the two basic predicates *S(a:) and G(x,y) is the set of sentences, thus
concrete linguistic expressions, while the range of y in B(x, y) was the
totality of all propositions.
While S(x) and G(x, y) can be defined in terms of B ( x , y ) (S(x): =
\JyB(x,y) and G(x, y):= \Jz(B(x, z) a B(y, z)), on account of the
difference in the ranges over which they are defined, B(x, y) cannot be
defined in terms of S(x) and G(x, y). Classes of synonymous sentences are
co-ordinated with the meanings of sentences in the realistic sense, how­
ever, and therefore it is now possible to regard or represent the meanings
of sentences as classes of this sort and to define a relation i?*(x, y) corre­
sponding to B(x, y) by way of i?* (x, y) : = y = XzG(x, z).
On this reconstruction of the meanings of sentences, they no longer
appear to be independent of language, as on the realistic interpretation,
but appear definable only with the aid of linguistic expressions, as on
the pragmatic interpretation.
To be sure, the problem of the meaning of sentences is not eliminated
without anything left over, for the question remains as to how the relation
of synonymity, which we usually understand as identity of meaning, is to
be defined as a basic relation intuitively without recourse to meanings.
But on the pragmatic interpretation this question can be answered by
saying that sentences are synonymous if they are used in the same way
(for the same purposes, in the same situations), and are thus exchangeable
for each other in all contexts. Of course that is only a preliminary answer,
since the concept of the use of linguistic expressions still requires much
clarification itself, but we will not go into that until Chapter II.4.3.

Indirect contexts, such as 6X says (wishes, claims, knows, believes,

desires, etc.) th a t...’ and 6X asks (doubts, inquires, etc.) whether...’ con­
stitute a further argument for accepting propositions. These contexts are
interpreted in realistic semantics, as we saw, in such a way that what we
have before us in these cases is a relation between a person and a proposi­
tion, for obviously nothing is said about sentences, about utterances of
X in them. If X says that it is raining, it does not follow therefrom that
X says ‘It is raining’. He can also say, for example, ‘Es regnet.’
But these contexts can also be interpreted in such a way that they do
make an assertion about a sentence which is invariant in truth-value (as in
76 CHAPTER II

content) upon the replacement of this sentence by any other synonymous

sentence.10 In order to make this explicit and clear, in the sentence
‘Fritz says that he was in Munich’ the predicate ‘saying th a t...’ can be
replaced by a predicate ‘saying*’..., which is then explicated as follows:
a says* 6A 9 if and only if x says that A .11 Then it is true of ‘saying*’, as
opposed to ‘saying’, that: A is synonymous with i?=>(x says* A=>x says*
B).12 On this interpretation the occurrence of such intensional contexts
does not force us to accept propositions - at least not propositions which
can not be defined as classes of synonymous sentences.
In a similar vein, Quine has proposed reconstructions of such contexts
according to which they are not assertions about sentences.13 He starts
with the attempt to replace the sentence (1) ‘Galileo said that the earth
moved’ with the sentence (2) ‘Galileo said, “The earth moves’” . That is
obviously inadequate, since Galileo did not speak English, so that (1) is
true and (2) false. Therefore Quine reconstructs (1) by means of sentence
(3) ‘Galileo said in Italian, “The earth moves’” . The predicate ‘saying in
Italian’ is on this account a new basic predicate, which can be explicated
as follows: ‘x says y in Italian’ means the same as ‘jc utters an Italian
sentence which is synonymous w ith /.
The like direction is taken by a proposal of Scheffler’s in [54], which
translates (1) as (4) ‘Galileo made a “The earth moves” - utterance’. Here
the expression ‘ “The earth moves” - utterance’ can be defined as the desig­
nation of an utterance synonymous with the sentence ‘The earth moves’.
It is crucial to both proposals that ‘synonymous’ be regarded as a well-
defined predicate which is not clarified in terms of ‘meaning the same
proposition’.14
The reconstruction of wish and belief sentences in Quine is less con­
vincing, however, since in these cases statements about wishes, etc., do
not come into play. Consequently, the general proposal offered above
for the understanding of such contexts taken altogether appears to be the
most acceptable one. Even it is of only limited interest, however, since
Quine’s rejection of propositions, as we shall yet see, proves to be un­
founded.

(2) We now come to Quine’s second and stronger thesis, that an unob­
jectionable definition of propositions is impossible.
A minimal requirement for a definition of propositions is that it must
 THEORIES OF MEANING 77

include a criterion for the identity of propositions. That is, a criterion for
the identity of sentence meanings, i.e. for the synonymity of sentences
must be given. Now Quine maintains that a sufficiently sharp classifica-
tory concept of synonymity, according to which two propositions are
either synonymous or not, can not be given.
Quine’s claim is directed not only against the realistic assumption of
autonomous propositions independent of language, but also against the
usefulness of the pragmatic reconstruction of sentence meanings as classes
of synonymous sentences, and against the scientific use of any talk of
sentence meanings and synonymy whatever.
To begin with, according to Quine the everyday pre-scientific concept
of synonymity can not be used as a scientific concept, for it is very im­
precise and a decision as to the synonymity of two sentences can be justi­
fied only in very simple cases. Furthermore, this concept is more a com­
parative than a classificatory concept, i.e. the synonymity we mean in
ordinary discourse is more a matter of more-or-less than of either-or. As a
rule there is only a greater or a lesser similarity of meaning between two
sentences, not any strict identity of meaning.15
And finally synonymity in the pre-scientific sense always holds only for
specified contexts and under specified conditions, never without quali­
fication. The pre-scientific concept, especially as a comparative concept,
is not suited, then, for the definition of propositions, for propositions
are either identical or they are not - they are not more or less identical.16
Indeed, as a comparative concept the pre-scientific concept is not even
capable of being used as the basis for introducing a classificatory scien­
tific concept by way of an explication to make it more precise. So the
scientific concept would have to be given an entirely new definition.
But Quine has already emphasized in [51] that there is no general
definition of ‘synonymous’. For example, if ‘G(A, B)’ were to be defined
by 6A = B is analytically true’, that would be circular. For analytic sen­
tences are defined precisely as just those sentences which are logically true
or can be derived from logically true sentences by substituting for some
expressions others that are synonymous.17
But if 6G(A, B)9is defined by 6A and B are definitionally equivalent’,18
that is noo narrow, for natural languages in any case, which lack explicit
definitional stipulations,19 but which do contain synonymous sentences,
such as ‘Fritz is a bachelor’ and ‘Fritz is an unmarried man’.
78 CHAPTER II

Finally, if reference were made in the definition of synonymity to

criteria of substitution, one would once again be involved in a circle.
For if ‘G(A, B)9 is defined as 6A can be substituted for B in all contexts
(sentences) without any change in their meaning’, one would have to know
beforehand what the meaning of a sentence is, or when two sentences are
synonymous. But that is precisely what is supposed to be defined. But
if ‘G(A, B)9 is defined as 6A can be replaced by B in all contexts salva
veritate9, then there are among these contexts either contexts that are also
intensional like 6A is analytically equivalent to B 9- and then one is no
wiser than before20 - or the contexts are restricted to extensional sen­
tences - and then the criterion is too broad and would have to be shar­
pened by the requirement of salva analyticitate instead of salva veritate.
Now naturally one could regard ‘analytic’ as a basic concept, i.e. forego
a definition of this concept and define ‘synonymous’ with its help. For
Quine, however, the concept of analytic sentence is problematic through
and through, as we shall see, and so that procedure is excluded for him.
The arguments advanced by Quine against accepting a precise defini­
tion of synonymity, especially one precise enough for defining proposi­
tions, are not sound, however. In the first place, a classificatory concept of
synonymity can be defined even with a comparative concept ‘x and y are
equally or less similar in meaning than u and v9: x and y are synonymous: =
x and x are less or equally similar in meaning than x and y. Lexicons,
dictionaries, and the criteria we use in determining the correctness of a
translation also bear witness against the claim that the pre-scientific
concept of synonymity is so imprecise as to be completely useless for the
analysis of meanings. And even if synonymity should exist only for speci­
fic contexts and consequently the meaning of a sentence would always
have to be established only for specific contexts, that would be no serious
hindrance to defining propositions on the basis of the everyday concept of
synonymity by way of abstraction, since Quine himself remarks that
every linguistic expression has a definite meaning only in definite con­
texts.21
On the basis of Quine’s arguments, then, there is no justification for
supposing that the pre-scientific concept of synonymity does not consti­
tute a foundation for a scientific explication of this concept, as linguistics
does indeed undertake to do. Naturally the concept, like all empirical con­
cepts, has a certain horizon of vagueness and is not defined with absolute
 THEORIES OF MEANING 79

precision; it may even require, more than many other empirical concepts,
supplementary delimitations, but it is not plain why such delimitations
should not be possible and why the concept should not function well in
practice. We do not forego talking of biological species or of the juristic
properties of the facts of a case because it may be impossible to define
them with ultimate clarity in all cases.
The following objection is more decisive: As we saw in Chapter II. 1.5.
Carnap has shown in [56] a way for the definition of intensions and mean­
ings by extensional concepts. This approach has been developed in recent
years, especially by Richard Montague into an intensional semantics,
which works with the extensional concepts such as ‘function’ and ‘class’
acknowledged even by Quine. Within its bounds the concept of propo­
sition can be exactly defined, as well as the concepts of the intensions of
predicates and proper names. Quine’s critique of the theory of meanings
has been largely overtaken by this development.22

(3) Quine’s third argument against accepting propositions goes:

If we assume that propositions are determinate entities, then identity
must be defined for them as well. If a definition is then given along the
lines of realistic semantics: ‘Two sentences are synonymous if and only if
their meanings are identical’, then what we have is a sharp classificatory
concept of synonymity which holds not only for sentences in one language
but can be applied to sentences in different languages. But according to
Quine there can be no such absolute concept of synonymity because of
the thesis of indeterminacy of translation.
Since we will bring forth arguments against this thesis below, we will
not go into Quine’s argument in any greater detail here.
Quine makes the same criticisms of accepting concepts as meanings of
predicates as he makes against accepting propositions. His arguments run
quite parallel to those discussed above, however, and therefore we will
not go into them more closely here.

3.2. Quine's Arguments Against Analytic Judgments

A distinction is usually drawn in philosophy between truths o f reason and
truths o f fact, between analytic sentences, whose truth results from the
meanings of the terms that occur in them alone, and synthetic sentences,
for which that is not the case.23 In his essay, ‘Two Dogmas of Empiricism’,
80 CHAPTER II

Quine attacked this distinction and he has further expanded on his argu­
ments in later publications.
Quine gives essentially two reasons for his thesis that there are no sharp
boundaries between analytic and synthetic sentences:
The first argument runs as follows: The concept ‘analytic’ is an artifi­
cial concept of science which has to be defined. The usual definition is:
A sentence is analytic if and only if it can be derived by the replacement
of terms in a logically determinate sentence (i.e. a sentence which is true
or false on purely logical grounds) by other synonymous terms. But accord­
ing to Quine, this definition can not be used, since, as was discussed
above, a concept of synonymity sufficiently precise for the definiens is not
available.
Our response to that has been that the ordinary concept of synonymity
is actually not as imprecise as Quine makes it out to be and that it can
be defined exactly within intensional semantics. But since we have al­
ready urged that the definition of analytic sentences, as Quine gives it, is
too narrow, even with a sufficiently precise definition of synonymity, one
would not obtain any satisfactory definition of analytic sentences.
Second, according to Quine’s behavioristic program, the term ‘analyt­
ic’ is to be defined so that we can tell by observing the linguistic behavior
in a population P whether a sentence is used in P as an analytic sentence
or not. It is problematic, however, if there are any sentences at all which
are used in a clearly analytic way.
The difference between analytic and synthetic sentences is by no means
as clear and unambiguous as it is often represented to be. One already
gets into trouble when one wishes to cite analytic sentences that are not
logically determined. The following examples are found:
(1) All bachelors are unmarried.
(2) All bodies are extended.
(3) What is green is not (all) red.
But with a little imagination for all of these sentences empirical condi­
tions can be found under which one would possibly regard them as false,
so that the analytic character of these sentences is doubtful. If, for example
because of some blunder in the legislative process, every man were declared
married? And if on the basis of physical theory it should prove to be
meaningless to apply the concept of spatial extension to elementary par-
 THEORIES OF MEANING 81

tides (as bodies)? Or if all men were to become color blind and so could
no longer distinguish between red and green? Would one still regard
sentences (1) to (3) as true in these cases?
Naturally, one could always say that sentences (1) to (3) hold afterwards
as they did before, for there would just no longer be any bachelors, accord­
ing to physical theory elementary particles would not be bodies, and after
as before no objects would be red and green at the same time, but we
would just have lost the capacity for distinguishing between red and green.
But we could also say that the sentences would then be false: A bachelor
would still be a man not living in matrimony; every object that may be
localized would still be a body; and ‘red’ and ‘green’ would designate
colors as perceived by men with normal sensory constitution as before.
In the face, then, of such remote contingencies we are uncertain what
to say. And in such cases even Carnap’s proposal is of no help, which he
illustrates with the following example: In order to determine whether a
person X regards the sentence (i) ‘All ravens are black’ as analytic or not,
it is sufficient to put the following question before him : ‘Mr. Smith caught
a white raven yesterday, which I will be glad to show you. In the face of
that fact, are you ready to withdraw your statement (i)?’ Now if X answers:
‘I would never have believed there are non-black ravens, and I still won’t
believe it until I have seen one with my own eyes. But in that case I shall
naturally have to regard (i) as false’, that makes it clear that X does not
regard (i) as an analytic, but as a synthetic sentence. But if X answers:
‘There can’t be any such thing as a non-black raven. If a bird is not
black, I do not call it a raven. So if Smith says he has found a white
raven (and he isn’t lying or joking) then he must understand either the
word ‘raven’ or the word ‘black’ in a different sense from the one in which
I understand it’, it is then clear that X regards sentence (i) as analytic.24
In view of the uncertainty in the face of very unusual situations such
an inquiry will neither elicit clear responses from individuals nor will there
be a convergence toward a common opinion in a population.
As D. Lewis has emphasized in [69] we have to distinguish two concepts
of analyticity here. On the one hand there is the predicate ‘sentence A is
analytic in language S ’ which may be defined b y 6A is true in all worlds on
which the interpretation of S is based’, on the other hand there is the
predicate ‘Sentence A is analytic in the population P ’ which may be
defined by ‘A is analytic in that language S which is the language of P ’.
82 CHAPTER II

In order to determine whether A is analytic in P we then first have to

ascertain which language is spoken in P. Lewis believes that the use of a
language in P is determined only for normal situations of communication
so that there may be a set S of precise languages (e.g. in the sense of inten-
sional semantics) such that all S in S are compatible with the normal
usage in P although they differ for remote and unusual occasions. The
languages in S may then diverge, especially for worlds representing such
unusual contingencies, contingencies that we normally do not contem­
plate. But sentences may be distinguished uniquely as analytical in P only
if they are analytical in all languages in S , and Lewis believes that it is
very well possible that there may not be such sentences at all. The inde­
terminacy of the concept ‘analytic in P ’ as against ‘analytic in S ’ therefore
has its basis in the fact that the use of language in P is compatible with
different logically precise languages.25
This analysis, however, is not quite adequate. It is not the case that all
members of P speak well defined though different languages S - then
Carnap’s test would give clear results at least for the individual speakers.
It seems to be more adequate to say that members of P speak essentially
(i.e. not counting the peculiarities of their idiolects) the same language S ,
but that this language is only well defined for normal situations of commu­
nication and vague or undefined for other cases. Natural languages are
not precise in the logical sense, and therefore the application of the con­
cept ‘analytic in S \ which is exactly defined in the logical case, to natural
languages yields no precise distinctions.
Quine’s third argument against the distinction analytic-synthetic is this:
According to Quine analytic sentences can not be sharply marked off
from synthetic, because they are supposed to hold true on the basis of the
meanings of their terms alone. But according to Quine the meaning of a
term is not independent of which of the synthetic sentences that contain
that term we accept as true. The meaning of a term changes with our
assumptions about the world; consequently, questions of meaning and
questions of fact can not be rigorously separated.
If my views about the world gradually change in such a way that I come
to believe the same things about cats that I now believe about dogs, and
about dogs the same things I now believe about cats - am I then supposed
to say that the meanings of the words ‘dog’ and ‘cat’ have remained con­
stant, and that only my views about the world have changed? But that
 THEORIES OF MEANING 83

would obviously be incorrect, for the characteristics and criteria of the

concepts have changed completely: A dog is now an animal that meows, a
cat an animal that barks, etc. It is more correct to say that together with
my beliefs the meaning of the words has changed for me: A dog is now
something other than it was before and therefore the definition of an animal
as a dog no longer has the same content it had before.26
Here is a further argument: We do not learn linguistic expressions all
by themselves, but in a context of sentences and situations. That holds
especially for predicates, which we learn by example in [logically] singular
contexts, such as ‘This apple is red’, ‘This lemon is not red’, and in general
contexts such as ‘All human beings die sometime’, ‘That is not a bird but
a butterfly; birds have feathers, but butterflies do not’.
It is not so, then, that we observe and state facts and, as an activity
entirely separate from that, establish the meanings of words, but rather
learning language and experiencing the world stand in a very close rela­
tionship : The facts we accept determine the meanings we connect with our
linguistic expressions.27
Finally, one can also argue in the following way: It is usually said that
meaning is first attached to single terms, proper names and predicates,
and that the meaning of a sentence is then a function of the meaning of
its terms and the way they are combined in the sentence. Fundamentally,
that idea is correct, especially where it is a matter of explaining how new
facts can be communicated in sentences; for that can succeed only if the
meaning of such sentences is determined by the meanings of the terms.
In the context of language learning, however, the situation is the other
way around in many respects: the primitive linguistic expressions are sen­
tences. We analyze these sentences into proper names and predicates. The
interpretation of the terms comes about, roughly speaking, somewhat as in
the case of so-called implicit definitions: Certain sentences are marked
out as true, certain others as false and an interpretation of the terms is
sought for that attaches their assigned truth-values to all of the sentences
in question. In that case, then, the meanings of the terms are determined
by an analysis of the meanings of sentences, not the meanings of the sen­
tences by a synthesis of the meanings of terms. And this determination of
the meaning of the terms depends in a decisive way on which sentences
we regard as true and which as false.
Now it is not so that the process of learning language ceases at a speci­
84 CHAPTER II

fic point in time and that we attach specific meanings to terms prior to
that time and that these meanings are then independent of our assump­
tions about the world. Instead, this process continues even after the
meanings of terms have become relatively independent, so that the depen­
dence of term meanings on these assumptions goes on being the case.
The dependence of the meanings of terms on sentence contexts can be
ascertained for ordinary language from the fact that within certain limits
words can change their meaning according to the context in which they
occur. Therefore it can not be simply said of a sentence in ordinary lan­
guage that the meanings of the terms determine the meaning of the sen­
tence, but rather the meaning of the sentence also determines the meaning
of the terms.28 What could perhaps be said is this: The meanings of the
terms pre-determine the meaning of the sentence within certain limits, but
the exact assessment of the sentence’s sense - and with it the exact assess­
ment of the meaning the terms have in the sentence - emerges only through
the interpretation of the entire sentence. Dependency of meaning on con­
text of this sort does not exist solely in connection with ambiguous words.

Before we show that, some remarks should be added here on the prob­
lem of ambiguity and on the terminology that is relevant to it. In this con­
nection we shall follow the ideas K. Heger has developed in [63] in part
using some of his examples.29
When it is said that one word is ambiguous or that several words have
the same meaning, the first thing that must be considered is how the ex­
pression ‘word’ is being used. If what one understands to be a word is a
specific expression together with its meaning, then it is immediately clear
that there can not be any ambiguous words, for whenever we have differ­
e n t meanings before us, we also have different words per definitionem. If,
on the other hand, the meaning is not regarded as an integral component
of the word, there still remain various possibilities for defining a word,
(a) In the first place it is required that identical words represent the same
(e.g. phonetic or graphic) expression, i.e. that - according to the realiza­
tion of the language to which one is referring - homophony or homography
is present. We have already emphasized that the two do not coincide,
and out of that arises an uncertainty in defining words, which can be
avoided, however, by specifying the word’s form of realization. Add to
this, nevertheless, the fact that there are words that can be inflected whose
 THEORIES OF MEANING 85

forms are constructed from different word stems (e.g. go - went, or in

French, beau - bel). We speak of allomorphy in this case and the question
is whether we wish to speak of two words here or of one. In the latter
case, not even the weak condition (a) would be fulfilled, (b) One can add
the further requirement that identical words also have identical syntactic-
grammatical functions (in the sense of logical grammar, for example).
On this view, the expression murder, as a verb and as a substantive,
represents different words. Equally so the French vers in the sense of
‘verse’ and of ‘contra’. Naturally the criterion of identical grammatical
function is precise only if account is taken of some specific grammar, (c)
Finally, it can also be required that identical words have identical etymo­
logical roots. In that case, according to Heger, the German word Kohl in
the sense of cabbage, derived from the Latin caulis, - and in the sense of
nonsense, twaddle, from the Hebrew gol, for example, represents two
different words. But this example also shows that the diachronic, i.e.
historical criterion of identical etymon is questionable within the scope
of a synchronic, systematic consideration of ambiguity. Those who use
the German language no longer feel that there are two different roots
concealed in ‘Kohl’ and therefore they see it not as two words, but as one.30
Let us assume that the concept ‘word’ can be clarified by such defini­
tions without reference to meanings. In that case there can be ambiguous
words. The two meanings of a word can have more or less in common, be
more or less related. If they have no features in common, or they are
unessential, one will speak of an ambiguous [mehrdeutigen] word (exam­
ples: ‘pipe’, ‘bat’). If the meanings do include a group of characteristics in
common, but typically different characteristics as well, then we speak of
an ambivalent [mehrwertigen] word (examples: ‘dark’ as in ‘a dark color’
or ‘a dark night’, ‘square’ as in ‘square in shape’ or ‘an area with a square
shape’).31 ‘Horse’, on the other hand, as a generic concept covering
‘stallion’, ‘mare’, or ‘pacer’, ‘trotter’, etc., is not an ambivalent word, for
as a rule we do not mean by ‘horse’ a pacer or a trotter, etc.; rather the
word means a generic concept and there are many general statements about
all kinds of horses, while statements about squares are as a rule either
about places [such as a market-square] or square shapes or square num­
bers. Finally, one meaning can include all of the defining characteristics
of the other, but not vice versa. Examples are: ‘neck’ (section of the verte­
brate body between the trunk and the head - any narrow segment in front
86 CHAPTER II

of the head of something) and ‘turn red’ (to change color to red - to
blush). In the former case we have a metaphorical extension of the original
sense of the word, in the second case a contraction of the word’s meaning.
As a rule we speak of homonymy in connection with words that have
more than one meaning and two (or more) different etymological roots,
of polysemy in connection with words with more than one meaning but
only one etymological root.
Now if we speak of meaning’s dependence on context, then a linguistic
context may be meant - the sentence or text in which the word occurs -
or an extra-linguistic context, i.e. the circumstances in which the utterance
is made. We shall return to the dependence of meaning on the circum­
stances of the utterance later on, here what we are concerned with at the
outset is the dependence of meaning of words on the sentence context in
which they stand, i.e. a consideration within the confines of langue, not of
parole.
Dependence on context is demonstrated by the following sentences:
(a) ‘Fritz is smoking his pipe’. Here ‘pipe’ is synonymous with ‘a small
tube with a bowl at one end, used for smoking tobacco, opium, etc.’ In
this case the ambiguity of the word ‘pipe’ is eliminated by the context.
(b) ‘The monument is in the square’. Here ‘square’ is being used in the
sense of a public place, and so the polyvalence of ‘square’ is eliminated
by the context, (c) ‘On hearing such praise, Fritz turned red’. Here ‘turn
red’ is used in the sense of ‘blush’ and so the context brings about a nar­
rowing of the meaning, (d) ‘The bottle had a crack in its neck’. Here
‘neck’ is used in its expanded figurative sense.
But there is a dependence on context to be seen even in the case of
words of which it can not be said that they are ambiguous or ambivalent.
J. Lyons offers some examples of this in [69], p. 452/: (e) ‘My dog has just
had pups’. Here ‘dog’ is being used in the sense of ‘bitch’ and can be
replaced by the latter word without any change in the meaning of the
sentence, (f) ‘I’m going to New York in my own car’; here ‘go’ is being
used in the sense of ‘drive’ and can be replaced by this word.
Grammatical ambiguities can also be resolved by the context. Katz and
Fodor in [63] offer the example, among others, of the two sentences: (g)
*Our store sells alligator shoes’ and (h) ‘Our store sells horse shoes’.
You know that (g) is ordinarily to be understood in the sense o f ‘Our store
sells shoes made o f alligator hide’ and (h) in the sense of ‘Our store sells
 THEORIES OF MEANING 87

shoes for horses’ if you know that alligators do not ordinarily wear shoes
and that shoes are not customarily made out of horsehide.
Katz and Fodor now bring it out that this selection of interpretations
for (g) and (h) does not follow any purely linguistic criterion; the other
possible interpretations are not excluded by linguistic rules, but are im­
probable on the basis of factual information. It cannot be said that in the
context ‘eating nuts’, ‘nuts’ always means nuts that grow on trees, follow­
ing purely linguistic rules. If there were a story of a humanoid metallic
monster who eats nuts and bolts, along with other more substantial fare,
we would be able to understand the expression ‘eating nuts’ differently in
this context. Both of these examples show, then, that general factual
information enters into the interpretation of linguistic expressions; that
dependence of meaning on context is not only a matter of linguistic regu­
larities but that we interpret sentences against the background of certain
things we know or assume.
Now if it is true that the meaning of linguistic terms depends on our
assumptions about the world, as Quine maintains, then there is no sharp
boundary between analytic sentences, whose truth-value is determined on
the basis of the meanings of their terms alone, and synthetic sentences, for
which this does not hold.
Actually there are still differences in this connection, however. The
sentence ‘All bachelors are unmarried’ is more to be regarded as analytic
than ‘Max is a bachelor’. Quine has no objection to a comparative con­
cept of analyticity.32 He represents the situation somewhat as follows: All
of our assumptions about the world, singular sentences and scientific
hypotheses and theories alike, analytic sentences (or meaning postulates)
and the theorems of logic alike, form an entire complex that can be con­
fronted with experience only as a whole. Now if experience is incompatible
with the complex of our assumptions, then it is permissible in principle
to propose modification anywhere, not only in the singular sentences and
hypotheses or theories, but in the meaning postulates, indeed even in
logical theorems as well.33 Basically, no assumption, no sentence in this
complex is incorrigible, i.e. immune to modification. Amendment of the
system follows various guidelines, such as simplicity and power of the re­
sulting new system, the principle of conservatism, to make the fewest
changes possible and to change less secure assumptions in preference to
those that are more secure, and so on. If we call the sentences we are
88 CHAPTER II

more prepared to change peripheral - within the entire complex - and the
sentences we are not so ready and willing to give up central, then meaning
postulates, along with universally accepted fundamental laws of nature
and logical theorems would be central sentences; and the central sentences
would be more influential in determining meanings than the peripheral
would be. But there are no sharp boundaries between modifiable sentences
and non-modifiable ones, between those that determine meaning and
those that do not.
The following comments on this presentation are to be made: For one
thing, it is better to leave logical sentences and the meaning of logical
expressions, mathematical as well, out of consideration here, since the
situation is fundamentally different with them. For logical operators can
be isolated and defined precisely without referring to empirical facts.
These definitions are stipulations and as such do not depend upon our as­
sumptions about the world. As against Quine, who denies that there is any
sharp boundary between logical and non-logical expressions, so that we
can only enumerate the operators used in our present logic, we can per­
ceive such a boundary in the strict definability of these operators.34
Another point is that Quine’s sketch remains very lacking in detail. In
his effort to show that there is no absolute boundary between analytic and
synthetic sentences, Quine obliterates the existing relative, but methodo­
logically very important boundaries:
Quine’s claim that there is no absolute boundary between analytic
and synthetic expressions can be assumed to be sufficiently demonstrated
by the arguments cited above. The insight that questions of meaning can
not be rigorously separated from questions of fact, because the meaning
of linguistic expressions depends on our assumptions about the world, is
also extraordinarily important for an understanding of the way in which
language functions. What is important for the methodological distinction
between analytic and synthetic sentences, however, is not its absolute but
its relative validity. When Quine says that even logical principles and
meaning postulates are not immune to correction by experience, that is
obviously true in quite a different sense than it would be in speaking of
the hypothetical truth of general synthetic sentences. For example, if at
a given point in time t the sentence G(a) follows from system S(t) of our
assumptions about the world, with the aid of classical logic and meaning
postulates, and if G(a) turns out to be false on the basis of observation,
 THEORIES OF MEANING 89

then it obviously makes no sense to change the logic or the meaning of

the predicate G(x) (say by giving up the pertinent meaning postulates
in 5 (0 ) in such a way that 5 ( 0 becomes compatible with i G(a), For the
result of that would be to change the meaning of the empirical sentence
G(a) so that possibly it would no longer reproduce the content of the
observations.
If one wishes to formulate the question of whether certain assumptions
fit with specific observations as a question about the compatibility of
sentences, then in doing so one must take a language into account, inter­
preted in a definite way, with logical and descriptive constants that have
a fixed meaning. Since the logical principles are based on the interpreta­
tion of the logical constants, a logic is also singled out accordingly.
An (interpreted) language and a logic thus constitute the reference system
within which the question is raised whether certain assumptions are com­
patible with specific observations or which assumptions must be modified
in order to achieve such compatibility. Therefore what comes into ques­
tion in adjusting the assumptions to experience is a modification only of
sentences that can be modified without changing the reference system, thus
without abrogating any laws of logic or any sentences whose truth follows
from the interpretation of language. Thus it would obviously absurd to
bring the sentence ‘This raven is white’ into agreement with the assump­
tion ‘All ravens are black’ by re-interpreting the predicate ‘white’ into
‘black’ or ‘all’ into ‘some’. It would be absurd because it would cancel out
the presupposition of putting the question, the acceptance of a specific
language and logic. Correction of assumptions on the basis of conflicting
experience can only concern non-analytic sentences, then, with the refer­
ence system held constant.
Naturally, one can change the interpretation of the language or choose
another system of logic. But that is a change that experience does not
make necessary - an incompatibility between observation sentences and
assumptions, as indicated, exists only within the limits of a definite system
of language and logic, and so cannot refute it. A change in the reference
system is on quite a different level and is motivated by quite different
considerations than is the abandonment of scientific hypotheses in the face
of conflicting observations. It can be based, for example, on the fact that
the language previously used has too limited an expressive capacity, that a
re-interpretation of some terms bestows upon it a greater fruitfulness for
90 CHAPTER II

the description of nature and permits general laws to be formulated more

simply, etc.
It is important, then, to hold fast to this distinction between changing
hypothetical assumptions and changing the reference system. Analytic
sentences exist only relative to such a reference system, because the mean­
ings of terms are fixed only within a reference system. The more precisely
the interpretation of the language is specified, the sharper is the boundary
between analytic and synthetic sentences with respect to this interpreta­
tion. Since the interpretation of descriptive terms in natural languages is
fixed far less precisely than the interpretation of logical and mathematical
terms, the analytic sentences that are not determined by logic or mathe­
matics are also much less precisely determined. But practically speaking,
the distinction between analytic - synthetic is sufficiently clear in most
cases.35
If Quine insists that the meanings of words change along with our
assumptions about the world, that is correct, viewed diachronically.
But at every moment we have a definite understanding of language (even
if it is of limited precision) which is the foundation of our statements
about the world and there is relative to it a distinction (of limited preci­
sion) between analytic and synthetic sentences. That we can say anything
about the world at all and communicate information about it depends on
the fact that there are statements for which meaning and truth-value are
independent of each other, that not all sentences have the character of
meaning postulates, then; that there are definitely synthetic sentences;
and that thus the boundary between analytic and synthetic sentences is
sharp in at least some cases.36

3.3. Quine*s Thesis o f the Indeterminacy o f Translation

We have portrayed the starting point of Quine’s ideas on the philosophy
of language, his pragmatic approach, by way of this quotation: “Lan­
guage is a social art. In acquiring it, we have to depend entirely on intersub-
jectively available cues as to what to say and when. Hence there is no
justification for collating linguistic meanings, unless in terms of men’s
dispositions to respond overtly to socially observable stimulations.” 37
From this view, that we learn language only by using it, there follows
for Quine one of the fundamental theses of his book, Word and Object,
the thesis of indeterminacy of translation.38 Quine formulates this thesis
 THEORIES OF MEANING 91

as follows: “The infinite totality of sentences of any given speaker’s lan­

guage can be so permuted or mapped onto itself, that (a) the totality of
the speaker’s dispositions to verbal behavior remains invariant, and yet
(b) the mapping is no mere correlation of sentences with equivalent sen­
tences, in any plausible sense however loose.” 39
Although the term ‘equivalent’ is not explained in greater detail any­
where, it obviously stands for an identity of meaning defined in some way.
What the thesis says, then, is this: The meanings of a language’s expres­
sions can not be determined unambiguously on the basis of its use; there
are radically different interpretations all compatible with the use of a
language.
This thesis can be reformulated in an especially perspicuous way if
instead of the mapping of a language onto itself, one considers the trans­
lation of one language into another, which takes into account nothing but
the criterion of word usage: “ ... manuals for translating one language
into another can be set up in divergent ways, all compatible with the to­
tality of speech dispositions, yet incompatible with one another.” 40
The thesis can best be made precise in this formulation by illustrating
it in terms of two simple languages, L 1 and L 2. We will assume that L 1
and L 2 are languages in elementary predicate logic that differ only in
their alphabet, not in their formal rules, i.e. how they draw the lines
between expressions of different syntactic categories, the terms (proper
names and predicates with various numbers of places) and sentences.41
Let the expressions in L 1, or L 2, be interpreted along the lines of inter­
pretation V 1, or V 2, which are defined over object domains yl9 or y2, so
that these sets include all of the objects it is possible to talk about in L 1 or
L 2.42 Now let X be a person who speaks language L 1 and learns the
(to him) foreign language L 2 from the use which the members of the
linguistic community - let Y serve as representative - make of the ex­
pressions in that language.
For the sake of simplification we will assume that X can read off the
syntax (the grammar) of L 2 directly from its usage (which we are assum­
ing to be identical with that of L 1). Furthermore, X is supposed to be
able to infer from the usage the semantic character of the logical operators
in L 2.43
In accord with these assumptions we can directly stipulate the following
conditions for a translation:
92 CHAPTER II

(1) A translation of L 2 into L 1 is a function 0, of which it is true

that:
(a) (j>coordinates to the terms and sentences of L 2 expressions of
the same semantic category in L l.44
(b) (j) is operationally faithful with respect to the form rules com­
mon to L 1 and L 2.45
(c) (j) coordinates to the logical operators of L 2 equivalent opera­
tors inL 1.
X can check these conditions (a) to (c) - this is our idealizing condition
- independently of any knowledge of the particular interpretation V 2 of
L2.
The specifically semantic problem of translation can now be formulated
as follows: What additional criteria of translation arise for X from Y 9s
use of language, what can X infer about V 2 from 7 ’s use of language?
Now what is decisive on this point is the sentences in L 2 that Y uses in
order to describe something which he and X can both directly observe,
since what is referred to is most distinct for such sentences.
We will assume along with Quine that X can establish whether a sen­
tence ^ in L 2, which Y utters at a specific moment t, (partially) describes
Z ’s field of perception W at t, i.e. refers to something X can perceive at f.46
X will then translate such a sentence <Pof L 2 into a sentence of L 1 that
likewise describes W.
We can then add in a last condition for the Quinean concept of transla­
tion:47
(d) It is true of all sentences <P of L 2 and all perceptual fields
W of X that 4> describes W if and only if $<!> describes IT.48
The thesis of indeterminacy of translation can then be expressed as
follows: For antecedently given and V 1, V 2 is not unambiguously de­
termined by conditions (a) through (d).49 Among other things, these con­
ditions are compatible with yl and y2 containing no common element and
interpretations V 2(a) and V 1(<£a) being different for all terms a of L 2,
so that the translations of expressions in L 2 have entirely different
meanings from those expressions.
Now the validity of this thesis obviously depends on condition (d).
Without this condition it would be a triviality. The question is, then, to
what extent does (d) restrict translation.
 THEORIES OF MEANING 93

Let us assume that Y uses a simple sentence such as ‘a has the property
F ’ to describe one of X ’s perceptual fields, that presents a rabbit sitting
in a meadow, for example. Now X does not know which object Y uses
V to designate or which property 6F \ The perceptual field, i.e. the set of
sensory stimuli X experiences at the moment in question, permits various
(partial) descriptions and Y can just as well be using his sentence to
describe the fact that the meadow is green as that the animal sitting in it
is a rabbit. Now T ’s subsequent use of language can be informative about
such differences of meaning, for not all rabbits are sitting in green mea­
dows and not every green meadow has a rabbit sitting in it. So if Y again
says 4a has property F ’ when he is looking at a green meadow with no
rabbit, the second possible interpretation is no longer open; but if Y says
4tf has property F ’ when he is looking at a rabbit with no meadow, then
the first possible interpretation is no longer open.50
But now if there are concomitant specifications, i.e. specifications that
always enter into the picture together, then there is no longer any possi­
bility of distinguishing between different meanings by way of language
use. Quine offers as examples of concomitant specifications ‘thing’, ‘thing
part’, and ‘thing stage’ (‘thing at time / ’).51
Quine accordingly sees a difficulty for the interpretation of all expres­
sions similar to the one Wittgenstein sees just for expressions in a private
language, more exactly, for expressions that stand for private experiences,
mental events and the like. Wittgenstein uses his beetle example to illu­
strate it:52 “Suppose everyone had a box with something in it we call a
‘beetle’. No one can look into anyone else’s b o x .... In that case it might
well be that everyone had something different in his box.... But if the
word ‘beetle’ were to have a use for these people all the same? - Then it
would not be the name of a thing. The thing in the box has no place in the
language game at all... you can ‘divide through’ by the thing in the box;
it cancels out, whatever it is.” 53
With Quine what corresponds to the box is the perceptual field and to
the beetle the thing in the perceptual field, or the quality or relation
designated by an expression of L 2. Quine’s thesis thus goes further than
Wittgenstein’s statements on the determinability of the meaning of predi­
cates by their use54 to such an extent that according to it not even the
meaning of proper names, indeed their references, can be unambiguously
determined, that there is not even one case in which we can learn from
94 CHAPTER II

the use of language what domain of objects a language is concerned

with.55
The examples of concomitants that Quine otfers do not make a very
convincing impression, to be sure - as we shall see, that is in the nature
of the case. But even if one does not assume that there are such concomi­
tants, there remains the following difficulty for translation: Every one
of X 9s perceptual fields admits of an infinite number of descriptions in
principle, for along with a sentence #, for example, the sentence # v ¥ is
also a description of W. Even the following simple geometric figure, for
example, can be (partially) described by

the sentences: ‘Figure ACE is an equilateral triangle’ or ‘Figures ABF,

BCD, DFB, FDE are equilateral triangles and together form an equilat­
eral triangle’, or ‘Figure ABF is an equilateral triangle, figure BCEF is a
trapezoid and together the two form an equilateral tiiangle’, or ‘Figure
ABF is an equilateral triangle’ or ‘Figures BDF and FDE have a side in
common’, or ‘Figure ABF is not a trapezoid’ - to cite only some of the
possible descriptions.
Since the set of perceptual fields W available to X for the analysis of
F ’s use of language is only finite - X is supposed to determine V 2 on the
basis of a finite number of observations of T’s use of language - that
means that of an infinite number of hypotheses about V 2 that are possible
according to (a) through (c), only a finite number can be ruled out on
inductive grounds. All of the remaining hypotheses are equally probable,
or better, equally improbable. This argument also shows, then, that in line
with Quine’s thesis V 2 is not unambiguously fixed by V 1 and </>, or that a
translation $ is not unambiguously determined by V 1 and V 2.
Shifted back from the case of translation to the case of learning a
language, what Quine’s thesis says is this: Since we learn any language
 THEORIES OF MEANING 95

only by using it, we can always interpret its expressions in different ways
as long as they are just in accord with usage.
But now it is nevertheless a fact that we can come to a very precise
understanding about the meaning of words. We weigh the nuances of
meaning of an English sentence and its German translation very precisely
and we discuss the relationships and differences of meaning of German
words in detail. There are not actually any competing English-German
dictionaries offering genuine alternatives and there is no sentence that is
not ambiguous in English as well for which there would be competing
translations in German that were in no sense ‘however loose’ equivalent
and between which we could not make a choice.
Quine too does speak of the different meanings of a word and distin­
guishes in language between his examples of concomitant specifications.
Furthermore, it is precisely a basic pragmatic thesis that concepts are
abstractions from predicates, that conceptual distinctions rest on lin­
guistic di stinctions. Thus wherever there is a conceptual distinction between
possible interpretations of a word, it is also communicable in language. It
can not be so, then, that the distinctions of meaning we ordinarily refer to
can not be justified in terms of language use. Language use must therefore
permit an essentially more precise determination of linguistic meanings
than it appears to according to Quine’s representations. In particular it
must be possible for the criteria of adequate translation to be sharpened
in essential ways, so that X can infer from Y’s use of language what
things, qualities and relations Yis talking about.
But how can that be explained? How can X know the meaning of the
terms of L 2 and the domain of objects y2 from how Y uses the sentences
of L 2 in concrete situations so as to apply them to observable things?
In such a situation there is a perceptual field given to X. Whether Yhas
the same or a similar perceptual field, X does not know. In principle it
could be the case that Yhas quite a different sort of sensory organization
than X and experiences entirely different sensible stimuli. But for X it is a
very natural and obvious, not to say self-evident, hypothesis, that Y has
essentially the same perceptual field as himself. Quine too has to presup­
pose that, for otherwise X cannot observe Y’s use of language at all. X
obtains information about the meaning of a sentence Y uses only when
he knows what that sentence refers to. But what Y uses his sentences to
refer to is not Y ’s perceptual field but his own. Here, as in what follows,
96 CHAPTER II

the exception proves the rule: for example, the one perceives sound or
light signals in frequency ranges that are not accessible to the other. But
all that we need to do here is to explain how we understand each other in
most cases, and not that we understand each other in all cases - which
is indeed actually not so.
Now the same perceptual field does admit of different descriptions in
general, however: F can mean the one description by his sentence <P and
X the other, as the rabbit example above showed. But there are not as
many descriptions of the perceptual field as you please, which X has to
take account of in his interpretation of #. F ’s perceptual field is no mere
‘swarm of sensations’, no unorganized mass of sensory stimuli which X
would first have to arrange into the form of things with qualities and rela­
tions. Instead, the perceptual field is structured to begin with. This fact is
supported by extensive evidence in the psychology of perception. The
aspects of the situations that strike upon the eye, the obvious descriptions
that X has in mind without further deliberation, even though they may
not be unambiguously determined, are only a few. Now it is another
obvious and natural hypothesis for X , which can be supported or shaken
by the knowledge F h a s of F, that Y not only has the same sensory organi­
zation as X himself, but that there are biological, psychological, socio­
logical features common to him and F which result in the same aspects of
the situation that strike on his eye also striking on F ’s, and that F with
his sentence $ expresses one of these aspects. The degrees of importance
the various aspects have for X are effective in the sense that X can assign
to his interpretations of $ different degrees of (subjective) probability -
taking account of his knowledge of F. So X can rule out many possible
interpretations with practical certainty from the beginning and he can
make a choice among the rest inductively by way of F ’s further use of
language. The greater the common features shared by X and F, the nar­
rower an antecedent selection X can make among his hypotheses about
the meaning of
Two examples of this: If a third person Z joins X, and F says to X with
a gesture in the direction of Z ‘That is a \ then X can be fairly sure that
the expression ‘a’ designates Z, and not just the volume of space that Z
exactly occupies, the third button on Z ’s vest, the sum-total of parts of Z
or a momentary Z stage - unless F is a philosopher, but that just isn’t the
rule. And if F, seeing a group of young girls, applies the predicate ‘F ’ to
 THEORIES OF MEANING 97

some of them who to X seem particularly attractive and denies it of others

who to X seem less attractive, X can be fairly sure that ‘F ’ is being used in
the sense of some variation on the meaning of ‘attractive’ - unless Y is a
misogynist, but that is not the rule either.
The interpretation of a predicate F of L 2, the use of which X has ob­
served in examples, in the sense of a predicate of L 1, then also has the
result that X can extend the predicate F beyond the examples Y has given
him. Finally, it is also an easy matter for X, on the basis of the above
reflections, to define the domain y2 of L 2 as the set of all such concrete
things as are designated by proper names of L 2 which X has already
learned. In the essentials, then, X will proceed upon the hypothesis that Y
is talking of the same things that he ordinarily has in mind himself.56
But how can X interpret 7 ’s sentences if he does not have mastery of
any language himself and consequently has no stock of concepts for
possible interpretations, if X is in the situation of the child just learning to
speak? The particulars of this learning process are the business of develop­
mental psychology. The only interesting aspect in our connection is the
question of the conditions of possibility of understanding language in this
situation. If it were so that to distinguish objects and their specific quali­
ties would be possible at all only through the medium of language, then
obviously one could never learn a language, since distinguishing linguistic
expressions and comprehending situations to which they refer do presup­
pose such determinations also. Even the child who does not yet speak
already has an environment in the sense of Jakob von Uexkull’s environ­
mental theory.57 This environment is divided into things and events which
are meaningful to the child - not in the sense of the meaning of a sign, but
in the sense of an importance, a role they play in the total context of his
life, a ‘feeling tone’, a ‘valence’ for the child. Even for the child the world
is already structured in this way and his attention is directed to particular
things and qualities in particular situations. The child incorporates a
comprehension of these things and qualities into the first words he learns.
There is certainly still a large margin of vagueness around these pre-lin-
guistic features. Only with growing experience and mastery of language
does the child learn to narrow them down. What is important is only that
in the existence of pre-linguistic qualitative characterizations we can find a
basis for the interpretation of linguistic expressions, for learning their use
inductively by way of particular cases of their application.
98 CHAPTER II

The existence of pre-linguistic qualitative characterizations can still be

established for the case of those who know a language as well. Thus we
can identify certain colors, sounds, smells, etc., without having any simple
descriptions in words available.
Quine also acknowledges the existence of pre-linguistic qualitative
characterizations, when he says “In effect therefore we must credit the
child with a sort of prelinguistic quality space.” 58 Above and beyond that,
however, what is decisive for the possibility of attaching subjective proba­
bilities to interpretive hypotheses and so arriving inductively at an unam­
biguous identification of meanings is the significance particular aspects of
the world have for us. This significance of things and qualities is what
first makes it possible to go essentially beyond the limits of the Quinean
thesis of indeterminacy of translation in attaching meanings. Now ad­
mittedly Quine has also assumed a “tendency to respond in different
degrees to different differences” - for example, “more tendency to disso­
ciate the ball from its surroundings than to dissociate its parts from one
another” 59 - and he describes it as a presupposition for learning a
language, but he did not take this tendency into account when he was
establishing his indeterminacy thesis in [60].

Now there is an obvious objection to the effect that we have not inter­
preted Quine’s statements in Word and Object correctly. For Quine would
not express thesis (I) as we did, when we said: “The meanings of the ex­
pressions in a language are not unambiguously determined by its use,”
since for Quine differences in meaning are always differences in use.60
Furthermore, Quine just does not assume in his reflections that languages
L 1 and L 2 have an identical or even just an analogous syntactic struc­
ture.61 He is not referring to languages that have wide-ranging cultural
features in common that would suggest a translation along the lines of
our discussion. Neither is Quine referring to such simple observation
sentences as we considered, but rather to theoretical sentences such as
“Neutrinos have no mass.” 62 And finally, he is just not investigating the
problem of translation in connection with such languages as English and
German, which have special relationships as he views them.63
If we have thus considerably diverged from Quine in the details, we have
nevertheless dealt with the central problem of the indeterminacy thesis.
For so far as our formulation of thesis (I) is concerned, this thesis becomes
 THEORIES OF MEANING 99

trivially false if meaning and use are identified in the pragmatic way and
non-equivalent sentences are consequently understood to have different
uses. This thesis can also not be interpreted pragmatically in such a way
that while the uses of the non-equivalent sentences are the same, the
uses of the terms occurring in them are not, for a difference in the use of
terms is nothing but a difference in the use of sentences in which they
occur.64
Now Quine argues for thesis (II) in such a way that not all of the argu­
ments can be carried over to case (I). But for Quine too (I) represents the
more interesting claim.65 To that extent the supplementary arguments for
(II) have less weight. Quine says for example that a synonymous transla­
tion is not to be achieved when the syntactic structures of the two lan­
guages are entirely different, when the sentences of one language do not have
the subject-predicate structure, for example, of the sentences of the other
language. One can certainly grant to Quine that languages which are
sufficiently different can not be translated into each other synonymously.
But this difference no longer applies in the case of thesis (I). In the same
way, strong cultural differences naturally result in considerable difficulties
in translation, and one will always have to understand the translated
sentences relative to a translation of the fundamental assumptions about
the world the alien linguistic community starts out from. But this difficulty
does not come into play in case (I) either.
For thesis (I), however, the following two arguments remain, in their
essential features: In the first place, the synonymity of two sentences is
never anything but stimulus synonymity for Quine, according to which
two sentences count as stimulus synonymous when they are affirmed under
the same stimulus conditions (we spoke of perceptual fields).66 But what
follows from stimulus synonymity for Quine is not synonymity in the
usual sense.67 What the argument we set forth above now comes to
exactly is that one can get from stimulus synonymity to synonymity of
sentences and with that to synonymity of the terms that occur within them.
Quine makes the further claim that for theoretic sentences stimulus
synonymity cannot be defined at all.68 But if, following our argument,
one accepts synonymity of observation terms, then synonymity for
theoretic terms69 can be defined, for example, in this way: The series of
theoretic terms tl9...9 tn of a theory T 1 is synonymous with the series
sl9..., sn of theory T 2 if the axioms of T 1 are transformed into the axioms
100 CHAPTER II

of T 2 by substituting st(i= 1,..., n) for tt and possibly the substitution of

other synonymous terms for observation terms.
For Quine, then, the decisive consideration is the claim that one can not
go from stimulus synonymity to synonymity in the usual sense, and we
have discussed this claim above.

NOTES

1 Among the good discussions o f Quine’s ideas on semantics are Harman [67] and
[68 ], for example.
2 Quine [60], p. IX.
3 In his reply to Chomsky’s criticism in Synthese 19 (1968/69), pp. 274-283, Quine
characteristically did not go into these points.
4 In many points, Quine’s remarks on the logical analysis of language (see e.g. [60],
Chapters III-V) coincide with the program of logical grammar. See Chapter 111.2 on this.
5 See Quine [64c].
6 See e.g. Tarski [35] and [44].
7 On this point see Quine [60], §43, pp. 206f. - Such an argument occurs in Grice
[56], p. 146
8 Quine brings in examples such as the expression ‘nothing’, which is used like a proper
name in the sentence ‘His briefcase contained nothing’, although it represents a quanti­
fier ("i Vx); and the expression ‘sake’, which is used in ‘for the sake o f’ like an inde­
pendent name, although it has no independent meaning. - While o f course ‘for the sake
o f’ is a fully idiomatic expression, that is not true of ‘meaningful’: This composite is
formed in a regular way out of components interpreted by themselves.
9 One can even get along with the basic predicate G (x, y ) by itself, for since the field
o f G (x, y) contains nothing but meaningful sentences and every sentence is synonymous
with itself, we can define S (*) by S (x ):= G (x , x).
10 On this point, see also Kamlah and Lorenzen [67], p. 85f.
11 Here again *A9 is a quasi-mention, i.e. to be understood as ‘the mention of A \
12 The transition from ‘say, that’ to ‘say*’ also has the advantage that the sentence
*x says* “Es regnet” ’, in contrast with *x says that es regnet’, can be regarded as a
grammatically well-formed sentence.
13 See e.g. Quine [60], § 44, and Quine [56].
14 For Quine, who as we shall see holds the concept o f synonymity to be not sufficiently
clear - especially for expressions in different languages - this point involves a funda­
mental difficulty for these attempts at translation. He seeks to avoid it by not explicat­
ing ‘say in Italian’ at all. - Davidson’s recommendation in [68 ] for translating (1) into a
direct sentence appears to be less felicitous. It proposes that (1) be replaced by (5) ‘The
earth moves. Galileo said this.’ - But what does ‘this’ identify? A sentence - then the
translation is just as inadequate as (2 ) - or a proposition - then we have not gotten
away from the realistic interpretation. - Quine makes still a further recommendation in
[60], p. 215f. According to it, (1) is rendered by (6 ) ‘Galileo said-that-the-earth-moves’,
where ‘saying-that-the-earth-moves’ is supposed to be a basic unanalyzable predicate.
But how is that predicate to be defined ? Surely by way o f ‘saying that the earth moves’.
Another reason for holding this procedure to be basically not usable is that it would
infinitely expand the number of basic predicates of the language. One could just as
 THEORIES OF MEANING 101

well interpret all sentences as unanalyzable sentential constants - because the concepts
‘predicate’ and ‘proper name’ are thought to be not clear enough, perhaps - only then it
would no longer be possible to give any explanation of how language works: how it is
possible to communicate an infinite number of states of affairs with a finite number of
linguistic conventions (with conventions concerning a finite number of linguistic
expressions), particularly when they are new states. - Quine’s remarks in [56] also be­
long in this context. What he wishes to do there is to replace the predicate ‘a said that
F(bY with ‘a said-F-of 6 ’ - a predicate that is not defined, despite the remarkable
properties Quine ascribes to it (for example, ‘a said that b had the property F ’ would not
be true, but ‘a said-F-of 6 ’ would be, if a ascribed the property F to someone he mis­
takenly took to be b). Finally, he wishes to replace it with ‘a said-‘F ’-satisfied-by 6 ’ -
a predicate that is also defined only for intensionalists, and for them by ‘a attributed to b
the property expressed by ‘F ’ ’ (in the reference language). - With such a procedure, we
must obviously ask what is supposed to be going on, when on the one hand talk of
intensions is characterised as not clear, but on the other hand in order to avoid it
predicates are introduced which either are not defined at all or are defined only by way
o f expressions that refer to intensions.
Quine’s suggestion in [69b] would be worth mentioning, finally: to reconstruct a
sentence like ‘0 believes that F (by along the lines of ‘a believes the class of possible
worlds in which F (b) is true’ - a suggestion that Quine himself treats with great reserva­
tions, for the range of worlds that come into the picture here is to be narrowed addi­
tionally by the fact that there is a wealth of further sentences that hold true in them
which hold true in the real world and enter into the definition of ‘F and ‘F ( * ) \ Also,
such classes of possible worlds, especially after the fashion o f Quine’s suggestion, are
really no more simply defined than propositions.
These and similar attempts at reconstructing intensional contexts are quite artificial,
however, and even Quine finally admits it, when he says: “There is, however, another
objection to taking linguistic forms as objects of the attributary and propositional
attitudes; viz., simply that that course is discouragingly artificial. With this objection
I sympathize.” ([58], p. 22).
15 Quine has no objection to a comparative concept of synonymity o f this sort. See
[60], p. 203. - N . Goodman also accepts a comparative concept of synonymity in [49]
and [52].
16 See Quine [60], p. 203.
17 This definition of analytic sentences is too narrow, to be sure, as the sentence ‘If a is
the father o f b, then b is not the father o f a 9 demonstrates, since it would surely be
ordinarily viewed as analytic. For this reason Carnap defines analytic sentences in more
general terms by recourse to meaning postulates. See also Stegmuller [69], p. 61.
18 Definitionally equivalent is used to characterize two expressions that result one from
the other upon replacement o f defined terms by the corresponding definitions (or
vice versa).
19 The statements about synonymity in a lexicon are statements asserting synonymity,
not stipulations about synonymity.
20 This example is not a happy choice, o f course, since in the sentence ‘A is analytically
equivalent to B ’ the sentences A and B are not used, but only mentioned. Instead we
could have used the example ‘It is necessary that A = B \
21 See Quine [60], p. 42.
22 For a discussion of Quine’s arguments against intensional semantics see D. Lewis
[69], p. 207.
102 CHAPTER II

23 The concept of analytic sentence is often taken more narrowly, so that only true
statements fall under it. - Kant’s definition in the Critique o f Pure Reason, which en­
compasses nothing but sentences o f the type A x (F (jc) => G (jc)), is completely unsuitable.
24 Carnap brings in this example in [63], p. 920. Arne Naess has given further precise
criteria by which synonymities and analytic propositions can be determined in [49] and
[53]. For tests of synonymity of predicates see Carnap [56], p. 238f. - The criterion
given by Carnap directly relates only to the analyticity concept of a single person, but
it can be expanded for language communities as well. For more of Carnap’s statements
on the problem of analytic propositions see Carnap [52], [55] and [63].
25 Cf. D . Lewis [69], p. 201 and Schnelle’s discussion in [73], p. 299 seq.
26 See also Harman [67] and Putnam [62b], where there are arguments along the same
lines.
27 To some extent, the converse o f this idea, that our language is also a co-determinant
o f our way of perceiving the world, will be discussed in Chapter IV.
28 In this vein, J. Stenzel says in [34], p. 16: “A word becomes determinate only as a
consequence of the context of meaning always to be presupposed in any concrete
language situation.” See also Chapter II.4.3.
29 Heger later modified these ideas again in [69].
30 In the phenomenon of words merging, there is also the case in which an expression
which has only one meaning has two etymological roots (e.g. French haut from Latin
alt us and Frankish hoh), or in the phenomenon of words splitting the case of two
expressions with different meanings having the same etymological root, e.g. French
compter and conter, from the Latin computare.
31 In this ambivalence, on which a polyvalence with respect to their grammatical
functions can supervene, is revealed their plasticity. See also Chapter III. 1.2 on this point.
32 See Quine [60], p. 203. - If a comparative concept of analyticity has already been
admitted (x is no more analytic than y), then o f course a classificatory concept (x is
analytic) could be formally defined as: 1 + 1 = 2 is at most as analytic as x.
33 Quine refers in this connection (e.g. in [51a], p. 41) to Duhem’s argument (see
Duhem [6 ], pp. 303-328), according to which it is possible to hold onto any scientific
hypothesis H whatever even when there are conflicting observations. For an observa­
tion sentence E follows not from H alone; E follows from H, rather, only relative to a
theory T (to which H belongs, or which says something about how the quantities in H
are measured, etc.) and distinctive limiting conditions A. But if T, H, A VE holds, then
what follows from ~\E is not “i H, but only “i ( T ^ H A/l). But that condition can be
satisfied by - \ T ^ H A A or T ^ H A - | A as well.
34 On the delimitation of the logical operators in Quine see e.g. [59], p. XIV. See also
Tarski [55], p. 418f.
35 Putnam, too, has urged that while there is no absolute distinction between analytic and
synthetic propositions, there is definitely a relative one [62b]. He sees the distinction in
somewhat different terms from the way in which we have characterized it, however: For
Putnam, an analytic proposition within a system o f assumptions, e.g. within a theory T,
is a basic proposition o f T, one that could not be given up without giving up the whole
theory T. Such propositions are, for example, Newton’s axioms of mechanics or
Maxwell’s equations for electrodynamics. - This can be said in response: Propositions
o f this sort can have in T an analytic character in the usual sense of being true by virtue
o f meaning if they (partially) establish the interpretation of theoretical terms in T ; but
they can also be synthetic propositions in the usual sense, so that Putnam’s characteriza­
tion o f analytic propositions deviates from the usual way of defining them. Only if T is
 THEORIES OF MEANING 103

universally accepted as true, so that the basic laws of T constitute central assumptions
about the world, for us an (at present) unproblematic foundation of understanding,
and explaining other propositions, can it be supposed that they are characteristic for
the meanings of the terms that occur in them. When Putnam in [62] offers the state­
ment ‘There is an infinite number of distinct, finite connected segments of space’ as an
example of a proposition which is analytic in the Euclidean geometry of classical
physics, but non-analytic in the Riemannian geometry of relativity theory, and remarks
that the analytic character of this proposition does not depend on the meaning of the
terms, which are the same in both physical theories, the reply is that this proposition is
not analytic either as a proposition of mathematical or of physical geometry. It is only
when the expression ‘segment of space’ is interpreted as ‘segment of Euclidean space’
that it becomes analytic.
36 On occasion, Quine’s arguments against the acceptance of analytic propositions are
interpreted as if Quine would not doubt the existence of a determinate set of analytic
propositions, but would say that this set is empty. But the point Quine is making from
the very beginning is that there are no sharp or systematically relevant boundaries
between truths o f fact and truths of meaning. - On the problem of analytic propositions
see also Bohnert [63], Gewirth [53], Kemeny [63], Martin [52] and [59], Mates [51],
Pap [58], Peach [52], Putnam [62], Wang [55] and White [50].
37 Quine [60], p. IX.
38 See Quine [60], Chapter II. See also Quine [58].
39 Quine [60], p. 27.
40 Quine [60], p. 27.
41 See Chapter II.4.4 on the logical and semantical concepts applied here.
42 It is sufficient to use an extensional semantics as a basis here, since Quine asserts not
only an intensional but also an extensional difference of some translations that are
possible in terms of linguistic usage.
43 Quine says: “Truth functions can be translated” ([60], p. 68 ). We can go on beyond
that and assume, for example, that the existence operator V x... (there is an *, such
that...) can also be translated. For the presupposition for that is not that yi and 72 are
identical (Quine [60], p. 60f.), or that X knows 7 2 , but instead V x... is already character­
ized as an existential quantifier in L2 if sentences of the form A [a] => VxA [x] are re­
garded as analytic (Quine says: “Stimulus-analytic sentences can be recognized,” [60],
p. 6 8 ; a stimulus-analytic sentence in this connection is a sentence that is affirmed in
every situation unless nothing is affirmed at all (see Quine [60], p. 55.)), and if a sentence
of the form VxA [x] =>C is always regarded as analytic when A [a ]^ C is regarded as
analytic (where a does not occur in C). Hintikka has proposed in [68 ] a semantic
instead of a syntactic characterization of quantifiers. It starts out from the assumption
that two languages can have as their basis a unitary domain of individuals, however,
which Quine expressly denies in [60]. Neither is our suggestion affected by Quine’s
objection to Hintikka’s proposal (Synthese 19 (1968/1969), 284-287) that it is possible
for it to occur in practice that someone asserts a sentence VxA (jc) without being pre­
pared to assert a sentence of the form A (a). This is, to be sure, not a verv weighty
objection.
44 I.e., (p correlates proper names to proper names, w-place predicates to w-place
predicates and sentences to sentences.
45 I.e.: If a term or sentence 0 in L2 is produced by combining terms or sentences
y/i, ..., y/n in L2 in accord with a formal rule, then (p0 should be produced by combin­
ing the expressions (py/1 ,..., (py/n according to the same formal rule.
104 CHAPTER II

46 If 0 describes W, that is to entail here that 0 is true.

47 Our definition represents a simplification and modification of Quine’s concept of
translation. Since all that we are concerned with here is the thesis of indeterminacy of
translations, we have taken up only those specifications of the Quinean concept in (I)
on which that thesis is based.
48 According to that it is not true of all sentences 0 in L2 that V 2 (# )= V 1 (<p0), but
only of sentences which describe a perceptual field W, particularly of observation
sentences, then. On the other hand (d) asserts more than this condition for the case of
such sentences. - If in (d) we speak of all perceptual fields of X, then there is an idealis­
ing assumption to the effect that in every perceptual situation X can check y ’s use of
language relative to that situation.
49 In Quine [60] the thesis is formulated in such a way that LI is a many-sorted lan­
guage, so that L2 can be mapped onto several component languages of L I. The claim,
then, is that for antecedently given VI and V2 the mapping (p is not unambiguously
fixed by conditions (a) through (d).
50 Quine says: “Observation sentences can be translated. There is uncertainty, but
the situation is the normal inductive one,” [60], p. 68.
51 See Quine [60], p. 53f., and Quine [58].
52 See Chapter II.4.6 on this point.
53 Wittgenstein [53], 293.
54 See Chapter II.4.4.
55 On the other hand, Quine’s thesis is weaker than what Wittgenstein says about
learning predicates when he assumes that we can learn the rule of use for observation
predicates inductively, thus that we can assign to a predicate F in L2 a predicate (pFin
LI, such that for all proper names a in L2 it is true according to (d) that V 2 (F (a ))=
=Vl(<pF(<pa)).
56 Naturally, we will not advance the hypothesis that translation of a foreign language
is merely a process of statistical choice. On the contrary, there enter into it constructions
that are creative through and through, similar to those of theory construction, especially
in setting up grammatical rules, of which we have not spoken at all here. What is impor­
tant, however, is that the translation hypotheses and theories can be inductively checked
in every particular.
57 See e.g. Uexkiill [24] and [28].
58 See Quine [60], p. 83, also Quine [69c], p. 123ff.
59 See Quine [57], p. 218.
60 See Quine [60], p. 26.
61 See Quine [60], p. 53.
62 See Quine [60], p. 76.
63 See Quine [60], p. 28, 76.
64 See Quine [60], p. 79.
65 Quine [60], p. 78.
66 Quine [60], p. 32f., 46.
67 See Quine [60], p. 46.
68 Quine [60], p. 46f.
69 On the concept of theoretical terms see e.g. Stegmiiller [70] or Kutschera [72], 3.3.
 THEORIES OF MEANING 105

4. WITTGENSTEIN’S PHILOSOPHY OF LANGUAGE IN THE

‘PHILOSOPHICAL INVESTIGATIONS’

After finishing the Tractatus in the period after the First World War,
Wittgenstein had at first entirely withdrawn from philosophy, since he
held the basic problems of philosophy to have been essentially solved and
disposed of with the Tractatus. Wittgenstein did not turn back to philos­
ophy until 1928. His attitude toward philosophical problems had already
changed in many respects since the Tractatus and so he believed he could
again accomplish creative philosophical work.1 Particularly since about
1933 Wittgenstein’s philosophical ideas changed radically. Above all he
was again concerned with the problems of language, but he now arrived at
views quite different from those he had advocated in the Tractatus. In the
Philosophical Investigations [53] he engaged in sharp criticism of the basic
ideas of the Tractatus. This work represents the pure pragmatic antithesis
to the pure realistic semantics of the Tractatus.
In what follows we will discuss first of all what Wittgenstein asserts in
[53] on the problem of meaning. But even these discussions will lead us to
the epistemological theme of Chapter IV, since Wittgenstein’s statements
about the semantic function are ultimately based upon epistemological
considerations.
The Philosophical Investigations are a collection of thoughts scarcely
connected with each other in any systematic way. Wittgenstein did not
wish to develop a semantic theory in them, but only to free us from ‘the
bewitchment of our intelligence by our language’, i.e. to fight false con­
ceptions of the function of language.2 But since we are here concerned
with systematic questions, our goal in what follows is not so much the
most faithful possible recapitulation and interpretation of Wittgenstein’s
statements as the clarification of the problems he has raised.3

4.1. The Abandonment of the Tractatus’ Ontological Presuppositions

In the Tractatus Wittgenstein had started from ontological presupposi­
tions according to which the world is divided up into complex and ulti­
mately into simple facts, which are once again composed of simple things
(objects and attributes).4 The task of the philosophical analysis of langu­
age was accordingly to provide an ideal language that is a true picture of
that reality; in which the simple terms stand for (simple) things and sen­
106 CHAPTER II

tence structure faithfully reproduces the structure of the facts pictured.

Since our ordinary language is no such language - its sentence structure
does not correspond to the way the facts pictured are objectively put
together, its names do not stand for simple objects, its meanings are not
precisely and exactly defined - it was accordingly the task of philosophical
analysis to analyze the sentences of ordinary language, which do not
satisfy the ideal of scientific exactitude, by means of sentences of the
ideal language.
In his later philosophy Wittgenstein turns against these basic ideas. The
most decisive factor is that he gives up the ontological presuppositions of
the Tractatus: There is now no longer any reality in itself for Wittgenstein,
no reality that is merely pictured by language, whose structures thus have
to accommodate themselves to the ontological structures. Instead it is in
the description in language that the world is first revealed to us. The world
is never given to us in and of itself, but only in an interpretation in lan-
guage^Thisepistemologicalideawill occupy us more extensively in what
follows. What we are immediately concerned with here is its far-reaching
consequences for Wittgenstein’s philosophy of language in the Philosophi­
cal Investigations:
(a) This idea cuts the ground out from under the realistic theory of
linguistic meaning. For if the entities, the things and attributes are not
given to us independently of language, then the meaning of linguistic
expressions can not be explained by saying that such entities are conven­
tionally assigned to them as meanings.6 In that case, we can not start from
the ontological structures either and attach isomorphic linguistic struc­
tures to them, but the ontological structures could rather be regarded the
other way about as projections of the linguistic structures given primarily,
in which we speak about the world.
(b) Furthermore, the idea cuts the ground out from under ontological
and with it linguistic absolutism and atomism. For what it means is that
there is no world built up in an unambiguous and definite manner (abso­
lutism) out of elements (atomism), that instead the world is articulated
for the first time in the linguistic description. How it can be articulated
depends on the linguistic forms available for describing it.7 The idea con­
sequently contradicts the notion of replacing ordinary language with an
ideal language, for precisely what had defined the ideal language was just
that it was supposed to picture the structure of reality in and of itself.
 THEORIES OF MEANING 107

But the structure of our world is only the structure of our language in
which we describe it, i.e. in essential features the structure determined
by the forms of ordinary language.
(c) With that, for Wittgenstein ordinary language now takes the place
of an ideal language, an artificial language modelled on a logical calculus,
in the center of philosophical attention. There is no more talk now of the
real logical structure of language and the real meanings of linguistic
expressions being hidden behind the inexact and vague formulations of
ordinary language. Instead what is said now is that in principle ordinary
language is in order just as it is - ‘ordinary language is all right’ - and
Wittgenstein no longer sees the task of philosophical linguistic analysis
in retreating from the formulas of ordinary language to sentences in an
ideal language, but conversely in retreating from the scientific use, espe­
cially the philosophical use - for Wittgenstein almost always a misuse -
of language back to the way words are utilized in ordinary language.8
This turn back in the direction of ordinary language marks the be­
ginning of a new philosophical movement, ordinary-language philosophy,
in the way it has been developed primarily at Oxford in Wittgenstein’s
wake as a counter-movement to the analysis of philosophical language,
using the instruments of modern logic, as propagated by logical empiricism,
for example. For it, as for Wittgenstein, most philosophical problems
stemmed from a misuse or a misunderstanding of ordinary language and
it is philosophy’s task to take up the “fight against the bewitchment of our
intelligence by our language” by attempting to understand better the way
ordinary language works.
(d) Finally, Wittgenstein’s epistemological idea of the world being put
into order by language speaks against the ideal of linguistic precision that
Wittgenstein had promulgated in the Tractatus with his demand for an
ideal language. Along with the idea that there are well-defined objects and
attributes in the world that come into question as exact meanings of
linguistic expressions, Wittgenstein also gives up the idea that there is
such a thing as absolute exactitude, particularly such a thing as absolute
precision in the meanings of words. The ideal of exactitude in the Trac­
tatus is for Wittgenstein a myth, a metaphysical fiction. Exactness is never
anything but an exactness sufficient for a specific context, there is no
absolute exactness.9 And there is no true sense of a sentence, to be worked
out with complete precision by means of analysis, one which is hidden to
108 CHAPTER II

some extent behind its formulation in ordinary language, what is really

intended by that formulation. Rather, the sense of any sentence is defined
only within certain limits and it remains open how these limits are to be
drawn closer together with a more precisely detailed definition.
The idea that reality is not given to us independently of its description
in language is thus the guiding idea for a radical criticism of the basic
ideas of the Tractatus and destroys their foundation completely.

4.2. Language Games

As we have already emphasized above, it is a basic characteristic of
realistic semantics to see the semantic function of language in its repre­
sentative or pictorial function alone and accordingly to identify the
meaning content of the expression with its descriptive content. According­
ly realistic semantics is almost exclusively oriented toward declarative dis­
course and completely severs semantic considerations from inquiries as
to how language is used in practice.
Now Wittgenstein pursues the pragmatic criticism of this realistic
approach. For him speaking is in the first place a human activity like
cutting wood, gymnastics, plowing, etc. This activity occurs in quite
diverse contexts of situations and actions and must therefore be seen and
analyzed against the background of these contexts. Wittgenstein speaks
in this connection of various forms o f life.10 For him language is always
part of a form of life, and since it can be part of various forms of life, its
function has to be defined anew for each form of life. As many forms of
life, as many contexts of action and situation as there are, there are that
many different ways of using language, that many different language
games as well. Language is used for giving orders, asking questions, de­
scribing objects, guessing, lying, telling jokes, making up stories, solving
riddles, giving thanks, pleading, greeting, praying, to cite some of the
examples Wittgenstein offers in [53], 23.11
Wittgenstein therefore wishes to study in the Philosophical Investiga­
tions the use of language in the pragmatic context and to relate all
analyses of meaning to these contexts of use. The talk of ‘language games’
is supposed to place special emphasis on the aspect of language use repre­
sented by the fact that it follows specific rules in each of the various life
contexts, that different systems of rules hold for the various language
games. Indeed the rule-governed character of language is the presuppo­
 THEORIES OF MEANING 109

sition of communicating by means of language: without firm rules for

using a word it has no firm meaning.
What is fruitful about this idea of fitting language into its pragmatic
context and studying it against that background is that the realistic
semantics’ overly abstract and overly narrow interpretation of language
as a mere instrument for describing things is corrected, that the variety
of modes of meaning is brought out and that the importance for semantics
of the pragmatic background is emphasized. On the other hand the
manner of speaking about language games may not be taken too literally.
For in the first place, we can not overlook the fact that we use the same
language in all action contexts, not different languages. That means,
however, that words and language forms are invariant within certain
limits as against the varying contexts of their use, and that their meanings
are not simply different from one occasion of their being used to another,
but remain more or less constant. And so analyses of meaning and com­
parisons of meanings of words can in fact often be carried out to a con­
siderable degree independently of the context of their use. Nor could we
lay down new rules of language use for every context of life, so that in
every new situation and activity we would be without any language to
begin with. Instead the serviceableness of our language depends precisely
on the fact that it can be used in the most varied contexts and in new
contexts as well.
According to Wittgenstein we should not let the external resemblance
of words and language forms in different language games blind us to the
fact that they have quite different functions in the various games, but
this external resemblance is nevertheless definitely not just a matter of
chance. Of course, there is a difference between the uses of language in
different contexts, which has to be noted.12 On the other hand, however,
there is also a relationship, which ought equally to be noted. How, for
example, is the use of the word ‘door’ in telling a joke differentiated from
its use in thanking someone?13
What concerns Wittgenstein above all in his later philosophy is to
stress differences, point out nuances and draw distinctions, to exhibit
the value of the particular case as against the generalization. But general­
izations - even if not too far-reaching - are necessary for a scientific
systematization of the variety of experiences.
Furthermore, particular language games, in contrast to other games
110 CHAPTER II

such as chess, football, etc., are not sharply set apart from each other and
overlap. Questioning, describing, suggesting, asserting, proving, for
example, are ‘games’ which are inseparably tied up with each other in
many situational contexts.
Taken literally, Wittgenstein’s assertions about words belonging to
particular language games are most readily applicable to such words as
‘Re’ and ‘hand’ which only occur in quite specific practical contexts (in
this case in the card game skat) and in that context are used according to
quite definite rules of play, but not to the mass of descriptive words in
our language.
But so far as the principle intended by the talk of language games
is concerned, expressing as it does the thesis that language is basically
always to be understood and analyzed in terms of its use in specific prac­
tical situations and that the variety of semantic functions corresponds
to the great diversity of such ways of applying language in practice, these
assertions of Wittgenstein’s contain a very important insight.14

4.3. Word Usage and Word Meaning

According to realistic semantics a word as a phonetic or graphic entity
becomes meaningful by virtue of having a proposition, an object or a
concept assigned to it as its meaning, so that along with meaningful
words there are always antecedently existing objective, even if not neces­
sarily concrete entities which are their meanings. If the ontological pre­
suppositions of realistic semantics are given up, especially the underlying
Platonism, this interpretation is no longer tenable. Concepts and propo­
sitions are in that case nothing but abstractions from predicates or sen­
tences based on synonymity, and so they can not be called upon to ex­
plain the meaning of predicates and sentences. But then a new definition
of the meaning of linguistic expressions or their synonymity must be given.
Wittgenstein holds that a word becomes meaningful by virtue of having
a definite function in a language game, by being used in a definite way in
that language game and for a definite purpose. According to Wittgenstein,
then, for a word to be meaningful there does not need to be something
which is its meaning. In this vein Wittgenstein says, “ Don’t look for the
meaning, look for the use”,15 i.e. if one wishes to understand and explain
the meaning of a word, one should not set out to look for concrete or ab­
stract entities which are assigned to the word, which it designates, but one
 THEORIES OF MEANING 111

should look into how the word is used. It is not just that we can obtain
information about the meaning of a word only from its use, but rather
the meaning is nothing whatsoever except the word’s use; besides the use
there is nothing else that constitutes the meaning of the word. According
to Wittgenstein, “for a large class of cases in which the word ‘meaning’ is
used - even if not for all cases of its use - the word can be defined as
follows: the meaning of a word is its use in the language.” 16
Wittgenstein illustrates this identification of meaning and use in [53],
454 by alluding to the meaning of non-linguistic signs, such as an arrow
pointing the way. The arrow ‘ -> ’ does not mean a fact, a piece of advice,
a direction, or the like, but is put up so that a person gets to where he
wants to go if he follows it. It is this use that makes the arrow a meaning­
ful sign; there is no need to assume a meaning-entity along with it, as­
signed by convention to the arrow as a sign.
This definition of meaning as use is simple and natural - simple,
because the entire cosmos of propositions, concepts and the like, which
remained doubtful in its ontological status, now becomes unnecessary.
And natural, because there is certainly no dispute about the fact that a
linguistic expression without firm rules for its use has no firm meaning
and that there is at least a gross one-to-one correspondence between use
and meaning.17 It has the further advantage that by reducing semantics
to pragmatics it brings into view the variety of meanings linguistic ex­
pressions have, the definitive role of the realization of language {parole
in Saussure’s terminology) for language as a system (langue), and the
cultural achievement that is expressed in a language, of which one could
scarcely speak if it were merely a system of assignments [of meanings].
This definition of meaning will also be essentially more in accord with
many linguistic-empirical facts - some of which we shall be discussing
below. Finally, this definition is also epistemologically much more satis­
fying than the realistic representative-image theory: since propositions
and concepts, as things given independently of language, disappear, the
problem of how these entities can be known also disappears and we
have access to the phenomena that confirm the dependence of thought and
experience on language.
The identification of use and meaning requires further explanations and
distinctions, however, before it can be accepted as a basic systematic
feature of semiotic. Although the term ‘use’ is a key word in Wittgen­
112 CHAPTER II

stein’s philosophy of language, as it is for the whole pragmatic semantics, it

remains quite vague for the most part and is used with a number of mean­
ings. And little effort has been taken to make this term more precise.18
The first thing to be emphasized is that Wittgenstein does not under­
stand the use of a word, along the lines of behavioristic semiotic, as a dis­
position to utter the word under certain circumstances, or to respond to
it in a certain way. For Wittgenstein, ‘use’ is not a behavioristic term.
The use of a word is also not defined by counting up the occasions on
which it is employed and reporting that so-and-so used the word on such-
and-such an occasion, in this or that situation. I.e., what is meant by ‘use’
is not a set o f instances o f use but the manner o f being used. This manner
or way of being used can be defined by general rules of use which say
that the word is used in cases of such and such a kind. It is one of the
important accomplishments of language that it can be used to communi­
cate new information. How language is employed for this purpose, i.e.
what meaning sentences have when they are so used, is not revealed by his­
torical instances of language use, however, which are indeed irrelevant to
new facts, but only by general rules of use applicable in these cases as well.
Talk about the use of words does not refer to some actually occurrent
use, either, but to their correct use in accord with a linguistic standard.
If one wishes to identify the meaning of a word with its use, then one must
refer to the rules for its correct use, i.e. to the generally accepted way in
which it is used, not to mutually incompatible modes of usage that
deviate from the general standard.19
Various objections have been made to the identification of meaning and
use. A discussion of them will make the content of that definition clearer:

(a) It has sometimes been said that the formula ‘meaning= use’ is false
because there are many ways in which words or sentences are used that
have nothing at all to do with their meanings.20 So, for example, we use
expressions to insult, to amuse, to flatter or for aesthetic reasons (e.g.
because they suit the rhythm of speech or because they are alliterative),
without these ways of employing them being relevant to the meaning of
these expressions.
This objection suggests an important distinction: The use of language
occurs concretely in single speech acts, in utterances in the sense of
Chapter 1.1. These speech acts are related to specific situations and non­
 THEORIES OF MEANING 113

linguistic activities. But in these individually very diverse utterances we

use the same language, the same words and sentences as expressions. What
is meant by talking about the ‘use’ of expressions, then, is not actions
but the outcomes of actions, phonetic or graphic expression-objects. One
can say, ‘I am using expression (-object) a’ instead of ‘I am performing
the expression (-action) which has a as its phonetic result’. The use of an
expression thus represents an utterance of that expression. Now we urged
in Chapter 1.1 that the meaning of an utterance of an expression a as
opposed to the meaning of a itself often included supplementary features
which emerge from the pragmatic context of the utterance. We called
attention to indexical expressions in that connection; but the meaning of
descriptive predicates can also be determined in more specific detail by
the context of an utterance. Just as the intended sense of an ambiguous
word can be picked out by its linguistic context (‘She married a crook’ -
‘She carried a crook’), it can also be rendered specific by the pragmatic
context of the utterance (‘This crook is a very smooth one’). Such supple­
mentary specification of meaning by the circumstances of utterance does
not come about only in the case of ambiguous words. For example, if
someone who wants to clip his hedge says to me: ‘I need a pair of shears’,
what kind of shears is clear from the pragmatic context here. In this con­
text, the utterances ‘I need a pair of shears’, and ‘I need a pair of hedge-
shears’, have the same meaning, even though the two sentences (as expres­
sions) are not generally synonymous. A third example is the dependence
of what J. L. Austin calls the illocutionary role of an utterance on its
context. The sentence ‘It’s raining’ can be employed as a report, as a
warning (‘Don’t go out without your umbrella’) and as a request (‘Close
the window so the rain won’t come in’). Which role it plays on a given
occasion does not reside within the sentence itself, but is a function of the
pragmatic context of the utterance.21
And so we must differentiate between the meaning of an expression and
the meanings of the corresponding utterances. The meaning of the ut­
terance is a function of the meaning of the expression and its pragmatic
context: What belongs to the meaning of the expression is only what all
the utterances have in common as regards meaning. The thesis ‘use=
meaning’ is thus to be understood in such a way that the manner of use that
is typical for all of the utterances determines the meaning of the correspond­
ing expression. Between the way an expression is generally used and the par­
114 CHAPTER II

ticular occasions of its use there can then still be more or less specific forms
of use which are typical for certain kinds of context. What corresponds
to them are more specific meanings of the expression in such contexts.
So if the claim is advanced that there are many specific uses of a word
that are irrelevant to its meaning, that is true enough, but it represents no
objection to the thesis in question: not all of the characteristic features of
a particular occasion of an expression’s use are relevant to its meaning;
often they do no more than define the meaning of the use on that occa­
sion - often they are semantically entirely irrelevant.
The argument against identifying meaning and use has still another
aspect, which is emphasized by Searle.22 He points out that besides
semantical criteria for the use of expressions there are also many pragma­
tic criteria for their use that are semantically quite irrelevant. In what we
say we also take into account whether it is suitable in the situation, polite,
appropriate, informative or expedient. How can we then distinguish
between semantically relevant and irrelevant criteria in order to save the
thesis meaning= use ?
Again the decisive difference will be that semantic criteria for the use
of an expression refer to all its applications while the other criteria are
only brought in for some utterances. Criteria for the use of obscene ex­
pressions and their equivalents in medicinal terminology, for instance,
are quite general and therefore we have here a difference in meaning in
spite of an identity of descriptive content. On the other hand the fact that
we cannot substitute the two expressions ‘stairs’ and ‘series of steps for
going up and down’ in all situational contexts is no argument for a
difference in meaning. That we do not express a warning ‘Don’t fall down
the stairs!’ by ‘Don’t fall down the series of steps for going up and down!’
is explained simply by the fact that in warning a person against an immi­
nent danger you have to be brief.23

(b) G. Pitcher, M. J. Charlesworth and P. Ziff have attempted to show

by examples that the identification of meaning and use holds as a rough
approximation at best.24
For example, Pitcher says that one can know the use of a word like
‘Amen’ without knowing its meaning, just as one can know the meaning
of a word, such as the Latin ‘ultor’ (avenger), without knowing how it is
used (in Latin). Charlesworth brings out the fact that an identification of
 THEORIES OF MEANING 115

meaning and use is impossible in the case of proper names, which definitely
have a use but no meaning.25
All these counter-examples are not very convincing, however. The word
‘Amen’ is not used descriptively, but as a formula of confirmation, and as
such what determines it as to its meaning is its use, not insights into its
origin and its earlier meanings. Further, do I know the meaning of ‘ultor’
if I do not know Latin and so I am unable to cite any contexts in which it
is used significantly and can take on various nuances of meaning? In reply
to Charlesworth there is this to be said, that in identifying use and mean­
ing for ostensive proper names what is to be understood by ‘meaning’ is
the semantic function of the name, i.e. its reference.
With the explications that have been given, then, these objections do
not speak against accepting the formula ‘use= meaning’ as a basic seman­
tic hypothesis.

(c) A further objection reads as follows: If the meaning of a predicate,

for example, is defined as the way in which it is used and this way of being
used is determined by rules for its use, then little has been gained over the
realistic position. Rules are no less abstract than concepts and therefore
the formulation: ‘The meaning of a predicate is the rule for its use’ is no
more or no less problematic than the formulation ‘the meaning of a
predicate is a concept’.
The pragmatic proposal, Wittgenstein’s proposal in particular, does
not amount to interpreting rules as the meanings of predicates, however,
but to replacing the predicate ‘x means y 9by the predicates ‘x is meaning­
ful’ and ‘x and y are synonymous’, along the lines of Quine’s theory.26
The first predicate is to be applied when the expression jc is used in such
a way that the way in which it is used is controlled by general rules and
the second predicate when x is used in exactly the same way as y . An ex­
pression is not meaningful when there is a rule which it means, but when
there is a fixed way of using it - there is no longer any need for a meaning,
an additional entity it stands for. The rule of use is not a hypostasized
something alongside of the rule-governed use, but consists in the fact
that we generally use the expression in such and such a way.27

(d) A fourth objection is as follows: In many cases, especially in the

case of words that are primarily used descriptively, the use of a word W is
116 CHAPTER II

defined most simply and appropriately by saying: W is used for the pur­
pose of expressing..., in the sense o f ..., or as a designation fo r.... But if
the use of W is defined by way of its meaning in this way, then it is circular
to define the meaning of W the other way around by way of its use.28
According to this view, the formula ‘meaning= use’ fits best with expres­
sions that have no descriptive meaning, particularly performative utter­
ances like ‘Good morning’, ‘Beg pardon’, ‘I ask’ (promise, assure, warn,
etc.)..., which do not speak of or describe an act (greeting, excusing,
asking, etc.) but with which we accomplish these acts.29 What can be said
here is ‘ “Good morning” is used as a greeting in the morning’, ‘ “Beg
pardon” is employed when one is asking to be excused’, and so on. That is,
in these cases the use of the expressions can be described without recur­
ring to their meanings, but with descriptive expressions that is not the case.
This objection has some justification in two respects. If one attempts
to give rules of use for descriptive words, one will for the most part refer
to the meanings of the words. Thus, for example, one will say (a) ‘the
predicate “rot” is used in German as an expression for the quality “red” ’.
Such descriptions of use by referring to the meaning of words are also by
far the simplest if appropriate synonymous expressions are already avail­
able and their meanings known. From the standpoint of the ‘use-theory’
one can say, however, that a description of the use of a word of this sort,
by citing its meaning, is nothing but the clarification of the use of that
word by using other words. That is, (a) says the same as (b): ‘The predi­
cate “rot” has a use in German corresponding to the use of “red” in
English’.30
This sort of locution, that use is defined in terms of meaning, or the use
of formula (a) instead of (b), is on this interpretation, viewed pragmatical­
ly, quite unproblematic.
On this understanding, the realistic sort of talk has its due also, and
meaning can be defined as a concept or a proposition, for example,
wherever linguistic expressions are available by which the concept or
proposition can be represented as an abstraction after the manner of
Chapter II.3.1. It is not so, then, that all of the statements made in realistic
semantics are proved to be inadequate and false by the pragmatic ap­
proach identifying meaning and use. Their title as practically very useful
forms of expression is preserved, rather - within the limits, to be sure,
within which they are translatable into statements made in pragmatic
 THEORIES OF MEANING 117

semiotic. E.g., (a) as against (b) shows the advantage of realistic formula­
tions: (a) avoids the difficulty of having to explain in greater detail the
‘corresponding usage’ of words in English and German. The terminology
of realistic semantics is actually also essentially better developed and
more precise than that of pragmatic semantics - for example, there is no
formulation of general semantic rules for grammatical compounds in
purely pragmatic language - so that for this reason alone it is indispens­
able in practice.
If we can adhere to the realistic way of speaking over wide areas, we
still can not do so everywhere. The basic realistic thesis, that a predicate,
for example, becomes meaningful by having a concept assigned to it,
loses its sense on the pragmatic interpretation in cases in which there is no
suitable predicate available for defining the concept in the first place.
This formulation breaks down, then, when it is a matter of saying how
linguistic meaning comes to be at the outset.
Because of the importance this problem of establishing predicate-
meanings has for understanding the thesis of the identity of use and
meaning, in the following section we will contrast Wittgenstein’s prag­
matic statements on the point with realistic statements and discuss them
extensively.
First, however, we wish to evaluate the second aspect under which the
objection that defining meaning by way of use is circular has a certain
justification. Even if the circular definition of the meaning of, for example,
the sentence, ‘It is raining’, (c): ‘ “It is raining” is used to express the fact
that it is raining’ is replaced by (d): ‘ “It is raining” is used to say that it
is raining’, the circularity is still there. For ‘to say that it is raining’ means
to use the sentence ‘It is raining’ or some synonymous sentence, so that
one can only understand definition (d) of the use of ‘It is raining’ if one
is already acquainted with its use - just as one can only understand defi­
nition (c) of the meaning of ‘It is raining’ if one already understands the
meaning of the words ‘It is raining’ that occur in the definition. But so far
no one has been able to specify the use of a descriptive sentence like ‘It is
raining’ without using that sentence or another one synonymous with it -
falling back on the known identical usage of a synonymous sentence does
not solve the problem of how to explain that use of the sentence in the first
place.31
This objection is based upon a mistaken understanding of the ‘use
118 CHAPTER II

theory’, however. That theory is not a naturalistic theory of language -

such as behaviorism, for example - which wishes to reduce language to
the non-linguistic and define linguistic concepts in terms of general
behavioral concepts, for example. It does not wish to derive what saying,
asking, or arguing is from other distinctions. On the contrary, the theory,
particularly as Wittgenstein has stated it, must be understood as empha­
sizing the independence and distinctiveness of language. It does so pre­
cisely by seeking to explain the function of an expression not in the
realistic spirit as being attached to an entity independent of language and
capable of being apprehended without the mediation of language, but by
saying: meaning is immanent to language, it is first constituted in language
and is not anything alongside of language. Meaning is only disclosed in
use, in the realization of language, i.e. the distinctive work done by lin­
guistic expressions, that of meaning something, lies in the particular way
in which they are employed. We understand language in its general func­
tions as in the special functions of individual words by learning to use it,
not from outside as it were.
The thesis, then, is not that meaning can be reduced by definition to a
use that does not presuppose meaning and the latter to a form of behavior
that does not presuppose language - such reductive explications are not
possible in most cases - but that linguistic functions only reveal them­
selves in language use, which can be taught and, when an understanding
of language can already be presupposed, can be explicated by means of
descriptions in language.
X ’s question, ‘What does the word a mean?’, is answered by teaching
him the use of that word or, to the extent that the necessary expressions
are known, explaining it to him in language. The general question, ‘What
is meaning?’, is answered by saying that the expression ‘to be meaningful’
marks the specific work done by linguistic expressions, which is based
upon the particular way they are used and is revealed only in use. That is
the thesis of meaning as use, not a claim of reducibility. It is thus much
weaker and, if you wish, more trivial than has frequently been assumed.
Finally, in reply to the argument that the use of a word is determined
by its meaning and so can not define the latter, the following comment is
to be made: When L. Antal, for example, says, “If words are used ac­
cording to their meaning, then meaning comes before use, in the same way
as the use of a language presupposes the knowledge of that language,” 32
 THEORIES OF MEANING 119

then it is obvious that what is being spoken of here is a specific use of an

already meaningful word, which is determined by its meaning and does
not define it. What Wittgenstein is referring to, however, as already
emphasized, with his identification of meaning and use is not instances of
use - his opinion is not that a meaningless expression would become
meaningful by being used one or more times - instead he says that the
meaning of a word becomes attached to it by way of general rules for
using it. That the individual cases of using a meaningful word are deter­
mined by its meaning is clear. But according to Wittgenstein all that
means is that they are determined by the general rules governing the
word’s use; a correct use is defined for the word, and it becomes meaning­
ful in the various instances in which it is used only when those rules have
been established.

4.4. Meaning and Use o f Predicates

Let us assume that a person X , who is acquainted with the use and the
meaning of a one-place predicate F unknown to us, wishes to teach us its
meaning and use. How will X proceed? If a suitable one-place predicate
G is already known to us, then X can say:
(a) F is used in the same way as G, or
(b) F means the same thing as (is synonymous with) G, or
(c) F means the concept that is expressed by (represented by) G.
The realistic accounts (b) and (c), viewed pragmatically, are only
alternative formulations for (a) and so quite harmless. Since furthermore,
as the realist sees it, the use of F is a use for expressing a concept, he will
interpret (a) as an equivalent version of (b) or (c), so that in this case the
two conceptions are compatible with each other.
But how is it when no such predicate G is available to us, none for which
(a), (b) or (c) holds true? Now the realist cannot say that X explains the
meaning of F to us by pointing out the concept for which F stands. For a
concept can be directly specified only as the meaning of a predicate, which
is not available in our case, but not by pointing, for one can not point to
concepts as abstracta. The method of teaching by examples constitutes
the only way for the realist as for the pragmatist: X teaches us the meaning,
or the use of F by telling us that F belongs to, or can be applied to these
objects and those objects, but not to these and those others. In this fashion
120 CHAPTER II

we obtain two classes of objects, K (F) and R (F), of which we know that
F belongs to all of the elements of the first and to none of the elements
of the second. Thus K (F) and R (F) include examples for employing F
correctly.33
Has the way in which F is used now been determined by means of such
an exemplary display of F ’s use? Obviously that is the case only when
the domain of objects D(F) for which F is defined at all, of which it can
be meaningfully affirmed or denied, then, coincides with the union of
K (F) and R (F). But now the cases of predicates F fo r which D(F) can be
given in the form of a finite list of objects are precisely the uninteresting
special cases. For in the case of predicates that are defined by the fact
that they belong to specified, enumerated objects and do not belong to
other specified enumerated objects, every application has an analytic
character. In order to make synthetic statements with F, and so be able
to ‘communicate new facts’,34 D(F) must be an open, i.e. an infinite
domain. But in that case, which we shall always have in mind in what fol­
lows, defining F for objects in K (F) and R (F) tells us nothing about
whether F can be applied to a new object not contained in K (F) or R (F).
Examples of using F do not distinguish any general criterion for the use
of F. If D (F) is infinite, then there are uncountably many concepts that
pertain to all of the objects in K (F), but to none of those in R (F ), or of
rules according to which F can be applied to all of the objects in K (F)
but to none of them in R (F).
The example Wittgenstein uses to illustrate this fact is the task of
determining the formation rule for an infinite series of numbers of which
only a finite number of initial elements are given. Such tasks are encoun­
tered in psychological intelligence tests, for example. The initial elements,
1, 4, 9, 16, say, of a series are given and the subject of the test is asked to
write down the next number in the series. If he now writes down 25, the
tester looks on him as intelligent, but if he writes 3, for example, he will
get bad marks for intelligence. But most unjustly! 25 is no more correct
than 3 or any other number, for there is no one correct continuation of
the numbers 1, 4, 9, 16 into a series, but there are just as many possible
continuations as there are numbers. A series is a function which assigns
a number to every natural number (1, 2, ...), and what is asked for is a
function f ( x ) for which / ( 1 ) = 1, /( 2 ) = 4 , /( 3 ) = 9 and /(4 )= 1 6 hold
true. One such function is f (x) = x 2, but another is given by the defini­
 THEORIES OF MEANING 121

tion ‘/ ( x ) = x2 for x < 4 and f(x) = 3 for x> 4 \ for example. Any arbitrary
continuation into a series is thus compatible with a finite initial section and
likewise any arbitrary continuation of the application of F is compatible
with the examples in K (F) and R (F). How, then, is one supposed to
infer from the definition by example which continuation is the correct
one? That obviously has to be possible, for the learning of predicates
by examples functions quite well enough in practice.
The only answer to the question of the correct continuation of a predi­
cate beyond a finite set of examples that is found in the Philosophical
Investigations is that the correct continuation is the one that is in accord
with the use of the predicate by the linguistic community.
This statement is certainly correct to this extent: Linguistic expressions
have meaning only by virtue of a convention which is expressed in the
case of words in ordinary language by general (correct) language use.
And so if the question is raised as to whether the predicate F can be ap­
plied to object a, what has to be done first of all is to fall back on the rule
for the use of F as a linguistic convention. It provides a criterion of
application R F; whether or not a satisfies this criterion is a factual ques­
tion. We will assume that a rule of use for F has the simple form:
(I) F may be applied to an object a if and only if R F(a).Sb
This statement gives us no answer to our question, however, for that
question is: presupposed that there is a general rule for the use of F, how
can this rule, in particular how can the criterion of application RF be
ascertained on the basis of individual cases?
Many of Wittgenstein’s statements in the Philosophical Investigations
suggest, however, that he sees more in the general use of language than
just the condition of the correct use of a word. They suggest, with (I) for
example, that he understands general linguistic usage to be the criterion
as well, so that R F is a criterion which refers to general usage. Thus he
says that you teach words to someone by example and practice. “And in
doing so, I am not telling him any less than I know myself.” 36 If we pay
attention to the fact that Wittgenstein makes a distinction between what
one knows and what can be said, e.g. between the distinctions one can
make and those that one can formulate in language, what this seems to
imply is that someone who knows how a predicate F is used is only
acquainted with the use of F in a set of examples (which is of course very
122 CHAPTER II

large in comparison with the learner’s), that the ‘general rule’ for the use
of F with which he is acquainted consists in the fact that he knows that
F is ‘generally’ used in such and such cases. Wittgenstein appears, then,
often to have in mind instead of a way of using F determined by a general
rule the instantial use of F over a large class of examples, and to see in it
‘linguistic usage’.
Furthermore, in the Philosophical Investigations Wittgenstein rejects
the possibility of private languages, i.e. of expressions introduced
privatim - Wittgenstein is thinking above all of expressions that stand for
a subject’s private experiences, mental events, etc. - and one argument
against them is that one could not speak of a correct use with respect to
the words of a private language, since the general criterion of correct use -
agreement with the usage of the linguistic community - is here in­
applicable.37 But if what one understands by this criterion is a general
condition of adequacy that concerns the way a word is used, the general
linguistic criterion (I), for example, then the usefulness of the criterion is
independent of how many members there are in the language community,
even should there be only one. Understood in this way, then, Wittgen­
stein’s claim cannot be applied as an argument against private languages.
Therefore it is natural to interpret it in such a way that Wittgenstein
understands general linguistic usage, the usage of the other members of
the language community, as the criterion for the use of words. For then
in the case of words in a private language, this criterion of use becomes in­
applicable, since there is no general usage, and one can now say with
respect to the correct use of these expressions: “ ... what is correct is
whatever is going to seem correct to me. And all that means is that there
can not be any talk of ‘correct’ in this case.” 38 Along these same lines,
what Wittgenstein means by ‘following a rule’ is also a general practice.
What is followed in it is what the language community consistently does.39
Now it is no doubt true that someone who wants to learn the use of a
predicate F, first turns his attention to its use as employed by the language
community, or his teacher X , and that he can be corrected and be ap­
proved in his own uses of F. But if he has grasped how F is used, i.e. the
criterion of its use R F has become clear to him, then he will be indepen­
dent of others in his uses of F an d will no longer need to pay any attention
to what the others say. The question, whether F can be ascribed to a new
object a, is no longer a question of language use for him, then - that has
 THEORIES OF MEANING 123

been fully established by his knowledge of convention (I)40- b u t a

question of fact, the question whether condition R F is fulfilled in a’s case.
If the language behavior of others were the criterion, every time F was
applied to a new object, there would have to be a vote on whether F
should be ascribed to a or not. All cases of using F for which F is deter­
mined, on which agreement has been reached, would then be components of
the definition of F, however, without empirical content and so no new
facts could be communicated by means of F.
But Wittgenstein did not intend his claim that even the criterion for
applying words is a criterion of intersubjective agreement in word usage
in this totally inadequate way. That is revealed by his comment that what
is the same is not “what all or most men agree in regarding as the same....
For in order to verify sameness, naturally I do not make use of the agree­
ment of mankind! - What criterion do you use, then? None at all. Using
the word without a justification does not mean using it wrongly.” 41
Now here the capacity for learning a predicate by examples, generally,
for following a rule makes its appearance as a basic human capacity that
cannot be analyzed further. And Wittgenstein addresses himself to the
problem in the Philosophical Investigations predominantly in this vein.
We just can actually learn rules by example, but no more can be said
about it. Why we follow the rule in this way and not another cannot be
supported by further reasons: “ ‘How can he know how he has to con­
tinue on his own?’ - Well, how do I know? - If that means: ‘Do I have
any reasons?’, then the answer is: My reasons will soon fail me, and then
I shall do without reasons.” 42 And: “ ‘... how am I able to follow a
rule?’ - If that is not a question about causes, then it is about the justifi­
cation of my acting in the way that I do. If I have exhausted my reasons,
then I have now reached bedrock and my spade turns back on itself. In
that case, I am inclined to say: ‘That’s just the way I do’” .43 And: “With
what right do I say: ‘Yes, that is red’? Well, I say it; and it cannot be
justified. And for this language game, too, ... it is characteristic that it
goes on with the peaceful agreement of all mankind.” 44
But it is quite unsatisfactory, as regards learning predicates from exem­
plary cases, which is after all a fundamental example of how language is
acquired and how it functions, that we are supposed to rest content with this
account, that what we have to do with here is a phenomenon not capable
of being analyzed further, a basic process that no light can be shed upon.
124 CHAPTER II

From the realistic standpoint, it could be explained as follows: Of

course it is true that a criterion R p for using F is not unambiguously
determined by examples of F ’s use taken from K (F) and R (F), but the
learner can consider various concepts or conditions R p such as are com­
patible with K (F) and R (F), i.e. that belong to all of the objects in K (F)
and to no object in R (F). By observing further instances of the use of F
he can inductively confirm or overturn these interpretive hypotheses and
so succeed in picking out a condition R F by induction. Of course at the
outset there is an infinite number of hypotheses that are compatible with
K (F) and R (F) and new instances of the use of F will confirm all of the
hypotheses that are not refuted, so that an initial selection must be made
from among the hypotheses which will make it practically certain from
the beginning that only a finite number of them will have to be taken into
account. But we have already explained in Chapter II.3.3 on what basis
such an initial selection can be made.
This realistic approach now appears at first look to be pragmatically
acceptable as well, if one is somewhat more sparing of realistic terminolo­
gy and does not call the criterion of use a concept marked out by the
illustrative classes K (F) and R (F), but possibly a distinction identified
inductively by means of those classes. On this view, we have at our dis­
posal a capacity, independent of language, for making certain distinc­
tions and these distinctions provide us with the criteria for using lin­
guistic expressions.
Wittgenstein emphatically rejects such an interpretation of learning
predicates, according to which the realistic way of speaking of concepts
that are independent of language is simply replaced by talk of distinc­
tions that are independent of language. It becomes clear in this connec­
tion that for Wittgenstein linguistic pragmatism is not just a different
fagon de parler from realism, but is based upon a fundamentally different
understanding of the phenomena. Befitting the pragmatic theory of the
priority of predicates over concepts, he refuses in the first place to explain
the use of predicates by resorting to concepts. The use of a predicate F ean
not be explained by a concept R p because concepts can only be obtained
by abstraction from predicates. But the same thing also holds for distinc­
tions. According to Wittgenstein we learn distinctions only by means of
language and the capacity for distinguishing things only by means of
predicates. He says: “ ‘How do I know that this color is red?’ - One
 THEORIES OF MEANING 125

answer would be ‘I have learned [English]’.”45 That is, in learning the

predicate ‘red’ I am not learning the abstract rule, “The predicate ‘red’
may be applied to a if and only if a is red,” but I am learning to orient my­
self in the world with the help of the word ‘red’, learning to distinguish
things into red and non-red.46
There is an obvious reply to be made to that: If we did not have the
capacity for making quality distinctions without language, then we
could not, as was already stressed in Chapter II.3.3, learn any language
either, for learning a language presupposes at least the possibility of dis­
tinguishing linguistic expressions one from another, and the situations in
which they are applied, too. And so there would surely be no possibility of
learning language entirely without the pre-linguistic or language-inde­
pendent drawing of distinctions. Even granted that many distinctions are
first conveyed by means of language or are made more precise by means
of it, language is nevertheless certainly not the single source of all dis­
tinctions.47
This objection is true enough, but it is nevertheless incomplete. In order
to understand that and to do justice to Wittgenstein’s statements, we
must have recourse to some epistemological considerations.48
Pre-linguistic distinctions, Quine’s pre-linguistic qualities and Uexkull’s
significances and valences to which we referred in II.3.3 are of quite a
different sort from most of the conceptual distinctions we formulate in
language. They are related to a subject’s private experiences, whereas
linguistic predicates for the most part mark out distinctions that relate
to things, to the world of objective, e.g. physical events.
Now what holds true objectively is basically also defined by the fact
that it holds true intersubjectively. As a rule we are only prepared to
accept a sentence about the world as true if an intersubjective agreement
on its truth can be reached. But if intersubjective agreement is thus
acknowledged to be a defining criterion of objective facts, then one must
also say that objective distinctions are not immediately available to us
directly on the basis of our own experience, but are obtained only by
means of language.
We can not go into this general thesis more closely here or give argu­
ments for it. Instead we will illustrate it using the predicate ‘red’ as an
example. We ascribe this predicate to things and we use it to express an
objective difference between things. The objective character of this dis­
126 CHAPTER II

tinction depends, however, on the fact that as a rule we agree in drawing

this distinction - if we did not, we would not ascribe the quality of red­
ness to things but to our own perceptions of things, varying from subject
to subject. This intersubjective agreement is manifested by the fact that
we agree in our use of the predicate ‘red’. It is only by learning this
predicate or another one with the same meaning that we come to be sure
of the distinction between red and non-red as an objective distinction and
know how this distinction is objectively applied.
In order for a person Y to be able to learn the predicate ‘red’ from the
use that others make of it, Y must be able to observe a sufficiently pro­
nounced similarity among the cases in which they apply this predicate. In
accord with our earlier discussion in Chapter II.3.3, there is an adequate
basis for that if we assume that Y can distinguish between things that
have for him a certain experiential quality or valence, redness y, and
other things. If completely disparate sensations in Y correspond to the
different cases in which ‘red’ was applied, then he would not be able to
learn the predicate’s use, because he would have no criterion for employ­
ing it. The subjective attribute, redness y, must be a sufficiently trustworthy
criterion for using the predicate ‘red’, then, so that it holds true as a rule
that ‘red’ can be applied to an object a, if a is redY. The correlation be­
tween redY and red does not hold in all cases, however, i.e. the quality
redY is neither a necessary nor a sufficient condition for red\ not every­
thing that looks red to Y is also red, and vice versa.
Thus while Y, before he has learned the use of ‘red’, has available to
him only the experiential quality redY, he learns the distinction between
red and non-red, i.e. the concept red, by means of the predicate ‘red’. The
foregoing quotation from the Philosophical Investigations can be under­
stood along these lines, so that we only learn to distinguish between red
and non-red things by way of the predicate ‘red’.
The statements of Wittgenstein’s that say that the criteria for using
predicates are criteria of linguistic agreement can also be better under­
stood now. Whether the predicate ‘red’ may be applied to an object a does
not depend directly on the subjective criterion for its use, whether a looks
red or not, but on whether a is objectively red, and what that means is -
since what is true objectively is essentially defined in terms of holding
true intersubjectively - on whether we agree in judging a to be red, and
what that means, once again, is whether we agree in attaching the predi­
 THEORIES OF MEANING 127

cate ‘red’ to a. And so in its epistemological aspect, conditions as to lin­

guistic agreement enter into even the criterion of use R F we had con­
sidered above as a factual criterion in contrast with the general linguistic
convention concerning the use of the predicate F.
In this vein Wittgenstein also says: “What is needed for communication
by the medium of language is not only agreement in definitions, but also
(strange as it may sound) agreement in judgments.” 49 “ ‘So what you are
saying is that human agreement decides what is true and what is false?’ -
What is true and false is what human beings say; and it is in language that
human beings agree.” 50 I.e., objective facts, like objective distinctions,
are mediated by language, for it is by means of language that we mark
out what holds true intersubjectively.
In the case of private phenomena, such as sensations and the like,
where there is no possibility of direct intersubjective observation, and
consequently agreement does not have the same relevance, the objective
character of the facts ceases to apply and in that area we can therefore say
along with Wittgenstein: “ ... what is correct is whatever is going to seem
correct to me. And all that means is that there can be no talk about
‘correct’ [in the usual, objective sense] here.” 51
If a deeper and more accurate understanding of Wittgenstein’s position
becomes possible on the basis of these very simplified and abbreviated
epistemological suggestions, they are yet to be supplemented in at least
two points.
In the first place intersubjective validity, as manifested in agreement in
linguistic usage, is not the sole criterion of objectivity. The domain of the
objectively valid or true is fundamentally defined also as a domain subject
to orderly laws that bind and regulate states and events and their conse­
quences. For that reason, it is also an important criterion for accepting
the truth of a sentence that it be in accord with generally accepted laws
of nature, that it fit into the entire system of our fundamental assump­
tions about the world.
Now since this entire system is not determined just by arbitrary con­
ventions but represents a system that has stood the test of organizing ex­
perience, the assertion that facts are marked out and distinguished by
intersubjective linguistic agreement in judgment is by no means to be
understood to mean that we define the set of facts by merely coming to an
agreement. Conventions play an important role in our statements about
128 CHAPTER II

the world; they are by no means established completely by convention,

however, but have an empirical content that can not be overlooked.
A second remark aims in the same direction. The idea that even the
criterion for using the predicate F, R p, is subject as a factual criterion to
the condition of intersubjective agreement in language, naturally does
not cancel out the distinction between the general convention (I) con­
cerning the use and the criterion for its use, R P. R P is established by way
of convention (I). But the fact that R P is satisfied in the case of a specified
object a, or not satisfied, is not a matter of free convention all over again -
there is no such convention contained in (I) ((I) does not determine how
F is to be applied in a’s case; the general rule for using ‘red’ does not
determine which things are red) nor does any new convention enter into
the scene here (no vote is taken on whether a should be called ‘red’ or not).
Whether R P is satisfied in the case of a is rather a question of fact, which
we state on the basis of experience. It is a sufficient condition for something
we state on the basis of experience to be counted a fact that as a rule it is
possible to achieve an intersubjective agreement about it. The judgments
in which we agree are not made on the basis of an agreement, however,
but the agreement is founded on the fact that as a rule our experiences
agree.
In this vein of thought, perhaps the one of Wittgenstein’s pertinent as­
sertions that comes closest to the mark is the passage in the Philosophical
Investigations where he compares the way in which a predicate F is used
with a method of measuring, and the application of F to a specific object
a to a measurement, and says: “It is one thing to describe the method of
measuring and another to discover and state the results of measurement.”
- Here, then, the general convention (I) concerning the use of F (the rule
for measuring, which defines a physical quantity, for example) is distin­
guished from the applications of F to particular cases, which are carried
out according to the criterion for applying F, R P (the criterion of the value
of the quantity to be measured given in the method of measuring), laid
down on the basis of (I). - “ But what we call ‘measuring’ is also deter­
mined by a certain constancy in the results of measurement.” 52 - I.e.,
criterion R P is capable of being used as a factual criterion only if we
generally agree in the judgments in language to which it leads us. But
agreement in the results of measurement is not a criterion for arriving at
the values measurement gives, but of their reliability. And so agreement
 THEORIES OF MEANING 129

in judging that F (a), or that ~\F (a) is not a criterion for arriving at this
judgment, but only a criterion of its reliability.

4.5. Speech Acts

Before going into two more specific themes concerning philosophy of
language in the Philosophical Investigations, we will first of all pursue the
fundamental theme ‘What constitutes the meaning of linguistic expres­
sions?’ still further. The path we take in doing so was suggested by Witt­
genstein, but J. L. Austin was the first to follow it up consistently and,
after him, J. R. Searle and others in their theory of speech acts. This
theory represents a decisive and important step beyond the use theory.
In the Philosophical Investigations Wittgenstein says: “By itself every
sign looks dead. What brings it to life? - It lives in its use. Does it then
have the breath of life within itself? - Or is its use its breath?” 53 And:
“ How does it come about that the arrow »>—>points'} Does it not already
seem to carry within itself something over and beyond itself?” - “No, it
is not the dead line; only what is mental, the meaning, can do that.” -
That is both true and false. The arrow is pointing only in the use that the
living creature makes of it.” 54 And in the Blue Book Wittgenstein says:
“But if we had to name anything which is the life of the sign, we should
have to say it was its use.” 55 As early as the Tractatus Wittgenstein is
advocating similar ideas, when he says: “What does not find expression
in the signs is shown by their use. What the signs suppress their use ex­
presses.” 56 The fact that in the Tractatus Wittgenstein understands the
sentence to be a fact, not an object, may be connected with this.57
If one wishes to understand what constitutes linguistic meanings -
perhaps one can interpret these suggestions so - then one may not start
out from expression objects, from sentences and words, but must begin
with the utterance performances, the concrete speech acts.58 What we
understand is not abstract objects, but concrete actions; the meanings we
attach to expression objects are secondary theoretical constructs, what
is primary are the meanings of the speech acts.
What pertains to the meaning of an utterance as a speech act is every­
thing we have to know in order to understand its use in the concrete
situation. In this linguistic sense of ‘meaning’ it coincides with ‘meaning’ in
that wider sense of the word we apply to non-linguistic actions as well, when
for example we speak of the ‘meaning’ of an action in a ceremony or a game.
130 CHAPTER II

What differentiates the function of an action from the action itself?

When I say ‘Good morning’, this action is a greeting and it has the func­
tion of a greeting; we can describe the action with the same word
(‘greeting’) as its function. But there are many concrete details belonging
to the action that are not essential to its function. The move in a game of
chess has the same function whether I make it quickly or slowly, with my
left hand or my right. Different actions can have the same functions, then.
The like holds for speech acts. Instead of ‘Good morning’, I can say, ‘A
good morning’, or ‘I wish you a fine good morning’, or, in case the
person I am addressing knows German, ‘Guten Morgen’. Phonetically
different speech acts can therefore have the same meaning. What we in­
tend by the meaning of a speech act coincides, however, with what we
called its ‘semantic aspect’ in 1.1. So we identify the meaning of an ex­
pression with its semantic aspect.59
The function of a linguistic utterance is based upon linguistic conven­
tions. These conventions are general rules, thus refer not to individual
actions but to ways of acting; not to utterances, then, but to expressions.
A convention of this kind does not say, for example, that Fritz can
greet Hans on January 1, 1973, in Munich by saying, ‘Good morning’,
but contains the provision that in the morning anyone can greet anyone
else (who understands English) with ‘Good morning’. On the basis of this
general convention a specific utterance of ‘Good morning’ then takes on
the function of a greeting. The meaning of the utterance is thus defined by
the expression and the circumstances of the utterance (who is speaking,
who is being spoken to, etc.).
Consequently, one must differentiate, as we have in Chapter II.4.3,
between the meaning o f the expression as a form of action: the common
function of all actions with this form, and the meaning o f the utterance:
the function of the particular act. The latter is a function of the former
and the circumstances of the utterance; and so it includes additional fea­
tures as compared with the former.60
In order to make what is here being understood by the ‘semantic aspect’
and ‘meaning’ of speech acts still clearer, we will briefly go into Austin’s
theory of the illocutionary force of utterances in [62a].
Austin points out there that a sentence like ‘It is raining’ can have quite
different functions in different situations. As we have already set forth in
Chapter II.4.3, it can be a report, but it can also be a warning (‘Look out!
 THEORIES OF MEANING 131

If you don’t close the window, it will rain in’.) or a recommendation

(‘Take your umbrella with you if you are going out!’). This illocutionary
force that particular utterances have (reporting, warning, recommending)
frequently can not be read off from the sentence itself. The context not
only indicates the reference of index expressions in such cases, but the
role of the sentence in communication is revealed only by the pragmatic
context in which it is used.
Austin is taking up in this connection Wittgenstein’s ideas about the
variety of ways in which language is used, the diverse language games. We
do not use language just to assert, to ask or to command; on the contrary,
there is a wealth of roles that language can assume, e.g. describing,
telling, judging, confirming, testifying, protesting, retorting, supporting,
suggesting, advising, explaining, elucidating, drawing conclusions, warn­
ing, demurring, recommending, promising, reporting, telling stories,
thanking, praying, confessing, conceding, agreeing, criticizing, praising,
blaming, greeting, proclaiming, excusing oneself, offering explanations,
deriding - these are all roles that declarative sentences can have. We can
use interrogative sentences not only to inquire but also, for example, for
casting doubt, for requesting or ordering (‘What are you waiting for?’),
and for asserting (in the form of a rhetorical question). We can use
imperative sentences not only for commanding, but also for wishing, pre­
scribing, recommending, guiding, requesting, appealing, and asking (‘Tell
me whether you broke the pitcher!’)
These distinctions are important if we wish to understand the commu­
nicative role of an utterance more precisely. They are by no means always
and perfectly expressed in the sentence but in many cases they become
clear from the context of the utterance. That led Austin to draw a funda­
mental distinction between illocutionary force and meaning*1 In his ac­
count, meaning is never anything but the (descriptive) meaning of the
expression. The illocutionary force of an utterance depends on the
meaning of the expression uttered; the meaning of a sentence limits the
set of illocutionary roles in which it can come into play, but it does not
fix upon any one such role unambiguously. Further, with Austin certain
results and presuppositions of a speech act are counted to its illocutionary
force. He does distinguish between illocutionary and perlocutionary act
(bringing about a result by means of a speech act), e.g. between expressing
a warning and bringing it about that the hearer has been warned (and
132 CHAPTER II

conducts himself accordingly). He urges, nevertheless, that certain results

belong inseparably to the character of many illocutionary roles. For ex­
ample, the obligation that follows from making a promise to do what has
been promised belongs to [the nature of] promising. It belongs to the act
of christening, accomplished by the formula ‘I christen this ship the
“ Hamburg” ’, that as a result the ship is now called the ‘Hamburg’, i.e.
that designating it in this way is binding. Since these results occur only
given certain presuppositions, such necessary presuppositions also belong
to the illocutionary act.62
In view of all that there is a distinct parallel between Austin’s illocu­
tionary force and our semantic aspect. But there are also two important
differences and for that reason we have chosen another terminology. In
order to fix the parallel still more clearly, we must distinguish between
the semantic aspect of an utterance and its performative mode as the act
type to which the utterance belongs. While performative modes are indi­
cated by such verbs as ‘assert’, ‘ask’, ‘warn’, etc., the semantic aspect em­
braces the speech act’s whole meaning. Austin’s illocutionary force of an
utterance, strictly speaking, corresponds not to its semantic aspect, but
to its performative mode, then.
The first difference when illocutionary force and performative mode
are being compared consists merely in the fact that we do not count the
results of an action and its presuppositions as part of its performative
mode. Many of the verbs by which we describe actions are achievement
verbs63, i.e. it is correct to apply them only if a certain result does occur
or if certain presuppositions are the case. Thus we can say ‘Hans knocks
Fritz down’ if Fritz does fall down after the blow, or ‘The doctor cured
Kuno’s gastritis’ if Kuno actually did suffer from gastritis and is healthy
after the treatment. I.e. we often characterize the action proper (‘Hans
gives a Fritz blow’, ‘The doctor treats Kuno for gastritis’) and its result,
or its presuppositions at the same time. Naturally it is also possible to
specify speech acts with respect to their presuppositions and results, but
in the particular case such specifications go beyond the linguistically
relevant aspect. The speech act is: promising, saying a christening
formula; it remains the same whether the promise is made sincerely,
whether an obligation is consequent upon it by virtue of general norms64
or whether it is kept; or whether the speaker had a right to christen the
ship or whether the ceremonial procedure is correct and successful. The
 THEORIES OF MEANING 133

function of the speech act does not encompass, then, everything that
Austin calls its illocutionary role, so that caution is appropriate in using
achievement verbs to characterize speech acts. In order to understand the
performative mode ‘saying the christening formula’ or ‘promising’, it is
necessary to know what ‘christening’ and ‘promising’ mean; one aspect
of that is knowing the presuppositions, results and obligations that nor­
mally accompany such actions. The meaning of the individual speech act,
however, is independent of whether such presuppositions are the case, the
results occur, or the obligations are fulfilled. In the same way, the meaning
of an assertion is independent of whether it is true or false, although it is
part of the definition of the performative mode ‘asserting’ that only true
assertions should be made.
The second important difference as compared with Austin is that we do
not regard performative mode and meaning as two distinct, even though
dependent parameters. Instead, we identify the meaning of an utterance
with its semantic aspect and understand the performative mode to be the
characteristic type of this aspect.65 What makes this possible is that the
concept of performative mode is narrower than Austin’s concept of illo­
cutionary force and that we do not, like Austin, start from the usual
descriptive concept of meaning (according to which sentence meanings, for
example, are propositions, which are coordinated with expression ob­
jects), but interpret meanings primarily as functions of action.
The performative mode can be explicitly indicated in language in the
sentence itself. In particular, that is the case with what Austin calls
explicitly performatory expressions, as for example:
(a) ‘I congratulate you’.
(b) ‘We invite you to settle your account’.
(c) ‘I confirm the statement by Mr. X .’
(d) ‘I am asking you if you will accept the nomination’.
(e) ‘You are requested not to smoke’.
In these statements performative verbs (‘congratulate’, ‘invite’, ‘confirm’,
‘ask’, ‘request’) occur which indicate the sentence’s performative mode.
They are in either the first person singular or first person plural active
indicative or, as in (e), in the corresponding passive, in which case the
name of the speaker or speakers can be absent as well. The act of the
speaker designated by the performative verb is accomplished in the utter­
134 CHAPTER II

ing of such a sentence. Uttering the sentences offered as examples does

not represent any description of speech acts, then - saying ‘I congratulate
you’ is congratulating, not describing or asserting.
Besides the performative verbs, however, there are also other expres­
sions that serve to specify - whether partially or completely - the per­
formative mode of a sentence. Conjectures are indicated by ‘supposedly’,
‘probably’, and ‘perhaps’, for example, rejoinders by ‘but’, ‘yet’,
‘however’, inferences by ‘so’, ‘therefore’, argumentation by ‘since’,
‘because’, etc. Further, there are verb forms (e.g. the subjunctive for
requesting, wishing, supposing), phonetic devices (e.g. intonation) and
so on. Punctuation marks (‘.’ after an assertion, ‘?’ after a question,
and T after a command) also characterize the performative mode of a
sentence.
Now every sentence can be represented in an explicitly performative
form, e.g. in the pattern P (a, b, A). P, the performative operator, is sup­
posed to indicate the performative mode. Thus P stands for a performa­
tive verb, which can be very general, however. The sentence ‘It is raining’
is largely undefined in its performative mode (we saw that it can be used,
for example, as a report, as a warning or as a request) so that it is not
possible to attach a specialized performative verb to it without changing
its sense. In the case of using words ironically or in jest, explicitly indi­
cating the precise performative mode would even destroy it.66 The sense
of a performative operator is often defined only within very wide limits,
then, and becomes determinate only in the context of an utterance, a and
b in P (<a, b, A) are supposed to be index expressions for the speaker or
speakers and the hearer or hearers. And A is supposed to be a sentence in
the form of a declarative sentence. Arguments a and b can even be mis­
sing. We rewrite the illustrative sentences as follows, then:
(a') Congratulate (I, you)
(b') Invite (we, you, you settle your bill)
(c') Confirm (I, what Mr. X said is true)
(d') Ask (I, you, you will accept the nomination)
(e') Request (I, you, you do not smoke).
Sentences (c') and (e') show how sentences in which, unlike (d), there
is no complete declarative sentence occurring as a subordinate clause can
be brought into the form P (a, b9A).
 THEORIES OF MEANING 135

We call the expression A in P (a, b, A) the descriptive component of the

sentence P (a, b, A). As such A represents only a part of the entire sen­
tence and A is not tied up with any speech act of its own. In and of itself,
then, A is meaningless in our present sense of ‘meaning’. The fact that
we give A the form of a declarative sentence, for the sake of simplicity,
does not imply that A would have a performative function (of its own)
and would represent, for example, an assertion. Nothing is asserted in
sentence (d'), even though an expression in the form of a declarative
sentence occurs in it.
Now how does the meaning of expressions as their performative
aspect - we shall also speak of their performative meaning - relate to
what are ordinarily called the ‘meanings’ of expressions as products of
phonetic acts? When, for example, it is said that the name ‘moon’ means
the moon, the predicate ‘is red’ means the property of being red, or the
sentence ‘This rose is yellow’ means the fact that this rose is yellow? We
will call this sort of meaning descriptive meaning. Having designated the
performative sense of “meaning” as primary and fundamental, we are
now concerned with characterizing descriptive meanings in terms of
performative and defining their place in the semantics of speech acts.
If, as discussed above, sentences are given an explicitly performative
form, then we obtain from the sentences:
(f) You are closing the door.
(g) Close the door!
(h) Are you closing the door?
the sentences
(f') I say (assert) that you are closing the door, or
Assert (I, you are closing the door)
(g') I tell (order) you to close the door, or
Order (I, you, you are closing the door)
(h') I am asking whether you are closing the door, or
Ask (I, you, you are closing the door)
These sentences have different performative modes, but a common
descriptive component (f). So the attempt will be made to define the per­
formative meanings of sentences in the form P (a, b, A) as functions of the
(still to be specified) meanings of components a, b, and A , with such a
function corresponding to each performative mode.
136 CHAPTER II

Let us consider first of all assertions in the narrower sense as descrip­

tions of facts.
The performative meaning of sentence (f') depends only on the descrip­
tive meaning of (f) and the circumstances of its utterance, and plainly in
a way that is typical of assertions. But we do not want to define the per­
formative meaning of a declarative sentence, starting from its descriptive
meaning. We want to go the other way and define its descriptive meaning
in terms of its performative meaning. The descriptive meaning of such a
sentence is also unambiguously determined by its performative meaning,
however: assertions are speech acts that are true or false, and their truth-
values depend only on the descriptive meaning of the descriptive compo­
nents and the circumstances of their being uttered.67 Therefore we can
characterize this descriptive meaning by indicating the conditions under
which the assertion is true. That leads to representing the descriptive
meaning of (f), for example, as a function which maps possible worlds and
circumstances of utterance onto truth-values. In this fashion descriptive
sentence components are interpreted by reverting to properties (true -
false) of the assertions as speech acts in which they occur.
Since we have given the descriptive components of all sentences the
form of declarative sentences and by so doing let them all correspond to
assertions, the descriptive meanings of all descriptive sentence compo­
nents are accordingly fixed and determined.68
Since the performative meaning of a declarative sentence depends
on its descriptive meaning alone, it is possible to regard descriptive
semantics as the semantics of declarative sentences - not in the sense that
the performative meanings of such sentences might be reduced to their
descriptive meanings, but in such a way that the descriptive meaning of a
sentence uniquely determines its performative meaning. The one-to-one
correlation of descriptive and performative meanings in the case of
declarative sentences makes it possible to define them in terms of each
other.
The concept of descriptive meaning can be interpreted in the semantics
of declarative sentences as an auxiliary concept that can also be dispensed
with. In that event the semantics would have to be directly interpreted as
the semantics of assertions, in which assertions as speech acts are said to
be true or false depending on states of the world and circumstances of
utterance.
 THEORIES OF MEANING 137

Since the semantic aspect of every speech act can be described, its
performative meaning can be characterized in terms of the descriptive
meaning of a declarative sentence. Consequently the semantics of all
speech acts can now be developed within the bounds of declarative sen­
tences, i.e. within the bounds of descriptive semantics.69
For that purpose we assign to every utterance a performative description,
which arises from a performative version of the utterance, i.e. from an ex­
plicitly performative formulation of the utterance, by interpreting it as an
assertion and replacing the index expressions for speaker and hearer that
occur in it with names, in case they do not coincide with the speaker and
hearer of the performative description, and by introducing names, or
index expressions for the relevant circumstances of its utterance.
Performative descriptions of utterances of sentences (f), (g) and (h) are,
for example:
(f") Fritz asserts (now), that Hans is closing the door.
(g") I ordered you (yesterday) to close the door [or, less idiomatic
in English, but more faithful to the example in the original: I
ordered (yesterday) that you are closing the door.]
(h") I shall ask you (tomorrow) whether you are closing the door.
In contrast to (g') and (IT), (g") and (h") are declarative sentences. They
are not explicitly performative sentences, because the subject of the per­
formative verb is not the index expression T , which stands for the one
who is uttering the description (as in (f ")), or because the time of the speech
act described does not coincide with the time of its description (as in (g")
and (h")).
Generally speaking, explicitly performative sentences, e.g. (IT), will not
be interpreted as performative descriptions. The description of a speech
act is a different speech act from the one described. You can not ask a
question and also describe it at one and the same time. To be sure, cases
can be cited in which an utterance with an explicitly performative pattern
serves as the description of another speech act occurring at the same time,70
but in that case it is not a description of itself. So we must always draw
a sharp distinction between the performance of a speech act and its
description; no speech act is a description of itself.
The following would be performative versions of (f "), (g") and (h"):
(f'") I say that Fritz says (now) that Hans is closing the door.
138 CHAPTER II

(g'") I say that I ordered you (yesterday) to close the door. [Or:
that you are closing the door]
(h'") I say that I shall ask you (tomorrow) whether you are closing
the door.
While an utterance is performed by a specific speaker under specific
circumstances, an expression, specifically a sentence, is a form of speech
activity, which can be performed by various speakers under diverse cir­
cumstances. Therefore performative descriptions of expressions are not
rendered by sentences, but by predicates. Performative descriptions of
sentences (f), (g) and (h) are along these lines:
(f"") saying (asserting) that the hearer is closing the door.
(g"") ordering the hearer to close the door [or: that the hearer is
closing the door].
(h"") asking whether the hearer is closing the door.
We can now regard the descriptive meaning of a performative descrip­
tion of an utterance as its performative meaning, and the same is the case
with sentences. If Hans says ‘Good morning’ to Fritz and someone who
hears him asks about the meaning of that utterance, what one will say is
that Hans is greeting Fritz. The descriptive meaning of the description
‘Hans is greeting Fritz’ is used to give information as to the performative
meaning of the utterance. And if someone asks what the expression ‘Good
morning’ means, what one will say is that saying ‘Good morning’ is a
greeting. I.e., the descriptive meaning of the description ‘greeting’ is used
to give information as to the performative meaning of ‘Good morning’.
If, as in Chapter III.2, one constructs a language L, in which all sentences
are interpreted as declarative sentences, then it is not necessary to make
the mode of assertion explicit in the sentences. The performative meaning
of a sentence A of L follows from the descriptive meaning assigned to
A in the semantics of L. In what way it follows from it is not fixed in
that semantics, but belongs to its application. Within the bounds of the
semantics, however, it is possible to correlate assertions and performative
meanings. To do so, one goes from a sentence A to a performative de­
scription of A (as an expression). This might read P (x9A), in which x is a
variable for the speaker and the performative operator P represents the
predicate ‘saying’ or ‘asserting’. The descriptive meaning of P (.x, A) is
different from that of A; A ’s descriptive meaning is something other than
 THEORIES OF MEANING 139

‘saying that^l’. Assertion sentences thus have a descriptive and a performa­

tive meaning, but they are different. Interrogatives, imperatives, etc., on
the other hand have no descriptive, but only a performative meaning.
That we understand even the performative meaning of a sentence A to be
a descriptive meaning is not capable of leading to confusion, for the per­
formative meaning of A is not the descriptive meaning of A itself (if there
is any), but the descriptive meaning of another expression A \ the per­
formative description of A 71
These interim and still very sketchy remarks about performative and
descriptive meanings, which we shall take up again in Chapter III.2,
should be supplemented by four further comments:
(1) The possibility of characterizing the performative meaning of a non­
declarative speech act in terms of the descriptive meaning of a description
of the speech act naturally does not exclude other possibilities for semantic
analysis. For example, questions can be semantically characterized in
terms of the descriptive meanings of possible answers to them, and so a
non-performative meaning can be attached to them analogous to the de­
scriptive meaning of assertions, one which uniquely specifies their per­
formative meaning.72 While the description-approach sketched above
has the advantage of simplicity and generality, approaches of this sort
offer more detailed information for the analysis of particular kinds of
speech act.
(2) One and the same utterance can have more than one semantic
function, so that its performative mode contains more than one compo­
nent. The sentence ‘You have deceived me for the last time’, for example,
contains along with the component of assertion (‘You have deceived me’)
the component of expressing a resolve (‘I will not permit myself to be de­
ceived by you again’) and possibly also the component of demanding
(‘Justify your behavior!’, ‘Say you are sorry’ or whatever). Descriptive,
expressive and evocative modes of meaning, as we explained them in 1.2
following K. Biihler in [34], can probably best be understood along these
lines as components of the performative mode.
(3) The fact that a linguistic expression is not just a vocal sign has been
the subject of frequent emphasis in linguistics, in agreement with W. von
Humboldt, and word and sentence have been understood to be a union
of sound and meaning. Most familiar is Saussure’s equation signe =
signifiant + signifie, i.e. the symbol consists of the (phonetic or graphic)
140 CHAPTER II

expression that signifies and the signification, the meaning. But this union
becomes intelligible only when one interprets expressions not as objects
but as speech acts: only in speech is there a union between the act’s
phonetic form and its function; it is only in uttering ‘Good morning’ that
greeting is accomplished by the production of sounds.
(4) So far we have been concerned only with the meaning of single sen­
tences and this theme will continue to stand in the foreground. At this
point, however, a few brief suggestions should be made about the problem
of textual semantics, but of course they must be limited to a few points in
passing.
A text is in the first instance a series of speech acts. The speaker can
be the same person throughout (as with a report) or various speakers can
enter in (as with a conversation). With texts, too, we differentiate between
the utterance aspect and the expression aspect (the text as what is spoken
in a specific situation by specific speakers as distinct from the form of
that text), the phonetic aspect and the act or product aspect after the
fashion of 1.1.
Now by what criterion do we regard certain sequences of sentences as
texts and others not? The first thing one will point to in this connection is
that the contents of the sentences have a connection with each other, but
that is a very vague characterization. It is probably more decisive that
the sentence sequence has as a whole a definite illocutionary role of its
own. Or in our terminology, that there is a performative mode which
characterizes it all together as a story, report, discussion, proof, etc.,
while smaller segments of the sequence, particularly the individual sen­
tences, can not be characterized in the same way. Now there are texts (e.g.
reports) in which one could tie all of the sentences together with ‘and’
into one long sentence, which would have the same performative mode
as the text. But not all texts are long sentences of this sort. Rather, in
many texts (e.g. in discussions) the individual sentences - or parts of the
texts - are characterized by their own diverse performative modes within
the comprehensive performative mood of the entire text (thesis, counter­
thesis, argument, rejoinder, challenge, etc.) Here for the first time the
Wittgensteinian image of the language game, of speech acts following one
on the other like moves in a game, becomes perfectly to the point. The
performative mode of the text indicates which language game is being
played, what rules govern the succession of individual moves, what moves
 THEORIES OF MEANING 141

are possible and consequently what function the individual moves have in
the game.
Just as what we did above was to start out from a semantics of (de­
scribing) assertion sentences and interpret sentences with other performa­
tive modes by way of performative descriptions of these sentences as
speech acts, we can now characterize all texts that do not have the charac­
ter of reports, i.e. of sequences of sentences capable of being conjoined
into one assertion sentence, in terms of performative descriptions, which
are represented once again as sentences, or as predicates. The descriptive
meaning of such descriptions can once again be regarded as the performa­
tive meaning of the texts as sequences of speech acts.
This is the way that looks in the example of a discussion: In an ad hoc
sense laid down as our basis, a discussion is supposed to be made up of
assertions, challenges to provide arguments for an assertion, and argu­
ments. An argument for a sentence A is itself a text made up of several
sentences of which A is the last and of premises (which count as new asser­
tions) and logical consequences of sentences that came earlier in the argu­
ment. There are also supposed to be rules for the discussion, as for example
the rule that when one of the participants asks for an argument to be
given, one must be provided, that arguments can be called for only to
support previous assertions by another participant and that an assertion
A by one participant may be challenged only by the assertion of ~iA.
The following predicates are defined in terms of such rules:
D (x ,y , C) - there is a discussion between x and y as de­
scribed by C
B (x, t, E, C) - x gives an argument at t for E, as described by
C
T (x, t, C) - x asserts at t that C
F (x, t, C) - x draws the conclusion at t that C
A (x, y, t, C) - x calls on y at t to give an argument for his
asserting that C.
A discussion that has the structure:
x asserts that C ; thereupon y calls on x to give an argument for this asser­
tion; x argues for his assertion by asserting that E and that C follows from
it; thereupon y responds by asserting that ~iE
can thus be described by the predicate:
142 CHAPTER II

D (x, y 9T (x , tl9 C) a ^ (y , x, r2, C) a 2?(x, f3, C, r(x , f4, £ ) a

a F ( x 9 tS9 C )) a T ( y 9169 - IE))

with variable x and y. In this description, tx through t6, after the fashion
of the relative temporal indicators, ‘thereupon’, ‘then’, etc., in the report
of the discussion, give the order of the various speech acts.73 If we use
<P(x9y) as an abbreviation for this predicate, then we interpret the de­
scriptive meaning of <P(x9y) as the performative meaning of the discussion
as a text form (form of speech activity). If the constants a9b designate two
participants in the discussion, then the descriptive meaning of <P(a9b) is
correspondingly the performative meaning of the discussion between
these two as a text utterance.

In conclusion, let it be emphasized once again that the theory of speech

acts, for which there are only scattered pointers in Wittgenstein, repre­
sents a distinct advance over the representation or picture theory and
over the use theory of meaning in the Philosophical Investigations. If one
starts out from expressions as (classes of) phonetic or graphic objects,
then the difficulty in explicating their meaning lies in the fact that taken
by themselves they do not mean anything more than stones or classes
of numbers do. Meaning must first be assigned to them, be it along realis­
tic lines or along pragmatic lines by way of rules of use. But interpreting
words as marks for things is just as unnatural and distorted as interpreting
them as pieces in a game. Speech acts, on the other hand, as conventional
activities have meaning in a very natural sense not restricted just to the
activities involved in speaking. We understand actions in a more imme­
diate sense than objects. And we do not ‘use’ a sentence by going back to
rules for employing it, which tell us that it can be used here, but we give
voice to the sentence because (on the basis of linguistic conventions) we
can accomplish what we want to by doing so. Performative meaning rec­
ommends itself as a more natural basic concept of semantics, with the
help of which descriptive meanings can then be introduced as theoretical
constructs.74

4.6. Private Languages

On one interpretation, a private language expression can be understood
to be an expression that someone has introduced for his own use, one that
 THEORIES OF MEANING 143

does not belong to the intersubjective language spoken by a linguistic

community. In this case we will speak of a private expression. But it can
also be understood to be an expression - whether private or in common
use - that stands for a private, inner experience of a subject, for a sensa­
tion, a feeling, a pain, say, to cite the example Wittgenstein brings up
most frequently. In this case we will speak of an expression for the private.
In the Philosophical Investigations, Wittgenstein denies the possibility of
introducing such private language expressions. In doing so he does not
draw any clear distinction between private expressions and expressions
for the private, but his arguments are aimed at both. This doctrine of
Wittgenstein’s, insofar as it concerns expressions for the private, is of
particular interest in connection with the philosophy of mind or con­
sciousness, with behaviorism and phenomenalism. Insofar as it concerns
private expressions it is also of significance in our context, as a thesis in
the philosophy of language, however.
Wittgenstein’s arguments against private expressions can be summarized
as follows:
(1) An expression of this sort has meaning only if there are rules for
using it correctly. But correct usage is not defined for private expressions,
since on Wittgenstein’s view the correct use of a word is always the use
that is in accord with the way the word is employed by a language com­
munity. But where a word is not employed by a community but just by
one individual, it is not possible to speak of that sort of correctness.
According to Wittgenstein, following a rule is a social practice and conse­
quently one can not follow a rule privatim.75
If we take into account the fact that according to Wittgenstein, as we
saw in Chapter II.4.4, not only is the general linguistic convention con­
cerning the criteria for using a word established by general usage, but
that conditions of general usage enter into these criteria of use themselves,
then the argument to the effect that there are no sufficient criteria for the
use of private expressions becomes still sharper.
(2) But even if the latter argument is ignored, the criterion for applying
a private predicate, for example, could not be anything but the fact that
a new object a appears to me to be similar to the objects in K{F) I used as
examples in introducing F. An ‘appears to be similar’ of that sort is no
criterion that can be used, however, because I can not put the correctness
of this appearing-to-be to any test: “One would like to say: what is correct
144 CHAPTER II

is whatever is going to seem correct to me. And all that means is that
there can not be any talk of ‘correct’ in this case.” 76 This argument takes
on particular force when F is a predicate for the private as well, e.g. for
my sensations. For while objects can be put alongside each other and
compared, that is not possible with sensations: earlier sensations can be
compared with present sensations only in memory. But memory is un­
reliable and can only be tested by other memories. Someone who tests the
correctness of one memory by means of another memory, however, is ac­
cording to Wittgenstein like someone who buys several copies of the
morning paper in order to assure himself that it is writing the truth.77
Thus while in applying a predicate F that is in general use one does
have a wider criterion for an ascription of Fbeing correct, viz. the linguistic
usage of others, with private expressions one is referred to nothing but
how things appear to oneself.
Wittgenstein’s main arguments against expressions for the private read
as follows, on the other hand:
(3) A predicate for the private, e.g. ‘feeling a pain’, since we can not
directly observe the pain sensations of others, can be learned only by
introspection, i.e. as it applies to our own pains. But then it is not possible
to apply the predicate we have learned in that way to other persons as well.
For if I learn “feeling a pain” only in terms of my own sensations, then
for me ‘pain’ is synonymous with ‘pain felt by me’ and in that case it is
nonsense to speak of another person’s pain.78 Nor can one say that some­
one else is feeling pains when he experiences the same sensations as I do
when I am feeling pains, for I have no criterion for applying ‘feeling pains’
to others.79
(4) Even if we managed to apply expressions for the private like ‘feeling
pain’ to others, it would nevertheless be impossible to embed them in the
intersubjective language. For we can not compare our private sensations
intersubjectively and so it would remain entirely unclear whether various
people do not each mean something entirely different by ‘feeling pain’.
The meaning of ‘feeling pain’, as an intersubjective predicate, would re­
main totally undefined, because we would have no common criteria for
applying that predicate.
How then is the meaning of such words as ‘feeling pain’ to be under­
stood, words that are nevertheless actually part of our common language?
According to Wittgenstein, one must first of all distinguish between state­
 THEORIES OF MEANING 145

ments such as ‘I have a pain’, which have no descriptive meaning, only

expressive - like ‘Ow’ or a groan - and statements such as ‘He has a pain’.
The possibility of introducing statements of the latter type into the com­
mon language has its basis in the fact that there is something like natural
pain behavior that all human beings exhibit when they have a pain, and
according to Wittgenstein statements about someone else’s pains are to
be understood along behavioristic lines as statements about such pain
behavior. If the mother observes such pain behavior in some third person,
she says to the child: ‘Now he has a pain’, and in this way the child learns
how to use the words ‘having a pain’ with pain behavior as the criterion
for applying it. The words ‘having a pain’ are thus used as applied to pain
behavior and therefore that behavior is what they mean. And so Wittgen­
stein can say, similarly as for the thing predicate ‘red’, “You learned the
concept of pain when you learned the language.” 80
Wittgenstein also uses the beetle example we have already cited in
II.3.3 to illustrate this idea: “ Suppose everyone had a box with something
in it we call a ‘Beetle’. No one can look into anyone else’s box.... In that
case it might well be that everyone had something different in his box....
But if the word ‘beetle’ were to have a use for these people all the same? -
Then it would not be the name of a thing. The thing in the box does not
belong to the language game at all... you can ‘divide through by’ this
thing in the box; it cancels out, whatever it is.” 81
The pain sensation is represented by the beetle here and what the ex­
ample says is: just as the word ‘beetle’ as generally used does not refer to
the possibly quite different things in the boxes, but only to the visible
boxes themselves, for instance, so the word ‘having a pain’ does not refer
to the possibly quite different pain sensations, but to the visible pain
behavior.
The case is similar with other expressions for the private: they too are
to be understood in a behavioristic way. To be sure, Wittgenstein does not
offer any discussion of this program, only a few paltry hints. But the
program has been carried out in detail by G. Ryle in [49], for example.

The following comments are to be made by way of a critique of Witt­

genstein’s arguments against private languages:
Ad 1 : As we have already urged in Chapter II.4.4, whether a conven­
tion concerning the use of a word can function or not does not depend on
146 CHAPTER II

the number who are subject to that convention. Even a single individual
can adopt a resolution that he can make use of. There is no reason why
one can not also follow a private rule. There is, to be sure, this much truth
in Wittgenstein’s argument, namely that absence of any way of checking
one’s own use of language against the linguistic usage of others would
result in some uncertainty in the employment of private expressions, but
that would not make the introduction of private words impossible. On the
other hand, we have seen in Chapter II.4.4 that linguistic intersubjectivity
is necessary for language to become a language about things, about ob­
jective facts. But this argument against the possibility of using private
expressions collapses in precisely the case Wittgenstein has in mind, the
case of private expressions for the private.
Ad 2: Faced with the question whether the use of a word is in accord
with earlier and with the stipulated ways of using it, the individual is
always ultimately thrown back on his own memories and impressions.
That is true of the question whether we are using a word in the intersub-
jective language in accord with general usage as much as of the question
of the use of private expressions. Ultimately we can never test the correct­
ness of our impressions and memories except by means of impressions and
memories, not of course by means of the same impressions and memories,
but by means of other ones. And Wittgenstein’s example of the morning
paper is off the mark to this extent: We do not buy several copies of the
same morning paper to convince ourselves of the truth of what is written
in it, but we buy different morning papers. And where we have no possi­
bilities for checking directly, that is an entirely sensible thing to do.
Furthermore, there is to be sure a distinct difference in the degree to
which our sensations and objective ‘public’ facts can be checked, but it is
just not the case that there are no distinctions within the realm of our
sensations that will bear any weight - since our ultimate criteria are al­
ways experiences of our own [Erlebniskriterien], otherwise distinctions
among things would not be possible either.
A d 3 : Even if we do learn predicates such as ‘having a pain’ primarily by
introspection, Wittgenstein’s argument is still not any valid objection to
the possibility of applying these predicates to other persons. We learn
predicates like ‘red’ also primarily by examples of red objects perceived
by us. But it does not follow from that that for each person X ‘red’ is
synonymous with ‘perceived as red by X \ Rather we apply this predicate
 THEORIES OF MEANING 147

to objects not perceived by us too and understand it when it is applied by

others to objects seen only by them. When a predicate F is learned from
examples to which G also belongs, it can be the case, but it does not have
to be, that F is understood in the sense of ‘F and G \
Ad 4 : Wittgenstein’s fourth argument holds not only for expressions for
the private, but, as the discussions of Quine’s thesis of indeterminacy of
translation in Chapter II.3.3 have shown, for all linguistic expressions. We
can say, as we did there: It is a natural assumption, to be checked by
further observations, that in like situations the other person has sensa­
tions like my own. If he hits himself on the thumb with the hammer, I can
reasonably assume that he has the same sensation as I do when I hit
myself on the thumb, namely pain. And this assumption is supported by
the fact that he exhibits the same reaction as I do on such occasions my­
self. If further the experiences he calls ‘feeling pain’ play the same role in
his whole life and behavior, then the assumption that he means something
different by this expression from what I mean by it has no sense. Wittgen­
stein’s beetle example has its basis after all in the fact that the beetle in the
box is an isolated phenomenon with no connection with any other visible
phenomena. That can not be said of our experiences of sensations, feelings
and the like, however, and to that extent this comparison is skewed.
This much is correct in Wittgenstein’s argument, however: we could
not introduce words for the private into intersubjective language unless
there were intersubjectively observable causes, effects and criteria for
private experiences.
Finally, it is to be granted Wittgenstein’s reconstruction of the mean­
ings of words for the private, that a distinction must be drawn between
their descriptive use and their expressive (Kundgabe) and evocative use
(Appell). But this distinction does not always coincide with the distinction
between employing them in the second and third person as against a first
person use: The sentence ‘I feel such and such pains’ said to a doctor is a
descriptive sentence and the cry ‘Ouch’ can not be substituted for it in
its semantic function. On the other side of the question, the sentence
‘Fritz is in pain’ may contain an appeal to the doctor to help Fritz.
Furthermore it does not follow from the fact that we often use behavi­
oral criteria in employing the words ‘feeling pain’ that because of this it
means those forms of behavior. When we say that someone has a pain
what we mean is that he is having a certain sensation, not that he is be­
148 CHAPTER II

having in such and such a way or that he has a disposition so to behave -

what behavior or what behavioral disposition is analytically necessary
and sufficient for applying this predicate? If we say that as a rule the atmo­
spheric pressure increases according to the rising of the barometer needle,
it does not follow from this that what we mean by the sentence ‘The atmo­
spheric pressure is increasing’ is that the barometer is rising.
In summary we can say, then, that Wittgenstein’s arguments against
private language expressions are not convincing. But they do illuminate
once again his endeavour to establish the connection between word
meaning and intersubjective word usage.

4.7. Family Resemblances, Type Concepts and Linguistic Fields

In the Philosophical Investigations Wittgenstein says that the objects to
which a one-place predicate F can be ascribed generally do not have any
quality in common (which all of the objects to which Fean not be ascribed
do not have), but that there are only certain similarities among them: “I
can think of no better way of characterizing these similarities than by
using the term ‘family resemblances’; for the various resemblances that
hold among the members of a family overlap and intersect each other in
the same way.” 82
Wittgenstein illustrates this in the case of the predicates ‘game’ and
‘number’. He says: “ Consider the proceedings we call ‘games’ for one
example. I mean board games, card games, ball games, games of combat,
and so on. What do these all have in common? - Do not say, ‘There has
to be something common to them, or else they would not be called
‘games’ ’ - but look and see whether they all have something in common.
- For if you look at them, you are surely not going to see something com­
mon to all of them, but you are going to see resemblances and relation­
ships...” And after having pointed out such common features and rela­
tionships within particular groups of games and having shown that none
of these features are true of all games, he reaches this conclusion: “We
see a complicated network of resemblances that overlap and intersect
each other. Resemblances writ large and writ small.” 83
“And the kinds of numbers equally constitute a family, for another
example. Why do we call something a ‘number’? Well, let us say, be­
cause it has a - direct - relationship to much that has previously been
called number; and by virtue of that, one can say, it acquires an indirect
 THEORIES OF MEANING 149

relationship to other things we call by the same name. And we draw our
concept out just as in spinning a thread we twist fibre onto fibre. And the
strength of the thread does not lie in any one fibre running throughout its
entire length, but in the fact that there are many overlapping fibres.” 84
Now these remarks of Wittgenstein’s obviously can not be understood
as saying that there is no property common to all games, for example -
for the very property of being a game is one such property. If concepts
or qualities, on the pragmatic interpretation, are determined by predicates,
then there must be a quality corresponding to ‘game’ and ‘number’ as to
other predicates. Nor can we interpret the remark in such a way that, as
Bambrough proposed in [60], there is no predicate other than F that ap­
plies to precisely all the objects to which F applies - n n f would be
such a predicate - or there is no observable characteristic that belongs
to precisely those objects - for the concept of observable characteristic
can not be applied to abstracta such as numbers and besides that without
being connected with any concrete scientific context it is much too vague.
More to the point is Bambrough’s comment that in saying these things
Wittgenstein is assuming a middle position between realism and nominal­
ism in the philosophy of language. According to realism a predicate F is
supposed to become meaningful by virtue of our assigning a concept inde­
pendent of F to it and ascribing it to precisely the objects that fall under
that concept. On that view, in Wittgenstein’s example, the use of the
predicate ‘game’ would be specified in such a way that we ascribe it to all
affairs that are games - the criterion for applying ‘game’ would thus be
the (antecedently given) property of being a game. For nominalism, on
the other hand, the objects to which we ascribe Fhave nothing in common
besides being called F, i.e. we do not use the word ‘game’ with an antece­
dently given conceptual criterion as a standard, but rather what is primary
is the use of the predicate; if need be the concept o f ‘game’ can be attained
by abstraction from the predicate on the basis of identical usage.
Now the realistic approach leads - apart from the problems we pointed
out earlier - to the difficulty that in fact we do not have at our disposal
any sharply defined concept of a game, according to which we employ
the word ‘game’. If we were asked to say what a game is in general,
we would be at a loss, even though we know how to use the word ‘game’
correctly. Nominalism, on the other hand, faces the problem that where
there is no criterion for applying F, then the use of F is arbitrary, and a
150 CHAPTER II

sentence saying that some affair is a game does not say anything about
that affair except that it is called a ‘game’, i.e. such predications have no
rational descriptive content.85
According to Wittgenstein, however, we apply a predicate F neither by
using an antecedently given concept as a standard nor arbitrarily, with­
out any criterion for applying it, but by using resemblances as a standard.
With his thesis of using predicates according to family resemblances
Wittgenstein wishes to do justice to the following linguistic phenomena:
(1) Although we know how to use such a predicate as ‘game’, we can
not, as was already stated, explain in general terms what a game is. A defi­
nition of ‘game’ such as might be in a dictionary, for example, is only a
later attempt to describe an antecedently given linguistic usage, which is
also determined by many accidental circumstances and arbitrary factors
for which no further reason can be given. The definition of the concept
is only a function of the word’s use. This latter is not guided by an ante­
cedently given concept, but concepts first arise, as the basic pragmatic
thesis would have it, where the use of predicates has been firmly estab­
lished.
But it is also an empirical fact, independent of pragmatism, that in
applying the word ‘game’ to a new sort of affair, we are not guided by firm
general criteria of games, but by relationships the new affairs manifest
with those we are accustomed to calling ‘games’.
(2) Predicates in ordinary language have a certain vagueness at their
margins, they are open or porous, as it is also put.86 If they stood for
determinate concepts it could not be the case that there are objects within
the domain for which they are defined of which one could either affirm or
deny the predicate with equal justification. If e.g. ‘red’ stood for a deter­
minate concept, we could never come to be in doubt, a doubt not capable
of being resolved by factual information, about whether we should call
an object ‘red’ or ‘orange’.
(3) Ordinary predicates are in many respects highly dependent on their
contexts for their meanings. We understand what a mosquito-eating bat is
only if we know that the word ‘bat’ can designate an animal as well as a
baseball or cricket bat - what occurs here is a typical ambiguity, different
ways of using the same word, quite distinct from each other. We under­
stand what a ‘pitch-black night’ is, on the other hand, on the basis of the
resemblance in meaning between the use of ‘pitch-black’ in the sense of
 THEORIES OF MEANING 151

‘having the black color of pitch’ and in the sense of ‘very dark’. Conse­
quently this way of using the expression requires no further explication.87
(4) The question of the synonymity of predicates is for the most part
not a matter of either-or, as it would have to be if predicates stood for
determinate concepts. Instead, it is a matter of more-or-less. We can
usually speak only of a resemblance of meaning between two predicates,
hardly ever of a strict identity of meaning.88
Wittgenstein can now explain these phenomena with his theory of
family resemblances:
For (1) that is obvious: If there is nothing but a family resemblance
among the objects that fall under F, this resemblance is the sole criterion
for applying F.
The marginal vagueness of F, as described in (2), arises from the fact
that an object can manifest resemblances both to objects to which F has
been ascribed and to objects of which it has been denied, and that these
resemblances may balance each other.
As to (3), the dependence of F on its context is made possible by the
fact that the F-objects spoken of in one context manifest specific resem­
blances to each other, and these then confer an additional meaning on F,
but they are different from the resemblances the F-objects spoken of in
another context manifest. This does not exclude the objects in both sets
from also being sufficiently similar to each other that they can all be
characterized by the one predicate F. Resemblances of meaning in differ­
ent contexts permit words to be used metaphorically as well, without such
use always having to be explained anew in the new contexts.
Finally, as to (4), light is shed on the fact that the synonymity of two
predicates F and G is a matter of degree by noting that on the thesis of
family resemblances an F-object is an object sufficiently similar to the ob­
jects ordinarily called F and likewise for G-objects. The resemblances
specified in this way hold to a greater or a lesser degree, however, and conse­
quently the description of an object as an F is more or less appropriate
than the description of it as a G, or more or less equivalent to it.
These examples already demonstrate the fruitfulness of the Wittgen-
steinian ideas. We shall see, however, that these ideas admit of still further
important applications in the philosophy of language. But before we go
into these any further, we will make an attempt at a more precise represen­
tation, since Wittgenstein’s statements have only illustrative value. A
152 CHAPTER II

scientific use of the term ‘family resemblance’ presupposes more exact

explication of the conceptual structures to which it is addressed.89
We will now give a logical model for the introduction of classificatory
concepts on the basis of resemblances.90 In doing so we shall refer, in
order to have a concrete case before our minds, to the example of intro­
ducing color predicates such as ‘blue’, ‘green’, ‘yellow’, and ‘red’. But
features peculiar to this example are not supposed to enter into the model;
instead it is supposed to be formulated in such general terms as to be
assured of broad applicability.
The color predicates, according to Wittgenstein, are not defined as ex­
pressing classificatory concepts of blue, green, yellow and red already
given in advance, but as being used by us on the basis of a resemblance
between the objects to which they are supposed to be applied and objects
that have already been distinguished as ‘blue’, ‘green’, ‘yellow’, ‘red’. We
will have to begin with a relation of color resemblance, then. Now this
color-resemblance is no classificatory two-place relation, such that two
objects from the class M of all colored objects, the field of the resemblance
relation, must either resemble or not - for a resemblance relation of that
sort would have to be reflexive, symmetric and transitive, i.e. an equiva­
lence relation, which would as such already define the domains of the color
concepts as equivalence classes, so that we would already have classifica­
tory color concepts.91
A two-place comparative relation makes no sense as a resemblance
relation. With a three-place comparative relation: a resembles b at most
as much as c resembles b, the resemblances of two objects can only be
compared with respect to a tertium comparationis b, not any arbitrarily
selected resemblances of pairs of objects. Consequently one will begin
with a four-place resemblance relation on M : a and b resemble each other
at most as much as c and d, symbolically a, b < . c, d.92
This relation is supposed to satisfy the following axioms:

(Al) a9b ^ . c , d v c9d < . 0, b

(A2) a, b ^ . c , d a c9d ^ . e , f 3 a,b < . e,/
(A3) a, b < . c, d 3 b9a < . c, d
(A4) a, b ^ . c, d 3 a, b < . J, c
(A5) a, b < . c, c
(A6) a,b = a9a 3 a ,c = b ,c
 THEORIES OF MEANING 153

We then define:
(dl) a9b = c ,d : = a, b < :,c,d a c , d ^ . a b
(d2) a, b < c, d : = c, d ^ .a , b
(d3) a ~ b : = A x(a, x = b, x)
(Al) to (A4) say that the relation {a, b) < . {c, d} : = a, b < . c, d between
pair sets {a, b} and {c, d} is a comparative concept along the lines of a weak
ordering. (A5) says that there are maximal elements for this relation
(identity implies maximal resemblance). And (A6) says that two objects a
and b with maximal resemblance are not distinguishable, i.e. a ~ b holds.
== and ~ are both equivalence relations - sameness of color in our
example - and the substitution principle a~b=> (A[a]^A[b\), where
A[d\ is a primary formula constructed with < . that contains a, holds.
On the basis of the relation < . on set M, color-resemblance in our
example, n classificatory color concepts are now to be introduced. We
represent them as classes Fu ...,F n. For that purpose we begin with
classes Bl9...,B n of objects in M given as examples of elements of
Fu ...,F n. We also call these classes B t(i = 1,..., n) example-classes. It is
supposed to hold true of them that:
(Bl) B, # A? i.e. the B t are not empty (let A be the null class),
and
(B2) a e B t A b e Bk z i —[ a ~ b for i ^ k , i.e. different example
classes do not contain any objects with the same color.
From that it follows immediately:
B t r\B k = A for i ¥* k.
The color classes Ft are now supposed to be defined as sets of objects
that resemble the elements of B t more than the elements of the classes Bk
with i # k . I.e. we define:
(D l) a e F i:= \ / x ( x e B t /k f \ y ( y e U**i Bk => y, a < x, a)).
[Jk*iBk is the union of all example-classes, eliminating the elements of
Br
According to (Dl), then, Ft is defined as the set of objects a, which have
a greater resemblance to one element of B t than to any element of Bk with
154 CHAPTER II

i ^ k . A definition: aeF t:= /\x y (x e B i a y e \Jk±i Bk^ y , a <•*, a), on

which an object a belongs to class Ft only if it resembles all of the objects
in B t more than it does any of the objects in Bk with &#/, would be in­
adequate precisely in view of our example of color classes, because a red
object lying on the Orange side resembles a yellow one more than it does
a red object lying on the Violet side.
(D l) satisfies the two conditions:
(Cl) B t czFi9
and
(C2) a e F i A b e F k =>- i a ~ b for i # k.
If the color classes Ft are defined in terms of the example classes B t in
this way, then the Ft depend entirely on the choice of examples. That
generates the following problem: Actually the introduction of color predi­
cates does not break down into two steps, so that first the example classes
are presented and then the Ft determined exclusively by referring to those
first examples. One is more inclined to say that every object in Ft can
serve as an example of Ft. But then this would have to be true: If you
substitute for the B t in (D l) any sets F *, for which B tciF*czFu then you
obtain the same Fi9 i.e. the Ft are invariant under any such expansion of
the example-sets. But this condition is not universally true.
This does not make definition (D l) inadequate, however. For invari­
ance with regard to expansions of the example classes actually only enters
in when there is already a large number of widely scattered examples that
indicate in particular the predicate Ft9s limits of applicability.
If we give this definition: Two series of example classes Bi9...9Bn and
B [ , ..., B rn are called equivalent if they yield the same classes Fl9..., Fn on
applying (Dl) - so that Ft are invariant upon substitution of any other
equivalent series of example classes for any given series - then it is clear
that invariance of Ft with respect to the selection of B t exists within very
broad limits.
On certain conditions comparative relations can also be defined along
these lines: a is at most as typical a case o f Ft as b is. The following is a
simple case, for example: If there is a sequence of objects bi9..., bn of Af,
such that the sequence {bx} ,..., {&„} of unit classes of these objects is
equivalent to the sequence of example classes Bu . . . , B n, then we can
 THEORIES OF MEANING 155

construct a definition as follows:

(D2) a ^ iC := a e F t a c e F t a a, < c, .
The bt can then be called pure cases of Ft.

We will now show that two further linguistic phenomena besides family
resemblances can be represented in our model of introducing classificatory
predicates by means of a resemblance relation and example classes.
(a) Ordinary predicates are in many ways not purely classificatory con­
cepts, but type concepts. A number is either a prime number or not. In this
case it makes no sense to say that 3 is more, or less of a prime number than
5. To that degree, ‘prime number’ is a purely classificatory concept. The
concept ‘pyknic’, on the other hand, is an example of a type concept.
Here, too, it can be said that a specific man is a pyknic or is not a pyknic.
But pyknics are not all equally pyknic; the pyknic, as a pure case, is
rather rare, in fact. Most human beings who can be so described manifest
other, atypical characteristics along with the pyknic features in their body
structure. In this case, then, it does make sense to say: a is a more typical
case of a pyknic than b, i.e. to be pyknic is not just a matter of either-or
but also a matter of more-or-less.
C. G. Hempel and P. Oppenheim first presented a logical analysis of
the structure of type concepts in [36]. They elaborated on the comparative
character of the type concept. But since classificatory and comparative
concepts represent different forms of concepts - a classificatory concept
is a one-place concept, a comparative at least a two-place concept - com­
parative concepts can not be used for purposes of classification directly
nor classificatory concepts for comparison. The usual type concepts
unite the classificatory and the comparative aspects, however, insofar as
they are used both in the way classificatory concepts F (x) are and in the
way comparative concepts defined on the set X xF{x) are. To that extent
the Hempel-Oppenheim analysis is not entirely satisfactory. Hempel’s
differentiation between ‘classificatory types’ (as purely classificatory con­
cepts) and ‘extreme types’ (as comparative concepts) does not do justice
to these two aspects of type predicates either.93
Our model for introducing predicates by way of (D l) can now be looked
upon as a model for introducing type concepts as well. For the Ft are
classificatory concepts; but with the aid of the resemblance relation used
156 CHAPTER II

in defining them in many cases a comparative concept in accord with (D2)

is also established, so that it makes sense to say that one object in X xF{{x)
is a more typical case of F%than another.
(b) The field concept in linguistics also impresses itself upon us in the
context of our model. The basic idea of linguistic field theory goes back
to W. v. Humboldt. The term ‘linguistic field’ was first introduced in
1924 by G. Ipsen, however, and later taken over by J. Trier in another
sense. We shall refer in what follows to Trier’s field concept, which is no
doubt the most relevant for semiotic.94
According to Trier certain groups of words form a wordfield. What is
characteristic of a word field is (1) that it is correlated with a semantic
region [Sinnbezirk\, i.e. there is some relationship of meaning among all
the words in the field, and (2) that the meaning of one word depends on
that of the other words in the field and can only be defined together with
and in distinction from their meanings.
One example of a word field that has often been used is the field of color
words. The correlated semantic region is the realm of color, and the
meaning of ‘red’, for example, depends on whether the words ‘orange’,
‘violet’ are also in the field or only ‘yellow’ and ‘blue’. Characterizing the
meaning of ‘red’ requires, then, that the whole field to which it belongs
be specified. Two different word fields correlated with the same semantic
region represent two different ways of dividing this region up linguistical­
ly. That will be plain on comparing our system of basic color words with
that of the Navahos, for example, in which one word takes over the func­
tion of our ‘blue’ and ‘green’, while two words correspond to our ‘black’.95
Other examples of word fields are the levels in a grading scale - the
predicate ‘good’ says one thing in the scale ‘very good’, ‘good’, ‘satis­
factory’, ‘adequate’, ‘inferior’, ‘unsatisfactory’, and something else in the
scale ‘good’, ‘mediocre’, ‘bad’ - or Trier’s standard example of words in
the region of intelligence. [The example contrasts the role of the word
‘wize’ in the Middle High German field of words *wize\ ‘witzig\ ‘sinic9,
‘bescheiden’, ‘kiinstic’, ‘listic\ ‘kundig\ 6karc\ etc., with the role of the
word ‘weise* in the New High German field of *weise\ 6klug\ ‘gescheit\
‘intelligent’, ‘schlau’, 6gerissen\ etc.96 For obvious reasons it can not be
translated into English without ceasing to be Trier’s example.]
Now our model of introducing classificator> concepts on the basis of a
comparative resemblance structure with example classes represents this
 THEORIES OF MEANING 157

process of establishing the meaning of predicates in a word field. The field

M of relation ^ . corresponds to the semantic region, which can be divided
up into subdomains Ft in quite diverse ways. Also, from the perspective of
(D l), we can apprehend the dependence of the Ft on the number n of
classes into which M is divided and on the examples from Bk of the other
color classes Fk with fc#/, i.e. on where the other color classes are settled
in M. I.e., we can see a dependence of the meanings of the Ft on each
other: if one Ft is altered, then the others will be altered along with it.97
Our model contains within it, then, the three linguistically fundamental
aspects of family resemblances, type concepts and linguistic fields, so that
it can be looked upon as a basic model for introducing predicates. This
model illustrates in a special case still another of Wittgenstein’s theses,
that we learn conceptual distinctions (here the classificatory predicates)
only in learning language, his thesis of the world being divided up by
language. If we consider the example again, it is clear that things do not
fall into red, blue, yellow, and so on by themselves, but that we first intro­
duce this differentiation with the assistance of language, by picking out
example classes of color-resembling objects and designating them by a
single word.
NOTES

1 See Wright [55].

2 See Wittgenstein [53], 109, 133.
3 For an interpretation of the Philosophical Investigations, see also Pitcher [64] and
[66], and Stegmuller [65], Chapter XI, Part II.
4 See Chapter II. 1.3.
5 See Wittgenstein [53], 101-104, and 373, 379f. and 384.
6 That is also true of those realistic theories according to which the entities to be
pictured are ideal structures, such as concepts, or mental data, such as ideas
( Vorstellungen). For mental reality, too, is revealed only by means of language and
concepts are not given to us independently of language, but represent nothing but
abstractions from linguistic givens.
7 See Wittgenstein [53], 60-64 and 90.
8 See Wittgenstein [53], 124.
9 See Wittgenstein [53], 87, 88, 98, 99.
10 See Wittgenstein [53], 7, 19, 23.
11 Lying, to be sure, is not so much a distinct language game as an infraction of the
rules of the game ‘communication’, which require that one assert only such declarative
sentences as one is convinced are true.
12 See Wittgenstein’s comparison o f words with tools in [53], 11.
13 See also Chapter II.4.3.
14 For criticism of the idea of language games see also Lenk [67].
15 Alston, for example, gives this formulation in [63b], p. 84.
158 CHAPTER II

16 Wittgenstein [53], 43. - On Wittgenstein’s ‘definition’ of meaning as use see also

Hallett [67].
17 Wittgenstein says in [53], 560: “The meaning o f the word is what is explained by the
explanations of its meaning.” In explaining a meaning what one explains is for the most
part the word’s use, however, and consequently the identification of meaning and use
comes naturally.
18 The vagueness and ambiguity o f the concept ‘use’ has often been emphasized in the
literature, See e.g. Alston [63a], Findlay [61], Ryle [60], Shwayder [60], Strawson [54]
and Xenakis [54].
19 See also the remarks in Pitcher [64] and Stegmiiller [65], as well as G. Ryle’s distinc­
tion in [53] between use (as correct, normal use) and usage (as the use that actually
occurs). - We have already referred in 1.1 to the difficulties of ascertaining by empirical
means the linguistic norms that are in force. - It has often been remarked that reasoning
in support of assertions about correct linguistic use is often quite questionable precisely
when it is advanced by ordinary-language-philosophers, among whom the armchair
method of a scientifically uncultivated spontaneous linguistic intuition more often than
not replaces empirical investigations. See e.g. Mates [58].
20 See e.g. W. P. Alston in [63a] and P. F. Strawson in [54].
21 See Chapter II.4.5.
22 See Searle [69], p. 146ff. Searle alludes to G. Ryle’s false analysis of the meaning of
the word ‘voluntarily’ in [49], which arises out of a mixture of semantic and non-
semantic criteria of use.
23 See Searle [69], p. 155, and for the last example Chomsky [68].
24 See Charlesworth [59], Pitcher [64] and Ziff [60], Chapter V. The example with
‘ultor’ is one of P. Ziff’s, see [60], p. 189. - Ziff’s further objections to the identification
o f use and meaning are based upon the error of understanding ‘use’ as nothing but
syntactic use, the distribution of the words.
25 Ryle’s remark in [60], that meaning and use can be identified only for words, but
not for sentences, because sentences are utterances, for which there is no general use
as such, is entirely off the mark, since sentences are also expressions and the use even of
new sentences is determined by the use of the words in them and the rules of grammar.
26 See Chapter II.3.1.
27 P. Ziff urges in [60], p. 34f. that it is better to speak of linguistic regularities than
o f rules, since rules are nothing but explicitly formulated norms for or descriptions of
behavior. We speak o f ‘rules’ here in a wider sense that does not imply explicit
formulation.
28 The fact that the use o f words is a use for the expression of something has been given
special prominence by M. J. Charlesworth in [59], p. 218 where he says: “linguistic use
is defined in terms of meaning, so that meaning cannot be defined in terms of use” - and
J. N. Findlay in [61], where he writes: “The reason why it is absurd to tell us not to
attend to the meaning of expressions but to concentrate on their use, is perfectly simple:
it is that the notion o f use, as it ordinarily exists and is used, presupposes the notion of
meaning... and that it therefore cannot be used to elucidate the latter, and much less
replace it or to do duty for it.” J. Wheatley expresses himself along similar lines in
[63], p. 643.
29 J. L. Austin first called attention to these forms of using language. See the article
‘Performative Utterances’ in Austin [61]. We shall return to the matter for more ex­
tended consideration in Chapter II.4.5.
30 W. P. Alston has advanced some arguments in [63a] against the possibility of sub­
 THEORIES OF MEANING 159

stituting explanations of use for explanations of meaning, arguments that represent for
him objections to the possibility o f identifying use and meaning. One argument, for
instance, goes: If you say “ V ’ is used in the same way as “Z>” ’ you do not presuppose,
as with ‘“a” means b’ that the other party knows the meaning, or the use of ‘6’ - That
may be. But instead o f saying the latter, one can also say ‘“ 0 ” means the same as “6” ’
or in the place o f the former, ‘The word “a” is applied to (a) b \ The possibility of
identification does not depend on such formulations. The rest of Alston’s objections
in [63a] are o f a similar character, so that we will not go into them further here.
31 D . Holdcroft has raised similar objections to explaining the meaning of an expression
in terms o f its use in [64]. - The idea comes easily to mind of sharpening this argument
as follows: Since an explanation in language always presupposes the understanding of
the language in which it is formulated, an explanation of language in language, e.g. an
explanation of what a statement is formulated in the form of a statement, is circular
and thus impossible. But the response to this would be: Naturally an explanation
(in language) is not directed to someone who cannot yet speak, and does not under­
stand the assertive function of a sentence, for example. The sense of such an explanation
is not to teach language, but to deepen an understanding of language that is pre­
supposed, e.g. a presupposed understanding of stating, by means of general definitions
or explications. And there is no circularity in that.
32 Antal [61], p. 217.
33 Wittgenstein says: “How do I explain to someone the meaning of ‘regular’,
‘uniform’, ‘the same’? - To one, let us say, who speaks only French, I shall explain
those words in terms of the corresponding words in French. But someone who does not
yet have these concepts, to him I shall explain the words by examples and by practicing
their u se” [53], 208.
34 Wittgenstein [22], 4.027.
35 Generally speaking, the use o f F will be characterized by a system o f rules with
the form ‘Under condition 7>i(«); F may be ascribed to a if (or: only if, or if and only
if) R f M ) is true’ (i = l ,. .. , n).
36 Wittgenstein [53], 208.
37 See also Chapter II.4.6.
38 Wittgenstein [53], 258. See also 202, ibid.
39 See Wittgenstein [53], 199.
40 Diachronically, i.e. considered in terms of its changes over time, language undergoes
certain transformations, and in the process the rules of use for langue can be changed by
some initially atypical instances o f use from parole. But our argument is based upon a
synchronic consideration of language, just as it is in a particular stage of development.
41 Wittgenstein [56], p. 184.
42 Wittgenstein [53], 211.
43 Wittgenstein [53], 217.
44 Wittgenstein [56], p. 184f.
45 Wittgenstein [53], 381.
46 Wittgenstein also attacks one o f the crucial presuppositions of the question, which
is supposed to be answered in terms of the notion of learning predicates inductively,
namely the presupposition that it is possible to speak o f a fully determinate meaning, or
o f a fully determinate general use o f a predicate of ordinary language.
Wittgenstein refers in this connection first to the openness o f these predicates: they
are not, as we often assume when we are idealizing for logical purposes, defined in such
a way for a determinate domain o f objects that for each of those objects it is settled -
160 CHAPTER II

though it need not be decidable in every case - whether the predicate is to be affirmed
or denied. Ordinary language predicates have a certain horizon of vagueness, rather, in
which there are no longer any binding criteria of use. Thus while for many things there
is no doubt as to whether they are to be described as ‘red’ or not, there are others that run
more into orange or violet, for which it is no longer possible to say with compelling
conviction whether they should continue to be described as ‘red’ or as ‘orange’, or
‘violet’. And for the things we usually encounter, we ordinarily think of, it is clear
whether or not they should be called an ‘easy-chair’. But if we encounter something
that has the shape and the solidity o f an easy-chair, but changes its size all the time and
appears and disappears inexplicably, then we would no longer know what we ought to
say (see Wittgenstein [53], 80). These obscure cases can be disposed of by new linguistic
conventions. In many cases, then, the question as to whether we can apply a word or
not is not a question of fact, but o f convention, of new stipulations about the use of
language.
For another thing Wittgenstein points out the fact that it is not possible to assign to
every predicate F a single property that belongs to precisely those things to which F
can be applied. Instead, there is a sort o f fam ily resemblance among the objects to
which a predicate can be applied, generally speaking, so that this is another reason that
the use of a predicate can not be settled by a sharp conceptual criterion of application
(see Wittgenstein [53], 65-67).
Neither observation, however, is a basis for any objection to the realistic notion of
learning predicates inductively. For family resemblance this follows from the comments
in II.4.7. Openness of predicates is understood from the realistic standpoint in the
following way: a predicate F, defined in the first place only over a domain D and stand­
ing for a co n cep t/in that domain, is supposed to be defined for an expanded domain
D'. N ow there are several continuations o f / o n D' and a new linguistic convention is
necessary, to determine which of these concepts / ' the predicate F defined on D' is
supposed to stand for. Wittgenstein’s easy-chair example depends in this way on the
fact that the predicate ‘easy-chair’ is defined initially only for the usual concrete things,
the things we ordinarily think of. But if some entirely new and unusual sort of thing
turns up, then that predicate is no longer defined for them and new conventions are
required for the expanded domain of application.
47 This objection can not be met by saying: in order to learn language, you must in
fact be able to make certain distinctions. But what being able to make distinctions
means is nothing but being able to learn distinguishing predicates. In this sense, the
capacity to learn language naturally must be present in order for language to be learned,
but that is a mere triviality. What the basis of the human capacity for being able to learn
language, however, what its neurological presuppositions are, etc., that is no philosoph­
ical problem, for clearing it up contributes nothing to the clarification of what we are
doing when we speak. (See e.g. the discussion o f this point by J. Mittelstrass in [68].)
The fact that we can make certain distinctions in the sense that we can learn them is
not what is relevant here. What is involved there is the fact that we can apply them so as to
separate the expressions we are supposed to be learning as well as the learning situations.
And a distinction of that sort cannot be a distinction made by means of language.
48 These ideas will be formulated and argued in greater detail elsewhere.
49 Wittgenstein [53], 242.
50 Wittgenstein [53], 241.
51 Wittgenstein [53], 258. - The qualification in square brackets is not part o f the
quotation.
 THEORIES OF MEANING 161

52 Wittgenstein [53], 242.

53 Wittgenstein [53], 432.
54 Wittgenstein [53], 454.
55 Wittgenstein [58], p. 4.
33 Wittgenstein [22], 3.262.
57 See Chapter II. 1.3.
58 See also Searle [69], 1.4.
59 For the concept of meaning in the theory of speech acts cf. Grice [57], Searle’s
critique in [69], 2.6, and the discussion in Lewis [69], IV.5.
60 The supplementary specifications of the meaning of utterances as against the mean­
ing of expressions are sometimes, but by no means always, the result o f conventions.
Thus Strawson in [64] stresses the point that the function of the sentence ‘This is
Mr. Smith’ has, in the corresponding situation, a conventional character as an intro­
duction (as opposed to an informative statement), while the function of the sentence
‘It is raining’ as a warning that you should take an umbrella along arises directly out
o f the circumstances o f the utterance, not from conventions.
61 According to Austin, meaning belongs to the locutionary act, within which he
distinguished three components: the phonetic, the rhetic (grammatical), and the phatic
(descriptive-semantic).
62 Austin’s theory o f infelicities is supposed to serve for the analysis o f these pre­
suppositions of success of illocutionary acts.
63 The term stems from G. Ryle, cf. [49], p. 238.
64 For criticism of Searle, who in [69], Chapter 8 propounds the notion that the theory
o f speech acts is a setting within which it is possible to derive norm sentences (such as
‘X ought to do F 9) logically from purely factual statements (‘X promised to do F 9), see
for example Kutschera [73], 1.12. - For Searle the preeminent accomplishment of the
theory of speech acts is that it provides a setting for a logical connection between
factual statements and normative propositions and for an adequate treatment of pre­
suppositions. (See [69], Part II). It is possible to have both of these in other ways as
well, as the discussions in Kutschera [73], 5.3 and [74] demonstrate. Thus we do not see
the significance o f the theory o f speech acts in these two points. We see it in something
still more fundamental, that it opens up an adequate access to semantics.
65 Searle, too, urges in [68], p. 412, as does Cohen in [64], that the illocutionary role
is part of the meaning.
66 While Austin stresses the impossibility of giving an explicitly performative formula­
tion to all propositions, Searle in [68], p. 417f. mistakenly draws the opposite conclusion
from his principle o f universal expressibility (that anything that can be meant can also
be said).
67 In setting out his differences with Austin, Searle in [68], p. 243 contests the notion
that speech acts can be called ‘true’ or ‘false’. But instead of saying ‘The sentence is
true, or false’ we can also say ‘An utterance of A is true, or false’. So this locution is
unproblematic.
68 This idea is developed in greater detail in Chapter III.2.1. There we shall go into the
descriptive meaning of terms, predicates, etc. - N .b : While the declarative sentence (f)
(as an act) has a performative meaning and (as phonetic expression object) a descriptive
meaning, in cases (g) and (h) descriptive meanings are defined only for the descriptive
components.
69 D. Lewis in [70], p. 54ff. takes a similar approach to the interpretation of non­
declarative sentences.
162 CHAPTER II

70 D . Lewis offers an example in [70], p. 60: Someone signs an order and talks about it
at the same time.
71 D . Lewis in [70] ascribes the same meaning to explicitly performative sentences as
performative versions of sentences and as performative descriptions of those sentences.
He does not look upon the difference in performative mode between the two interpreta­
tions as a difference in meaning, but as a difference in the sentence’s use. But these
differences as differences of meaning are just what the semantics of performative modes
is about.
72 See e.g. Keenan [72].
73 Relative temporal indicators o f this sort can also be avoided by giving an w-tuple
<,...,> instead of the arguments of the functors D and B for the sequence of speech acts
that make up the discussion or debate.
74 As the use-theory of meaning was a step beyond realistic semantics, and the theory
of speech acts an improvement on use-theory, a decisive step forward has again been
taken by D. Lewis in [69] in his analysis of linguistic conventions. Speech acts, accord­
ing to Lewis, are ways to solve a problem of coordination between speaker and hearer,
and they are only effective on the basis of a convention. The analysis of the coordina­
tion problem to be solved by them gives a deeper insight into the nature of their func­
tion and meaning than has been possible in the theories propounded by Austin or
Searle. The theory of linguistic conventions, therefore, has to be regarded as the proper
fundament of the theory of speech acts.
75 See Wittgenstein [53], 202.
76 Wittgenstein [53], 258.
77 Wittgenstein [53], 265.
78 See Wittgenstein [53], 302.
79 See Wittgenstein [53], 350.
80 Wittgenstein [53], 384. - Wittgenstein in [53], 312 also alluded to the following
thought experiment: If for everyone there was always a pain sensation tied up with
touching certain things, sensations we had only upon touching those things, then it
would be conceivable that we would use an equivalent predicate ‘to be painful’ instead
o f the predicate ‘to feel pains’. We would use it to describe the things in question and
we would be able to say ‘These things are painful’. This predicate would then describe
not a sensation, but a property of things. (See also Strawson [54], p. 47f.) What
Wittgenstein wishes to say is that ‘to have pains’ is only a psychological and not a
physical predicate because the occurrence of pains depends on factors that are different
from one individual to another, not because it stands for private sensations. - But in
similar cases we very definitely are used to differentiating between properties of things
that cause pain and the pain caused, e.g. between the heat of the fire and the pain
sensations it can produce. Also, it does not follow at all from the fact that two predi­
cates have the same extension that they have the same meaning, much less does it
follow in the case before us, then, where the predicates’ ranges of use are different.
81 Wittgenstein [53], 293.
82 Wittgenstein [53], 67.
83 Wittgenstein [53], 66.
84 Wittgenstein [53], 67.
85 Besides that, there is an infinite regress here: If precisely those things are supposed
to have the property F which are called \F ’, then the question arises, which things have
the property of being called ‘F \ Obviously, the things that are called ‘called “F ” ’, and
so on.
 THEORIES OF MEANING 163

86 F. Waismann spoke in this connection of an open texture that empirical concepts

have, in contrast with logico-mathematical concepts. See also Wittgenstein [53], 69-71,
as well as 84, 87, 88 and 142. - See also note 46.
87 On dependence of meaning on context and the polyvalence of words see Chapter
II.3.2.
88 See Chapter II.3.1 also.
89 That such precision is not superfluous is already apparent from the problem of what
similarities we attend to in applying a predicate. Obviously not just any similarity,
for there is always a property with regard to which two objects a and b are similar to
each other, so that all predicates would be universally applicable if we were not to
specify the similarities more narrowly. We surely do not call an object ‘red’ because it is
similar in size to other objects that have been so designated. On the other hand, the
similarities are not defined by means of a classificatory concept, either: Whether a is
called ‘red’ is not determined by its being similar in its red color to other objects so
designated, else one would have to define the similarities by the concepts, and they were
precisely what was supposed to be fixed and determined by similarities.
90 On what follows see also the more comprehensive and detailed presentation in
Kutschera [72a].
91 For a similar reason, we cannot use Carnap’s procedure in [28] for representing
family resemblances. Carnap starts out there from a two-place relation of partial
similarity Ae, that is supposed to hold between two objects if and only if a and b
approximately agree in a quality. Then the qualities are defined with Ae as a starting
point. Even if one ignores the serious shortcomings in the definition, that have been
pointed out by Carnap himself and by N. Goodman in [51], such a procedure still has
no more than formal significance: Intuitively, the relation Ae is no basic predicate,
but is defined in terms of an approximate resemblance between two objects in a quality,
i.e. the qualities are presupposed as being already known. Nor, for the practical handl­
ing of ordinary language predicates, can one assume that Ae is extensionally defined by
citing all the pairs of objects to which Ae applies, as Carnap does.
92 For comparative concepts see for instance the presentation in Stegmuller [70],
Chapter I, and also Kutschera [72], 1.2.
93 See Hempel [65], pp. 165-171. The ‘ideal types’ that Hempel discusses there do not
belong in the context of our discussion.
94 See Trier [34b] on this point, where Trier takes up a position critical of the field
concepts of Ipsen, Porzig and Jolles. On Trier’s field concept see also Trier [31], [32] and
[34a]. Other field concepts are presented in Oehmann [51], p. 72ff.
95 See Hoijer [54b], p. 96.
96 Of course, it is problematic whether both groups of words belong to the same
semantic region. The semantic region o f the intellect itself is marked out from more
inclusive regions, and Trier’s way o f drawing its boundaries is based upon the vocabu­
lary of New High German. Just on the basis of his analyses, it can not be said that
there would be a distinct linguistic field in Middle High German corresponding to that
semantic region, for the word ‘wize’ belongs to a comprehensive theory of virtue, a
theory of ideal human types, not to a terminology for distinguishing pure intellectual
capacities. See Trier [32], p. 422. Generally speaking, there is considerable difficulty in
drawing the boundaries of regions of meanings for different word fields. The relation­
ships are as simple as in the two model cases of colors and pitches only in a very small
number o f cases.
97 Trier would like to apply the field account principally within the realm of the non­
164 CHAPTER II

physical [undinglich], see Trier [32], p. 420. He writes there: “Its view of [semantic
regions as] a whole and of the proposed analysis and classifications of that whole and of
the variations of the internal boundary lines will find the most to consider in the realm
o f objects that are not concrete things. Indeed, the field account must inevitably present
itself as a simple, logical modification of a previously held referential account of
onomasiology wherever a question is raised about referents in such a dom ain.... There
is no referential account of cleverness as there can be a referential account of the sickle;
a referential account o f cleverness can exist only within a total and integral account of
intellectual terminology.” - But what is decisive for the question as to whether or not a
linguistic field is to be assumed is surely not whether the semantic region falls within the
realm of the concrete or of the abstract - abstract mathematical concepts are not field
concepts, while color concepts, for example, do belong to a word field - but whether a
resemblance field is linguistically broken down into classificatory concepts or not. Of
course, the domain of the concrete is particularly rich in phenomena that are constant
and already divided up beforehand in perceptual terms, phenomena to which classi­
ficatory concepts can very easily be attached: plants and animals offer a better basis for
classificatory concepts than phenomena in the psychological realm.
 CHAPTER III

T H E O R I E S OF G R A M M A R

A language in which there is not just a finite number of messages that can
be formulated must solve the task of determining an infinite set of possible
messages with a finite number - practically the fewest possible - of con­
ventions concerning the message content of particular linguistic expres­
sions. The only way in which this can be done is by starting with a finite
set of the minimal linguistic units, with their meanings determined by con­
vention, and then building complex linguistic expressions by combining
these minimal units in accord with a finite number of rules, which deter­
mine the meanings of the compounds from the meanings of their com­
ponents.1
Now it is the task of the grammar of a language L to give the rules for
forming all of the meaningful expressions in L by combining basic units
of L . We limit ourselves in this connection to investigating how single
sentences, particularly declarative, i.e., assertion sentences, are construct­
ed. At the present time, the grammatical analysis of texts is still in its very
beginnings and there do not yet appear to be any sufficiently general and
thoroughly discussed theoretical approaches to that topic.
A grammar includes three components: a syntactic component, i.e., a
system of rules according to which the set of grammatically correct sen­
tences of a language L can be constructed from the words of L listed in
its lexicon; a semantic component, i.e. a system of rules by which the
meanings of sentences are determined from the meanings of the words,
which are given in the lexicon; and a phonological component by which a
phonetic interpretation can be assigned to every sentence of L . Here we
will ignore the phonological component entirely, however, since it scarcely
has any philosophical relevance.2
Now there is a close connection between syntax and semantics. Since
the sentences have a communicative function, they must be formed syntac­
tically in such a way that the rules for semantic interpretation always
assign a meaning to the syntactically constructed expressions. In arti­
ficial logical languages this is achieved by having a rule for semantic
166 CHAPTER III

interpretation corresponding to every syntactic formation rule. It deter­

mines the meaning of the composite expression on the basis of the mean­
ings of the expressions that compose it. No such close parallel between
syntax and semantics exists in natural languages. Despite that, a grammar
should formulate syntactic and semantic rules in such a way that the
connection between them becomes clear, that the semantic function of the
syntactic structures, which is ultimately the point, is elaborated.
The philosophically interesting problem concerning grammar is the
general question of how the combination of semantic elements in lan­
guage works at all, not how the combination of the smallest units of
meaning - the so-called morphemes, which we shall often identify with
words in what follows, as a simplification - works in some specific lan­
guage or how the formation rules for its sentences look in detail. But
since it is by no means established beforehand that there are universal
structures for compounding of meanings common to all existing, or even
to all possible, languages,3 we shall refer to specific languages, like
English or German mostly.
The rules of composition are not explicitly formulated at the outset in
the case of natural languages, but there are customs, regularities in­
tuitively grasped and practiced, just as there are, for example, moral con­
ventions and rules of social conduct. It is then the task of a grammar to
make the regularities of linguistic usage explicit by means of a system of
rules. If one assumes that the set G(L) of grammatically well-formed sen­
tences of a language L is fixed, then a grammar G for L (a system of syn­
tactic and semantic rules) can be called syntactically or semantically
adequate, if the set of expressions that are syntactically discriminated as
sentences in G coincides with G(L), or if over and beyond that G also
interprets these expressions semantically as we do intuitively.4 Naturally
there can be more than one adequate grammar for L . In that case, the sim­
plest of them will be given preference.5 In particular, such grammars can
also differ in the lexicon with which they begin, i.e., in their lists of gram­
matical elements (words, or morphemes), for what counts as grammatical­
ly simple depends on the particular grammatical analysis in each case.
Now the assumption that G{L) is a determinate set with precise bounda­
ries and that every sentence in this set has a determinate meaning repre­
sents an idealization. Actually the boundary between sentences we intu­
itively apprehend as well-formed and other expressions is not a sharp one.
 THEORIES OF GRAMMAR 167

There is rather a spectrum of expressions that are neither definitely well-

formed nor definitely not well-formed, but only more or less well-formed.
In this sense there are degrees of grammaticity.6 In the same way, it is not
always true of meaningful sentences that they can be described as having
the same meaning or different meanings; instead, there is often only a
greater or lesser similarity of meaning.7 As a rule, however, we are in
agreement as to whether an expression is a grammatically well-formed
sentence or not and whether two sentences (in a certain context) can be
regarded as synonymous or not. The idealization is thus reasonable and
defensible in our present context.
As already stressed in 1.1, along with the formulation of grammatical
regularities in the form of explicit rules there is some sharpening as well:
in grammar the set of well-formed sentences is defined more precisely than
it is in linguistic usage and the like holds for the semantic interpretations.
Grammar thus incorporates an explication of linguistic facts: the im-
precisions of linguistic usage are put to good account in order to draw
more precise boundaries appropriate for systematic purposes.8 And so
from the very outset it can not be expected that a grammar only describes,
that it settles nothing over and above linguistic customs. On the contrary,
if it is scientifically exact, it represents a clarification, which is forbidden
to conflict with the intuitive regulative provisions of usage only where the
latter are unambiguous.
There are essentially three types of theory of grammar: traditional,
logical and generative grammar.9 We will discuss these types in what fol­
lows. We begin our discussion with the traditional grammar, which
nowadays, despite all the criticism directed against it and despite all the
alternative approaches, still represents the dominant type of grammar.
And so far no comparable representation of grammatical detail has been
achieved elsewhere, so that at the present time the other grammars still
have to refer back to it.

NOTES

1 In the same way, we obtain names for all of the infinitely many natural numbers by
constructing them out of a finite number of figures and specifying how the meaning o f
the groups of figures is a function of the meaning o f the individual figures.
2 See also 1.1.
3 We will go into the question of the universality of grammatical structures in Chapter
IV.
168 CHAPTER III

4 Instead of syntactic and semantic, Chomsky speaks o f strong and weak adequacy.
5 Making the concept of simplicity precise generates considerable difficulty, o f course.
See Kutschera [72], 4.2.
6 See Chomsky [61b] on this point.
7 See Chapter II.3.1.
8 On the concept of explication see Kutschera [67], 6.3.
9 We regard the corpus-grammatica of American structuralism as the antecedent of
generative grammar.
 THEORIES OF GRAMMAR 169

1. T r a d i t i o n a l grammar

What is called ‘traditional’ or ‘classical’ grammar nowadays is grammar

as based upon the first grammatical distinctions made by Protagoras,
Plato, and Aristotle in Greco-Roman antiquity, further developed in the
medieval and modern periods and still taught in the schools today. The
6Duden-Grammatik\ Grebe [66], represents a good example of such a tradi­
tional grammar for German, in which the old grammar’s basic scheme is
preserved in spite of many modern approaches and modifications. We
shall repeatedly refer to this work above all in presenting and criticizing
traditional grammar.
Since the traditional grammar is well-known, it will be sufficient here
to refer briefly to its basic features.
Classical grammar is divided, if we ignore the modern addition of
phonology once again - the traditional grammar was above all a grammar
of written language1 - into a theory of words (Wortlehre) and a theory
of sentences (Satzlehre). The theory o f words embraces a division of words
into categories - the theory of word types, a theory of the inflection of
words and a theory of word formation.2The theory o f sentences or syntax -
the word is understood in a grammatical sense here, not in the sense
defined in 1.3 - embraces a theory of sentence parts and sentence structures
and a theory of the forms and positions of words in a sentence.

1.1. The Theory o f Word Types or Parts o f Speech

In the theory of word types or parts of speech, the words of a language
are supposed to be exhaustively classified into word categories. These
categories are:
(1) Substantives: Substantives name objects (in the broadest sense of
the word, thus including persons, living creatures and the like).3 A further
distinction is drawn between:
(a) Proper names, which designate individual objects, such as ‘Soc­
rates’, ‘the Zugspitze’, ‘England’, ‘Mars’, etc.
(b) Class names, which designate classes of objects and ‘at the same
time each individual being or thing in the class’,4 such as ‘human being’,
‘flower’, ‘chair’, etc. Subordinate groups of class names are:
(a) Collective terms, which stand for aggregates or concrete collec­
tions, such as ‘forest’, ‘herd’, ‘fleet’, etc.,
170 CHAPTER III

(/?) Substance or mass terms, such as ‘water’, ‘gold’, ‘pepper’, etc.

(2) Verbs: Verbs “ say what is going on or what is the case.” 5 They
express activities, events or states of affairs.
(3) Adjectives: Adjectives tell us “what is the nature of a being or thing
or how an event is taking place.” 6 Their basic function is to express the
position taken by the speaker concerning beings or things (substantives),
to situations or events (verbs), to properties themselves (adjectives or par­
ticiples) or even to the circumstantial features (adverbs), to indicate the
impression that the beings, things, events, properties make on the speaker.
(4) Articles: Articles are modifiers of substantives, with the task of
identifying the object named by the substantive as a specific individual
object or as representative of a class.7
(5) Pronouns: Pronouns represent a substantive, refer to it or accom­
pany it.8 Personal pronouns represent a substantive; possessive pronouns
indicate a relationship of ownership or attachment; demonstrative pro­
nouns a relationship of time or position; relative pronouns indicate that
a verbal expression pertains to a substantive; interrogative pronouns a
question about a being or thing;9 indefinite pronouns, finally, “ designate
a being or thing in an entirely general and indefinite way, if the speaker
does not wish to or cannot describe it in more specific terms. We also
count number words that express a number or amount in an entirely un­
specified way within this group.” 10 Indefinite pronouns of this type are
‘all’, ‘another’, ‘both’, ‘one’, ‘some’, ‘a few’, ‘something’, ‘each’, ‘someone’,
‘no one’, ‘none’, ‘many’, ‘several’, ‘nothing’, ‘whole’, ‘much’, ‘little’,
‘which’, ‘who’.
(6) Numeralia: Number words are modifiers of substantives that give
numerical support to their capacity for differentiating between unity and
plurality.11 Number words are words for specific numbers (cardinal
numbers, ordinal numbers, fractions, etc.), distributive number words
(‘every three’, ‘every three times’), multiplicative number words (‘single’,
‘double’), number words of recurrence (‘once’, ‘sometimes’, ‘never’, ‘an­
other time’) and classificatory number words (‘same’, ‘diverse’).
(7) Adverbs: “It is the adverb’s task to characterize the states or situa­
tions ( Umstande) named in the sentence in terms of their most general
outlines.” 12 Classifying in greater detail: adverbs of place (‘here’, ‘to the
right’, ‘inside of’), of time (‘yesterday’, ‘never’, ‘at the same time’), of
modality (‘gladly’, ‘in vain’, ‘somewhat’, ‘twice as’, ‘firstly’, ‘very’,
 THEORIES OF GRAMMAR 171

‘enough’, ‘not’, ‘never’, ‘perhaps’, ‘probably’) and of reason (‘therefore’,

‘consequently’, ‘therewith’, ‘otherwise’, ‘nevertheless’).
(8) Prepositions: For the most part prepositions designate relationships
between the being or thing named by the substantive and an event or
state (verb), another being or thing (substantive) or an attitude or opinion
(adjective).13 “It is the principal task of the preposition to attach a sub­
stantive subject to it to another word.” 14 They are differentiated as prep­
ositions of space (‘on’, ‘toward’, ‘over’), of time (‘until’, ‘since’, ‘from’),
of modality (‘exclusive of’, ‘except’, ‘notwithstanding’) and of reason (‘on
the occasion of’, ‘thanks to’, ‘through’, ‘under’, ‘by means of’).
(9) Conjunctions: Conjunctions connect sentences or parts of sentences
and express a relationship that holds between words or sentences in
thought.15
(10) Interjections: Interjections express feelings or commands. They
stand outside the sentence frame.16 Since they are not components of sen­
tences, but self-contained exclamations, they are of no grammatical in­
terest and for that reason we shall ignore them in what follows.

Now these specifications of verbal categories cannot be used for con­

structing a grammar with precision, for in the first place they are too
vague for us to be able to attach any exact sense to them or to achieve
any well-defined classification of words with them. To be sure, they are
illustrated by examples and counter-examples, but frequently these are not
in accord with the general definitions. In the second place, the classifica­
tion is not disjunctive, i.e., one and the same word can belong to different
categories. And thirdly, the classification is not based on any one unitary
point of view and words in the same category do not have either the same
semantic nor the same syntactic function.
The following points suffice to establish the first claim:
Ad (1): Only proper names, but not class names, can be said to name
objects. ‘Human’ is neither a name for the class of humans (otherwise the
word would be a proper name) nor a common name for all humans (you
can not say ‘human is mortal’ instead of ‘Socrates is mortal’) nor a name
for indeterminate humans, for there are no creatures in addition to specific
individual humans who would be indeterminate or unspecific humans.
Also, what object is the substantive ‘unicorn’, for example, supposed to
name? Can we say further that substantives like ‘truth’ or ‘beauty’ name
172 CHAPTER III

objects? And finally, how do things stand with substantives like ‘event’,
‘action’, ‘state’?
Ad (2): Verbs do not say ‘what is going on’; only sentences can do that.
What kind of an event is “ sleep” really supposed to express, for example?
A d (3): It is not only adjectives that say ‘what the nature of a being or
thing is or how an event is taking place’. That can also be expressed by
substantives, verbs or adverbs. Furthermore, purely descriptive adjectives,
such as e.g. the word ‘red’, in the sentence ‘This rose is red’, do not entail
any taking of a position on the part of the speaker. That holds true only of
evaluative adjectives like ‘beautiful’, ‘delightful’, ‘unpleasant’, etc. Finally,
the word ‘presumably’ in ‘Presumably Hans is coming today’ expresses the
speaker’s attitude towards an event without that making it an adjective.
Ad (7): What is a state or situation? In the sentence ‘Fritz ran fast’
does ‘fast’ express a situation, or ‘not’ in ‘Fritz did not run’? A state or
situation is a fact and facts are expressed by sentences. And what are the
‘general outlines’?
A d (8): ‘Loves’ also expresses a relation between two persons, without
being a preposition.
Ad (9): That a word connects other words or parts of a sentence is
something that can be said of almost any word. Likewise that it expresses
‘relationship in thought’.
In support of the second statement it is sufficient to point out that Grebe
himself says there are no firm boundaries to be drawn between articles,
pronouns and number words.17 But they also overlap with adverbs, as
shown by words like ‘somewhat’ and ‘twice’, for example.
As to the third point; according to Grebe, the classification of words
into categories is determined by the particular way “in which they enter
into manifesting the world by means of language, or by the presence or
absence of a realm of forms.” 18 But those are two entirely different view­
points: a semantic viewpoint, according to which words are classified
according to their meanings - it is scarcely possible to understand any­
thing else in this context by the very vague and general expression ‘mani­
festing the world by means of language’ - and a syntactic viewpoint.
But a strictly semantic or syntactic definition of these word categories
is not possible either, because many particles,19 for example, do not have
any independent semantic function and because the characterization of
words in terms of certain syntactic functions is only the outcome of a
 THEORIES OF GRAMMAR 173

grammatical theory. Even the distinctions drawn among cases are in

many respects a theoretical construction, since there are often no mor­
phological or semantic distinctions corresponding to them.20 In Chapter
III.2 we shall see that on a very simple and natural definition of the syn­
tactic functions, words with the same syntactic function occur in entirely
different categories and words in the same category have entirely different
syntactic functions. For example, the adverbs ‘not’ and ‘presumably’ are
quite as much as the conjunctions ‘and’ and ‘therefore’ expressions with
which one can construct new sentences out of sentences; while the adverbs
‘fast’ and ‘gladly’ represent expressions with which one can create sen­
tences out of predicates and names just as much as do the verbs ‘love’ and
‘believe’.
In the theory o f the inflection o f words distinctions are made between
gender, number and case in connection with forms of substantives, ad­
jectives, articles, pronouns and numerals, as well as person, number,
mood, tense and voice or diathesis (in [English and] German: active and
passive) in connection with verbs, and positive, comparative and super­
lative forms of adjectives. An account is given of how these forms are
constructed for various classes of words as well. These inflections them­
selves are peculiar to a language and are of no interest to us here. What is
interesting to us, on the other hand, is the function these forms have.
What traditional grammar has to say about them is once again most un­
satisfactory. For example, in addition to the grammatical functions of
case, which of course are not always represented by their own special
case forms, a distinction is often drawn between the subjective function
(of indicating the subject of the sentence, i.e., the object of the sentence is
saying something about (nominative)), the objective function (indicating
the direct object of the activity expressed by the sentence predicate (usually
accusative)), the indirect-objective function (indicating the indirect object
which is collaterally a target of the activity asserted by the sentence pred­
icate (usually dative)), the possessive function (indicating the relation of
possessing (with the genitive)), the instrumental function (indicating a
means used in connection with the activity expressed by the sentence
predicate, as in 'Fritz beat Hans with a stick’ (the ablative in Latin, for
example)), the agentative function (to indicate something that causes an
event, but which is not the subject of the sentence, ‘Hans was visited by
Fritz'), and the comitative function (to indicate someone accompanying
174 CHAPTER III

the active subject, as in ‘Hans met his wife and visited Fritz with her’).21
It is already becoming clear that one and the same case can have en­
tirely different functions in different contexts. Thus, for example, the
German dative can have an indirect-objective function, an instrumental,
an agentative and a comitative. But these functions can scarcely be ascribed
to the case alone, for the same function can be expressed by different
cases. So there is no clear connection between these functions and case.
Furthermore, a number of such function systems have been advanced
and whatever claim to completeness any of them can make is solely on
account of the vagueness with which they define their functions. The
distinction between direct and indirect object, for example, has no basis
in the proceeding sketched out by a sentence like ‘Fritz gave Hans the
book’, so far as the facts of the matter are concerned. Its basis is rather in
the grammatical distinction between accusative and dative object and
therefore it cannot be the basis of the latter.
Attempts to start from the individual case and attach various semantic
functions to it also miscarry, as e.g. the discussion of functional analyses
of the genitive in Groot [56] demonstrates convincingly.22 And so we can
say that no unambiguous and independent functions can be attached to
the morphologically distinct cases.
The theory o f word formation by way of combination (‘underarm’,
‘dogfood’) or derivation (‘fly’ - ‘flight’, ‘idea’ - ‘ideal’, ‘kind’ - ‘unkind’,
‘do’ - ‘outdo’) studies how it is possible in language to obtain new words
from words. It belongs, however, - even from the standpoint of the older
grammar - partly to lexicography, i.e., to the description and analysis of
the vocabulary rather than to grammar as a theory of combination. For
the newly formed words often have an independent semantic function not
derivable from the meanings of the component words.23 For example,
the word ‘Junggeselle’ (bachelor), as far as meaning is concerned, has
nothing to do with either ‘jung’ (young) or ‘Geselle’ (fellow). [‘Cockpit’,
unlike ‘airplane’, has nothing to do with either of its components.] And
even where the meaning of a compound word is defined by the meanings
of the component words (as in ‘underarm’ - ‘the part under the arm’,
‘birthplace’ - ‘the place of birth’, ‘skyblue’ - ‘blue like the sky’), these
combinations do not always follow any general rules, but each one of
them is a unique formation exhibiting only more or less loose analogies to
other formations.
 THEORIES OF GRAMMAR 175

In other cases, however, there are semantic rules which do permit the
introduction of new formations into the language without their having
first to be defined.24 And so we can say that word formation belongs in
part to lexicography, in part to grammar.

1.2. The Theory o f Sentences [or Syntax]

The traditional theory o f sentences says that there is one basic form of
sentence, which can be expanded by means of supplementations, and that
sentences constructed in this way can be combined into sentence se­
quences.
The basic form of every sentence consists in subject and predicate. The
subject indicates the object or objects of which something is supposed to
be asserted in the sentence. What is supposed to be said of that object or
those objects is the content of the predicate.
The subject without any supplementation is a substantive (‘Fritz is
coming’, ‘Men are mortal’), a pronoun (7 am sick’, ‘They are lying’), an
infinitive (‘To err is human’), a clause (Satz) (‘That Fritz is coming makes
Hans happy’) or a series of words (‘Fritz and Hans and M ax are coming’)25
[or a participle (‘Smoking is forbidden’)]. The unsupplemented predicate
is a verb (‘Fritz turns'), or an auxiliary verb [(‘to be’, ‘to have’ in their
various forms)], combined as copula with an adjective (‘Fritz is sick') or a
substantive (‘Fritz is a jurist ’, ‘Fish are vertebrates'),26 or a series of pred­
icates (‘Fritz laughs and sings').
The supplementations are supplements to the subject, to the predicate
or supplements to other supplements. The following supplementary com­
ponents of a sentence are differentiated in the traditional grammar:
object, attribute and adverbial modification.
The attribute involves a more detailed characterization of a substantive,
pronoun, adjective or adverb. What is concerned can be single words
(‘Two trees were felled’, ‘The tree was felled’), word groups (‘Half o f his
great wealth he inherited’, ‘His pleasure at the gift was great’) or clauses
(‘The tree that stands in the garden was trimmed’).
An object indicates which object or objects the subject’s (or subjects’)
activity asserted in the predicate relates to (‘The horse is pulling the cart',
‘Hans strikes Fritz', ‘Hans gives the book to Fritz', ‘I make sport o f him');
and to that extent it is regarded as a supplement of the predicate.
An adverbial modification states where, when, how or why an action
176 CHAPTER III

takes place, and so characterizes the content of predicative components

of sentences as to form and manner (‘The ocean is especially cold today9,
‘He is trying with all his might9, ‘Fritz skis just as well as Hans9, ‘Out o f
diffidence he probably will not protest’).
Sentence sequences arise from the combination of complete sentences,
e.g., by means of coordinating conjunctions like ‘and’, ‘or’, ‘but’, ‘al­
though’, ‘therefore’, etc.
Finally, the traditional theory of combining words in a sentence and
imbedding them in it embraces the theory of word position and con­
gruence and of the forms of sentence components. To a large extent, it
foregoes general statements and essentially restricts itself to describing
the regularities of actual linguistic usage. The role of congruence is seen in
its marking out expressions as belonging together; word position can
determine sentence structure and, as e.g. in ‘Hans loves Mary’, make the
distinction between subject and object where that distinction is not made
by case endings and prepositions.
The traditional model of syntax, its theory of sentence components in
particular, is inadequate for the following reasons:
In the first place, the definitions of subject and predicate are unsatis­
factory : The grammatical subject of the sentence does not always designate
or does not always designate only the object or objects of which something
is said in the sentence. Thus in ‘Fritz loves Bertha’ something is said about
both Fritz and Bertha, in ‘Fritz comes to Bertha with Hans’ something is
said about Fritz, Hans, and Bertha, and in the sentence ‘No one loves
Kuno’, ‘Not a thing will be revealed by Kuno’ it would surely seem to be
Kuno more than anyone or anything else who is being spoken of. To the
extent that what a sentence speaks of is definitely settled, then - and this is
a thoroughly problematic question27 - we can not say that sentences speak
of the objects their grammatical subjects designate.28 It is a similar matter
with the predicate: If there is already some obscurity about the subject-
matter of a sentence, this is also true of what is asserted of it. Furthermore,
the objects and adverbial specifications contained in it also make essential
contributions to determining what the sentence says, as the following
examples demonstrate.
The most striking flaw in the traditional theory of sentences surely
consists in the fact that the subject-predicate kernel of a sentence is very
often not a meaningful sentence. The kernel ‘Regensburg lies’ in ‘Regens­
 THEORIES OF GRAMMAR 177

burg lies on the Danube’ is not a reasonable sentence, for by itself ‘lying’
means something entirely different from ‘lying on’ (it expresses the op­
posite of ‘standing’ or ‘sitting’, and these expressions can not be applied
to cities.) Likewise the sentence kernel ‘Hans gives’ in ‘Hans gives Fritz
a book’ is not a meaningful sentence.
For this reason the object or objects of the sentence are often counted
as part of the predicate. But that is not enough, for adverbial specifications
can also essentially alter the content of a sentence. Thus ‘singing’ in ‘Fritz
sings’ [used as what is sometimes called the narrative present in English]
means something different from what it means in ‘Fritz sings for the
pleasure of it’, for the former sentence makes an assertion about an ac­
tion going on at that moment, but the latter makes an assertion about a
disposition. The ‘supplementations’ are by no means additions that can
always be omitted, but often are just as important elements of the sentence
as subject and predicate.
For this reason the attempt has often been made to replace the one
subject-predicate schema with a number of basic forms of sentence, sen­
tence models or sentence designs, and to define supplements relative to
them.29 But this attempt is not very convincing, either.
So far no unity has been achieved as to the identification of certain basic
sentence designs. The reason for this is the inadequacy of the so-called
method o f reduction (Abstrichmethode)zo for investigating such designs.
Every assertion is taken to be an assertion about ‘a something-or-other’,
of which an assertion ‘with a relatively definite character’ is made. Ac­
cording to the method of reduction it is supposed to be possible to free
this kernel of the assertion by striking out the assertion’s supplements, in
such a way that ‘whatever posit is intended’ by the sentence is still con­
tained in this kernel. So in the sentence ‘The farmer plows his field early
in the morning’, we can strike the expression ‘in the morning’ as a supple­
ment that can be dispensed with, but not ‘his field’, because in this case
plowing-a-field is what is meant to be posited and just not plowing.31
What ‘posit’ the speaker intends can not be inferred from the sentence
as such, however. It may plainly and simply be the case that the primary
intent does lie in the specification of nothing but plowing (the emphasis
then falls on ‘plows’ and says that the farmer is not about the business of
harrowing, say), or in specifying that the plowing is going on in the early
morning, not just before noon, say (the emphasis then falls on ‘early’).
178 CHAPTER III

In general, one will have to assume that the speaker wishes to say every­
thing he does say in the sentence, i.e., that nothing specified in the sen­
tence is non-essential.
It is a further basic flaw in this method that it defines sentence designs
in terms of the traditional categories (e.g. subject-predicate-supplement-
type-prepositional object), to which, as we have seen, there correspond no
single grammatical functions of expressions (and so the sentence design
identified embraces such different sentences as ‘Fritz laughs heartily at
Hans’ and ‘No one laughs openly at himself’). But sentence designs could
only have a determinate grammatical function if they are composed in the
same way of components that have the same function. Even the character­
ization of these sentence structure functions proves to be correspondingly
imprecise, then.32

In summary, then, we can say this: The traditional grammar does bring
out a wealth of linguistic regularities, but for the most part those regu­
larities do not hold rigorously, but hold only in typical cases that are
pointed out as examples. In other cases, also pointed out only in examples,
cases that are very numerous, they do not hold. In most cases, statements
intended to gather together in one general linguistic rule the regularities
of linguistic usage illustrated by examples either remain too vague or do
not hold without restriction. And so this grammar is far from providing
a rigorous system of rules for forming sentences.
Something else is lacking, too, however: an exact functional analysis
of words and parts of speech that defines what is accomplished by them
semantically and tells us how the meaning of the sentence is determined
by the meanings of the words. What is particularly wanting is a sufficiently
precise conceptual instrumentarium. For functional analysis, the tradition­
al grammatical categories are inadequate and too vague. The problem of
functional analysis is taken up in logical grammar. Since a solution of
this problem, as will yet be demonstrated, is also a presupposition of the
feasibility of generative grammar, we turn first to logical grammar in the
next chapter.
At this point in the discussion, however, we can already say the follow­
ing: the advantage that has often been claimed for traditional grammar as
against other grammars is its proximity to language. This consists in the
fact that it analyzes grammatical composites (sentences) according to the
 THEORIES OF GRAMMAR 179

words (or morphemes) of which they are constructed. The elements of its
analyses are thus the expressions out of which the sentences are com­
pounded in a purely external syntactical manner. These ‘natural building
blocks’, however, usually do not have any fixed semantic function in
natural languages, one that is independent of context. Natural languages
are not constructed atomistically: The meaning of a sentence is on the
one hand determined by the meanings of the words of which it is com­
posed, to be sure - without the whole being determined by the parts in
this way it would not be possible to construct an infinite number of sen­
tences with a finite set of words - but conversely the specific meaning of
the individual words in the sentence is determined by the context, by the
relationship in which they stand to the other words in the sentence. Words
have a certain indeterminacy of meaning, then, which is eliminated only
in their context. That we not only understand the sense of the sentence
on the basis of the meanings of the words, but conversely understand
word meaning on the basis of sentence meaning is also shown by the fact
that we actually learn and use words only in context. This plasticity of
words, as we will call it, is also illustrated by the fact that we very com­
monly have to respond to the question ‘What does this word mean?’ by
countering with the question ‘In what context?’ It is a property that dif­
ferent words possess in different degrees and may not be confused with
the ambiguity of words - although there are no firm boundaries in that
respect, of course. On the whole, it contributes significantly to the ex­
pressiveness of natural languages.33
The plasticity of words involves a certain difficulty for the construction
of a grammar as a rigorous system of rules, however: precise and suf­
ficiently simple grammatical rules will always have to be formed in such
a way that they determine the meaning of compound expressions as a
function of its components and ignore an additional determination of the
meanings of the components by their context. I.e. they will have an
atomistic character. The elements of analysis for such grammars will in
many respects not coincide with the ‘natural’ building blocks of language,
words (morphemes), therefore, but must be semantically determinate word
groups or parts of words or even constructs that do not themselves occur
in the language. But in virtue of that the grammatical analyses become
‘artificial’, grammar becomes a theoretical construction, fundamentally
different from the mere description of superficial, i.e. directly observable
180 CHAPTER III

linguistic facts and generalization from them. The words of Roger Bacon
‘philosophus grammaticam invenif hold true of this interpretation of
grammar as a theoretical construction.34 To be sure, this theoretical
construction must have an empirical linguistic content and it must prove
itself to be an adequate and fruitful theory in terms of that content.
As a consequence, for natural languages there is a certain incompati­
bility between the demands of proximity to the language and of precision,
and the root of the vaguenesses and inadequacies of traditional grammar
probably lies in the fact that it wants to relate a precise system of rules -
the construction of which must be the goal of a scientific grammar - to the
natural building blocks of language, which in many ways just have no
firm, context-independent function.

NOTES

1 See also Lyons [69], p. 38ff.

2 The theory of how the vocabulary is divided up, e.g. into word fields, belongs to the
theory of meaning.
3 See Grebe [66], p. 65, 134.
4 Grebe [66], p. 135.
5 Grebe [66], p. 64.
6 Grege [66], p. 65.
7 See Grebe [66], p. 66, 152.
8 See Grebe [66], p. 66, 246.
9 See Grebe [66], p. 246.
10 Grebe [66], p. 272.
11 See Grebe [66], p. 66, 282.
12 Grebe [66], p. 302.
13 Grebe [66], p. 67.
14 Grebe [66], p. 315.
15 See Grebe [66], p. 67, 334.
16 See Grebe [66], p. 67.
17 See Grebe [66], p. 247.
18 See Grebe [66], p. 64. - This definition, which is generally the customary one in
traditional grammar, goes back to Dionysius Thrax, by whom the substantive is
defined as a word that has case and designates persons or things, the verb as a word
that does not have case but does have person, number and tense, which designates an
action or passion, etc.
19 Particles are the non-inflected words, i.e. adverbs, prepositions, conjunctions and
interjections.
20 See also Lyons [69], p. 291f. as well as Beckmann [63].
21 See Lyons [69], p. 295f.
22 See also Fillmore [68], p. 7. - Fillmore makes a study of semantically determined
depth case in this work.
23 See also Grebe [66], p. 348.
 THEORIES OF GRAMMAR 181

24 See also Brekle [70a] and Marchand [69].

25 See Grebe [66], p. 527, 543.
26 While the older traditional grammar views the auxiliary verb ‘to be’ in the last three
examples as a dependent component of the predicate, as we indicated, in Grebe [66],
for example, there occurs the notion that “to be” is an independent predicate here,
while ‘sick’ in the first example is an adverbial supplement and ‘a jurist’ or ‘vertebrates’
in the last two sentences are adjunctions in the so-called ‘nominative of identification’
(See op. cit., p. 540ff., 472f., 531). But this notion is hardly tenable. For the unsup­
plemented component expression in the first sentence ‘Fritz is’ is not a complete and
meaningful sentence. To be sure, the word ‘is’ can also be used in the sense of
‘exists’, but in that case the supplement is impossible. We can not say: ‘Fritz exists’, -
‘How does he exist?’ - ‘Sickly!’ Likewise, there can be no talk of identification in the
second and third cases. ‘To be’ can not be interpreted as having the meaning of the
predicate ‘to be identical with’ in those cases, for Fritz is not identical with a jurist
(which one? !), all jurists or the class of jurists and the same holds for the third sen­
tence. Therefore the older conception of the auxiliary as copula is more adequate. -
Auxiliaries that are used in constructing verb forms (‘Fritz has slept') count as part
o f the verb.
27 See Goodman [61], for example.
28 See also Lyons [69], p. 343f., as well as the discussion of sema and rhema, or of
topic and comment. There are references and bibliographical information in Lyons
[69], 8.1.2 and in Brekle [70a], 3.5 and 4.2.1.3.
29 For the terminology and the idea of sentence design see Weisgerber [63a], p. 264. —
30 See Weisgerber [62], V. II, p. 372, and Grebe [66], p. 468ff.
31 The example and terminology are taken from Grebe [66], p. 468ff.
32 A quotation on this point: “While the subject of the unsupplemented sentence was
something the statement touched on directly [‘Nobody is perfect’?], and the subject of
the identification sentence was something simply held up for comparison [‘Fritz is an
unknown’?], here [i.e. in a sentence with an accusative object] the subject is the agent,
whose action expressed in the predicate is purposefully directed to another being or
thing [‘Snow covers the earth’?].” Grebe [66], p. 473.
33 See also Lyons [69], p. 406, 410, as well as Chapter II.3.2.
34 Cited in Robins [51], p. 76, Note 2.
182 CHAPTER III

2. L o g i c a l grammar

In modern logic, as built on foundations laid down by G. Boole, A. de

Morgan and especially G. Frege from the middle of the 19th century on,
artificial languages are constructed in order to avoid the vagueness,
ambiguities and lack of rigorous and simple general grammatical struc­
tures that are characteristic of natural language. The syntax of these artifi­
cial languages follows simple and exact rules and, following Leibniz’s idea
of a characteristica universalis, is built up in such a close relationship to
semantics that the syntactic form of expressions reflects the structure of
their meanings. There is a grammar, a syntax and a semantics for such
artificial languages; they meet all the requirements of exactitude. It is
natural to ask whether such logical models of grammar can not be applied
to natural languages as well.
Now such an application cannot proceed in such a way as to employ
the syntactic and semantic concepts for a logical language L in the
analysis of a natural language S directly - S just is not a language con­
structed “with the edge of a logical ruler” (Frege). The attempt, rather,
will be to present an interpretation M of L and an analyzing relation
R (X , Y) which correlates to every expression Y of S a grammatically
well-formed expression X of L that has the same meaning. In doing so,
several expressions of L will be correlated to an ambiguous expression in
S . The domain consisting of the items that have the first position in R , i.e.
the set of expressions A of L for which there is an expression A' of S with
R(A, A'), is thus a subset of the grammatically well-formed expressions
of L and the relation R is many-many, since there cannot only be more
than one A with R(A, A') for one A' but also more than one A' with
R(A, A ') for one A - the same proposition can often be expressed by more
than one sentence in S . If we are successful in arriving at a general defini­
tion of R then the grammar of L gives us the two fundamental grammatical
specifications of S as well:
(1) An expression A' of S is grammatically well-formed if and
only if there is an expression A of L with R(A, A').
(2) An expression A' of S has meaning a if and only if there is an
expression A of L with R(A, A') that has the meaning a.
In this way, using L and R, we obtain a logical grammar for S.
 THEORIES OF GRAMMAR 183

In what follows we shall do no more than construct a logical language

L with sufficient power of expression and offer some examples of the
analysis of expressions in natural language. We cannot attempt to define
R for a natural language or even an interesting fragment of a natural lan­
guage1, for that requires detailed and lengthy analyses, and all we are
concerned with here is presenting the basic ideas of logical grammar. Of
course part of what belongs to that enterprise is a sketch of the structure
of a logical language L that has a rich stock of expressions and in carrying
this out we must also burden the reader with some technical details and a
knowledge of elementary logic is useful for understanding them. What
justifies making this demand of the reader is that logical systems like L are
still not very well known and that they have an eminent interest for the
philosophy of language since they offer the possibility of linguistic analyses
that satisfy modern criteria of exactitude.
The works of R. Montague form the basis of the following exposition,
especially his essay ‘Universal Grammar’ of 1970, in which a powerful
and intuitively satisfying semantics of intensional logic was developed for
the first time. Our language L coincides in essential respects with the
language L 0 presented there. For other approaches see, for instance, D.
Lewis [70] and Cresswell [73].

2 . 1. The Structure o f Logical Language L

2 . 1.1. The syntax of language L. L is constructed in two steps. First the
syntax of L is specified. The main task is to define the notion of well-
formed expression, or term, as we shall say. Then in the semantics we
define the concept of an interpretation of L. Instead of the traditional
word categories we define logical categories as follows:
(Logical) Categories:
(a) a, v are (basic) categories.
(b) If t, p are categories, then r(p) is also a category.
(c) If t is a category, then i(t) is also a category.
<r is the category of sentences, v that of proper names (names of objects).
There is no need to assume any further basic categories besides a and v,
although that could easily be done. t(p) is the category of a functor that
generates expressions of category t out of expressions of category p. i(t) is
the category of intensions of expressions of category t.
184 CHAPTER III

<j(v), then, is the category of a one-place predicate (such as ‘ is human’)

which generates sentences out of names, v(v) is the category of a one-place
functor (such as ‘the father of’) which generates a new name out of a
name, <t(g) is the category of a sentential operator (such as ‘not’) which
generates a new sentence out of a sentence. A two-place predicate (such as
‘hits’) which generates a sentence from two names, has the category
(<j( v))( v) - we can also write <j ( v , v) for this - a two-place predicate (such
as ‘believes that’) which generates a sentence out of a name and a sen­
tence has the category (cr(v))(cr) - we can also write cr(v, a) for that;
<j(<j( v)) is the category of an expression (such as ‘all’ and ‘some’) which
generates a sentence out of a one-place predicate, v(<j( v)) is the category
of an expression (such as ‘the thing which’) which generates a name out
of a predicate, etc.
The alphabet of L consists of the basic symbols 2, p, = , 5, (,) and a de-
numerably infinite number of constants and variables for each category t.
As the occasion arises, we shall indicate the category of an expression by
putting an index above it.
An expression of L is a finite sequence of basic symbols of L. If * is a
symbol that does not occur in L, and if A [*] is a finite sequence of basic
symbols of L and this symbol, then A [a] is the expression generated from
A [*] by replacing all occurrences of * by a. If the symbol * does not occur
in A[*]9 then A[a\=A[*]. The set of grammatically well-formed expres­
sions or terms of L is specified by the following rules:

2.1.1.1. Terms o f L
(a) Constants of category t of L are terms of category t of L.
(b) If F is a term of category t( p) (t ^ z) and if a is a term of category p
of L, then F(a) is a term of category t of L.2
(c) If a and b are terms of L of the same category, then a=b is a term
of L of category a.
(d) If A[a\ is a term of category t, a a constant of category p, x a vari­
able of category p of L which does not occur in A [d\9 then Xx(A [x]) is a
term of category t(p) of L.
(e) If A is a term of category t of L then p(A) is a term of category z(t)
ofL.
(f) If A is a term of category z(t) of L, then 5(A) is a term of category t
of L.
 THEORIES OF GRAMMAR 185

Sentences of L are terms of L of category a, proper names of L are terms

of L of category v.
Parentheses that are not necessary to delimit the scope of an operator
A, )U, or 5 unambiguously may be omitted in what follows.
Let L x be that part of language L in which the operators p and <5 do
not occur and the only constants and variables belong to those categories
constructible according to rules (a), (b) alone.

2.1.2. Extensions. We shall build the semantics of L up in several stages,

and consider first of all the case in which the expressions of L have nothing
but extensions assigned to them. In this connection, we restrict ourselves
to sublanguage L ±.
Let A B be the set of functions defined over domain B and taking values
from range A .3

2.1.2.1. Let EtU be the set o f possible extensions of the terms of L x of

category t, relative to object domain U. We set:
(a) E v>u= U
(b) Ea,v = { tJ }
(c) E t(p),v = E t,
‘t 9represents the truth-value ‘true’ and the truth-value ‘false’.

2.1.2.2. An extensional interpretation of L t over the (non-empty) object

domain U is a one-place function M, for which:
(a) M (a)eE xU for all constants a of category t.
(b) M(F(a)) = M (F) (M(a)) for all terms constructed in accord with
2.1.1.1b.
(c) M (a = b) = t iff M(a) = M(b) for all terms constructed in accord
with 2.1.L1.C.
(d) M (XxA[v]) is that function/ from E z(p) u for which: f(M '(b)) =
= M ’(A[b\) for all M '= bM. The term XxA[x] is to be constructed from
A[a] in accord with 2.1.1-l.d, and the constant b of the same category as
a should not occur in A[a\.
M ' = bM says that the extensional interpretation M ' coincides with M
with the possible exception of the values M(b) and M'(b). I.e., M ' is based
upon the same object domain as M and M(c) = M '(c) is true for all con­
stants c ^ b .
186 CHAPTER III

The term AxA[x] stands for what Frege called a Wertverlauf (value-
distribution), i.e. for the function which takes the value A [a] for the argu­
ment a, taken as a class of pairs consisting of an argument and the corre­
sponding value of the function. Therefore we have AxA[x](a)=A[a\. If
A [a] is a sentence, then AxA [x] is a one-place predicate. Concepts may be
represented, according to Frege, as functions with the range {t , f j . This
allows a strong simplification in the formalism. Classes as extensions of
concepts correspond then to the Wertverlaufe of such functions. There­
fore predicates are in effect assigned classes by 2.1.2.2.d, as usual in
extensional semantics.
The following definitions show how rich the type-theoretical language
L x is.4 Here A is to be the universal (‘For all things...’), V the existential
quantifier (‘For some things...’), “ i the negation, a the conjunction, v
the disjunction, => the implication sign, and e represents the relation of
being an element of a set.

2.1.2.3. Definitions. We can now define the usual logical operators in

terms of the operator = , for example:
(a) Ax \A ): = AxzA = A x \x x= x x)
(b) —iA: = A = f \ x a(xa).
(c) A a B:= Ax a{a\B = { x a^ \ A ) = x a^ \B )))
(d) A v B : = —i(- i A a ~i B)
(e) A idB:= - iA v B
(f) V x TA: = Ax T-iA .
(g) apeba{p): = ba{p\ a p).

It is often suitable to operate with names that do not name any (real)
object, and so have, in the sense specified in II. 1.2, no referent, names such
as ‘Odysseus’ or names for persons no longer living, such as ‘Socrates’ or
‘Eisenhower’. While in many contexts the use of such names makes no
sense (What, for example, does ‘Socrates is sick’ mean, uttered at a time
in which Socrates is no longer alive?), in other contexts (such as ‘Hans is
dreaming of Odysseus’, ‘Nixon remembers Eisenhower’ or ‘Eva believes
that Socrates is a living philosopher’) they can be used in a perfectly in­
telligible manner. But if a is such a name, one that has no referent, then
the principle A[a\=> \fxA[x\ should not hold and AxA[x]^>A[d\ should
be equally invalid.
 THEORIES OF GRAMMAR 187

Now if we regard U as the set of ‘possible objects’5 and single out a

non-empty set Uf within U as ‘existing objects’, then we can use a con­
stant E of category <r(v), for which we have
(e) M (E )= U'
and define
(h)
(i) V jcM [xv]: = -i A -x '-iA [xv].
These new quantifiers then refer only to existing objects so that we have
A[a\AE(a)=> V [ x ] ) but not ^ V [^]j and a E(d)
zdA[a], but not A .xA[x]ziA[a\.

2.1.3. Intensions. We shall now consider the assignme ntof intensions,

too, to the expressions of L and to do so we move from L t to L . L can be
characterized as a type-theoretical language o f modalities, then. The con­
struction of a semantics for such an intensional language is to be seen
as the decisive step in determining a logical grammar, since natural lan­
guages are very rich in intensional contexts, so that an extensional logical
language does not represent any suitable instrument for their analysis.
On R. Carnap’s understanding of it the intension of an expression is
that function which establishes its extension relative to possible worlds.6
A world is defined in terms of the set of objects that belong to it and their
attributes. So if what attribute a predicate means is a settled matter, then
its extension for every world is settled as well. But the converse also holds
if we understand attributes in such a way that they can only be distin­
guished if they have different extensions in at least one world. Understood
in this way, attributes are the intensions of predicates.
Apparently not all objects existing in one world also exist in all others.
Are there any objects at all which belong to different worlds? According
to what criteria may an object a in world i be said to be identical with an
object b in world j l Or can we only speak of similarities or correspon­
dencies between objects of different worlds?
These problems have been extensively discussed in the literature. There
are three proposed solutions.
(1) The simplest approach, which we shall take in what follows, is to
take objects as individuals which may be identified or distinguished in­
dependently of their attributes. Then there may be the same objects in
different worlds. We shall even assume that all the worlds have the same
188 CHAPTER III

set U of possible objects. This still leaves the chance to assign the worlds
different sets t/f of objects existing in them.
(2) The second approach arises from the idea that we can identify
objects only by means of the attributes they have. In fact we should say
that two objects that have different essential properties cannot be identical.
So what is, say a planet in one world cannot be a rabbit in another. In
correspondence to the Leibniz-principle according to which (in one
world) an object a is identical with b iff they have the same properties, we
should therefore have a criterion for trans-world identity.
We can, however, still use the first approach even if we consider coin­
cidence of essential properties as a necessary condition for identity. For
this coincidence may be understood not as a defining criterion for identity
but as a principle for choosing the admissible worlds. Then essential
predicates would have the same extension in all worlds. The identity of
the objects in different worlds may furthermore be guaranteed by the
coincidence of other essential properties besides those expressed in L .
(3) The third approach is to give each world its own domain of objects
so that no object occurs in two worlds at once. In place of trans-world
identity we then have a counterpart relation, a relation not of identity but
only of correspondence between objects in different worlds. This makes
for higher generality but the simplicity of the first approach is to be pre­
ferred as long as this generality cannot be put to use.7
In every case there must be a relation of direct or trans-world identity
or of correspondence between the objects of different worlds so that the
definition of intensions makes sense. If the name a were to refer to quite
different things oci9 which have nothing to do with each other, in different
worlds i then the function / ( 0 = ai cannot be understood as an intension
of a in any sense, however loose.
Now let / be a set of worlds with the common domain U of possible
objects and domains Ut of existing objects (ie l)9 such that £/*(=£/. We
shall interpret proper names as standard names in the sense of II. 1.5, i.e.
they are to have the same reference in all worlds.
According to Carnap’s idea, for every term A of L there will be an
extension E t(A)9 and the intension of A - represented in the object lan­
guage by p(A) - is the value distribution XHE^A). ‘A*’, as distinguished
from the symbol ‘2’ of the object language, is to be the metalinguistic
symbol for functional abstraction.
 THEORIES OF GRAMMAR 189

The extension of a function expression, such as ‘It is necessary that

p \ for example, often depends not only on the extension but on the inten­
sion of the argument. We assign to the argument place in the functor ‘It
is necessary that -’ the category i(g) for propositions (as intensions of
sentences), then, instead of the category a for sentences. But since we can
express the intension of *p9in L by ‘p(p)\ we can also say that extensions
of function expressions always depend only on the extension of their
arguments, along Fregean lines, if we write ‘It is necessary (/i(p))’ for ‘It
is necessary that p \ The word ‘that’ can often be interpreted as the ordi­
nary language equivalent o f‘/i’ (applied to sentences). Following this proce­
dure, however, one must introduce special extensions of type i(t).
We complete the specifications in 2.1.2.1 with this:
(d) E i { x ) i u : =E Zi u I
and give the following definition:
2.1.3.1. An intensional interpretation of L over the (non-empty) world
domain / with object domain U and (non-empty) domains Ut with £/
is a two-place function M t(a) such that for all iel:
(a) M t (a)eE zU for all constants a of L of category t.
(b) M i(a) = Mj(a) for all constants a of category v and all j e l .
(c) = M t(F) (M t(a)) for all terms constructed in accord with
2.1.1.1b.
(d) Mi(a= b) = t iff M t(a) = M t(b) for all terms, in accord with 2.1.1.1c.
(e) M i(E )= U i.
(f) Mi{XxA[x\) is that function/ from E z(p)iU, for which: f(M ((b)) =
= M[(A [6]) for all M ' with M ' = bM and M[(b) = Mj(b) for all j'e l;
the term AxA [x] is to be constructed in accord with 2.1.1.Id and the
constant b of the same category as x is not to occur in AxA [x],
(g) M M A ) ) = k*iM {A)8
(h) M t8iA )) = M j A m -
Now what M ' = hM says is that the intensional interpretations M, M '
are at most distinguished with respect to the values M^b) and M((b) for
any number of iel.
M t(A) is the extension of the term A in world /, Ai*Mt{A) its intension.
Definition 2.1.3.1 requires some explanation:
(1) If the variable x in Ax(A [x]) does not appear within the scope of an
occurrence of p, then we can also give the following definition: M t(AxA [x])
190 CHAPTER III

is that function / from E t(p))U for which /(M /(b)) = M[(A [Z>]) is true for
all M ' with M ' = bM ; since in that case for all M \ M" with M ' = bM 9
M " = bM and M[(b) = M?(b): M [(A[6])=M ” %(A[b]).
But that does not hold true if AxA [x] is e.g. the expression AxvG*(l(<r))
( p(Fffiv)(xv)))9 for p(F(a)) can also depend on the intension of a, i.e. on
the values Mj(a) withj # /, even if it does not have to depend on these values
in every case. Now if there is such dependence, then the expression
AxG(p(F(x)))9 which is supposed to be a function taken from Ea(v)tU9 has
no sense. But the formation of terms AxA[x] cannot be restricted to
cases in which x does not occur within the scope of an occurrence of p.
For there are also interpretations of G and F9 for which the truth value
of G(p(F(a))) does not depend on the intension of a9 but only on its
extension. Deontic contexts, such as 0(p(F(a)))(F(a) is obligatory), are
of this sort, for example.9 In such contexts, we can not renounce the
formation of terms like AxA [x]9 f\x A [ x \ \fxA[x\. And so the formation
of terms AxA [x] in general must be syntactically permitted, and they must
then be so interpreted as to have their normal sense when the dependence
in question is not present. But condition (f) provides for that. The problem
can be solved better when partial interpretations are used.10
(2) The operator 5 is required in order to be able to construct the term
Ax*(v)G<T(*(<T))(juC/7,<r(v)(5(jc*(v)))), for example, and with it the sentence
/\xG (fi(F (d(x)))-for all intensions of type i(v) it holds that G9 applied
to the intension of F, applied to the extension of x 9 is true. While
AxG(fi(F(x))) is a function belonging to Ea^ )tXj9 AxG(p(F(3(x))) is a func­
tion belonging to Effil(v))>u.
We can now supplement the definitions in 2.1.3 by defining modal
operators in L, as for example:
N A : = fi(A) = /i( A x v(xv—x v)) (It is necessary that A)
M A : = N ~iA . (It is possible that A)

2.1.4. Pragmatic referents. So far we have constructed a semantics of

expressions. It is the intent of this section that a semantic characterization
be given of utterances as well. Semantics can not be concerned only with
expressions, but must also take utterances into account, for the meaning
of an utterance, as we have already pointed out in Chapter 1.1 and II, by
no means always coincides with the meaning of the expression that is ut­
tered, but is often specified more precisely by the pragmatic circumstances
 THEORIES OF GRAMMAR 191

of the utterance. That becomes clear, for example, in the fact that a deter­
minate extension can not be correlated with the expression but only with
the utterance. Our earlier illustrations of this phenomenon concerned
index expressions and descriptive predicates as examples.
According to Montague the relevant pragmatic parameters of an ut­
terance can be condensed in the index j (J is thus an «-tuple of parameters),
the point o f reference of the utterance. Let J be the set of these points of
reference. One can then interpret utterances as pairs <A J } of an expres­
sion A and an index y, following Y. Bar-Hillel, and assign extensions and
intensions to such pairs; or the parameter j is included as an additional
argument in the interpretation M, or M t, of expressions. The two ap­
proaches are of equal value.
We give the following definitions:
A pragmatic extensional interpretation of L x over the (non-empty)
object domain U and the (non-empty) index set / is a two-place function
M j{x\ such that for all j e J Mj(x) is an extensional interpretation of L x
over U as defined in 2.1.2.2.
Pragmatic intensional interpretations are defined in a corresponding
manner.
We designate M itj(A) as the extension o f the utterance <A , j ) relative
to i and A*iMifj(A) as the intension o f the utterance <^4,y >. And we call
X*jMitj(A) the extension o f the term A relative to i and X*ijMitj{A) the
intension o f the term A.
If a term A does not depend on its pragmatic context, then Mj(A) =
Mr {A), or M itj(A) = M UJ. (A), respectively for all j J 'e J . On this condition
the extensions, and likewise the intensions of all of the utterances of A
are the same. The extension of A is the function with the constant value
M(A) = Mj(A), and analogously for intensions.
The problem of applying this logical schema for pragmatic interpreta­
tions concerns whether and how meaningfully boundaries can be placed
around the set of pragmatic parameters to be taken account of by the index
y. The speaker, the person(s) addressed, the person(s) or (objects) spoken
of (as referents of 3rd person personal pronouns or of demonstrative pro­
nouns), the place and the time of the utterance surely belong to it, but as
a rule that is not enough, as is shown by the example of the dependence
of the meaning even of descriptive expressions on pragmatic context that
was discussed previously.
192 CHAPTER III

This characterisation of an intensional logical language will suffice here.

For an application of intensional logic in analyses of sentences of natural
languages we would have to supplement the semantics of L , for instance
by introducing partial interpretations for dealing with syntactically well-
formed but meaningless expressions, or by bringing in anormal (logically
impossible) worlds to determine a more narrow concept of meaning than
that of intension. But since we do not intend to give systematic analyses
our definitions are sufficient in so much as they illustrate how logical lan­
guages may be constructed syntactically and semantically, and that is all
we wanted to do here. For more details we have to refer to the literature.11

2.2. The Logical Analysis o f Sentences in Natural Language

We shall now illustrate by a few examples how to analyze natural lan­
guage sentences in L. We confine ourselves to an analysis of the logical
form of such sentences, i.e. we do not refer to a specific interpretation of
L, but only coordinate analysing sentences B of L to English sentences A,
so that for a suitable interpretation of L B becomes synonymous with A.
This means that we only describe the structure of A with the help of the
logical categories introduced in the last chapter.12
So that we do not have to specify the categories of constants of L all
over again in every example, we stipulate that the symbols in the left hand
column of the following table belong to the categories indicated in the
right hand column:
a, b , c, d constants and
x, y , z variables of categories v and i(v)
p, q constants and
r, s variables of categories c and 1(0)
F, G, H , I constants and
f, g , h variables of categories ct( v ),
a(v, v),..., <t ( i ( v ) ) , . . . , a(a),
a(a, a),..., a(v, a),...,
U, V constants and
u, V variables of categories v(v),
v(v, v),..., v(i(v))
M, N constants of categories o{o (v)), o(v, o-(v)), 0(0, v(v)),
ffOOO)))
T, W constants of categories v(o-(v)), v(v(v)), v(v, cx(v)),...
 THEORIES OF GRAMMAR 193

These provisions are sufficient to identify the categories of the constants

in the contexts that follow. We shall write, as discussed in III, 2.1, ‘cr(v, v)’
for ‘(cr(v))(v)’ and ‘F(a, by for \F{a)){by and use analogous abbreviations
for other many-place categories and predicates.
In order to bring the following examples into some system, we shall go
through the categories of traditional grammar referred to in III. 1 and
show how they can be represented from the point of view of logical gram­
mar. For the sake of perspicuity we shall also make use of the logical
operators defined in III.2.1.
Let us first consider the traditional theory of words. If we study word
types with regard to the logical categories that occur within them, what
emerges is the following:
(1) Substantives: Proper names are names in the logical sense. Articles
in front of proper names are part of these (‘the Zugspitze’, ‘the moon’).
Class terms, on the other hand, are predicates, mostly one-place predicates
like ‘man’ (category cr(v)), but also predicates with more than one place,
such as ‘gift from - to’ (category cr(v, v, v)), predicates of predicates, such
as ‘type of sport’ (category <x(i(cr(v)))), sentential predicates such as ‘sur­
prise’ (category a(i{a))) and function expressions such as ‘content’ (cate­
gory v(v)). Collectives are one-place predicates that are applied to names
of aggregates of objects and even mass terms are one-place predicates.
From the above it follows that the analyses for the following examples
are the expressions of L given after each of them:
(1) Socrates is a man - F(a).13
(2) This vase is a gift from Fritz to Hans - F(a, b, c).
(3) It is a surprise that Fritz is coming - G(n(F(a))).
(4) Skiing is a type of sport - M ( 11(F)).
(5) This bar is gold - F{a).
(6) The content of this bottle is water - F(U(a)).
(2) Verbs: Verbs are predicates, too, primarily predicates of one or
more places belonging to category <r(v,..., v), such as ‘run’ (<r(v)), ‘hit’
(cr(v, v), ‘lie between - and’ (cr(v, v, v)), but also predicates of categories
<x(v, i(g)) and <r(v, i(<r(v))) such as ‘believe’ and ‘can’.
Reflexive verbs such as ‘enjoy oneself’, ‘bestir oneself’, ‘perjure one­
self’, etc., which have come to be firmly tied to the relative pronoun,
count as one-place predicates. Impersonal verbs such as ‘rain’, ‘thunder’,
194 CHAPTER III

etc., which occur only in company with the pronoun ‘it’, can not be con­
sidered one-place predicates, but only as sentence constants. The pronoun
‘it’ does not play the role of a proper name here, for it makes no sense to
ask ‘What is raining?’ or ‘What is thundering?’131415
(7) Fritz runs - F(a)
(8) Klais lies between Garmisch and Mittenwald - F(a, b, c)
(9) Fritz believes that it is raining - F(a, pip)).
(10) Fritz can play tennis - M(a, p(F)).
(11) Fritz is hunting a rabbit - \/x(F(x) a G(a, x)) or M (a, ju(F)).
The first reading says that there is a particular rabbit Fritz is hunting. The
second reading says, on the other hand, that Fritz is on a rabbit hunt, i.e.
he does not want to bag a particular rabbit, but just some rabbit or other.
In the latter sense ‘hunt’ is a predicate of category a(v, z(c(v))).
(12) Fritz wants to catch a fish and eat it - I(a, p( \/x(F(x) a
A G (a,x)A H (a9x W *
Here ‘want’ is a predicate of category n(v, z(n)). The relation between
‘it’ and ‘a fish’ is established by the variable x . If one were to regard
‘want’ as a predicate of category a(v, z(cr(v))) and the sentence as the
conjunction of ‘Fritz wants to catch a fish’ and ‘he wants to eat it’, this
relation would remain open if Fritz does not want to catch a particular fish.
(3) Adjectives: Adjectives are also predicates, for the most part one-
place predicates of category cr(v), such as ‘big’, ‘heavy’, ‘red’, or of cate­
gory a(v, v), such as ‘larger than’, ‘friendly to’, but there are also adjectives
which are predicates of category <t( z(<t)), such as ‘pleasant’ (applied to
facts), or cr(z(cr(v))), such as ‘fast’, ‘intentionally’.
(13) Hans is bigger than Fritz - Fia, b)
(14) It is pleasant that the sun is shining - G(p(F(a))).
(15) Hans is running fast - Vf ( M (/) a f(a) a N ( f )).
The adjective ‘fast’ does not refer to the fact, that Hans is running, in this
case - and so we can not write G(p(F(a))); it is not the fact that is fast but
the activity of running. But we can not write M (p(F)) either, since part
of the sentence’s content is that Fritz is running. The sentence is thus to
be analyzed along the lines of ‘Fritz is engaged in an activity of running
(M) and this activity is fast (N )\
 THEORIES OF GRAMMAR 195

(4) Articles: Articles have very different functions depending on the

context in which they stand: Definite articles are parts of names (‘the
Zugspitze’); [in inflected languages], they express the number and case of
the substantive (this is becoming especially important in German, with
its declined endings in the process of being reduced or eliminated)16;
they serve as index expressions or so as to express identification [by way
of a definite description] (‘the father of Fritz’, meaning ‘that particular
person who is father of Fritz’) and in such cases are expressions belonging
to category v(cr(v)) or, in the case of descriptive identifications of one-place
concepts, to category n(v, of ct(v))). Or they may serve to express generaliza­
tion (‘The lion is a mammal’ as meaning ‘All lions are mammals’) and
they are then expressions belonging to category ofofv)).
The principal function of the indefinite article is to express particulariza­
tion (‘A locksmith came’) - they are expressions belonging to category
<7(cr(v)) in such cases - but they can also serve to express generalization
(‘A lion is a mammal’ as meaning ‘All lions are mammals’) - then they
have the same category - and together with the copula they serve to
express the predicative use of substantives (‘Fritz is a gardener’) - in such
cases they have no semantic function of their own, but serve as connectors.
(16) The dog Fritz rescued is injured - F(ix(H(x) a G(x 9a))).
(17) The (a) lion is a mammal, or: (All) lions are mammals -
A x(F(x ) =>(?(*)).
(18) A locksmith is coming - \/x(F(x) a G(x )).
(19) Fritz is a gardener - F(a).
(5) Pronouns: Personal pronouns stand for proper names that were al­
ready introduced in the context concerned or which are determined by
the context of utterance (‘Fritz is a student and he is studying law’,
‘I am tired’) - used in this way they serve as proper names and can
be replaced by proper names - or they serve, like demonstratives, as
index expressions, or they have the same function as variables in L, i.e.
they occupy the places for arguments in an expression which as a function
of these arguments is the argument of another expression.17 The same
holds true for reflexive pronouns, to the extent that they are not, as is the
case with reflexive verbs, fixed components of verbs. The personal pro­
noun ‘it’ can also, as was already mentioned above, occur as a fixed com­
ponent of an impersonal verb.
196 CHAPTER III

Reciprocal pronouns (‘one another’, ‘each other’, ‘mutual’) that occur

in sentences such as ‘Fritz and Inge love one another’, ‘Hans and Fritz
are standing next to each other’, ‘Max and Hans have given utterance to
mutual denunciations’, serve as abbreviations of sentences with the same
predicates - in unabbreviated form, the sentences read ‘Fritz loves Inge
and Inge loves Fritz’, ‘Hans is standing next to Fritz and Fritz is standing
next to Hans’, ‘Max has denounced Hans and Hans has denounced Max’.
Demonstrative pronouns (‘this’, ‘that’, ‘that particular’, ‘such’, ‘so’, ‘him
(her, it, etc.) self’ [in their reflective use]) serve first and foremost as index
expressions. But demonstrative pronouns also function like personal
pronouns as substitutes for proper names, as for example in the sentence
‘Bei dem Kampf zwischen Fritz und Hans wurde jenem das Nasenbein
gebrochen, diesem der Unterkiefer’. [Literally: ‘In the fight between Fritz
and Hans, that one suffered a broken nose and this one a broken jaw’.
In English we would have to use ‘the former’ and ‘the latter’ in the place
o f ‘jener* (‘that one’) and ‘diesef (‘this one’), since the demonstrative pro­
nouns themselves do not accomplish the discrimination of referent fixed
by their German usage. That is accomplished in English by referring ex­
plicitly to the order of previous reference. In the sentence immediately
preceding this one, it should be noted, the word ‘that’ does have the func­
tion discussed in the text. We could, with only some subtle change of
emphasis, use the personal pronoun ‘it’ instead. The same point could
have been illustrated with ‘this’ in the place of ‘that’. In each of the three
cases, ‘this’, ‘that’ or ‘it’ would be standing in for ‘the discrimination of
referent ...’ An example (two, in fact) closer to that (here = ‘the one’) in
the original text might be: ‘Hans gave Fritz a black eye but that (1) was
not the worst of it. He also shamed him in front of Inge. That (2) was the
real blow that ended their friendship’.] The (non-reflexive) pronouns
formed with ‘-self’, ‘himself’, ‘herself’, ‘itself’, serve principally to place
emphasis on the fact that one particular thing and no other is meant, as
in ‘Hans said so himself’, ‘Do it yourself’. When they have these func­
tions, they contribute to the sentence’s expressive aspects (‘my information
is authentic’, ‘I will not do it for you’).
Relative pronouns (‘who’, ‘which’, ‘that’) function like variables in def­
inite description expressions (‘the man who met me’) or like personal
pronouns (‘Anna, who entered the room, saw Fritz’ - in this case the at­
tributive sentence is to be understood conjunctively, i.e. as meaning ‘Anna
 THEORIES OF GRAMMAR 197

entered the room and Anna saw Fritz’). The expressions ‘who’ and ‘what’
have a generalizing character in, for example, the sentences ‘Who dares,
wins’ (meaning ‘Everyone who dares wins’) and ‘What is red is not green’
(meaning ‘Everything which is red is not green’).
Interrogative pronouns (‘who’, ‘what’, ‘which’) function like relative
pronouns in declarative sentences. For example, the interrogative pro­
noun in the sentence ‘Hans believed what Fritz told him’ can be replaced
by a relative pronoun: ‘Hans believed that which Fritz told him’. In
grammar expressions such as ‘all’, ‘each’, ‘none’, ‘nothing’, ‘one’, ‘some’,
‘something’, ‘many’, ‘few’, ‘a few’, ‘several’ are counted as indefinite
pronouns. Like the logical quantifiers, all of these words belong to cate­
gory o^OO). Further, the expressions ‘another’ [‘someone else’] and
‘both’ are numbered among the indefinite pronouns. The expression ‘an­
other’ or ‘someone else’ can be rendered logically by means of a par­
ticularization (so, for example, the sentence ‘Fritz praised Hans and some­
one else censured him’ can be rendered by ‘Fritz praised Hans and there
is a person who is distinct from Fritz and who censured Hans’). And so
that expression belongs to category (7((j(v)). The expression ‘both’ stands
for two proper names, as in ‘Fritz and Hans are studying medicine and both
are in their fifth semester’ (meaning ‘Fritz is studying medicine and Hans is
studying medicine and Fritz is in the fifth semester and Hans is in the
fifth semester’).
The possessive pronouns (‘my’, ‘your’, ‘his’, ‘her’, ‘its’, ‘our’, ‘their’)
viewed logically stand for function expressions (e.g. of category v(v)) or
identifications (definite descriptions) with or without supplementary con­
stants expressing a property, origin, or some other sort of relationship
Qhis father’ can be rendered by ‘the person who is father in relation to
him’, 6his hat’ can be replaced by ‘the hat that belongs to him’). The only
personal pronouns these function expressions contain are pronouns for
which, according to the preceding discussion, proper names, or variables,
are to be substituted.

(20) Fritz is a student and he is an engineer - F(a) a G(a).

(21) It is only a coward who takes flight - / \ x (F( x ) zd G(x )).
(22) Max praises himself - F(a, a)
(23) Max and Manfred hate each other - F(a, b) a F(b, a)
(24) Who dares, wins - A*(F(x)=3G(x))
198 CHAPTER III

(25) The man who married Eva is a cousin of Max - F(ixG(x, a), b)
(26) Hans believes what Max told him - F(a, irG(b, r)) - here r is
a variable belonging to category i(a).
(27) Nothing is perfect - i \JxF(x).
(28) Some logicians are philosophers - \Jx(F(x) a G(x ))
(29) Fritz is an engineer and his father is a doctor - F(a) a G(v(a)).

(6) Numeralia: The principal occurrence of words for cardinal numbers

(‘one’, ‘two’,...) is in combination with substantives, as in the sentence
‘There are two books on the table’. In such a context cardinal number
words belong to category cr(cr(v)).18 In mathematical statements such as
‘2 + 2 = 4 ’, on the other hand, cardinal number expressions serve as names
of numbers, i.e. they belong to category v. Words for ordinal numbers
(‘the first’, ‘the second’, ...) also occur primarily in connection with sub­
stantives and have in such cases the category v(p(y))9 as in the expression
‘The first man who set foot on the moon’. Recurrence number words
(‘once’, ‘twice’,...) indicate the number of events of the same kind, as in
the sentence ‘Max failed the examination twice’, and so they belong to
category cr(<7(v)) in the example. Number words of indefinite recurrence
(‘often’, ‘sometimes’, ‘never’, ‘several times’) belong to the same category.

(30) Two books have been damaged - M(Xx(F(x) a G{x ))).

(31) The third contestant is a Berliner - F( T(XxG(x))).
(32) Sometimes it rains - M{XxF(x)).

(7) Adverbs:
Adverbs o f place and time: Indications as to time refer to events, they
express temporal relations between events (such as ‘simultaneously’,
‘while’, ‘when’, ‘earlier than’, ‘before’, ‘later than’, ‘after’, etc.) - Adverbs
of time, which stand for such temporal relationships, are of category
a{a9a)9 then - or they indicate either points or periods of time in which
events are occurring or have occurred (‘now’, ‘today’, ‘yesterday’), or
relations between such points or periods of time (‘before’, ‘after’, etc.) -
in such cases they can be represented either as sentential operators of
category <?(&) or as names for temporal points or periods (category v),
which function as the arguments of predicates with which the events in
question are expressed or as the argument of a relation between points of
time and events (category <r(v, a)).
 THEORIES OF GRAMMAR 199

In representing indications of time as names for points in time adverbs

of time can also function as quantifiers, such as ‘always’, ‘never’, ‘some­
times’, ‘frequently’, ‘often’, ‘since’, ‘before’, etc. and they then have cate­
gory c(a(v)). They can also be represented as sentential operators (cate­
gory < t(g)), however. Which form of representation is adopted depends
on general applicability and systematic simplicity and on how far one is
willing to diverge from the language. Representation by means of tem­
poral parameters, as is usual in the physical sciences, for example, seems
to be at a considerable remove from natural language, but it is systemati­
cally simple and it can be applied universally. Representation by means of
sentential operators, on the other hand, is closer to language but generally
more complicated.19
The like holds for adverbs of place. At the same time it should be noted
of them that indications of place do not refer to events only but can refer
to objects as well.
Adverbs o f modality: They have very diverse functions and serve as
sentential operators (category o(o)\ such as ‘perhaps’, ‘fortunately’,
‘possibly’, ‘not’. They can also serve as expressions belonging to category
<7(<r(v)), such as ‘at least’, ‘at most’, or as sentential operators of category
<t(<t, <t), such as ‘even’, ‘also’, ‘rather’, and as independent components of
comparative statements, such as ‘so’, ‘very’, ‘all too’, ‘especially’, or
finally as predicates. For example, ‘gladly’ is a predicate belonging to
category cr(v, o(v)), for in the sentence ‘Fritz gladly goes to the theatre’,
this word does not characterize the activity of going to the theatre,
but an attitude of Fritz’s toward the activity of going to the theatre,
so that this sentence could also be rendered as ‘Fritz likes to go to the
theatre’.
Adverbs o f reason: These are sentential operators that belong to cate­
gory 0 (0 , cr), such as ‘therefore’, ‘consequently’, ‘accordingly’, ‘otherwise’,
‘nevertheless’, ‘besides’, or to category (t(g(v)9v), as the instrumental
‘with’.20
(33) Hans came yesterday - F(a, b)
(34) Max visits Fritz before he visits H ans-G (ix(F{a, b )\ fi(F(a, c))).
(35) The car was stolen while Fritz was asleep - N(XxF(a, x),
XxG(b, x)).
The sentence says that the time interval in which the car was stolen falls
200 CHAPTER III

within the time interval in which Fritz was asleep. So ‘while’ is a predicate
of category <x(<t(v), <r(v)) in this case.

(36) ,
Perhaps Fritz is sick - F(a ju(G(b))).

Although ‘perhaps’ is an expression with which new sentences can be con­

structed out of sentences, this word is not used descriptively (it is not the
fact that Fritz is sick that is characterised by ‘perhaps’), but serves an
expressive purpose and indicates the sentence’s performative mode. And
so we offer a performative version of the sentence (‘I regard it as possible
that Fritz is sick’).

(37) At least four books are missing - M(Ax(F(x) a G(x ))).

(38) Fritz gladly goes to the theatre - M (a, p{f)).
(39) Fritz is not coming because he is sick - H(p(-nF(aj)9p(G(ajj).

8. Prepositions: They are either dependent components of predicates,

such as ‘on’ in ‘Regensburg is situated on the Danube’, or like adverbs
they express space and time relationships, as do ‘within’, ‘around’, ‘out­
side’, ‘behind’, etc., or sentential operators belonging to category a(a, cr),
such as ‘despite’, ‘without prejudice to’, ‘except’. The instrumental ‘with’,
as we shall demonstrate at length below, is a functor of category cr(v, cr(v)).

(40) Regensburg is situated on the Danube - F(a, b).

(41) The building is closed outside of business hours - A*(“ iF(x) =>
=>G(a, x)).

9. Conjunctions: Conjunctions are predominantly sentential operators

belonging to category g( i{g))9 such as ‘necessarily’, ‘probably’, ‘hardly’,
‘presumably’, or operators of category g(g, o) such as ‘and’, ‘besides’,
‘likewise’, ‘not only - but also’, ‘or’, ‘either - or’, ‘otherwise’, ‘but’, ‘if -
then’, ‘if and only if’, ‘insofar as’, ‘so’, ‘consequently’, ‘therefore’, ‘be­
cause’, ‘all the more since’, ‘if only’, ‘for’, ‘in order that’, ‘so that’. They
are also expressions for temporal relations, such as ‘as long as’, ‘while’,
‘after’, or dependent components of comparatives like ‘just as - so’, ‘as’,
‘all the more’, ‘the - the’, ‘the same as’, or connectives like ‘that’ and
‘whether’.
In this connection we need to check whether the conjunctions are (also)
used after the manner of performative operators (such as ‘presumably’)
 THEORIES OF GRAMMAR 201

or not (e.g., ‘necessarily’); in the former case what should be logically

analyzed is an explicitly performative version of the sentence.
(42) Eva is young and beautiful - F(a) a G(a)
(43) Fritz is going either to Rome or to Naples - I(F(a, b)9F(a, c)).
(44) Max asks whether he may go - F(a, n(M(a9ju(G)))).
(45) Max comes whenever Fritz calls him - A x(F(a9b9x) 3 G(b9x)).
We see, then, if we make a review of the various types of words or parts
of speech in relation to the logical categories of the expressions that fall
under them, that word classes and logical categories are not at all parallel.
These brief reflections underscore still more emphatically the absence of
any homogeneity of grammatical function among the expressions tradi­
tionally grouped together under one and the same word type.
Let us now consider inflection from the point of view of logical analysis:
As a rule, the gender of a substantive does not have any independent
logical function. The grammatical gender of words does not coincide with
the natural gender of the objects they designate, or which belong to the
set of concepts they express. Grammatical gender probably arose originally
by transference of the distinction ‘masculine - feminine’ onto the world of
things, but it also conforms sometimes to pure syntactical principles. For
example, in German all words that end in ‘-/mg’ are masculine, all words
that end i n 6-keif or ‘-keif are feminine. Grammatical gender can even be
contradictory to natural gender, as exemplified by the expression ‘das
Weib\ And so the principal grammatical function of gender consists in
making clear which expressions in the sentence belong together, the func­
tion, then, of an (often unnecessary) connective.21
As connectives we here designate those linguistic means which indicate
connections between expressions in a sentence, or determine the scope of
functors, distinguish their arguments etc. - which, in other words, in­
dicate the structure of a sentence. In L these are brackets and commas,
besides the order of the expressions.
On occasion, however, gender does still have a semantic function, when
it alone characterizes human beings or animals as masculine or feminine.22
In these cases it can be rendered by means of special constants or what it
specifies semantically can be assimilated into the meaning of the sub­
stantive.
In German the forms taken by number are singular and plural. Besides
202 CHAPTER III

them Indogermanic formerly had the dual as a special form for expressing
two in number. In accordance with their meanings names are always in
the singular. In the normal forms of sentences, however, common nouns
are also always in the singular. We must therefore inquire whether sen­
tences that contain a plural can always be translated synonymously into
sentences that contain only singular forms.
The linguistic plural expresses the fact that the predicate relates to a
number of things. It occurs, for example, in the following sentences:
(a) Fritz and Hans are football players.
(b) Some Austrians are Tyrolians.
(c) The French were engaged in battle with the Prussians at Jena.
(d) Fritz and Hans are brothers.
(e) Erna and Karl love each other.
In sentences (a) and (b) the predicate is distributive, i.e. it concerns each
and every individual thing mentioned in the sentence. With that in mind,
we can translate these sentences into the following expressions, in which
only the singular occurs:
(a') Fritz is a football player and Hans is a football player.
(b') For (at least) one man it is the case that: He is an Austrian
and he is a Tyrolian.
In sentences (c) through (e), on the other hand, the predicate is not distri­
butive, i.e. these sentences can not be reformulated into the expressions:
(c') For every individual x and for every individual y it is true th at:
If jc is a Frenchman and y is a Prussian, then x was engaged in
battle with y at Jena.
(d') Fritz is a brother and Hans is a brother.
(e') Erna loves herself and Fritz loves himself.
What is meant by (c) is rather that the French army was engaged with
the Prussian army at Jena, i.e. we must replace the plural ‘the French’,
or ‘the Prussians’ with the singular ‘the French army’, or ‘the Prussian
army’ and by so doing we obtain as the correct translation into the singu­
lar the sentence:
(c0 The French army was engaged in battle with the Prussian
army at Jena.23
 THEORIES OF GRAMMAR 203

In sentences (d) and (e), however, there are two-place predicates, which
can be linguistically analyzed by using conjunction.24 These expressions
also involve abbreviations. Therefore we can render the sentences as fol­
lows:
(d") Fritz is brother of Hans (and Hans is brother of Fritz).
(e") Erna loves Karl and Karl loves Erna.
These translations show that the plural has no independent logical
function.
The principal function of case is just to establish the relationship be­
tween the individual words in the sentence. Congruence, which has to do
with gender and number, determines which adjectives belong to which
substantives and which verbs to which substantives. In particular, case
establishes whether a substantive is functioning as subject or object of the
sentence and distinguishes its various objects. In the analyzing terms of
sentences this function is taken over by the position of the terms in the
sentence. For example, while in German one can say not only ‘Erna liebt
Hans' but also ‘Den Hans liebt die Erna\ so that it is necessary for case to
establish who loves whom; after specifying that in the two-place predicate
‘loves (x, y)’ V represents the subject, ‘/ the object of the love, the only
form that remains a possibility is ‘love (Erna, Hans)’, but not ‘love (Hans,
Erna)’. For that reason case is also superfluous here. In the expression,
‘father’s mother’, case determines which part is the argument and which
part the function constant, while this distinction is effected in the normal
form ‘mother (father)’ by word position and parentheses.
Viewed logically, then, case has the function of specifying the connec­
tion of the expressions in the sentence, i.e. case belongs among the con­
nectives.25
Along with that, however, case can also have a semantic function. For
example, ‘So-and-so’s picture’ can mean the same as ‘the picture that
belongs to so-and-so’, ‘the picture that so-and-so painted’, or ‘the picture
that represents so-and-so’. Here the genitive has three different semantic
functions, which are rendered by relational constants in the analyzing
expression of L.26
With adjectives the formation of comparative forms is also counted as a
matter of inflection: From certain adjectives, e.g. ‘tali’, ‘light’, ‘pretty’,
‘old’, ‘bright’, but not from other adjectives such as ‘red’, ‘written’, ‘dead’,
204 CHAPTER III

‘silent’, comparative forms can be constituted, such as ‘taller’, ‘lighter’,

‘prettiest’, ‘very useful’, and so on. We shall take the adjective ‘tall’ as an
example.
Viewed logically, what we are concerned with in the case of the positive
(the basic form) is a one-place classificatory predicate belonging to cate­
gory <x(v). In grammar, however, the form of comparison ‘just as tall as’,
which is a two-place predicate in category <r(v, v), is also counted as falling
within the positive. In the case of the comparative form ‘taller than’ we
also have to do with a two-place predicate. ‘Less (more) tall than’ is also
a comparative form. The superlative ‘the tallest’ or ‘tallest’, as well as the
elative ‘very tali’, are once again one-place predicates of category cr(v). In
contexts such as ‘Hans is the tallest student’ or ‘Fritz is a very tall boy’,
on the other hand, superlative and elative have category cx(<x(v)).
There is no logical formation of comparative forms as complex predi­
cates with the positive as argument corresponding to the formation of
comparative forms to the positive in language, i.e., what we are concerned
with here is not the formation of new expressions from others by com­
bining them logically.27 From a logical point of view, the formation of
comparative forms belongs, then, to the lexical part of the theory of word
formation.
In the case of verbs, besides person (the speaker, the one spoken to, the
one spoken of) which can be read off from congruence with the subject,
and number, which was discussed above, there are still the following to be
considered as inflected forms: tense, mood and voice (active and passive).
Tenses (in German: present, preterite, perfect, pluperfect, (1) and (2)
future) serve primarily to indicate time in terms of past, present and
future, and can be rendered by time parameters or sentential operators.
In this connection it is to be noted that these time indications are index
expressions, for on each occasion they relate to the point in time at which
the statement is made as the present.28 If this point in time is indicated by
the sentence itself, then the time indications as given by tense have a rela­
tional character. The present can serve as the basic form for the logical
analysis, since it is also used for the representation of facts that are time­
less or do not depend on time.29 Also polyadic temporal relations are ex­
pressed by tense, particularly by the pluperfect and (2) future, as for
example in the sentences ‘Fritz had just entered the room, when the tele­
phone rang’, ‘When you arrive in Rome, Max will already have left the city’.
 THEORIES OF GRAMMAR 205

In addition, the distinction of preterite - perfect also marks a distinc­

tion in the speaker’s relationship to the event described in terms of aspect
or participation. A sentence in the perfect contains a reference to the
speaker and his present situation that a sentence in the preterite does not
contain. Generally speaking tense can also have a modal character, the
future in particular often expressing expectations and aims.30 Such ex­
pectations and intentions are to be rendered by performative predicates.
Finally, tense forms can also be used to express the beginning (Aorist),
duration, frequency and repetition of events. In such cases they are to be
rendered, as in example (35), by quantifying over points in time, for instance.
The basic mood that utterances have is the indicative. The subjunctive,
which has become quite rare in present-day German,31 primarily serves
the purpose of Kundgabe [expressing the speaker’s attitude] and can be
represented in that function by performative predicates. It serves to ex­
press obligation (‘Be that issue left unresolved’ - ‘That issue ought to be
left unresolved’), exhortation (‘May we attain to peace and calm’ - ‘We
desire to attain peace and calm’), concession (‘Be it so, then!’ - ‘It should
be so!’). Besides these functions, however, the subjunctive also serves to
express irrealia, as for example in the sentence ‘If Hans had put on the
brakes in time, the accident would not have happened’ - or to express the
content of a position that is maintained, as in the sentence ‘Fritz claims
not to have been responsible for the accident’. [The subjunctive is not
used for this purpose in English; it is used, however, to express the content
of a question, e,g. ‘Fritz asked if it were the case that Hans had not put on
the brakes’, or of a demand, as in ‘Fritz demanded that Hans be charged
with criminal negligence’].
The imperative, finally, serves to express command, and so in logical
analysis it too is rendered by a performative operator.32
The infinite (i.e., indeterminate as to person and number) verb forms,
the infinitive and participle, finally, have the following function: The in­
finitive is used when a verb assumes the role of argument (‘Max loves to
ski’). [In English, the participle is used for the same purpose (‘Max loves
skiing’).] The (1) and (2) participle are used especially in attribution as
forms of an adjectival use of the verb (‘the blossoming tree’, - ‘the fallen
tree’).33 The gerundive (‘the action to be done’ - ‘the action that should be
done’) can be expressed by combining the passive with the sentential
operator, ‘should’.
206 CHAPTER III

The formation of the passive from transitive verbs (‘Hans strikes Fritz’
- ‘Fritz is struck by Hans’) is usually interpreted as the formation of the
converse relation.34 That interpretation, however, does not do justice to
the fact that there is an impersonal passive for intransitive (i.e., one-
place) verbs (‘lacherC - ‘Es wird gelachf [Literally, ‘to laugh’ - ‘It is
laughed’; there is no such use of the passive voice of intransitive verbs in
English]), while there is no converse for monadic predicates. Of course,
one can say: ‘Es wird gelachV means the same as ‘Jemand lachf [‘Some­
one is laughing’], so that the impersonal passive takes on the role of
particularization. There is still a question, however, as to whether this is
an adequate interpretation or whether impersonal passive sentences are
not to be interpreted as independent sentential constants (in analogy with
‘Es regneV [‘It is raining’]) and passive forms of transitive verbs as inde­
pendent monadic predicate constants. On that view, the expression ‘by
Hans’ in the sentence ‘Fritz is struck by Hans’ would be an adverbial
qualification of ‘Fritz is struck’. Depending on how one interprets the
passive, then, one arrives at quite different analyses of the structures of
sentences in the passive mood.35
Now how does the traditional theory o f sentences look from the stand­
point of logical grammar?
The syntax of the logical language L is differentiated from traditional
syntax in virtue of the fact that it replaces the traditional sentence schema
subject - predicate, which can be expanded by supplementations, with
the substitution of arguments of categories xu ..., t m(rn> 1) in a predicate
of category Tm) as a basic logical schema. This logical predicate
replaces the traditional grammatical predicate. It is differentiated from
the latter by the fact that it can contain expressions the traditional gram­
mar counts as supplementation (e.g., prepositions, as in ‘to lie between...
a n d ...’), and that it does not always contain the constants that traditional­
ly count as predicate (in (17) the traditional predicate is ‘are mammals’;
the logical predicate is ‘all’). Furthermore, the expression that serves
as analysans can contain constants that do not occur in the original
sentence themselves and these new constants can even play the logical
role of predicate (cp. example (17)).
The traditional sentence subject is usually viewed logically as one of the
arguments of the logical predicate, but in logical analysis it loses its dis­
tinctive role in relation to the other arguments, for example the sentence
 THEORIES OF GRAMMAR 207

objects. But in many cases the subject of the sentence does not even have
the role of an argument of the predicate, as for example in the normal
forms of ‘All lions are mammals’ and ‘Hans ran fast’.
Objects o f sentences for the most part function as arguments of the logi­
cal predicate (‘Eva loves Kuno9, ‘Max gives Fritz the book9, ‘Hans is help­
ing Fritz9, ‘Hans is caring for the wounded9), but they can also be com­
ponents of the argument, as for example in (34).
As viewed by logic, attributes too can have quite a variety of functions:
They are conjunctive sentence parts (as in ‘The pale winter sun stands low
in the sky’ - ‘the winter sun is pale and stands low in the sky’, or in ‘Hans,
who was coming in the door, saw Fritz’ - ‘Hans was coming in the door
and saw Fritz’), or identifying descriptions (as in ‘The red car belongs to
Maria’ - ‘The car that is red belongs to Maria’, or in ‘The rejoicing over
the victory did not last long’ - ‘That rejoicing which was a rejoicing over
the victory did not last long’) or predicates of predicates (as in ‘The rose is
pale rose’ - ‘The rose color this rose has is pale’, or in ‘Kunigunde is
frightfully ugly’ i.e. ‘The quality of ugliness Kunigunde possesses is
frightful’), or predicates of sentences, or parts of predicates of sentences
(as in ‘Max travels by train very often9- ‘It is very often the case that Max
travels by train’, or in ‘Hans lived in Munich for five years9- ‘During a
period that lasted five years it was the case that Hans lived in Munich’)
or as argument of the logical predicate (as in ‘To suffer injustice is better
than to commit injustice\ or in ‘To the right of his house there stands a
church’).
Finally, adverbial modifications, like adverbs, when viewed logically
are predicates of sentences or predicates of predicates, but they are also
predicates of sentences with arguments or the like as well.
Taken all in all, then, no uniform logical function can be assigned to
the traditional categories of sentence components.
Since the treatment of adverbial modifications in translating sentences
in ordinary language into logical formulae is an excellent example of how
complicated logic can be, it would be well to offer three more typical
examples.
Let us consider first of all sentence (15) ‘Hans runs fast’. In that sen­
tence the word ‘fast’, as we have already insisted, does not characterize
the proposition that Hans is running, and so it is not a sentential operator
of category o(o); instead it specifies the nature and manner of the activ­
2 08 CHAPTER III

ity of running, which is described as fast running. Therefore sentence

(15) can also be formulated in the following way: ‘Hans is engaged in an
activity of running and that activity is fast’ or ‘There is an activity / of
which it is true th a t:/is an activity of running and Hans is engaged in/
and / is fast’ - as a formula of logic written: V /(£ * (/) a f( a ) A S ( f ) 9
where V stands for ‘Hans’, ‘S ’ for ‘fast’ and ‘L*’ for ‘activity of run­
ning’.36 In this case L *(/) is a predicate belonging to category a(a(y)).
The predicate ‘running’ - L(x) in symbols - on the other hand, belongs
to category cr(v). These predicates are connected as follows: A *(£(*) =
= V /(£ * (/) a f ( x ))). Thus L(x) can be defined by way of L *(f) but not
conversely,37 and to that extent no expression that contains the pred­
icate L(x) can be assigned to (15) as analysans. It also becomes clear
from this that adverbial modification is not, as it seems to be at first
glance, a supplementation added on to the sentence ‘Hans runs’, but that
‘running’ is a different predicate here.
The situation is similar with adverbial modifications that indicate in­
struments used or ends pursued, as the following two examples demon­
strate:
(46) Hans strikes Fritz with a stick - There is a thing x, of which it
is true that: x is a stick and there is an activity / , of which it is
true th a t:/is an activity of striking and/ applies to Hans and
Fritz and / is accomplished by means of x - \/x(F(x) a
a Vf(M(f) Af(a9b) a N( f 9x))).
(47) Max works to earn money - There is an activity /, of which it
is true that: / is an activity of working and/ applies to Max
and / is an activity of which earning money is the purpose -
Vf ( M{ f ) Af ( a) AN( f 9G)).
The examples given for the logical forms of natural language sentences
cannot be more than indications of how to analyze such sentences in L ;
how an analyzing relation R(X9 Y) between the sentences of a natural
language S and sentences of L would look.
These examples, however, suffice to illustrate the following points:
(1) The logical analysans of a natural language sentence A is not uni­
quely determined. This became apparent, for instance, in the analysis of
temporal adverbs. For the same sentence A there may then be several
sentences Bl9...9Bn of L with R(Bl9 A)9...9R(Bn9 A) even if A is not am­
 THEORIES OF GRAMMAR 2 09

biguous. Therefore it would be expedient to standardize the logical analyses

so that we may speak of the analyzing term of unambiguous sentences.
(2) The logical structure of an analyzing term of L for a sentence A of
S is often very different from the grammatical structure of A. If we call
the former the deep structure of A, the latter its surface structure, then
deep structure can be very far from the way a sentence is constructed in
language. A consequence of this is that there are no simple and systematic
connections between logical categories and the traditional grammatical
categories that aim at surface structures.
(3) Deep structures are theoretical constructs, in our case constructs
in the framework of a logical theory of grammar. Instead of L we could
also have taken another logical language L' which would then determine
different deep structures. If we try to diminish the distance between sur­
face and deep structure, then also other, less orthodox logical languages
gain in interest.38 And it may very well be the case that different logical
languages are suitable for the analysis of different natural languages.39
(4) The deep structure of a sentence is often much more complicated
than its surface structure. That is illustrated for instance by our examples
(46) and (47).40
(5) The plasticity of natural languages cannot be modelled by the rigid
logical languages. The same expression of a natural language may have
quite different syntactical and semantical functions in different contexts.
They must be assigned different logical categories in these contexts. The
same expression can also function as a (descriptive) constant in one sen­
tence and as a connective in another. The constants of L, moreover, not
only correspond to words of S, but may represent parts of words or
groups of words in S. We already mentioned, for instance, that ‘to lie’
occurs as a 1-place predicate in ‘The cat lies on a mat’, as a 2-place pred­
icate in ‘Regensburg lies on the Danube’, and as a 3-place predicate in
‘Klais lies between Garmisch and Mittenwald’. In fact there is no semantic
connection between ‘lie’ as a 1-place and as a 2-place predicate, at most a
partial homonymy: while ‘lie’ is the opposite o f‘stand’ or ‘sit’, ‘lie on’ ex­
presses a situation, and may be applied also to objects, like cities, of
which we cannot say that they lie, stand or sit. Finally the analysans may
have to contain constants or functors which do not correspond to expres­
sions in the analysandum. In the analysans of the sentence ‘Lions are
mammals’, for instance, we have to put in an expression for ‘all’.
210 CHAPTER III

Natural languages are just not constructed ‘with a logical ruler’41, their
regularities have not arisen from systematic considerations - least of all
by those according to which L is constructed - but have developed in all
their peculiarities and accidental features in a long historical process.
The examples point out the astonishing semantic and syntactic poly­
valence of the words and forms of natural languages - astonishing pri­
marily, because the number of constants of a logical language like L with
equal expressive power is reduced considerably, without making the
statements less intelligible as a rule.
(6) At present the value of logical grammar, in the absence of a general
definition of analyzing relations R for natural languages, consists less in
the fact that it is possible to present a syntax and semantics for such lan­
guages within the framework it provides than in specific analyses of sen­
tences in natural language. For example, if what we are concerned with
is to describe the ambiguities of sentences in a systematic fashion and not
just list the various possible interpretations by means of paraphrases, the
concepts of traditional grammar are often insufficient. Let us consider an
example of a sentence that is structurally ambiguous: ‘Hans had a book
stolen’.42 In this case the ambiguity of ‘had’ is at work (‘caused to’ and
‘happened to’), so that it is possible to interpret this sentence along the
lines of ‘Hans was the victim of a theft of a book’ (which in its turn could
be read either as ‘of one of his books’ or as ‘of a book that was in his pos­
session but that did not belong to him’) or ‘Hans caused someone to steal
a book for him’ (which is again ambiguous in a way roughly parallel to
the ambiguity of the former interpretation. I.e., Hans might have caused
someone to steal one of his own books for him - presumably for the sake
of recovering from an insurance company - or, the more natural inter­
pretation, he caused someone to steal a book he did not own, but wanted,
for him.) On the former interpretation of ‘had’, Hans is an innocent
victim; on the latter, he is an unscrupulous instigator of either a theft or
a fraud. Since the ‘had’ of ‘Had such-and-such happen to’ is of category
a(v, i(o’)), while the ‘had’ of ‘caused someone to’ belongs to category
<r(v, v, i(ff(v))), this semantic ambiguity is also the source of a structural
ambiguity. [The foregoing example has been modified to fit the author’s
presentation of it as closely as possible while using English rather than
German idioms. Hence it is not precisely the example chosen by the author
for his German text, since that can not be rendered literally in English,
 THEORIES OF GRAMMAR 211

nor is it Chomsky’s original example in English (although it, too, turned

on the ambiguity of ‘had’), since Kutschera’s treatment is significantly
different in some respects.]
And so the sentences in L that provide the analyses are:
V x(F(pc) a M (tf, ii(VyG(y, x, a)))),
and
V x(F(x) a \JyM (a, y, ju(AzG(z, x, a)))).43
(7) After many grammatical details have disappeared in logical anal­
ysis (such as accent and position, case, and the distinction between sub­
ject, adjective, verb), the final question arises, whether logical semantics
is capable of reproducing all of the discriminations of meaning in natural
languages. For reasons discussed below we cannot maintain that all
semantically relevant differences many be represented in L. What we
will do, however, is to illustrate by an example that the difficulties are not
as great as they are sometimes represented to be: The two sentences (a)
‘Hans patted his friend on the shoulder’ and (b) ‘Hans patted his friend’s
shoulder’ both have the same normal form (at least at the first look).
Weisgerber urges that in the first sentence there is exemplified a sentence
type or sentence structure [in German] that he calls an ‘advertive action
sentence’ [‘zugewandten Betatigungssatz’].44 He regards the function
specifically accomplished by such sentences as lying in the fact that the
event to which they refer is not identified as a goal-directed activity but
as an activity of the subject adverted to the person cited as dative object.
The role of the dative as being that of a case that expresses ‘the inward
concern of the person referred to by the verbal concept’ and as ‘more sub­
jective, warmer, more inward than the genitive, which simply states ob­
jectively a relation of possession’,45 is therefore essential to this sentence
structure, according to Weisgerber.
But apart from the fact that Weisgerber’s statements are very imprecise
and, as he himself confesses, remain incomplete and that these sentence
structures also occur in cases in which there can be no talk of an ‘adver­
tive activity’ (as in ‘The stone hit his friend on the head’, ‘Hans slapped
his friend in the face’ alongside of ‘Hans gave his friend a slap in the face’)
[These examples depend on the dative case being used in the first and the
last, and so the point can not be made as clearly in English.], the difference
in meaning between the two sentences can be represented as follows:
212 CHAPTER III

‘patting someone on the shoulder’ is a conventional form of expression,

which is only conditionally tied up with the predicate ‘patting’, since the
prevalent component of meaning lies in the expression of appreciation or
approval, not the expression of the activity of patting. But this constant
does not occur in the second sentence (b), which consequently does not
also express appreciation or approval, but only the activity of patting and
so has a different meaning from (a). Similarly with Weisgerber’s other
examples, such as ‘Hans looks Fritz in the face’, etc.

The difficulties of a semantical analysis of natural language sentences

with an interpreted logical language L did not appear in our examples
because we aimed at a purely syntactical analysis with the help of an
uninterpreted L. The following remarks may be made on this point:
(8) We have analyzed only assertions. Although we have indicated in
II.4.5 how to interpret other types of sentences in the framework of de­
scriptive semantics with the help of performatory operators we still
should have to develop the procedures for the analysis of imperatives,
questions, etc. in L.
(9) In the semantics of L we have treated meanings as intensions. For
the reasons discussed in II. 1.5 such a meaning-concept is too weak. We
have indicated how to narrow intensions by admitting logically incon­
sistent worlds, but at present there is no really satisfactory and technically
well defined procedure to handle meanings in intensional semantics.
(10) Natural languages contain many expressions which are gram­
matically wellformed but meaningless; which are assembled from mean­
ingful words or morphemes according to the syntactical rules but derive
no meaning under the semantical determinations. Let us take five typical
examples of such wellformed but meaningless expressions:
(a) Incompletely defined functors: Many predicates are not defined for
all syntactically permissible arguments. Thus the verb ‘to run’ is defined
for animals with locomotive appendages, for humans, machines, fluids
and for noses, not however for plants, minerals or numbers. And the
German verb ‘lachen’ is defined only for humans and the sun. The sen­
tence ‘Der Mond lacht’, though constructed grammatically just as ‘Die
Sonne lacht’, has unlike this sentence no meaning.
(b) Non-existent objects: Sentences about objects which do not exist or
no longer exist form a significant sub-category of example (a). The sen­
 THEORIES OF GRAMMAR 213

tences ‘Odysseus is (now) shaving himself’ and ‘Eisenhower is (now) sick’

are meaningless but not the sentences ‘Professor Snell is dreaming of
Odysseus’ or ‘Nixon remembers Eisenhower’. Thus many predicates are
defined for non-existent objects while others are not. Since the question
of whether a human being is alive or dead is purely empirical, syntax
cannot refer to this distinction.
(c) Invalid presuppositions: A presupposition of a statement or utter­
ance A is a state of affairs which is not itself asserted in A , but which must
be the case if both A and the (colloquial) negation of A are to be meaning­
ful. Thus the sentence ‘John gave up smoking’ presupposes that John
previously smoked. ‘Jack knows that there is a university in Regensburg’
presupposes that Regensburg does indeed have a university. The utter­
ance ‘As a doctor I realise how dangerous this symptom is’ presupposes
that the speaker is a physician. These presuppositions are not part of the
content of the sentences but rather preconditions to them being meaning­
ful at all. Such presuppositions, being again matters of empirical fact,
cannot be accounted for by syntax.
Invalid presuppositions also appear in the following special cases:
(d) Definite descriptions with unfulfilled normality conditions: Descrip­
tion terms as ‘Russell’s book’ or ‘George Vi’s son’ have no meaning
because the describing predicate fails to apply to exactly one object as
the normality condition of descriptions requires. Whether this condition
holds or not is again an empirical question, not a syntactic one.
(e) Empty generalisations: In ordinary discourse the sentence ‘All of
John’s children have red hair’ is meaningless if John does not have any
children. In general a sentence of the form ‘All A ’s are B ’ is only meaning­
ful if there are A ’s. Such a sentence thus presupposes the sentence 6A ’s
exist’. This should not be understood in every case to mean that there
must exist ‘real objects’ which are A ’s - sentences like ‘All Greek Gods
were assimilated into the Roman Pantheon’ indicate to the contrary that
they can also be possible objects. These presuppositions of descriptions
and generalisations were first noticed by P. F. Strawson.
In the semantics of L we have assigned all well-formed expressions A
of L an intension which for every world i determines the extension of A
in /. To analyze expressions in natural languages which for some worlds
have no extensions, or for functors which are not defined for all syntacti­
cally permissible arguments we have to define partial interpretations of L.46
214 CHAPTER III

(11) Natural languages contain many vague and ambiguous expressions

or regularities. This is true in the syntactical domain - there are degrees
of grammaticalness, of well-formedness of sentences, the grammatical
categories apply more or less to an expression - and much more so in the
semantical domain. A logical analysis, especially a semantical analysis of
natural language sentences with the exact concepts of logic and precise
logical interpretation therefore often implies an explication more than an
interpretation, and hence a modification of these sentences.
Of course the first aim in developing a logical grammar is to have a
functioning syntactic and semantic system at all. And for that the price of
a certain modification may be very well worth paying. But in a second
step one will have to try to replace some classificatory concepts by com­
parative ones; for instance comparative concepts of well-formedness or
of synonymity. With these concepts the vagueness of natural languages
can be mirrored exactly, since it is based on the comparative or typological
character of grammatical categories and rules. Because it is possible to
describe in a logically exact way even what is vague and not exact, there
is no justification to think that a logical analysis is inadequate in prin­
ciple, then. At present, however, as far as I know, there has been no at­
tempt to develop such comparative concepts in the framework of logical
grammar.
We can sum up these remarks then by saying: Logical grammar is a
step in the right direction. The logical analysis of natural language is
fruitful even now, but there is still a lot to be done before we can claim
that an adequate logical analysis of natural languages is possible.

NOTES

1 See e.g. Montague [70] and [70a].

2 For F (a i)...(a n) we shall sometimes write F (a i, ..., an).
8 (AB)C can be represented by A BxC, but not A(B°): (AB)C is the set of functions h o n C
with h (x )= g for x e C andg e A B.I f y e B , theng (y ) eA , and so h (x) (y) eA , and for that
we can also write h(x, y) eA . A(B°), on the other hand, is the set o f functions h, which
assign values drawn from A to functions g of C into B: h (g )e A .
4 See Montague [70], p. 387. The definitions o f negation and conjunction were given by
A. Tarski in [23].
5 This form o f expression is very problematical, since ‘possible’ is not an adjective, but
a sentential adverb. The procedure sketched in what follows simply amounts to assign­
ing a referent to the term that has no referent, but making a distinction between existing
objects existing in one world and others.
6 See Chapter II. 1.5.
 THEORIES OF GRAMMAR 215

7 See D . Lewis [68] on counterpart theory.

8 Since A*iMi(A)^A*i(A*iMi(A)), the exp ression s//^ )an d n(n(A)) do not have the
same extension. The fact that there may be no relevant distinction in natural language
corresponding to this formal difference in meaning is no objection to the interpretation
o f//(A ).
9 See Kutschera [73a], Chapter 1.
10 Montague in [70] interprets variables, too, along the lines o f condition (a) and in (d)
requires instead of M 't(b)= M 'j(b) for all j e l , that M 'j(b)= M j(x) should hold for all
j # i. But then such simple logical principles as AxA [x] =AyA [y] are no longer valid.
11 For partial interpretations cf. Kutschera [74], for anormal worlds Montague [70],
for intensional logic in general Cresswell [73].
12 Logical categories of this sort were first developed by K. Ajdukiewicz and S.
Lesniewski. They are used for grammatical purposes also in Bar-Hillel [53] and [60b],
in Curry [61] and in Lambeck [61].
13 The predicative character of class words is linguistically clear in sentences in which
such a word is a part of the grammatical predicate, as in (1), but not in sentences in
which the class term is the subject of the sentence, as in ‘Man is mortal’. Superficially
considered, ‘Man’ has the same place here as “Socrates” in (1). For that reason, the
substantive has sometimes been denied any predicative character generally and (1) has
been interpreted as an identification of Socrates with a man, so that ‘is’ (in the sense of
identity) would be the predicate all by itself (see III. 1.2, Note 26). But this conception,
as we have seen, is untenable. It is necessary to take the opposite course and interpret
class terms serving as subjects o f sentences predicatively also, as becomes clear in the
following formulation o f the second sentence: ‘For every thing it is true that if it is a
man, it is mortal’.
14 See Grebe [66], p. 457ff., 472.
15 Montague discusses this example in [70a].
16 See Grebe [66], p. 176. On p. 152ff. there is also an explanation of how the article
arises through the blurring o f declined endings. For example, if someone says *Ich ziehe
Wein Wasser vor [I prefer wine to water/water to wine]’ that is ambiguous, since the
arguments taken by ‘vorziehen [prefer]’ are distinguished by case as dative and accusa­
tive object, but the accusative and dative of “wine” and “water” read the same. Con­
sequently, articles are used here to distinguish between the sentences *Ich ziehe dem
Wein (das) Wasser vor’ and ‘Ich ziehe (den) Wein dem Wasser vor9. [In English, one way
o f resolving the notoriously ambiguous oracle: ‘The Medes the Persians will subdue’,
would be to insert the nominative form of the pronoun in apposition before whichever
noun was intended to be agent of the subduing and so the subject of the sentence.]
17 Cf. sentence (21) below or the sentence ‘Hans saw a lion and Fritz killed him’.
18 See the representation of enumerative statements by means of quantifiers and iden­
tity, e.g. in Kutschera [67], 3.1. The statement that there are (precisely) two things -
symbolically V =2xF (x) - is represented as follows: V x y (F (x )/\F (y )A x ¥ zy V
A xyz(F (x) a F (y) a F (z) =>x —y v x = z v y = z ) . In our context it is o f no importance
how the operator V =2 is defined in terms of other logical operators, only that what is
involved is an expression that generates sentences out of predicates belonging to the
category a (v) that can be substituted for F(x), one which belongs to category a (a (v)), then.
19 It becomes clear here, for example, that the logical analysis of expressions is not
unambiguous. - On the logical systematization o f statements o f time and place see also
Rescher [68] and the literature cited there.
20 See example (46).
2 16 CHAPTER III

21 Johann Werner Meiner also makes a point of that in his Versuch einer an der
menschlichen Sprache abgebildeten Vernunftlehre oder Philosophic und allgemeine
Sprachlehre, Leipzig 1781, in the Preface, p. XLVIIIf.: “On the other hand, when a
language does not assign any fixed and definite position to the modifying word, as in
the Latin and Greek languages, where it is equally correct to place the modifying word
before or after the substantive; it then becomes doubtful which of the two substantives
that a modifying word stands between it belongs to. It is then necessary to make such
[relationship] clear by means of visible and audible signs. That is the true reason for
gender in language, and not at all the two genders of animals, as heretofore believed.” -
H. Brekle called my attention to this passage.
22 See Lyons [69], p. 287. Lyons offers examples such as le chat (tomcat) - la chatte
(cat); ragazzo (boy) - ragazza (girl).
23 It becomes clear in this case that the plural is not always used to express plural
number but also to express collectives. See also Lyons [66], p. 28If.
24 This is a method that is often used - particularly in primitive languages - for render­
ing relations in terms of conjunctions and representing the sentence ‘a loves b \ for
example by ‘a and b love’, or the sentence ‘a gives b a book’ by ‘a gives and b receives
and a book is given’.
25 Lyons says in [66], p. 218, along these same lines: ‘Case is not present in “deep
structure” at all, but is merely the inflectional “realisation” of particular syntactic
relationships’. - On the concept o f ‘deep structure’ see III.3.1.
26 According to Ch. Fillmore in [68], semantic or ‘deep cases’ play a corresponding role.
27 Logically, it would be better to regard the concepts ‘just as tall as’ or ‘taller than’ as
more basic than ‘tali’, since ‘tall’ can be defined with those concepts as starting points,
e.g. as ‘taller than a \ where a is a comparison object or standard, or by ‘taller than
most things’, or in other ways. The superlative ‘tallest’ can be represented with the aid
o f a definite description as ‘that object which is taller than all others’. And the elative
‘very tall’ could be defined e.g. in terms of a reference to standard object b as ‘taller than b \
28 See also Lyons [69], p. 305.
29 The use of the present along with time parameters also occurs in the historical
present (‘Socrates dies in 399 B.C.’) or the use of the present to express the future
(‘I come tomorrow’).
30 It is variously urged that tenses primarily express point of view and express objective
temporal relationships only secondarily. See Lyons [69], p. 311, 313ff., Weinrich [64],
Weisgerber [62], V. II, p. 327, Grebe [66], p. 103f., as well as IV.3.
31 See Grebe [66], p. 97.
32 Reichenbach in [47], 57 takes the indicative along with the subjunctive to be a means
o f expressing the speaker’s attitude toward the sentence. The indicative is supposed to
express an affirmation, just as the subjunctive expresses denial (in irrealia), wish, warn­
ing or abstention from judgment (‘Fritz said he was innocent’). But the assertive
character is common to all declarative sentences, even if the indicative does not occur
in them. Thus the sentence ‘If he were your friend, he would help you’ is an affirmation,
too, one that Reichenbach characterises as a non-affirmation. And conversely, the
indicative enters into sentences that have no assertive force as well, as for example in
the dependent clause, ‘Hans claims that he is innocent’. To that extent, to describe the
indicative as the basic form o f statement is surely more correct.
33 In compounds with auxiliary verbs the (2) participle is the component of a verbal
form all the same, as in the sentence ‘The tree has been felled’.
34 The converse relation to F (x, y), F -1(x, y) is defined by F _1(x, y ) \= F ( y , x).
 THEORIES OF GRAMMAR 217

35 See Weisgerber [63b] and Chapter IV. 3.

36 One could also try to transform sentence (15) into the sentence: ‘That activity/ , of
which it is true that / is an activity of running and Hans is doing it, is fast’ -
5 (i/(jL *(jc)A /(x)A /(fl))). But this description is meaningful only when Hans is
running, otherwise the meaning o f the new formulation is undefined while the existence
proposition will be simply false. Thus the representation of (15) as an existential
proposition is more adequate.
37 Of course it is true that £ * (/)= > A x ( f W D L(x)), but the converse is not true, else
every empty predicate would express e.g. an activity of running.
38 See D. Lewis [70] and Cresswell [73].
39 To be sure, logical grammar is often referred to as universal or rational grammar,
see e.g. the title of the Port Royal Grammar by Arnauld and Lancelot [60] and the
work by Montague cited. E. Husserl even speaks of an a priori grammar. What is
bound up with these expressions is the idea that there is only one grammar for all
languages, just as there is only one logic. Thus Roger Bacon, for example, says that the
grammar of all languages is substantially the same, even if it may be subject to acciden­
tal variations: “Grammatica una et eadem est secundum substantiam in omnibus
linguis, licet accidentaliter varietur.” ([57], p. 278). And K. Becker writes ([42], p. 5):
“To derive these basic relationships common to all languages out of the idea of lan­
guage as an organised instrument of human nature and to demonstrate them by com­
paring differing languages is the task of universal grammar.”
40 Wittgenstein says in the Tractatus: “Ordinary language is a part of the human orga­
nism and no less complicated than it is. It is humanly impossible to get the logic of
language out of it directly. Language disguises thought. And in just such a way that it is
impossible to draw any conclusions from the external form of the costume, the disguise,
about the form of the disguised thought; because the costume’s external form is made
for purposes quite distinct from that of making known the form of the body. The
implicit adjustments for understanding ordinary language are immensely complicated.”
([22], 4.002)
41 Thus Frege in an unpublished letter to Husserl dated Nov. 1, 1906.
42 Chomsky brings this example in in [65], p. 21.
43 A structural ambiguity exists only where there is a semantic ambiguity. The converse
does not hold, however. The sentence ‘Hans sings’, for example, has the same logical
structure whether the verb ‘to sing’ is understood in the sense of ‘to give forth words
with a melodic intonation’ or in the sense o f ‘to make statements to the police concern­
ing the participation of others in a crime in which one had had a part one-self’, for in
either case ‘to sing’ belongs to category cr(v). - There is another example of semantic
ambiguity in the sentence ‘Later on the letter by Hans was read’ [A literal rendering o f
the idiomatically correct example in German]. (This example is brought in by Bierwisch
in [66], p. 102.) In ‘the letter by Hans’, ‘by’ indicates the letter’s origin, and so stands
for the predicate ‘written by’, while in ‘by Hans was read’ the word identifies the sub­
ject performing the activity [of reading]. Chomsky in [65], p. 22 cites the example of the
two sentences in English, ‘I persuaded John to leave’ - ‘I expected John to leave’, which
on superficial inspection have the same structure, although ‘persuade’ is a predicate
belonging to category cr(v, v, and ‘expect’, on the other hand, belongs to
category cr(v, i(p)).
44 See Weisgerber [63a], p. 262ff. See also Grebe [66], p. 491.
45 Thus A. Weiss, cited in Weisgerber [63a], p. 291.
46 Cf. Kutschera [74].
218 CHAPTER III

3. G e n e r a t iv e gram m ar

3.1. Generative Syntax

On its syntactic side, which we will take up first, generative grammar
starts out from the problem of constructing the grammar of a language L
as a system of rules by which the set of grammatically correct sentences
of L can be engendered. Hence its name. It lays particular emphasis on
the requirement that the syntactical rules form a system of really exact
rules and in this respect it intends to go beyond the traditional grammar
which, as we have seen, does not provide any precise and complete rules,
but only illustrates regularities of sentence structure by example and
counterexample without precisely delimiting the range within which
these rules are valid.
The conceptual-technical presuppositions necessary for making genera­
tive systems of rules precise were created around 1930, through the devel­
opment of the theory of recursive functions and its formalization in terms
of Semi-Thue systems, Turing machines, etc. Generative grammar aims
at making use of this mathematical apparatus in carrying out its task.
The first requirement imposed on a generative syntax is that it should
provide a procedure for the recursive enumeration of the set G{S) of syn­
tactically well-formed (grammatically correct) sentences of a language S ,
i.e. a routinely applicable general procedure firmly fixed in every detail
by explicit rules, with which one after the other every expression in G{S)
can be effectively generated. Since all recursively enumerable sets of ex­
pressions can be generated by Semi-Thue-Systems, we can represent a
generative syntax as a Semi-Thue-System, for example.1
A Semi-Thue-System T is defined over a finite alphabet A = {Si9..., S n}
( n ^ 1), i.e. over a set of basic symbols Su ..., Sn. We define as an expres­
sion or word on A &finite sequence of signs drawn from A and possibly
the empty sign (symbolized by □ ), which when it is put into a particular
place indicates that there is no sign in that place. Where W is a word it is
then true that WQ = □ W — W. The definition of Tis now accomplished by
giving a finite set of pairs ( R h of expressions on A, T s defining relations.
Let Wi W2 be the expression that arises from the concatenation of Wt
and W2.2 We then say that in T a word W is directly transformable into
a word W f - symbolically W=> W ' - if there is a defining relation <R h R[y
of T and (possibly empty) words U and V such that W = URtV and W ’ =
 THEORIES OF GRAMMAR 219

= UR[ V both hold. And we say that in T a word W is transformable into

a word W ' - symbolically W -> W '~ if there is a finite series of words
W09 Wl9..., Wm such that W0= W and Wm= W ' and Wk=> Wk+1 all hold
for k = 0 9..., m —1. Such a sequence is also called a derivation of W ' from
W. If some of the words of A are laid down as axioms (one is sufficient),
then those words that can be derived from an axiom can be designated as
provable.
But it is not only required that the set G(S) of well-formed sentences
of S be recursively enumerable, but that it also be decidable, i.e., that
there be an algorithm, a routinely applicable procedure firmly fixed in
every detail, by which for any expression it can be determined in a finite
number of steps whether or not it belongs to G(S). The communicative
function of the sentences in a language surely requires at a minimum that
it be possible to know definitely whether or not they are grammatically
correct. If we start out from Semi-Thue-Systems, then, we can limit our­
selves to such systems as are decidable.
But now the set of Semi-Thue-Systems that produce decidable sets of
sentences is not recursively enumerable, so that it is impossible in par­
ticular to offer any fixed schematism for decidable Semi-Thue-Systems
of that sort.3
With a view to the remarks that follow, it is reasonable to delimit de­
cidable Semi-Thue-Systems by imposing the following requirements on
the systems considered:
(I) It should hold true of all defining relations <i£i} Rl) that R[ is at
least as long an expression as R t. Such systems are decidable, as can easily
be seen.4
Systems that satisfy requirement (I) can now be formulated in such a
way that:
(II) All defining relations have the form <UAV, U ZV }9 where A is a
single symbol and U and V9 but not Z can be empty.
For a relation { C 1 ...Cm9 D ^ ^ D ^ n ^ m ) the relations ( C t ...Cm9
D1C2...Cm>, <(D1C2...Cm, D 1D 2 C3 ...Cmy 9 . . . 9 ( D i ...Dm_ 1 Cm9 D 1 ...Dny
can then be substituted.
A further restriction on Semi-Thue-Systems going beyond (II), or (I),
contains the condition:
(III) All defining relations have the form (A , Z ) , where A is a single
symbol and Z is not empty.
220 CHAPTER III

Defining relations of this sort, in contrast to context-bound relations

<XA 7, X Z Y ) with X Y not empty, according to which Zcan be substituted
for A only in the context X AY, are identified as context-free A
A further restriction vis-a-vis (III) is:
(IV) All defining relations have the form <A , XB) or (A , Y >, where
the set of words on A is to be composed of two disjunct parts, the terminal
vocabulary VT and the non-terminal vocabulary Vn, and A, B are single
symbols from Vn, X and Y expressions on Vt , of which X but not Y can be
empty.6
Derivations that satisfy (II) can be written in tree form as well.7 For
example, if
ABCDE
AB'B"CDE
A 'B fB"CDE
A'B'B"CD'D"D'"E
A'B'B"CD'D+D ++D'"E
is a derivation in T, this derivation can also be written in the form

A B C D E
I / \ / l \
A' B' B" D'D"D'"

/ \
D* D++
But syntax is not just a question of providing rules for the recuisive
enumeration of sets G(S). The fact is that the sentences of G(S) are sup­
posed to be interpreted in semantics, and this semantic interpretation also
must take the form of a system of rules. Since we wish to interpret all of
the sentences of G(*S),8 we shall consequently find it useful, as has already
been urged in the introduction to this chapter, to frame the syntactic rules
so that the way they are set up syntactically will correspond to their
semantic structure.
Perhaps it is this principle that is behind the somewhat vague require­
ment of generative grammar that in the syntax sentences are to be assigned
a structure that is supposed to correspond to the ‘intuitive structure’, in
terms of which the sentence is understood by speaker and hearer.9 Syn­
tax, then, should not only make it certain that a sentence such as ‘Hans
 THEORIES OF GRAMMAR 221

had a book stolen’, can be generated, but that there should be as many
syntactic structures coordinated to it as there are possible interpretations
of that sentence.
In grammar this requirement that sentences be given a syntactic struc­
ture is now often interpreted so as to mean that every sentence must be
analyzed into connected parts which can then be classified as expressions
of a definite type, e.g. according to word or sentence categories. Those
expressions can then be analyzed further and the parts classified once
again. Lees describes this as follows: ‘... the linguist... has assumed that
the sentences of a language may each be analyzed into a linearly con­
catenated sequence of immediate constituents, and that this bracketing
or parsing operation may be performed at various levels of generality to
yield a hierarchical branching-diagram, such that any unit at any level is
just a certain continuous string within some sentence or else a class of
such strings drawn from different but grammatically equivalent sen­
tences.’10
Thus, for example, the sentence ‘The man hits the brown dog’ will be
analyzed as follows:

Sentence
^Noun^Phrase ^ V e r b Phrase^

A rtic le ^ ^ S u b s ta n tive Verb Noun Phrase

Article A djective Substantive

The man hits i

the i
brown J
dog.

Such a hierarchical analysis of a sentence in terms of its components

is also called a phrase-marker or structural description (‘SD’ for short).
Now the basic idea of generative grammar is the characterization of
the syntactic structure of a sentence in terms of its derivation within a
Semi-Thue-System of Type (II). In such a system the derivations become
SD’s if we proceed as follows: Category symbols such as S (for sentence),
N (for noun), V (for verb), At (for article), Aj (for adjective), Av (for
adverb), Pr (for preposition), Np (for noun phrase), Vp (for verb phrase),
etc., are introduced into the alphabet A of system T. These category sym­
bols constitute the non-terminal vocabulary VN of A and count as single
symbols as intended in (II). The words (or morphemes) of language S', the
222 CHAPTER III

sentences of which are to be generated, constitute the terminal vocabulary

of A , Vx. The symbol S counts as an axiom of T. Rules of structure such
as <S, NpVp), <Np, AtN>, <Np, AtAjN), <Vp, VNp>, for example, and
rules of substitution such as <At, the), <N, m an), <N, dog), <V, hits),
<Aj, brown), for example, are given as defining relations. The expres­
sions of VT occur only as second constituents in the defining relations, so
that every derivation in T ends with a sequence of words of S.
In that case the SD given above for the sentence ‘The man hits the
brown dog’ can be constructed as a derivation in T.
One naturally arrives, then, by way of the desired SD, to systems of
syntactic rules presented as Semi-Thue-Systems of type (II). For this rea­
son, within the domain of generative grammar such systems are also
called general SD-grammars, contrasted to systems of type (III), which are
called simple SD -grammars.
The detailed elaboration of simple SD-syntax leads to formidable com­
plications and difficulties, however. Thus, for example, gender, number
and case are to be provided for every substantive, person, number, mood,
tense and voice for every verb. For example, it is not generally possible to
substitute ‘man’ for N, but depending on the context either ‘man’s’ or
‘men’. For V it is not possible always to substitute ‘jump’, but ‘jumps’,
‘jumped’, ‘would jump’, and so on, too. Consequently, further specifica­
tions must be added to the category symbols and the rules of substitution
will read, for example, not <V, jum p), but <V(3. person, plural, indicative,
present, active), jumps). On top of that we have the fact that congruence
must be taken into consideration in connection with the category rules.
That is, we have to replace the rule <S, NY), for example, with <S, N
(nominative, plural), V(3. person, plural, indicative, present, active)),
and so on.
Further complications arise when one takes into account the fact that
not all predicates are defined for all substitutions. For example, ‘hard’ is
defined only for concrete and tangible things [Konkreta], ‘smile’ only for
human beings and the sun, ‘run’ only for the higher animals, machines,
water faucets and noses, ‘whinny’ only occurs in connection with horses,
‘meow’ in connection with cats, and so on. That is, we would now have to
add specifications of a semantical character for every substantive and
verb, that say whether it is defined for concrete or abstract items, animate
or inanimate, animal, human, etc., and we would have to pay attention
 THEORIES OF GRAMMAR 223

to the agreement of these specifications in making substitutions.11 But

that is definitely doomed to failure by the fact that the number of such
semantic categories can not be delimited in any rational manner. But as
a consequence, the program of constructing the syntax in such a way that
the set of well-formed sentences in S coincides with the set of semantically
well-interpreted sentences in S is also compromised.
There are other arguments besides, quite fundamental ones, that count
against the feasibility of a pure SD-syntax, and they have led Harris and
Chomsky to a modification of that type of grammar. The two most im­
portant considerations are these:12
(1) The basic idea of SD-syntax is that all of the grammatically con­
nected expressions in a sentence form a closed contextual unit. But that is
often not the case, as the example ‘The girls called Hans up’ shows. Here
‘call up’ is the verb, which does not constitute a connected unit, however,
but is divided in two by the object of the sentence, ‘Hans’. Now the SD-
syntax does not contain any possibility for classifying disconnected parts
of a sentence as one sentence component. Nor does it contain any mech­
anism that would permit us to deal with transformations of the sentence,
i.e. first of all to form:

The

in order then to transform the latter sentence so that it will be gramma­

tically correct. Therefore the only way in which it can analyze the sen­
tence is:
S

/ \ / i \
At N V N Pp

The
I .1
girls
I I I
called Hans up

But that is inadequate, since ‘up’ is not an independent preposition here

but a constituent of the verb ‘call up’. Furthermore, the future is expressed
224 CHAPTER III

as ‘The girls will Fritz call-up’ [‘Die Madeln werden Fritz anrufen’. The
example, so far as the future tense is concerned, depends on the fact that
the German verb ‘anrufen’ has a separable prefix. The point can not be
made in English in the same way.], so that we would obtain different
structural descriptions of the sentence for present and future.13
(2) SD-syntax can not adequately render the phenomenon of reflexive
reference, i.e. the structures of sentences like (a) ‘Hans is sleeping and so
is Fritz’ and (b) ‘Fritz is sleeping and he is snoring’ are not correctly rep­
resented, since there is no way of expressing the reference of ‘so’ back to
‘is sleeping’ and of ‘he’ back to ‘Fritz’. The SD’s of these two sentences do
not make clear the connection between the sentence ‘Hans is sleeping
and Fritz is sleeping’ and (a) or ‘Fritz is sleeping and Fritz is snoring’
and (b).
To these two objections we could add still others, although in most
cases they come to the same thing, that SD-syntax can not give an ade­
quate expression to that structure of the sentence we are concerned with
in semantic interpretation, because it cuts sentences into parts that are
attached to each other in surface structure instead of analyzing them in
terms of the expressions’ semantic function.14
On account of these inadequacies of the SD-syntax, Harris and Choms­
ky have developed a syntax with transformation-rules, T-syntax for short.
In it transformation rules are permitted along with the SD rules for con­
struction. These rules generate a sentence out of one or more SD (such as
do not end with a well-formed sentence of S).15 The syntax of these trans­
formationally generated sentences is supposed to be represented by the
transformational structure, i.e. by the derivation of the sentences from the
SD using rules of transformation.16
One can now distinguish two stages in the development of this T-syn-
tax: in the first stage, represented by Chomsky’s works [57] and [61a], for
example, the rules of transformation are given a significantly larger do­
main of application than in the second stage, which is exemplified by [65].
Above all, in the first stage even sentence sequences are generated by trans­
formation rules and the embedding of subordinate clauses in the primary
sentence is accomplished by means of transformation rules. That requires
transformation rules with more than one argument.17
At the second stage only transformation rules with one argument are
used. Different tasks are now assigned to the SD part and the transforma­
 THEORIES OF GRAMMAR 225

tion part of syntax: For every sentence, besides its syntactic surface struc­
ture,, which is grasped by traditional grammar, for example, there is a
depth structure, which determines the sentence’s semantic interpretation.18
Now it is the task of the SD part to generate the depth-structure of sen­
tences, which need not be grammatically well-formed sentences, and the
task of the transformation rules is to generate wellformed sentences out
of the terminal expressions of the SD.19 For that purpose one-place trans­
formation rules, which generate one expression out of another, are suf­
ficient.20 The transformational structure is no longer of any importance
now, for the task of the SD was above all else to present the semantic
structure of a sentence (so that with the aid of the SD, for example, one
can differentiate between the possible meanings of ambiguous sentences).
The transformation rules, however, leave the interpretation of the ex­
pressions invariant and to that extent one can identify the depth-structure
of a sentence with the SD of those expressions from which the sentence
can be obtained by means of transformation rules.
By means of a transformation rule of the form <X - finite present or
preterite form of ‘to call up’ - object - Y, X - finite present or preterite
form of ‘to call’ - object - ‘up’ - Y} we can then produce from the last
lines of the SD
S

At N V N
i i i i
The teacher called up Hans

the sentence ‘The teacher called Hans up’.

Besides the approaches to a generative grammar sketched here, SD-
Syntax and its supplementary transformation rules, there are still further
systems, but they have not attained the same importance in the discussion
[of this matter].
Many of these systems have been shown to be inadequate and for others
proofs of equivalence have been given. Thus in Bar-Hillel [60b] it is
demonstrated that SD-grammars (or (Ill)-systems) are equivalent to the
categorical grammars of Bar-Hillel21 (i.e. that the same sets of sentences
can be produced by them) and in particular to restricted categorical
226 CHAPTER III

grammars. Furthermore Chomsky proved in [59a] that the so-called

finite state systems22 are equivalent to the (IV)-systems and the latter
equivalent once again to deterministic and to non-deterministic finite
automata,23 We had already made reference to Harman’s system in
[63].24 We will not go into these systems here, but make a critical comment
on T-syntax instead.
It appears to us to be a fundamental defect of this syntax that the SD
of the sentences make use of the traditional grammatical categories,
which, as we saw in III.l and III.2 are neither adequate nor precisely
defined. Now since the depth structure of a sentence is supposed to deter­
mine its semantic interpretation, as far as we are concerned nothing seems
to count against identifying the depth structure of a sentence with the
logical structure of a sentence in L that analyzes it and so describing
depth structure in terms of logical categories.25
If that is done, then the SD-part of generative grammar simply amounts
to giving the rules for generating sentences in L. Since the sentences of L
are unambiguous with respect to their semantic structure, it is not neces­
sary to attach structural descriptions to them, in the form of derivations,
for example.
The transformational part of generative grammar is determined by the
analyzing relation R for S: A sentence B of S can be generated out of a
sentence A of L i t and only if R(A, B) is true. If it is possible to define R
inductively so that for every A the set of B with R(A, B) is recursively
enumerable, then the transformations are also defined thereby, with a
precision appropriate to the standards of generative grammar.
In specifying the analyzing relation, likewise the transformation rules,
all the grammatical distinctions such as gender, case, construction of in­
flected forms and the like naturally play a role that can be ignored in
logical analysis. Experience will have to tell the extent to which we have
to introduce new categories beside the logical ones in this connection.
Generally speaking, all of the distinctions relevant to sentence con­
struction in S that are included in the so-called substitution-categories
may be important, which we can define as follows:
Two expressions X and Y in S are said to be isogenic if and only if for
all expressions A[Z] it is true that: A[X] is a sentence in S if and only if
A[Y] is a sentence in S'.26
The equivalence classes pertaining to the relation of isogeny are then
 THEORIES OF GRAMMAR 227

substitution categories,27 which per definitionem cover exactly the distinc­

tions that are syntactically relevant in S.
As against generative grammar, besides the advantage of superior
precision, logical grammar also has the advantage of greater simplicity,
for the replacement of the derivations in Semi-Thue-Systems by sen­
tences in L makes the SD-part essentially simpler and would hardly make
the transformational part any more complicated. Since the end-products
of the SD have to render unambiguously the semantic structure of the
sentence, they will hardly be less complicated or closer to surface struc­
ture than the analyzing sentences of L ; i.e. the transformation rules will
look just about as complicated as the definition of the relation R . Besides
that, as we shall see immediately, a workable generative semantics can be
constructed on the basis of this syntax, in contrast with T-syntax.
It is possible, furthermore, to attain yet another decided simplification,
if we give up the requirement that the set of well-formed expressions that
can be constructed in the syntax coincide with the set of semantically
interpretable sentences in S . We have seen above that Chomsky’s recom­
mendation that semantic categories be included in the SD leads to great
difficulties. But it does no harm if we permit within the syntax of S sen­
tences such as ‘Prime numbers laugh’, or ‘The moon is hitting the sun’,
for example, which are well-formed in terms of grammatical categories
but have no sense. For the nonsensical character of these sentences is a
purely semantic phenomenon based solely on semantic considerations.
Those sentences can be singled out and set aside in the semantics, then.28

3.2. Generative Semantics

A semantics within the confines of generative grammar was first developed
by Katz and Fodor in [63] and by Katz in [64a] and by Katz and Postal
in [64b].29
This semantics starts out from a T-syntax of a language S . Let G be the
class of those syntactic categories (or category symbols) that are used in
this syntax for describing expressions in S , and let K be a finite class of
semantic categories (or category symbols) by means of which the mean­
ings of words in S are supposed to be characterized. The following can
serve as semantic categories of this sort, for example: activity, state, con-
creteI-abstract entities, living creature, inanimate thing, animal, human,
male, adult, etc. Now in the lexicon of S there should be associated with
228 CHAPTER III

every word one, or in case of ambiguity more than one, set of category
symbols {wl9...9wn} drawn from G and K. These constitute a syntactic
and semantic characterization of the word. In addition, there is a distinc-
tor, i.e. a statement concerning the meaning of the word that distinguishes
its meaning from all related meanings insofar as the semantic categories
do not do that already, and also a selector, which specifies the range with­
in which the word is defined. For example, the following readings are
ascribed to the English word ‘ball’:
(a) ball - {N, Social Activity, Large, Assembly} [For the purpose
of social dancing]
(b) ball - {N9Physical Object} [Having globular shape]
(c) ball - {N9 Physical Object} [Solid missile for projection by
engine of war].
These readings thus define the different possible interpretations of a
word. The distinctor stands within square brackets.
Now let X be a sentence with which an SD (depth structure) B is as­
sociated. X will then be assigned an interpretation by way of applying the
following rules.
(1) A rule of substitution E coordinates every word W with the set of
those readings of W in the lexicon that have syntactic characteristics com­
patible with the syntactic characteristics of W in B. If this set is empty,
then W and consequently X is meaningless.
(2) One now goes through the SD B from bottom to top and specifies
by means of a projection rule R the set of readings anew at each branching.
We can illustrate this best by means of an example that Katz and Fodor in­
troduce in [63]: The starting point is the following structural description:

At Aj N

The man hits the colorful ball,

In the lexicon the following readings will be attached to the words other
than ‘ball’ that occur in this sentence, with the selector in pointed brackets.
 THEORIES OF GRAMMAR 229

(d) The - {At} [Some contextually definite]

(e) colorful - {Aj, Color} [Abounding in contrast or variety o f
bright colors] (Physical object or social activity)
(f) colorful - {Aj, Evaluative} [Having distinctive character, vivid­
ness, or picturesqueness] ( Aesthetic object or Social activity)
fg) man - {N, Physical object, Human, Adult, Male]
(h) hits - {V(transitive), Action, Instancy, Intensity} [Collides
with an impact] (Subject: Higher animal or Improper part or
Physical object; Object: Physical object)
(i) hits - {V(transitive), Action, Instancy, Intensity} [S W te vwYA
a Wow or (Subject: Human or Higher animal; Object:
Physical object; Instrument: Physical object)

The next step is to construct the set of readings o f ‘colorful ball’. For that
there is a projection rule concerning certain connections between ex­
pressions, among them the attributive connection, by which we obtain
the following readings from the sets {e, f } and {a, b, c}:

(j) colorful ball - {Np, Social Activity, Large, Assembly, Color]

Abounding in contrast or variety o f bright colors] [For the
purpose o f social dancing}
(k) colorful ball - {Np, Physical object, Color] [.Abounding in con­
trast or variety o f bright colors] [Having globular shape]
(l) Colorful ball - {Np, Physical object, Color] [.Abounding in
contrast or variety o f bright colors] [Solid missile for projec­
tion by engine o f war}
(m) Colorful ball - {Np, Social activity, Large, Assembly, Evalua­
tive] [Having distinctive character, vividness or picturesqueness]
[For the purpose o f social dancing]

Of the six possible readings, then, the combinations (f)—(b) and (f)-(c) are
inapplicable, since the selector of (f) is incompatible with the category
symbol Physical Object. The attributive rule thus immediately picks out
the mutually compatible readings and eliminates the nonsensical attribu­
tive combinations.
Within the compatible readings we then go on to combine the semantic
categories and likewise the distinctors.
By means of a projection rule R 2 we can connect the article and sub-
230 CHAPTER III

stantive expressions. We thus obtain from readings {d} and {g} the reading
(n) The man - {Np} [Some contextually definite] {Physical object,
Human, Adult, Male}
and from {d} and {j, k, 1, m} the readings
(o) The colorful ball - {Np} [Some contextually definite] {Social
activity, Large, Assembly, Color} [Abounding in contrast or
variety o f bright colors] [For the purpose o f social dancing]
(p) The colorful ball - {Np} [Some contextually definite] {Phy­
sical object, Color} [Abounding in contrast or variety o f bright
colors] [Having globular shape}
(q) The colorful ball - {Np} [Some contextually definite] {Phy­
sical object, Color} [Abounding in contrast or variety o f bright
colors] [Solid missile for projection by engine o f war}
(r) The colorful ball - {Np} [Some contextually definite] {Social
activity, Large, Assembly, Evaluative} [Having distinctive
character, vividness or picturesqueness] [For the purpose o f
social dancing}
By means of a projection rule R 2 we can then connect a transitive verb
with its object. We obtain accordingly from readings {h, i} and {o, p, q, r}
the readings
(s) Hits the colorful ball - {Vp, Action, Instancy, Intensity} [Col­
lides with an impact] [Some contextually definite] {Physical
object, Color} [Abounding in contrast or variety o f bright
colors] [Having globular shape} {Subject: Higher animal or
Improper Part or Physical object)
(t) Hits the colorful ball - {Vp, Action, Instancy, Intensity} [Col­
lides with an impact] [Some contextually definite] {Physical
object, Color} [Abounding in contrast or variety o f bright colors]
[Solid missile for projection by engine o f war} {Subject: Higher
animal or Improper Part or Physical object)
(u) Hits the colorful ball - {Vp, Action, Instancy, Intensity}
[Strikes with a blow or missile] [Some contextually definite]
{Physical object, Color} [Abounding in contrast or variety o f
bright colors] [Having globular shape} {Subject: Human or
Higher Animal) Vp,
 THEORIES OF GRAMMAR 231

(v) Hits the colorful ball - {Vp, Action, Instancy, Intensity}

[Strikes with a blow or missile] [Some contextually definite]
{Physical object, Color} [Abounding in contrast or variety o f
bright colors] [Solid missile for projection by engine o f warj
(Subject: Human or Higher Animal)

Now the combinations (h)-(o), (h)-(r), (i)-(o), (i)-(r) are inapplicable

here on account of the selection specifications in (h) and (i). Moreover,
as is already the case with R 2, the semantic characteristics and the dis-
tinctors are not connected up with each other as with R±.
Finally, by means of a projection rule R 2 we can connect subject with
predicate. This gives us from readings {n} and {s, t, u, v} these readings:

(w) The man hits the colorful ball - {S} [Some contextually def­
inite] {Physical Object, Human, Adult, Male} {Action, In­
stancy, Intensity} [Collides with an impact] [Some contextual­
ly definite] {Physical object, Color} [Abounding in contrast or
variety o f bright colors] [Having globular shape]
(x) The man hits the colorful ball - {S} [Some contextually def­
inite] {Physical object, Human, Adult, Male} {Action, In­
stancy, Intensity} [Collides with an impact] [Some contextual­
ly definite] {Physical object, Color} [Abounding in contrast or
variety o f bright colors] [Solid missile for projection by engine
o f war]
(y) The man hits the colorful ball - {S} [Some contextually def­
inite] {Physical object, Human, Adult, Male} {Action, Instancy,
Intensity} [Strikes with a blow or missile] [Some contextually
definite] {Physical object, Color} [Abounding in contrast or
variety o f bright colors] [Having globular shapej
(z) The man hits the colorful ball - {S} [Some contextually def­
inite] {Physicalobject, Human, Adult, Male} {Action, Instancy,
Intensity} [Strikes as a blow or missile] [Some contextually
definite] {Physical object, Color} [Abounding in contrast or
variety o f bright colors] [Solid missile for projection by engine
o f war}.

Now the set {w, x, y, z} is the set of possible readings of the sentence
‘The man hits the colorful ball’. If this set were empty, the sentence would
232 CHAPTER III

be meaningless, if it included only one reading the sentence would be

unambiguous. But since four readings are now associated with the sen­
tence, it is ambiguous in four ways, i.e. there are four possible meanings
for the sentence. If the sets of readings for two sentences X 1 and X 2 con­
tain no reading in common, then these two sentences are distinct in
meaning; if they contain n readings in common, then the two sentences
are synonymous in n ways. If they contain precisely the same readings,
then the two sentences are fully synonymous.
Since the transformation rules on the second stage of the development
of T-syntax are so defined as to leave the meaning of the expressions in­
variant, the semantic rules always apply solely to the SD of a sentence,
which gives its depth structure.

Even this short sketch of Katz and Fodor’s semantic approach clearly
reveals its fundamental defects.
In the first place, how the set of semantic categories is to be determined
is left completely open. Now the authors’ intent, to be sure, is not so much
to provide a semantics for a specified language as to set forth the general
outlines of a semantic theory, but even in such outlines it would be neces­
sary to say how these categories are to be understood: Are they charac­
teristics of the concepts to be characterized or properties of those concepts?
The two accounts have entirely different logical status: A characteristic
(Merkmal) G of a concept F of category <r(v) is a concept of the same
category such that f\x(F(x) = G(x) a ...). A property (Eigenschaft) of F(x),
on the other hand, is a concept M ( f ) of category cr(cr(v)) such that M(F)
is true. Both characteristics and properties occur in Katz and Fodor’s
work. For example, the word ‘bachelor’ is supposed to be defined by way
of the category Human, and so in terms of a characteristic, but ‘light’ (as
an adjective) by way of the categories Color, or Weight, i.e. by way of
properties. Furthermore, if the categories are always taken to be one-
place concepts, in specifying them for transitive verbs, for example, they
will have to be properties for the most part. Finally, there can be no ques­
tion of anything but properties in the case of proper names and function
expressions.
Let us now consider the projection rules, first of all R ±. This rule has
the form WL- {K,} [Z)J ( S , ) , W2 - {K2} [D2] <S2>=* Wt W2 - {Ku K2)
< s 2>.
 THEORIES OF GRAMMAR 233

Now the question is, whatever is the expression in the conclusion sup­
posed to mean: If Wx and W2 are one-place predicates and { K j and {K2}
groups of characteristics, then {Ku K2} could represent the conjunction
of those characteristics. So in that case, W1W2 would be interpreted as
the conjunction of two one-place predicates. This interpretation would
surely enter into the question for many attributive combinations, such as
‘red ball’ in the sense of ‘x is red and x is a ball’, but this would by no
means be true for all cases of applying R u among which Katz and Fodor
in fact also count the construction adverb-verb, for example. But one can
not interpret ‘to sing badly’ as ‘to be bad and to sing’. Furthermore, even
in the case of constructions in the form adjective + substantive - insofar
as they can be interpreted conjunctively at all - this interpretation of
{Kx, K2} will not hold up generally, for we saw that not only charac­
teristics but also properties occur in {KJ and {K2}. One can not infer
M{Xx{F(x) a <j (X))) from M{F), however. For example, it is true that
‘red’ is a color predicate, but ‘red ball’ is no color predicate. Finally, in
III.2.2 we saw that by no means all attributive constructions are capable
of being interpreted as conjunctive connections.
We lack any reasonable interpretation of {Kl5 K2}, then. But we miss
still more an interpretation of [.DJ fD2]. Such an expression is never
defined. If it were supposed to be equivalent to [Dx a D2], that again
would require Dx and D2 to be characteristics, and again that works only
with conjunctive constructions. Otherwise the expression remains com­
pletely unintelligible.
It becomes still worse with the other rules. Here, in the conclusions
there occur expressions like WXW2 - { K j [D±] {K2} [Z>2], which remain
entirely undefined. A reasonable interpretation t)f such expressions is no­
where in sight, not even is it distinguished from expressions of the form
{Ku K2) [[Zh] [D2J. It is simply left up to the reader, who knows the inter­
pretation of Wx W2 anyway, once again to figure it out from the varie­
gated succession of semantic fragments that are offered. Naturally that
sort of thing no longer has anything to do with scientific precision. Com­
pare, for example, readings (y) and (z): You are able to learn semantic
information from them only because you already know what they are sup­
posed to mean, and because you do not permit yourself to be led astray in
what you know by even the most extraordinary way of writing.
The most elementary semantic insights are also absent here: things,
234 CHAPTER III

concepts of different types, function expressions and sentences are not

differentiated semantically, there is no account of how concepts apply to
objects, etc.30
Although the semantics of Katz and Fodor is only a prototypical form
of generative semantics that has been superseded in the meantime, it is
nevertheless well suited to illustrate the shortcomings of that semantics.
So far a really precise semantics, capable of accomplishing anything,
exists only within the confines of logical grammar. There, too, as we in­
dicated in III.2.1, such fundamental problems as the partial definition of
predicates in natural language, for example, admit of satisfactory solu­
tions.31 Therefore at the present time logical grammar constitutes the best
approach to the development of exact grammar-models for natural lan­
guages.

3.3. Innate Ideas

In connection with generative grammar, we will, in the form of a short
excursus, go into a hypothesis that really does not belong to the theory of
grammar, but to epistemology. However, it should be mentioned at this
point because it is based upon generative grammar.
This hypothesis o f innate ideas, abbreviated IH (Innateness Hypothesis),
has been advocated by Chomsky especially.32 It states that the occurrence
of language learning in humans is differentiated from other learning pro­
cesses, e.g. learning to play chess, by the fact that certain items of in­
formation about language that enter into the process of learning lan­
guage are built into the human mind (or human brain). On this view, we
have an innate linguistic capacity, but not an innate capacity for playing
chess. Everything we know about chess we have learned from experience,
but we know more about language than we have learned from experience-
indeed, according to Chomsky, more than we could ever learn by way of
experience.
More exactly, the information about language that is independent of
experience is supposed to be the content of general language theory.33
What belongs to general language theory are the general properties of
grammars: that (a) every grammar contains an SD component and (b)
transformation rules, that (c) every grammar contains certain syntactic
and semantic categories, ‘proper name’, ‘sentence’, ‘verb’, etc., for
example, that (d) certain general semantic rules belong to every grammar
 THEORIES OF GRAMMAR 235

and that (e) the phonological component of every grammar interprets

sentences over a certain finite basic stock of phonetic signs. Further, it
includes an evaluation of different grammars that generate the same set
of sentences, in terms of their degree of simplicity, for example.
Now what is explicitly formulated in general language theory, the
general properties of grammars, for example, is supposed to be built into
the way the human Language Acquisition Device works - LAD for short -
so that the LAD functions in such a way that we select one specific gram­
mar from all those possible, namely the simplest of those grammars that
fit the linguistic data presented in experience. I.e., the LAD provides us
with no information concerning the language we consciously use, but it
steers the process of language acquisition and our linguistic behavior with­
out our having been aware of the way in which it works nor being able to
know how it works by introspection. The talk of ‘innate ideas’ is to be
understood as nothing but a historical allusion then - actually Chomsky’s
hypothesis is essentially different from the presentations of rationalism
in the 17th and 18th centuries from Descartes to Leibniz.34 Chomsky does
not offer any more precise account of the manner in which the LAD func­
tions, but in his IH merely hints at a cybernetic model of this LAD.
According to Chomsky, what the IH accomplishes consists in the fact
that it explains:
(1) The extensive similarity among all human languages (on the basis
of the specifics common to all grammars, the so-called (formal) linguistic
universal), as they are stated in (a) through (e).35
(2) The species-specific character of languages (apes can not learn
them, even though they do possess a general learning ability and intel­
ligence).
(3) The independence of the mastery of language from intelligence
(children learn language at an age in which their general intellectual
capacities are still quite undeveloped) and - most extensively - from the
linguistic data presented, which can be quite different from one individual
to another.
(4) The ease with which language is learned (language is an extraordi­
narily complicated system of rules - think of Chomsky’s generative syn­
tax - which is learned by children in a relatively short time, however).
(5) That language is learned at all: an infinite set of sentences, as marked
out by a grammar, on the basis of merely finite linguistic data. According
236 CHAPTER III

to Chomsky, that cannot be explained in terms of a general capacity

for learning, for in learning a language not only the surface structure but
the depth structure, too, which is not apparent in the data at all, must be
analyzed.

There is the following to be said in criticizing this hypothesis, however:

(A) The content of the hypothesis is not rendered sufficiently precise so
long as we are not told exactly which items of information the LAD con­
tains, or what it does.
(B) As to what the LAD accomplishes, this much can be established:
In the first place, the IH cannot be directly tested. To do that one would
have to produce a language that does not conform to the LAD schema
and would have to show that human beings could not learn it or at least
not learn it as an initial language, i.e. one would have to give children
practical trials with this language exclusively and show that they could
not master it within a reasonable time, even though it was essentially no
more complicated than our own. But no such language has yet been pro­
duced and according to Chomsky it may be that no such language can be
produced by us at all.36
All that is left, then, is to judge the IH in terms of its explanatory value*1
How does it stand in that respect?
This is to be said in response to (1): the general validity of formal uni­
versal, for which the IH is supposed to provide a basis, can also be ex­
plained independently of the IH:
(a) It is not correct that every syntax contains an SD-component and
a T-component: the generation of sentences in logical grammar is not
something done with an SD-grammar. Nor do SD-grammars show up in
other systems of generative grammar. All we can say, then, is that for
every language there is an SD-grammar, supplemented by transformation
rules, that generates all the sentences of that language. But that is trivial,
because the transformation rules - in the general terms, as Chomsky
considers them - have the form of the defining relationships of general
Semi-Thue-Systems, so that every Semi-Thue-System can also be written
in the form of a (Il)-system with transformation rules. But since the sen­
tences of any language have to be recursively enumerable, they can be
generated by means of such systems.
(b) That e.g. the syntactic category ‘proper name’ occurs universally is
 THEORIES OF GRAMMAR 237

in the first place empirically questionable; but if it actually is the case,

that can be explained as a consequence of the power and usefulness of lan­
guages with proper names.38 But the category ‘proper name’ together with
the actually universal category ‘sentence’ already suffice to define all of
the logical categories.39
(c) The semantic rules, not specified in any greater detail by Chomsky,
emerge simply out of the demand for defining semantic interpretations par­
allel to the fundamental syntactic rules.
(d) The finite stock of basic phonetic signs is explained, according to
Putnam, [67], by the identity of human linguistic organs.
Finally, according to Putnam many features common to languages are
easily explained by taking all human languages to have descended from
the same source.
In response to (2), this is to be said: The species-specific character of the
ability to learn languages is also explained by a species-specific higher in­
telligence in humans.
And to (3): There is not a strict independence of language from intel­
ligence. Less intelligent human beings learn language less perfectly than
the intelligent, their vocabulary and their grammatical abilities lag behind
the others.40 The extensive independence from linguistic data arises from
the fact that those data have to constitute a very large and representative
set of linguistic examples; otherwise one can certainly not count on any
independence.
In response to (4) Putnam urges in [67]: The language of which the
normal adult is master is the one that every normal adult can learn. But
what every normal adult can learn can not be described as ‘complicated’.
Especially, when we consider that one needs 9 to 10 years in order to
master a language to some extent, the grammar would have to be essen­
tially still more complicated than it actually is, for learning it to be aston­
ishing.41 Besides, it is not permissible to identify the complexity of lin­
guistic habits with the complexity of explicitly formulated grammatical
rules that give those linguistic habits some precision and surely not at all
with the complexity of Chomsky’s generative grammar.
In response to (5), finally, we can say this - and this is the main objec­
tion to Chomsky’s argument: As long as there are no precise, fully worked
out theories in learning, we may not claim that general learning capacity
is incapable of explaining language learning 42 That the mastery of an in­
2 38 CHAPTER III

finite number of sentences can not be attained from a finite amount of

data is false, for we can learn just a finite number of rules. The chess
player, too, learns how to conduct himself correctly in (practically) in­
finitely many situations on the board with the few general rules of play.
Despite that one does not assume any innate knowledge of chess.
Taking all of that into account, the IH has no explanatory value either:
It explains nothing that can not be more simply explained in other ways.
But since the hypothesis can not be directly tested either, there is then no
reason of any kind for accepting it and building upon it farreaching views
about the theory of knowledge.

NOTES

1 On the concepts o f recursive function, recursive enumerability and Semi-Thue-

System see e.g. Hermes [61] and Davis [58].
2 In view of the associativity o f concatenation W i(W 2W3) = (W iW 2) Wz it is also
acceptable to put together a sequence of several expressions without parentheses.
3 It is generally true that the set o f decidable sets of words is not recursively denumer­
able. For along with a recursive enumeration of these sets you would also have an
(effective) enumeration M i, M 2,... for them. If so, where A i, ^ 2,... is an effective
enumeration of words, it would be possible to define a decidable set M as follows:
A ie M = -\A ie M i. But then, in contradiction with the assumption, M would not be one
o f the enumerated sets, since for M = M k we obtain Ak^Mk=~~\Ak^Mk.
4 If T is a system of this sort on A and X is an expression on A, then there is a finite
number of expressions Y o f the same length as X, from which X can be derived in T.
It can be determined which possibilities for Y there are. Y is investigated in the same
way, etc. In every case in which you do not encounter one of T s axioms anyway, an
expression Z which is shorter than X must turn up as a premise. In this way, the length
o f the expressions from which X can be derived is reduced step by step, so that
it is possible after a finite number of steps to reach a decision about the derivability
o f X.
5 Chomsky proves in [59a], Theorem 4, that (III) represents vis-a-vis (II) a genuine
limitation, i.e. that there are word sets that can be generated by way of systems of type
(II), but not by systems of type (III), - See also Postal [64b].
6 In [59a] (Theorem 7), Chomsky proves that there are word sets that can be generated
by (Ill)-systems, but not by (IV)-systems. - In particular, those word sets that are self-
imbedding, i.e. that contain rules o f the type A->XA Y, are of this kind (Theorem 11).
7 On the other hand, this is not true of other Semi-Thue-Systems, e.g. the systems given
by Harman in [63], in which the defining relations have the form {A , X ) and <AB,
X B Y >, where A and B are single signs and X, Y can also be empty.
8 We will later relax this requirement.
9 See e.g. Chomsky [65], p. 24. Chomsky calls a grammar of this sort ‘descriptively
adequate’.
10 Lees [57], p. 385.
11 This is Chomsky’s suggestion in [65], Chapter H, § 2.
 THEORIES OF GRAMMAR 239

12 See e.g. Postal [64a], p. 73ff. as well.

13 Harman’s systems in [63], cited above in note 7, have been tried for the purpose o f
making such transformations possible. As Chomsky has stated in [65], such systems no
longer provide any structural descriptions, however. For example, we obtain the
derivation:

Np Vp

At
/ \ N V N
I
The
I
teacher
I called up Hans.
I
v---------- J t '
______ A«*_____
Called Hans up.

The words in the expression ‘called Hans up’ are no longer classified as to category in
this case and have to be identified as substantive and verb components by supplemen­
tary specifications.
14 See also Postal [64b]. - Many other objections often raised against SD-syntax are of
no great weight. The objection, for example, that the connections between active and
passive, assertion and related question, and similar connections are not expressed.
It is not the task of syntax to bring out clearly every semantic connection, although
a syntax with rules corresponding to the rules of semantic interpretation does at least
make the basic semantic relations much clearer than SD-syntax does.
15 It is often required that a transformation rule generate not only a sentence, but an
appropriate SD. But if this SD is supposed to satisfy the general requirement that it
adequately represent the sentence’s grammatical structure, then all of the sentences
generated by transformation can also be generated in the SD-syntax without transforma­
tion rules. In that case the only point of transformation rules would be to call attention to
relationships and connections between SD, e.g. to generate questions from assertions in a
simple way, etc. But calling attention to such semantic connections is not, as already re­
marked above, the primary goal of syntax. The essential part of the justification for intro­
ducing transformation rules is that there are sentences which syntax has to generate, to
which no SD can be attached, e.g. in the case of discontinuous expressions. Therefore,
there will just not be adequate structural descriptions for every sentence that can be gen­
erated by transformation rules and hence it will not be possible to lay down a general
requirement that transformation rules generate SD. On the other hand, you then run into
difficulties in repeated applications of transformation rules, for the arguments of these
rules are supposed to be not sentences, but SD!
16 Now if the arguments of a transformation rule are always supposed to be structural
descriptions - in the usual sense as constituent structures and transformation structures
- then the problem arises of defining these transformation rules in such general terms
that they are defined for all of these arguments. That problem has never been solved, as
far as I know.
17 See the example Chomsky cites in [66a], p. 30ff., for one, as well as [62a].
18 The distinction between deep and surface structure can be illustrated by examples of
ambiguous sentences in which identical sentences have different deep structures (‘Hans
240 CHAPTER III

had a book stolen’ and ‘Later the letter by Hans was read’ [See above note 43 to
Chapter III.2]), or in sentences such as ‘Fritz laughs often’ and ‘Fritz laughs happily’,
where sentences with the same surface structure (proper name - verb - adverb) have
different depth structures. We have already discussed these examples in Chapter III.2.2.-
See also Postal [64c].-O n the history o f the concept of depth structure see Chomsky
[66b].
19 See e.g. Chomsky [64], p. 85.
20 This avoids the difficulties in defining transformation rules that were mentioned
above. It is also possible to incorporate into the definition of the transformation rules
specifications that refer to the deep-structural classification of the expressions in their
arguments. These specifications do remain invariant in relation to transformations.
21 On their definition see Bar-Hillel [53].
22 On their definition see Chomsky [59].
23 See Bar-Hillel, Theorems 1 and 2.
24 For discussion of other generative grammars see also Postal [64a], Chomsky [62b]
and Chomsky [56].
25 Bierwisch also emphasizes in [66], p. 145f. the relation between depth structure and
logical structure. He does not take them to be identical, to be sure, but to be closely
related. These ideas have been more extensively developed first by E. Bach in [68], by
G. Lakoff in [70] and by J. McCawley in [68].
26 The representation A [Z] was explained in III.2.
27 These and similarly defined substitution categories were introduced by Bloomfield,
Harris and Bar-Hillel. See e.g. Bar-Hillel [50].
28 Curry also expresses himself along these lines in [61], while Putnam takes the posi­
tion in [61] that there are no sharp boundaries between syntactic and semantic irreg­
ularities. - Perhaps there are none from the point of view o f language, but that does not
speak against drawing such boundaries in grammar.
29 See also the presentation in Katz [66].
30 This lack is still more obvious in Abraham and Kiefer [66]. Here distinctors and
selectors are omitted in the readings and what shows up as a single rule of projection is a
rule that corresponds to Katz and Fodor’s rule R i in so far as it combines character­
istics (?) of the components of the new expression conjunctively. At best, however,
it can be used to identify conjunctive combinations of one-place predicates, but not
names, sentences, polyadic predicates, functors or functor-argument constructions.
31 See Chapter III.2.1.6 on this point.
32 See Chomsky [66b] and [67].
33 On the definition of general language theory see e.g. Chomsky [61a], p. 120 or
Postal [64a], p. 3f.
34 Chomsky takes up the historical parallels to his hypothesis in [66b].
35 Chomsky draws the distinction between formal - substantial categories in [65],
Chapter I, § 5.
36 See also Goodman’s discussions of the point in [67].
37 Even if the IH would explain certain phenomena for which we have no other satis­
factory explanation, it would not follow from this, as Goodman emphasizes in [67],
that it is correct or even just acceptable. The latter would be true only if it did not raise
more problems than it solved.
38 See the remarks in Chapter IV. 3.
39 See also Putnam [67].
40 The question is, o f course, whether we are not obliged to say that persons o f lesser
 THEORIES OF GRAMMAR 241

intellectual gifts do master another language, namely the everyday language they speak,
just as well as others master cultivated speech.
41 Putnam says: “Nine or ten years is enough time to become pretty darn good at
anything ” [67], p. 20. Of course, we have to draw some distinctions here: As a rule, a
child has mastered simple sentence structures by five years, but needs more than ten
years to learn more complex sentence forms; here again there is the question o f whether
it is not a matter of different language systems.
42 Along these lines, Putnam in [67], too.
 CHAPTER IV

L A N G U A G E A N D R E A L IT Y

1. T h e t h e s is of t he r o l e l a n g u a g e p l a y s in e x p e r i e n c e

A thesis concerning the role language plays in experience was first set
forth on a scientific basis by Wilhelm von Humboldt, whose ideas con­
tinue to have an effect up to the present, especially in German linguistic
theory.
For Humboldt language is not only a means of expressing or communi­
cating what is in thought, but thought and language form an inseparable
unity: “The most obvious, but the most limited view of language is that
which considers it a mere means of understanding each other. ... But lan­
guage is through and through not just a means of understanding, but the
imprint of the speaker’s mind and his view of the world; association in
society is the essential vehicle of its development, but it is by far not the
sole purpose it works tow ard....” 1 He speaks of the “inseparability of
human consciousness and human language” 2 and says: “Language is
the formative organ of thought. Intellectual activity, mental through and
through, passing by all inwardly and somehow without a trace, becomes
external and perceptible to the sense by virtue of sound in speech. It and
language are therefore one and inseparable from each other. But in itself,
too, it is bound to the necessity of entering into a connection with the
sound of words; else thinking can not attain clarity, the idea not become
a concept.” 3 “The concept may ... no more loose itself from the word,
than the man can lay aside the features of his face. The word is its own
particular shaping and, if it wishes to depart therefrom, only in other
words can it find itself again.” 4
The concept is first formed by means of the word; concept and sound
pattern are not brought together in the word: “From the first element on,
production of language is a synthetical process in the most genuine sense
of the word, where the synthesis engenders something that does not lie
within any of the connected parts by themselves.” 5
Accordingly, it is only with the help of language that we apprehend
 LANGUAGE AND REALITY 243

the world, the things, distinctions, properties and relations within it; and
we apprehend them in a particular way with a particular language: Every
language contains a view of the world [Weltansicht] and its structure, a
distinctive ontology: “The mutual dependence of the thought and the
word on each other makes it clear that languages are not really instru­
ments for representing the truth already known, but far more of revealing
the previously unknown. Their differences are not a matter of sounds and
symbols, but differences in their very world views.” 6 “Here too we find
that the way of representing language as if it did no more than describe
objects perceived in and of themselves is far from exhausting its full and
deep content. Just as concepts are not possible without language, so there
would be no objects for the soul without it, since every external object
obtains its essential determination only through the mediation of a con­
cept. The entire manner of the subjective perception of objects merges
into the structure and use of language. For it is out of this perception that
the word arises and it is not an imprint struck from the object in and of
itself but from the image of it created in the soul. Since subjectivity is
unavoidably mixed in with all objective perception, even independently
of language each human individuality can be viewed as a unique and
distinctive standpoint in viewing the world. It becomes still more so,
however, because of language, since the word, vis-a-vis the soul, makes of
itself an object once again and introduces a new distinctive feature sepa­
rate from the subject, so that now the concept includes a threefold aspect,
the impression of the object, the way in which this is taken up by the
subject, the influence of the word, as speech sound.” 7 “Subjective activity
constitutes an object in thought. For no class of ideas can be regarded as
merely a passive reception and contemplation of an object already present.
The activity of the senses must be synthetically combined with the mind’s
inward functioning, and out of that combination the idea emerges a
separate thing, becomes, vis-a-vis the subjective faculty, object and re­
turns as such, newly perceived, into the former. For that, however,
language is indispensable.” 8 “The concept first attains its clearness and
distinctness by means of being mirrored from another intelligence.
It is created, as we saw in the foregoing, by loosing itself from the turbu­
lent mass of ideas and constituting itself as an object vis-a-vis the subject.
Yet it does not suffice for this rupture to occur within the subject alone,
objectivity is not brought to fulfillment until the thinker of the ideas be­
244 CHAPTER IV

holds the thought really outside of himself, which is possible only in

another being, like him, thinking and having ideas. But between one
intelligence and another the only intermediary is language, and so its
necessity for the complete realization of the thought is a consequence of
this too.” 9
“But language is not world view just because, since every concept must
be grasped by it, it must equal the entire compass of the world, but also
because it is only its transformation of the objects that makes the mind
capable of insight into the connectedness that is inseparable from the con­
cept of the world. For it is only by transducing reality’s impression on
the senses and the sensation over into its own realm of articulated sounds
made ready beforehand as organ of thinking that it enables us to unite
objects with clear and pure ideas, and so shedding light upon the way the
world hangs together. Man primarily lives with objects just as language
brings them to him, and his sensation and action depend on his ideas,
even exclusively so. By the same act by which man spins language out of
himself, he spins himself into it, and every language draws a circle around
the nation to which it belongs, a circle that can be crossed only insofar
as one at the same time steps over into the circle of another language.” 10
Humboldt speaks of the “transformation of the world into language” 11
and says: “The distinctive feature of language is that, mediating between
men and outer objects, it fastens onto sounds a world of thought.” 12
“ But the path it takes is always different and its constructions arise from
the interaction of external impressions and inner feeling, tied to the
general goal that language pursues, binding subjectivity with objectivity
in the creation of an ideal, but neither entirely inner nor entirely outer
world.” 13
For Humboldt it is true that languages: “ ... remodel, each with the
power that dwells within it, the common land that lies before us into the
property of mind.” 14
This world view that belongs to a language bears the stamp of the life
will, or, as we also say, the cultural will of the nation that speaks that
language; conversely, the peculiar nature and the character of a culture
manifests itself in the potency o f language, for language and culture,
language and life are bound up in the most intimate way:
“Language lives and moves in nationality and the deepest secret of its
being manifests itself precisely in the fact that it comes forth out of the
 LANGUAGE AND REALITY 245

apparently confused mass of individualities, within which no one of them

need be singled out. It receives its whole form from this natural working,
dark and obscure, of unconsciously concurring tendencies, since whatever
comes of individual aim, no matter how well calculated, does nothing
but play around their outskirts, so to speak, in visible impotence.” 15
“From every language, therefore, conclusions can be drawn about the
national character.” 16 For Humboldt, then, a language is a “ spiritual
individuality” 17 and this individuality is that of the nation: “Basically the
language is ... the nation itself, and strictly, genuinely the nation.” 18
“The language, however, has just this power to thank for its origin, or
more correctly: the particular national power and strength can be
externally expressed only in the particular national language, these
sounds, these linkages of analogy, these symbolic hints, these specific
principles of inward development. It is this that we may well call, but
always improperly and loosely, the making of language by the nation.” 19
“We can, as generally acknowledged, assume that the diverse languages
constitute the organs of the nations’ distinctive ways of thinking and
feeling, that a large number of objects [of thought] are first brought forth
by the words that designate them, and have their being only in [those
words] ...” 20
If culture is the maker of language, then it is a power over the individual
that determines his thinking and his experience in the form of its world
view and so has a part in shaping his experience:
“The idea that the various languages do nothing more than describe
in different words the same vast number of objects and concepts existing
independently of them and that the words are arranged according to
different principles, which apart from their influence on understanding
have no further significance, is too natural to man for him to be able
easily to free himself from it before he has reflected more deeply on lan­
guage. He disdains what appears in the particular to be so small and
negligible as to be mere grammatical fussiness, and forgets that the
cumulative mass of those particularities, all unknown to himself, con­
strains him and rules him all the same. Always living, moving, acting in
[a world of] objects he takes subjectivity too little into account; the con­
cept comes hard to him of a power given by nature itself, a power that is
not fortuitous nor whimsical nor arbitrary but so transforming all, ac­
cording to its own inward principles, that the apparent object itself is
246 CHAPTER IV

only a subjective notion, yet one that has every right to lay claim to
universal validity.” 21 “But thinking is not just dependent on language in
general, but to a certain extent it is determined by each individual one as
well.” 22 “ For the influence of the individual view of the object on the
moulding of the word also determines, as long as it remains a living force,
the way the word recalls the object.” 23 “But language, as the work of the
nation, and of time past, is something foreign to the individual; on the
one side he is bound by it, but on the other he is enriched, invigorated
and stimulated by all that earlier generations have deposited in it.” 24
“When one considers what a binding influence over each generation of a
people everything their language has experienced throughout all the
preceding centuries has... it will be clear how slight is the power of the
individual as against the mighty force of language.” 25 And therefore it
is true: “(1) that language attains an individuality through the influences
that work upon it, an individuality that really becomes its nature, to the
extent that it now reacts upon it and can be used willingly only within
its limits. (2) that its reactive force is the more definitive as what is brought
to bear upon the individual by means of whole epochs and nations is
working within it, whose very individuality, inclined in the same direction
by the likeness of the influences affecting it, is scarcely capable of resist­
ing.” 26 “But man does not speak because he wants to speak that way,
but because he has to speak that way; the way he speaks is a compulsion
of his intellectual nature; it is free, to be sure, because it is his own,
original nature, but there is no bridge leading him in a unifying conscious­
ness from each momentary appearance to that unknown fundamental
presence.” 27
This influence of language on thinking is the result of the conceptual
system or vocabulary, but also of the forms of grammar: “It is only on
more precise consideration, but then clearly and distinctly, that we find
the character of the diverse world views of peoples adhering to the value
and weight of words. I have already remarked in the foregoing... that it
is scarcely possible for different individuals uniformly to take up some
word into their idea, unless it be momentarily used merely as a material
sign of its concept. We can say, therefore, quite flatly that there is some­
thing in each one that can not be differentiated once again by means of
words and that the words of diverse languages, even if they designate
roughly the same concepts, are never true synonyms.” 28 “Yet it would be
 LANGUAGE AND REALITY 247

one-sided to think that the distinctive quality of the nation’s spirit and
character is revealed in concept formation alone; it exerts a similar large
influence on the way discourse is structured and can equally be known in
it.” 29
For Humboldt, grammatical form is even the primary respect in which
different languages are distinguished from each other.30
In view of all this, the usefulness of comparative studies of language
lies in the fact that by learning new languages we become acquainted with
new world views and ways of thinking and thereby widen the limits of
our own horizon. It is precisely through the study of the most diverse
languages that the broadest perspective rises, for: “Difference in structure,
even where it is quite substantial enough, is nevertheless often insufficient­
ly recognized and appreciated as long as we are concerned with a small
number of languages not totally different one from the other.” 31
But Humboldt sees very clearly, too, the difficulties that stand in the
way of comparing languages or the world views they are based upon.
In the first place, according to him, every language, no matter how
‘primitive’ it may be in particular detail, has a universal character, i.e.
every thought or concept can be expressed in it, even if it is with difficulty
and only in approximations. I.e., it would never be possible to say: This
concept or that thought can not be rendered here at all, this and that do
not enter into the ontology of this language:
“But in both the language and the concepts of any people, no matter
how culturally undeveloped, there resides, and this is by far the more de­
cisive point here, a totality commensurate with the full range of the
unrestricted human cultural potential, out of which can be engendered,
with no outside aid, every detail that humanity encompasses.” 32 “Ex­
perience with translation from very disparate languages, and with the use
of the roughest and most undeveloped of them in teaching the most arcane
doctrines of a revealed religion [in the case of Bible translation], does
show that, even if with greatly variable success, any sequence of ideas can
be expressed in every one of them. But this is not merely a consequence
of the general relatedness of everything, and of the flexibility of the con­
cepts, and their signs. For the languages themselves and the influence
they have on the nations proves nothing but what follows from them
naturally; not what they can be compelled to, but what they invite and
encourage.” 23 “In translating phrases in such languages with such and
248 CHAPTER IV

such a form, therefore, it is necessary to keep very much in mind the fact
that those translations, so far as they have to do with grammatical forms,
are almost always false, and their grammatical aspect is quite different
from the one presented by the speaker in saying them. If we wished to
avoid this, we would have to use in translating only the grammatical
forms present in the original language; but then we would come up against
cases where we would have to give up translating at all.” 34 “Nor is it
simply indifferent whether one language accomplishes by periphrasis what
another language expresses in one word; not with grammatical forms,
since in the case of periphrasis, they no longer present the appearance of
modified ideas, but that of specifying the modification, contrary to the
concept of a mere form; not in the designation of the concepts, either.
The principle of segmentation necessarily suffers when what presents
itself as a unit in concept does not appear to be so in expression, and all
of the word’s vivid effectiveness, as an individual, falls by the wayside for
the concept that lacks any such expression.” 35
There are always translations, then, but they are for the most part
inexact. But since there are no generally valid and sharp criteria for
precision of translation, the differences in meaning are often very difficult
to assess.
For Humboldt language is furthermore an organism, i.e. a structure,
the individual parts of which must always be viewed in terms of their
function in the whole. And it has an inner dynamics that cannot be
adequately caught by descriptions of momentary linguistic situations.
This implies an additional difficulty for grasping the peculiar character of
languages and for making comparisons between them.
Humboldt says: “There are no isolated phenomena in language; each
element in it proclaims itself to be a part of a whole.” 36 And: “ Language
is not at all something spread out there, a stuff made up of the mass of
words and rules it offers, but something ongoing, an ideal process, as
life is a corporeal one.” 37 “The direct, immediate exhalation of an organic
being in its sensual and intellectual status, it shares the nature of every­
thing organic, that each thing within it exists only through the other and
all of them only through the one all-pervasive power that permeates the
whole.” 38
The diversity of languages does not lie in their details, which viewed in
isolation can always be interpreted in such a way that they are similar
 LANGUAGE AND REALITY 249

to each other, but in the languages’ entirety. Only with regard to the
diversity of the whole structures do the details also reveal themselves
to be diverse. But it is very difficult to attain a perspective on languages
in their entireties and to catch their characteristic and distinctive features
in statements that are general and yet sufficiently precise: “But the spirit
that forms language knows how to master the technical grammatical
instruments and to give them differing value so that neither their presence
nor their absence will lead to generally definitive and certain conclusions
about the essence of the language form. Therefore if we run through any
given language, it is hard to find a single point it would not be possible
to conceive in another way, with no damage to the essential nature of its
language form, and we will be forced to return to the overall impression.
Here the exact opposite comes into play; the most decided individuality
strikes the eye clearly, impresses itself compellingly on our feeling. If one
goes back to the material and the technical aspect of language from here,
there is scarcely anything else to do but take all and everything together,
just as concrete as it is, as making up the language form, understanding
this in a sense that would really exclude any possibility of alteration
within the same language form. Languages can be compared, most ade­
quately, to human physiognomies in this respect. Individuality is ob­
trusively present, similarities are recognized, but there is no measuring,
nor any description of component parts as to particular detail or their
connection with each other that is capable of summing up into one con­
cept the distinctive character. It is based on the whole and on the way it is
apprehended, once again an individual matter, and thus every physiog­
nomy is sure to take on a different look for every individual. Since lan­
guage, in whatever form you may take it up, is always a spiritual exhala­
tion of an individual national life, both must be realized in it. No matter
how much you isolate it, pin it down, embody it, there is always something
more, precisely the most important thing about it, wherein there is the
unity and the breath of life of a living being.” 39 “Within language, the
character of the whole sweeps the individual along with it every time.
If you forget to hold fast to this basic principle in arriving at judgments
about languages, you are mistaking their very nature and every genuine
difference among them as well. For they do not ever diverge from one
another to such an extent that there should be no particular resemblances
even between the most disparate of them.” 40 “Language offers us an
250 CHAPTER IV

infinitude of particular details, in words, rules, analogies and exceptions

of every kind. We are more than slightly perplexed as to how we are to
bring this mass of detail - which seems to us a confusing chaos still,
despite the order into which it has already been brought going un­
noticed - into some judgment comparing it with the unified image of
human intellectual capacity. Even if we find ourselves in possession of the
necessary lexical and grammatical details of two important language
stems, e.g. Sanskrit and Semitic; that will still not take us very far in our
effort to sum up the character of each of them in such a way that we can
make fruitful comparisons between them and specify their proper places
in the general enterprise of language creation, as measured by their rela­
tion to the spiritual-intellectual capacity of the nations. What is still
required is a special seeking out of the common sources of the individual
particularities, the gathering together of the scattered features into the
picture of an organic whole. That is the only way to enable us to get a
firm hold on the particulars.” 41

Current discussion of the thesis that language plays a role in experience

has been prompted, at least in Anglo-Saxon world, more by the works
of Edward Sapir (1884-1935) and Benjamin Lee Whorf (1897-1941) than
by Humboldt.
Sapir, one of the most influential American linguists, expressed the
thesis of language’s role in experience as follows: “Human beings do not
live in the objective world alone, nor alone in the world of social activity
as ordinarily understood, but are very much at the mercy of the particular
language which has become the medium of expression for their society.
It is quite an illusion to imagine that one adjusts to reality essentially
without the use of language and that language is merely an incidental
means of solving specific problems of communication and reflection.
The fact of the matter is that the ‘real world’ is to a large extent uncon-
ciously built up on the language of the group.” 42 “The relation between
language and experience is often misunderstood. Language is not merely
a more or less systematic inventory of the various items of experience
which seem relevant to the individual, as is so often naively assumed,
but is also a selfcontained, creative symbolic organization, which not
only refers to experience largely acquired without its help but actually
defines experience for us by reason of its formal completeness and because
 LANGUAGE AND REALITY 251

of our unconscious projection of its implicit expectations into the field of

experience.” 43
Whorf, who took up these ideas of Sapir’s and developed them further,
formulates his principle of relativity in this way: He refers to the realistic
conception and says: “Natural logic says that talking is merely an inci­
dental process concerned strictly with communication, not with formula­
tion of ideas. Talking, or the use of language, is supposed only to ‘express’
what is essentially already formulated nonlinguistically. Formulation is
an independent process, called thought or thinking, and is supposed to be
largely indifferent to the nature of particular languages. Languages have
grammars, which are assumed to be merely norms of conventional and
social correctness, but the use of language is supposed to be guided not
so much by them as by correct, rational, or intelligent thinking.” 44
Against this realistic interpretation Whorf now urges: “It was found
that the background linguistic system (in other words, the grammar) of
each language is not merely a reproducing instrument for voicing ideas
but rather is itself the shaper of ideas, the program and guide for the
individual’s mental activity, for his analysis of impressions, for his
synthesis of his mental stock in trade. Formulation of ideas is not an
independent process, strictly rational in the old sense, but is part of a
particular grammar, and differs, from slightly to greatly, between different
grammars. We dissect nature along lines laid down by our native languages.
The categories and types that we isolate from the world of phenomena
we do not find there because they stare every observer in the face; on the
contrary, the world is presented in a kaleidoscopic flux of impressions
which has to be organized by our minds - and this means largely by the
linguistic systems in our minds. We cut nature up, organize it into con­
cepts, and ascribe significances as we do, largely because we are parties
to an agreement to organize it in this way - an agreement that holds
throughout our speech community and is codified in the patterns of our
language. The agreement is, of course, an implicit and unstated one, but
its terms are absolutely obligatory; we cannot talk at all except by sub­
scribing to the organization and classification of data which the agreement
decrees.” 45
Further: “We are thus introduced to a new principle of relativity, which
holds that all observers are not led by the same physical evidence to the
same picture of the universe, unless their linguistic backgrounds are
252 CHAPTER IV

similar, or can in some way be calibrated.” 46 Or: “From this fact pro­
ceeds what I have called the ‘linguistic relativity principle’, which means,
in informal terms, that users of markedly different grammars are pointed
by their grammars toward different types of observations and different
evaluations of externally similar acts of observation, and hence are not
equivalent as observers but must arrive at somewhat different views of the
world.” 47
In many passages Whorf emphasizes this influence of language on
thinking and experience so strongly that he speaks of determination or
compulsion by language: “This study shows that the forms of a person’s
thoughts are controlled by inexorable laws of pattern of which he is
unconscious. These patterns are the unperceived intricate systematizations
of his own language - shown readily enough by a candid comparison and
contrast with other languages, especially those of a different linguistic
family.” 48 “Thinking follows a network of tracks laid down in a given
language.... The individual is utterly unaware of their organization and
is constrained completely within its unbreakable bonds.” 49
No more than Humboldt do Sapir and Whorf take the view that the
world view of a culture is determined by its language alone, however.
They urge rather that though the individual bears the stamp in his
thinking and his world view of the language that he learns and accepts -
although he also has a certain share in the formation of the language -
the many individuals as a social culture create the language through their
own particular way of coming to grips with the world in which they live.
- “Which was first: the language patterns or the cultural norms? In main
they have grown up together, constantly influencing each other. But in
this partnership the nature of the language is the factor that limits free
plasticity and rigidifies channels of development in the more autocratic
way. This is so because a language is a system, not just an assemblage
of norms. Large systematic outlines can change to something really new
only very slowly, while many other cultural innovations are made with
comparative quickness. Language thus represents the mass mind; it is
affected by inventions and innovations, but affected little and slowly,
whereas TO inventors and innovators it legislates with the decree imme­
diate.” 66
Whorf does not view the connection between language and culture and
language and world view as being as close as it is in Humboldt’s view.
 LANGUAGE AND REALITY 253

In particular, he does not believe that one would obtain all the informa­
tion about a culture and its world view out of language, or conversely
that the basic structures of the language would already be implied in the
world view. So Whorf denies “that there is anything so definite as a ‘cor­
relation’ between culture and language” 51 and insists: “there are connec­
tions but not correlations or diagnostic correspondencies between culture
norms and linguistic patterns.” 52
For Whorf, as for Humboldt, the general significance of comparative
linguistic research lies in the fact that it widens our horizons by showing
us the relativity of our world view, which is mediated by our language:
“ One significant contribution to science from the linguistic point of view
may be the greater development of our sense of perspective. We shall no
longer be able to see a few dialects of the Indo-European family, and the
rationalizing techniques elaborated from their patterns, as the apex of
the evolution of the human mind, nor their present wide spread as due to
any survival from fitness or to anything but a few events of history - events
that could be called fortunate only from the parochial point of view of
the favored parties. They, and our own thought processes with them,
can no longer be envisioned as spanning the gamut of reason and knowl­
edge but only as one constellation in a galactic expanse.” 53
In the next two sections, we will discuss some arguments for the rela­
tivity thesis drawn from linguistics, in order to make clear their concrete
content, and then in Chapter I V.4 we shall go into the aspects of that thesis
that have to do with the philosophy of language and theory of knowledge.

NOTES
1 Humboldt [03], V. VI, p. 22f.
2 Humboldt [03], V. VI, p. 16.
3 Humboldt [03], V. VII, p. 33.
4 Humboldt [03], V. VII, p. 100.
5 Humboldt [03], V. VII, p. 94.
6 Humboldt [03], V. IV, p. 27.
7 Humboldt [03], V. VI, p. 179.
8 Humboldt [03], V. VII, p. 55.
9 Humboldt [03], V. VI, p. 160.
10 Humboldt [03], V. VI, p. 179f.
11 Humboldt [03], V. VI, p. 28.
12 Humboldt [03], V. V, p. 110.
13 Humboldt [03], V. VII, p. 115.
14 Humboldt [03], V. IV, p. 420.
15 Humboldt [03], V. VI, p. 189.
254 CHAPTER IV

16 Humboldt [03], V. VII, p. 172.

17 Humboldt [03], V. VI, p. 151.
18 Humboldt [03], V. VII, p. 641.
19 Humboldt [03], V. VI, p. 127.
20 Humboldt [03], V. VII, p. 640.
21 Humboldt [03], V. VI, p. 119.
22 Humboldt [03], V. IV, p. 22.
23 Humboldt [03], V. IV, p. 29.
24 Humboldt [03], V. V, p. 27.
25 Humboldt [03], V. VI, p. 182.
26 Humboldt [03], V. IV, p. 423f.
27 Humboldt [03], V. VI, p. 127.
28 Humboldt [03], V. VII, p. 190.
29 Humboldt [03], V. VII, p. 92.
30 Cp. Humboldt [03], V. VI, p. 249f.
31 Humboldt [03], V. VI, p. 121.
32 Humboldt [03], V. VII, p. 28.
33 Humboldt [03], V. VI, p. 16f.
34 Humboldt [03], V. IV, p. 293.
35 Humboldt [03], V. IV, p. 20f.
36 Humboldt [03], V. IV, p. 14f.
37 Humboldt [03], V. VI, p. 146.
38 Humboldt [03], V. IV, p. 3.
39 Humboldt [03], V. VI, p. 245.
40 Humboldt [03], V. VI, p. 266.
41 Humboldt [03], V. VII, p. 44f.
42 Sapir [29], p. 209.
43 Cited by P. Henle in [65], p. 1.
44 Whorf [56], p. 207f.
43 Whorf [56], p. 212f.
46 Whorf [56], p. 214.
47 Whorf [56], p. 221.
48 Whorf [56], p. 252.
49 Whorf [56], p. 256.
50 Whorf [56], p. 156. - P. Henle also has this to say along thesame lines: “In neither
case have we claimed, nor would wewant to claim, thatlanguage is the sole influence
or even the primary influence. In neither case have we claimed that the causal relation­
ship does not also run in the other direction as well. Because of the enduring character
of languages and the fact that a population changes in time, it may well be that language
considered in the large is molded by environmental conditions, social organisation, and
prevalent modes of thought. This would not prevent language being an influence on
thought in the development of the individual, and this is all we have claimed.”
([65], p. 16f.)
31 Whorf [56], p. 139.
32 Whorf [56], p. 159.
33 Whorf [56], p. 218f. - For presentation and discussion of the Whorfian relativity
theory, see also the contributions in Hoijer [54a] and Henle [65]. - For references to
standpoints similar to what we have presented with respect to Humboldt’s and W horf’s
statements in this section, see Basilius [52].
 LANGUAGE AND REALITY 255

2. T h e r o l e of v o c a b u l a r y

If one wishes to give to language a role in experience, one can make

reference to vocabulary or to grammar. Vocabulary’s part can be under­
stood more easily, but from the point of view of Humboldt and Whorf
[it is] altogether less important than that of grammar.
We will treat the relativity thesis for vocabulary in terms of an example
of diverse arrangements of one semantic region into word fields, one which
has played an important role in the discussion of this thesis: the example
of diverse word fields in the differentiation of colors.1
Most languages have only a few basic words available for the classifi­
cation of the extraordinarily large number of distinguishable shades of
color. Now to a considerable degree the classifications of colors realized
by means of these basic words partially diverge from each other. For
example, the language of the Navaho Indians has two color words for our
‘black’, which designate the black of night (darkness) and the black of
coal (black color), say, but they have only one word in the place of our
‘blue’ and ‘green’, on the other hand.2
Now if we assume that the distinction blue - green cannot be repro­
duced in Navaho in some simple way, e.g. by auxiliary adjectives (cold
blue-or-green = blue, warm blue-or-green=green, say, or sky blue-or-green
= blue, grassy blue-or-green= green), according to Whorf that would mean
that the Navahos typically do not distinguish between blue and green in
color experience, and that the division of the color region in language
determines the differences in the way we perceive colors.
Max Black criticized W horf’s claim that a classification example of
that sort implies an influence of language on experience in [59]. What
he says there is: The Navahos are able to distinguish colors, e.g. blue
and green, just as well as we are, i.e. they have at their disposal the same
concepts as possibilities of differentiation as we do. It is only that among
the various distinctions that can be made, the ones they can express in
linguistically simple ways are different from the ones we can express in
that way. There are no empirical grounds of any kind for taking the view
that different color vocabularies have an influence on how colors are
perceived. What this case has to do with is simply a difference in classifi­
cation, not a difference in perception.3
But one must make a distinction here: Whorf does not deny that the
256 CHAPTER IV

Navahos have the capacity for distinguishing colors just as precisely as

we do, in particular that they are in a position to distinguish between
blue and green. He does not say that the Navahos are color blind with
respect to the distinction blue - green. But that does not make it possible
to maintain that ‘they have at their disposal the concepts’ blue and green
- a way of putting it that Black takes over from H. H. Price. You do not
already have a concept at your disposal when you are capable of learning
a distinction, but only when you have learned it. We can scarcely say of
a layman in mathematics, for example, that he has at this disposal the
concept of perfect number, even though he does have the capacity for
learning the distinction between perfect and imperfect numbers. He has
that concept at his disposal only when he has learned that a perfect
number is a number n, for which the sum of its whole number divisors
= 2 •«, i.e. when the concept, or the use of the predicate ‘perfect number’
has been explained to him. Beforehand, however, he does not make that
distinction and is entirely incapable of making it. He does not just lack
the means of making the distinction.
To begin with, then, it is not a matter of the capacity of being able to
make distinctions in the sense of being capable of learning them, but of
the distinctions that are actually made.4 It is also not a matter of distinc­
tions that are made on occasion, but of the distinctions that are made
systematically and often.
Now what Whorf claims is that distinctions for which there are no
simple and stereotyped forms of linguistic expression are not that sort of
common and systematic distinctions. This surely has the highest degree
of plausibility, for one would construct simple forms of expression for
common distinctions for reasons of economy.
Perceptions, observations, generally speaking, experiences, furthermore,
are not isolated from other activities, but are intimately entangled in the
totality of all we d o : Distinctions we establish in observation cause us to
conduct ourselves differently, and conversely, we do not make just any
and every distinction in observation, but direct our attention to the
distinctions that are relevant to what we do. But since our life and activity
is very firmly linked up with society and since we use language to coor­
dinate our common activities, distinctions within experience are not
things that would be only partially and imperfectly expressed; on the
contrary, every functioning language must be capable of expressing all of
 LANGUAGE AND REALITY 257

the distinctions that are practically relevant for the members of its linguistic
community. The variety of the distinctions that are systematic, regular and
important, even when they have not first been learned through language, will
essentially coincide with the language’s diverse possibilities of expression.
But that is to say: there is a close connection between means of making
distinctions in language and the contents of experience. In the color
example, then, it is not so that we only classify and express the same color
perceptions in different ways in language, but rather different forms of
color perception correspond to the various ways of dividing and arranging
the semantic region of color into word fields.
Of course the example of color word fields is not suited for the illustra­
tion of more radical differences of apprehension or understanding. Those
become much clearer when one considers other world fields, from the
psychological or intellectual domain, for example. Here we make distinc­
tions between e.g. ‘sorrowful’, ‘melancholy’, and ‘dejected’, or ‘wise’,
‘intelligent’, and ‘clever’, for which in other languages there are no simple
stereotyped possibilities of expression, simply because those distinctions
are not made as a rule in the relevant cultures. And it would be extremely
peculiar to say that the people who belong to those cultures perceive,
just as we do, that a man is intelligent but not wise, or melancholy but
not sorrowful, but just can not express it in words or do express it in
words that when translated back into our language mean something
quite different - why should their linguistic behavior be so awkward? It
is surely more sensible to say: They do not express those distinctions
because they do not make them.
Furthermore, we can extract from a people’s juridical language the
distinctions they make in matters juridical and from the vocabulary of a
language a great deal about a people’s cultural forms and living condi­
tions. For example, the Nootka Indians of the northwest coast of America
have a large number of very specific words for sea animals, while other
inland tribes have available instead detailed designations for berries and
edible plants. The Paiute Indians, who live in the desert, have very exact
terms for topographical particulars - a very important matter for orienta­
tion in a uniform landscape, and so on.5
It is in this sense that E. Sapir says that the vocabulary of a language is
a “complex inventory of all the ideas, interests, and occupations that take
up the attention of the community.” 6
2 58 CHAPTER IV

The alleged role of language in experience is not, then, a matter of

which distinctions could be made or happen to be made on occasion, but
of which distinctions are typically made; nor is it a matter of what can be
expressed in a language at all, but rather of what can be expressed in
simple and stereotyped locutions.7 In the latter sense, however, we can
regard a close connection between vocabulary and experience as having
sufficient evidence in its favor.
Nevertheless, the thesis of the role of language in experience is not
proved thereby. For the fact that the distinctions that are typically drawn
in the experience, as generally within the life of a linguistic community
are expressed in its language, that the various cultures’ ways of under­
standing and experiencing the world are reflected in their languages has
hardly been contested, but what has been contested is that language con­
versely has an influence on experience, on the apprehension of the world,
and, if it is not the only influence, is nevertheless a codeterminant of the
form and shape of experience. And only that gives us the full content of
the relativity thesis.
On this point, however, Humboldt, and Whorf as well, as we saw above,
draw some distinctions. Neither one of them advocates the thesis that a
people gets a language from somewhere or other, which then determines
its typical forms of experience, action and life, nor would it be the least bit
plausible. Rather, language is shaped out of the confrontation with the
material, social and religious requirements and circumstances of a people.
By and large, language is a part of culture and is formed by the powers
that form that culture also. Equally, by and large, the human beings who
belong to a people form its culture, and not vice versa.
If, on the other hand, one considers the individual, it makes sense to
say that he is shaped by the culture into which he is born, and it also
makes sense to say that the language he acquires participates in deter­
mining his way of experiencing. Along with words he learns certain differ­
ences, certain forms that determine the experienced world in a way that is
relevant and important for the society and culture to which he belongs,
and consequently for him as well. Along with these differences he learns
to orient himself in the world and to interpret the objects of his experience.
The world reveals itself to him in these differences. Because we learn
distinctions and determinations already made, language influences the
way and manner in which the individual perceives and experiences, and
 LANGUAGE AND REALITY 259

what he perceives. Perception does not consist just in having the senses
stimulated, but in the perception of something as specifically character­
ized in such and such a way.
Naturally, not even for the individual is there total dependence on
language, such that he would not be in a position to make other distinc­
tions besides those that can be simply expressed in language. For in the
first place an adequate and expressive language is so plastic that new
distinctions can be formulated in it as well. For another thing, the individ-
dual can introduce new expressions and mark new differences in language
which expand or alter the language. For example, that is what occurs
constantly in scientific work, where when new domains of objects are
revealed, a new terminology for describing them must be developed. On
the large view, however, the individual follows the language as he finds it
rather than shaping it anew, and therefore it can be said, following
Humboldt, Sapir and Whorf, that our language determines the form and
manner of our experience.8
But it is also an error when M. Black says in [59] that there are no
empirical arguments for a correlation between perception and language.
In a very careful empirical-psychological study - one, moreover, which
in its general statements about the Whorfian thesis belongs among the
works on that theme that are most worth reading - R. W. Brown and
E. H. Lenneberg have proved in [54] that there is a connection between
the recognition of colors and linguistic codability. This is defined in terms
of the simplicity (brevity) of the color designation, its availability (the
speed with which it is named) and its reliability (individual and inter­
personal agreement in using it). It is shown that colors are recognized the
more easily, the greater their codability. And E. H. Lenneberg and J. M.
Roberts proved in [53] that the Zuni Indians, who have in their language
only one designation for yellow and orange, confuse these two colors more
frequently than English-speaking Americans do. Interestingly, the success
of bilingual Zuni, who also know English, in recognizing these two colors
falls between the success of monolingual Zuni and that of Americans who
speak English only.
These studies lend support to the intimate involvement of language
with experience and the role played by language in experience by showing
that we recognize colors - and we may say by way of generalization:
objects - not so much by way of a comparison between present sense-
260 CHAPTER IV

impressions and earlier ones as by comparing earlier designations with

present ones, and so with the help of linguistic classifications. Success in
recognition within some specific domain depends upon the means of
linguistic classification for the domain and for that reason success can
vary conspicuously from language to language.

NOTES

1 See also Oehmann [51], p. 123ff., and Gipper [69], Chapter 5 on this point.
2 See e.g. Hoijer [54b], p. 96.
3 See Black [59], p. 231.
4 P. Henle also makes statements along these lines in [64], p. 7f. and H. Hoijer in
[54b], p. 96. See also C. F. Hockett’s assertions in Hoijer [54a], p. 122.
3 See Henle [65], p. 5.
6 Sapir [121, p. 228.
7 Humboldt also makes this point in the passages cited.
8 In Whorf, it is true, there are also many statements, as was emphasized in Chapter
IV. 1, according to which language does not just suggest a certain world view to us but
forces one upon us. In this connection he refers primarily to grammar, which on his
view leaves a much stronger imprint on our experience than vocabulary. - P. Henle
gives a very weak formulation for the relativity thesis, when he says: “Neither finally
have we argued that there is any compulsive influence of language upon thought, that
language makes impossible all but certain modes of perception and organisation of
expression. Since perception and experience are ordinarily manifested only through
language, the point being made here may be made in another way. In natural languages,
the elements we have been considering - vocabulary, inflection and modes of sentence
structure - do not make it impossible to express certain things, they merely make it
more difficult to express them.” ([65], p. 17). But that lies on the border of triviality and
need not be discussed. What is expressible in a simple way in a language naturally
depends on the language’s expressive resources. Whorf says, for example, that we not
only always express ourselves in the simplest way that is possible in our language, but
that we see the matter in the way the forms of our language suggest.
 LANGUAGE AND REALITY 261

3. T h e r o l e of g r a m m a r

In arguing for his relativity thesis Whorf places particular emphasis on

linguistic background phenomena, i.e. on the basic grammatical structures
of language and their role in determining experience. For the most part
we do not notice these basic structures, because we take them for granted
as obvious and use them continuously. Whorf says: “that the phenomena
of a language are to its own speakers largely of a background character
and so are outside the critical consciousness and control of the speaker.”1
“ ... we all hold an illusion about talking, an illusion that talking is quite
untrammeled and spontaneous and merely ‘expresses’ whatever we wish
to have it express. This illusory appearance results from the fact that the
obligatory phenomena within the apparently free flow of talk are so
completely automatic that speaker and listener are bound unconsciously
as though in the grip of a law of nature. The phenomena of language are
background phenomena, of which the talkers are unaware or, at the most,
very dimly aware - as they are of the motes of dust in the air of a room, though
the linguistic phenomena govern the talkers more as gravitation than as dust
would. These automatic, involuntary patterns of language are not the same
for all men but are specific for each language and constitute the formalized
side of the language, or its ‘grammar’ - a term that includes much more
than the grammar we learned in the textbooks of our school days.” 2
We shall show in what follows, however, that the influence of grammar
on experience is much more difficult to confirm than is that of vocabulary.
Out of the many examples of the diversity of basic linguistic forms that
Whorf gives, we cite three:
(1) When we start out from our own language, the categories sub-
stantive - adjective - verb appear to us to be fundamental grammatical
categories, and we customarily associate those word types with basic
ontological categories, by distinguishing between things (or persons),
attributes (or states) and activities (or processes). These ontological dis­
tinctions, we say, produce the distinctions among types of words and
therefore we expect those same distinctions to be expressed in every
language. But that is not the case. In the language of the Hopi Indians,
for example, all of the predicates by which events of short duration are
expressed are verbs. Thus the substantives ‘lightning’, ‘cloud’, ‘flame’,
‘meteor’, ‘smoke cloud’, etc., do not exist, only verbs corresponding to
262 CHAPTER IV

them. What is said, for example, is not ‘The flames are making a bright
light’, but ‘It’s burning brightly’, and so on. In Nootka, on the other hand,
all predicates take the form of verbs.3 According to Whorf that shows
that nature itself is not divided into things, states and processes, but that
we have imposed those distinctions on the world with our language, so
that those categories are only forms of interpreting reality.4
Similarly, for us the distinction between class terms and substance
terms is essential and we also distinguish between things and the substance
of which they are composed. But the fact that in Hopi, for example, sub­
stance terms are treated like class terms - so that depending on the con­
text one can say, for example, ‘a water’ for ‘a glass of water’ or ‘a lake’,
or ‘a gold’ for ‘a golden object’ - shows that this distinction, too, is only
induced by language.5
The fact is that the distinctions substance - attribute - verb on the one
hand and class terms and substance terms on the other prove not to be
fundamental in the logical analysis of grammar nor do those distinctions
actually follow any clear criterion: One can just as well say ‘Lightning
flashes’ as ‘It lightens [Es blitzt]\ ‘A strong wind is blowing’ as ‘It is
storming’, and so on. And the expressions ‘golden object’ or ‘made of
gold’ as class terms accomplish the same thing as ‘gold’ as a substance
term. However, as for the most part we use a substantive to express the
complex of basic enduring features that belong to an object existing over
a long period of time and an adjective, on the other hand, for particular
attributes or states, often accidental or momentary, the use of a substantive
instead of an adjective suggests a different interpretation: If instead of
saying ‘This rose is red’ you say ‘This rose has a share of redness’, that
can be understood just as well as the former case. But that the latter
formulation suggests another meaning is clearly shown by the philosophi­
cal questions that have been joined to that formulation.6 If you speak not
of temporal relationships (‘earlier than’, ‘simultaneous with’) but of ‘the
time’, and if you say, e.g. ‘The time has gone by’ instead of ‘It is late’,
these changed locutions raise entirely new problems.7 That is to say,
despite their equivalence in particular cases, the use of substantives,
adjectives or verbs still commonly suggests a different interpretation.
(2) Whorf contrasts the time concept as it is expressed in our temporal
adverbs and prepositions and the tense forms of verbs, and the idea of an
‘objective, steadily flowing time’, with the time concept of the Hopi. He
 LANGUAGE AND REALITY 263

writes: “After long and careful study and analysis, the Hopi language is
seen to contain no words, grammatical forms, constructions or expres­
sions that refer directly to what we call ‘time’ or to past, present, or future,
or to enduring or lasting.” 8 According to Whorf, the Hopi have only three
basic forms of statement. One form refers to what is objectively real or
realized, whether in the present or the past, another to the subject realm,
to wishes, intentions, expectations and ideas, and the third form is the
form of statements that are universally and timelessly true. Statements
in the future are rendered by statements of the second type, and so have
to do with expectations and intentions of a subject concerning something
that is yet to be realized. Statements about the past that no longer have
any connection with the present, and so are in the preterite, are rendered
by means of statements of the first type with an adjunct that characterizes
them as from memory. Thus according to Whorf our division of time
has only a very indirect counterpart in the Hopi language. There is only
the difference between facts and expectations as viewed by the individual
subject and between facts that refer to what is experienced in the present
and those for which such a connection exists only by way of recollection.
Therefore Whorf says that the Hopi lack a time category as well as time
expressions corresponding to our own and that therefore the Hopi’s
world picture is basically different from ours.
Unfortunately W horf’s statements concerning the distinction between
the realms of objective and subjective phenomena are not extensive enough
to permit any very precise idea of what it is. What is to be regretted most
of all is that Whorf does not even succeed in characterizing precisely the
time concept on which the Indoeuropean languages, or our world picture
is based. When he speaks of a “kinetic one-dimensional uniformly and
perpetually flowing time” 9 that is anything but illuminating, so that one
asks oneself the question whether ‘Hopi time’ was really defined precisely,
if there was no such success even for ‘European time’.
Add to this the fact that the function of time cannot be represented
simply by way of reference to physical time even for European languages.
Weinrich in particular showed that in [64].10 Although an interesting
problem has been laid out here and interesting points made, it still does
not provide any firm support for the linguistic relativity principle.
(3) The third example of Whorf’s illustrations of his thesis concerns the
fundamental structure of European languages, namely the subject-pred­
264 CHAPTER IV

icate structure of sentences. According to Whorf every language carries

out an “artificial chopping up of the continuous spread and flow of
existence in a different way,” 11 in that whenever something is specified in
language certain events, things and states are isolated out of the stream
of sense-data, and the subject-predicate structure of the European lan­
guages indicates a particular form of this organization.
M. Black has rightly criticized this formulation:12 Reality is not ‘sliced
up’ in being formulated in language, it is only that partial aspects come
to be represented. Furthermore, according to Whorf there is no ‘reality
in itself’ at all, but what is real is always relative to a language. And finally,
the ‘stream of sense-data’, ‘the welter of sensations’ is itself an abstraction
from our already organized experiences.
But this can be said: In our experiences what is experienced is always
interpreted in some way, and what Whorf says is that the subject-predicate
structure is a specific form of interpretation. According to Whorf, we
first of all isolate particular events and interpret those events in the form
that there are one or more things (or persons) which are involved in an
activity (or a state). Even when to do so is scarcely adequate considered
in terms of the situation, we apply this schema: We speak of things where
we have no relatively firmly delineated bodies such as stones, tables or
creatures before us and so lack any sharp criteria of identity, as the words
‘heaven’, ‘wave’, ‘cloud’, ‘marsh’, ‘plain’, and so on will witness, and we
consistently understand all events in such a way that ‘things’ enter in as
agents or bearers of states and processes. Even sentences such as ‘It is
raining’, which are constructed with impersonal verbs and have no genuine
subject, have at least the grammatical form of subject-predicate sentences.
The Hopi, in contrast, can use verbs without subjects, i.e. grammatically
impersonal constructions are possible, too, and the set of impersonal
sentences is essentially larger as against the European languages.13 But
the Nootka language, according to Whorf, lacks subjects and predicates
entirely: there complex sentences are formed by means of suffixes from
simple sentences serving as basic constants of the language.14
Without a more exact knowledge of Nootka grammar than W horf’s
statements provide, no more exact representation of the views that are
basic to that language can be given. This point - the linguistic analysis of
events into things and states - is so important, however, that we will come
back to it again in more detail below.
 LANGUAGE AND REALITY 265

These three examples of the way in which Whorf supports his relativity
thesis give an impression of the arguments with which the problem of the
influence of language on experience is discussed in linguistics. They offer
a series of quite interesting suggestions and pointers, but they are incapa­
ble of lending support to the thesis adequately in all particulars. That also
becomes clear from the discussion of the following criticisms, which do
not concern W horf’s particular examples, but his argument’s basic ideas.
One objection takes this form: If Whorf were correct in saying that our
language fixes our interpretation of the world, then we would be totally
incapable of establishing that there are languages with radically different
interpretation schemata and ways of understanding the world, as Whorf
says we can, for we would remain always imprisoned within the world view
of our own language and would therefore be totally unable to grasp and fully
work out an understanding of a radically different language. In particular,
every attempt to translate the sentences of that language into our own
would lead to a totally inadequate result. A translation of such languages
would then either be false or would contradict the relativity thesis.
There is actually a serious difficulty here, but one which both Humboldt
and Whorf have seen. Whorf writes as follows: “In order to describe the
structure of the universe according to the Hopi, it is necessary to attempt
- insofar as it is possible - to make explicit this metaphysics, properly de-
scribable only in the Hopi language, by means of an approximation ex­
pressed in our own language, somewhat inadequately it is true, yet by
availing ourselves of such concepts as we have worked up into relative
consonance with the system underlying the Hopi view of the universe.” 15
And: “They [these abstractions] are not, as far as I can consciously avoid
it, projections of other systems upon the Hopi language and culture made
by me in my attempt at an objective analysis.” 16
This difficulty also becomes clear in all of Whorf’s attempts to express
in the English language the peculiar nature of the quite different Indian
languages which are supposed to provide him with the primary examples
in support of his thesis. Here the reader who does not know those lan­
guages actually has the feeling most of the time that they are matters of
more or less superficial syntactical differences, but not basic differences
of language form and interpretation of the world.
For example, when Whorf says that the Nootka do not have any
knowledge of the subject-predicate structure, but can form sentences
266 CHAPTER IV

only out of sentences and then illustrates this claim by citing a Nootka
sentence that means the same as ‘He invites people to a feast’ and reads in
literal translation ‘boil-ed-eat-ers-go-for-he-does’, that is not at all con­
vincing, since even in the literal translation what are being combined into
a sentence are proper names and predicates.17
But the reason for this is that single examples are not sufficient to
grasp the peculiar nature of a foreign language. For that a thorough study
of the language, especially of its grammar, is indispensable. For a single
linguistic phenomenon, a grammatical form or a word field, as Humboldt
emphasized, should always be viewed against the background of the
entire language and also against the background of the whole culture to
which the language belongs and its world picture. Really very different
languages can never be compared in their detail, in the structure of single
sentences, but only as a whole.
But this difficulty does not imply any fundamental objection to the
relativity thesis. For our language does not compel - as Whorf sometimes
formulates it overly sharply - a conception of the world upon us, but
suggests it to us and we are able - by learning a new, quite different
language, for example - to open up other conceptions of the world and so
come to know the relativity of our own conceptions.
The following objection weighs more heavily: A language’s basic
grammatical structures remain constant as it develops in an essentially
higher degree than its vocabulary, and they survive for long periods of
time, during which the culture, the apprehension of the world and the
interpretation of reality often change radically. The influence of those
linguistic structures on experience relates, then, either to quite fundamen­
tal components of the conception of the world, which remain invariant
during the change in the culture - as Whorf does say - or it has only an
indirect and limited effect, i.e. there are no unambiguously well-defined
forms of interpretation corresponding to the grammatical forms.
The three examples we introduced above and with which Whorf wishes
to confirm the role of grammar in experience, prove only such an indirect
influence of grammar, for example. We have already seen above that in
German [or English] substantives, adjectives or verbs can often be re­
placed by expressions belonging to other categories without any great
shift in the statement’s sense. Whether one says ‘The leaves are yellowing’
or ‘The leaves are (are becoming) yellow’, ‘Fritz lives in Munich’ or
 LANGUAGE AND REALITY 267

‘Fritz is a resident of Munich’ does not make such a great difference

that one could speak of diverse interpretations in this case. The fact is
only that in most cases we have synonymous adjectives [or descriptive
nouns] for the verbs and that there are always nuances of meaning in
which the substitutions, where they are possible, differ. The efforts, by
L. Weisgerber18 for example, to catch these differences generally and
exactly show, however, how hard it is to arrive at binding assertions here.
What is to be said of the time concept is that our time concept nowadays
is to a great extent that of physics, although we express it linguistically
in forms in which the subjective aspect they originally contained is still
clearly demonstrable. And the subject-predicate structure of such sen­
tences as ‘It is raining’ does not mislead us into bringing a thing which
performs an activity or in which a process is realized into the picture.
In addition, there is the following: Just as, when we consider vocabula­
ries, comparing different word fields often makes for difficulties, because
the semantic regions related to them do not coincide - we had brought
that out above in Trier’s example for the region of the intellect - so
comparing grammatical forms of different languages makes for serious
difficulties: If in doing so one starts out from syntactically determinate
forms related by descent from a common origin, then one discovers that
the same forms often have quite different semantic functions - as will
become quite clear below in the case of the passive. On the other hand, if
the set of forms is defined in terms of their semantic function, or of the
results they effect, in Weisgerber’s sense, then it will be trivial that
semantically equivalent forms bring about the same result for the inter­
pretation of experience. Grammatical forms, too, are to be analyzed in
each case with the background of the whole language behind them. It is
clear, then, that in some respects the functions of syntactically determinate
forms are changed very much during the course of a language’s develop­
ment, just as the meaning of particular words can change. There is, then,
no direct influence of syntactic forms on our experience, because the
interpretation of them can shift, no more than there is an influence of
syntactic word patterns on our experience. In the one case there is only
the influence of the concepts as interpreted words and in the other there
is only the influence of conceptual forms as interpreted grammatical
forms. In particular, a considerable phase shift between the semantic and
the syntactic change in these forms can come into play.
2 68 CHAPTER IV

These critical remarks are no arguments against the relativity thesis

itself; they only restrict it as against many of Whorf’s assertions and point
out the difficulties in proving it, as Humboldt had already seen, more
clearly than Whorf.
Because this thesis has such an import and interest just for philosophy,
we will discuss somewhat more extensively in what follows two cases of
grammatical interpretation schemata that may be especially important
for epistemology: the interpretation schema active-passive and the schema
subject-predicate.
In discussing the interpretation schemata of active and passive sentences
we shall draw upon the studies of L. Weisgerber.19 Weisgerber begins
with the fact that the range of forms of passive sentences encompasses
significantly more than sentences with verbs constructed with passive
endings. Along with the simple passive forms that can be constructed
from transitive verbs of action especially, where the accusative object can
appear without restriction as the subject in the related passive forms
(as in ‘I praise you’ - ‘You are praised by me’), within the range of the
passive there are also active verbs that can be substituted for passive
forms (‘He was given a cuff on the ear’ - ‘He got a cuff on the ear’, ‘The
bill was read’ - ‘The bill came to a reading’, [and in German only] ‘Der
Ring wurde gefunden’ - ‘Der Ring hat sich gefunden\ ‘Die Tur wurde
geoffnef - ‘Die Tur offnetesich\ and so on). Furthermore, the impersonal
passive, which can be constructed with intransitive verbs as well Fritz
lachf - ‘Es wird gelachf) is very close to the impersonal active QEs wird
getanzt’ - ‘Man tanzf [‘A ballet is danced’ - ‘One dances a ballet’], *Es
wird iiber Apollo 13 gesprocherf - ‘Man spricht iiber Apollo 13\ 6Es wird
geklopff - ‘Man klopft’). The converse also holds: Where no ‘one’-form
[6marf-Form] of the active is possible, neither is an impersonal passive
possible (neither 6Man bliiht’ nor 6Es wird gebliihf). Correspondingly,
event verbs used only in the third person with a non-personal subject,
such as ‘happen’, ‘gelingen\ ‘misslingen\ ‘occur’, ‘take place’, etc., do
not form a passive either. Finally, we note in considering forms that not
all transitive verbs form a passive, e.g. ‘have’, ‘become’, etc. [a number of
examples from the German are omitted.]
If one now attempts to define the semantic function of the passive, the
obvious interpretation on first glance - since we usually think first of all
of transitive verbs of action in connection with the passive - is that the
 LANGUAGE AND REALITY 269

passive sentence is the converse in subject-object relationship of the active

sentence, i.e. that the sentence ‘Hans was struck by Fritz’ has essentially
the same content as ‘Fritz struck Hans’, only that the event was seen more
from Hans’ standpoint in the first case and more from Fritz’s in the
second.20 J. Wackernagel says in this vein: “The passive has been cor­
rectly described as a linguistic luxury, because the passive sentence repre­
sents nothing but the converse of the normal active sentence.” 21 But this
interpretation, expressed in the designation ‘passive’, will not do justice
to all of the linguistic data and grammatical facts that were pointed out
above.
Weisgerber opposed to this conception the following account. On the
whole, it is quite convincing: He distinguishes between two types of sen­
tences: The first type interprets an event or state as the act of an agent-
subject, as an occurrence brought about or supported by such a subject or
as a state caused by him. In particular, those active sentences that have
a definite personal or non-personal subject belong to this type, such as
‘Fritz is laughing’, ‘Fritz is watching Hans’, ‘The ball rolls’, etc. The
second type does not interpret an event or state in this fashion, i.e. even
where a definite subject or object is involved, it does not interpret it as
agent-subject. Here is where impersonal active sentences such as ‘It is
raining’, or ‘One sings the Marseillaise’, and active sentences such as
‘A mishap occurred’ or ‘Fritz owns (has, gets) a bicycle’, especially belong.
If sentences of the first sentence type are designated as action sentences
and those of the second type as action-free sentences, then the action
sentence is typical for Indogermanic. In German, action sentences pre­
dominate in a ratio of about 10:1, i.e. we have at our disposal many
more linguistic forms for interpreting an event along the lines of an action
event. Now in Weisgerber’s view the passive, which was developed only
relatively late in the various Indogermanic languages, is a form to com­
pensate for this predominance of action sentences, a form for transform­
ing action sentences into action-free sentences, so that the predominantly
active stock of words can be used for expressing action-free sentences as
well. For this reason the passive is constructed on such verbs as function
in action sentences, and so on verbs of action especially, but not on forms
that are only formally active, which do not express any action, such as
‘happen’, ‘have’, etc. Further, as a consequence the passive stands very
close to the impersonal active sentences: ‘Man lachf coincides in meaning
270 CHAPTER IV

to a broad extent with 'Es wird gelacht'. Weisgerber sees the transforma­
tion of an active sentence into a passive one in the following way: ‘Fritz
lacht' -> 'Man lacht' -> 'Es wird gelacht' [literally, ‘Fritz is laughing ->
-> ‘One is laughing’ -» ‘It is laughed’] ,... etc. I.e., the definite agent-sub­
ject is eliminated and in the passive even that indefinite reference to an
agent-subject often still concealed in the word ‘man' [‘one’] is set aside.22
In connection with sentences with transitive verbs the transition to
the passive is then to be understood as follows: ‘Fritz schlagt Hans' ->
-> 'Man schlagt Hans' -> 'Hans wird von Fritz geschlagen' [literally,
‘Fritz strikes Hans’ -►‘One strikes Hans’ -►‘Hans is struck by Fritz’].
I.e. in this case one starts out from an event interpreted as an act of the
agent-subject Fritz directed at Hans, from there one goes to an agent-free
interpretation of the event, and from there to a specification of the event
as an event caused by Fritz. I.e. ‘Fritz’ functions in the passive sentence as
part of a specification of the circumstances (like ‘stick’ in ‘Hans was struck
with a stick’). The addition of ‘by Fritz’ is only an adverbial supplement
to the sentence, then, which in citing a subject causing has the practical
effect of rescinding the elimination of the agent-subject that was achieved
by means of the passive. This interpretation shows how difficult it is to
understand the passive from this case, according to Weisgerber a typical
case of a passive sentence.
Now this interpretation of Weisgerber’s has much in its favor, since
in particular it explains the near relationship between the active and the
passive impersonal construction and about covers the range of forms
sketched above. Of course the expressions ‘action sentence’ and ‘action-
free sentence’ certainly require still more extensive and careful definition.
What is the outcome of this study for the relativity thesis now? In the
active and the passive sentence - more generally, in the action sentence
and the action-free sentence - it points to two forms of interpretation of
our experience. These forms of interpretation are not entirely bound up
with the related linguistic forms, however, in such a way that whenever
we use an action sentence, for example, we assume a (human, animal
or superhuman) subject and interpret what happens as the act of that
subject. Thus we say ‘The one ball bumps into the other’ without regard­
ing the ball as an active subject. But the language form nevertheless often
suggests such an interpretation. That is apparent, for example, in the
old way of talking in physics about the 'vis viva\ the ‘live force’ (kinetic
 LANGUAGE AND REALITY 271

energy) within the first ball, with which it works on the second, or in the
interpretation of biological phenomena by assuming an entelechy. To
that extent the strong predominance of the world of active forms in
Indogermanic generally is a clear indication of a world view clothed in
linguistic tradition, according to which reality is represented predomi­
nantly as something going on that is dominated and determined by agent-
subjects.
We said above that there can be a phase shift between linguistic forms
and their interpretation, so that the forms of expression can lose their
original meaning. Something of that sort is certainly true of action sen­
tences insofar as they have no human or animal subjects, since nowadays
we view the world more as passive, i.e. in the form of states and changes
of state, than in the form of subject-determined events. That complicates
the proof of a connection between language form and resulting interpre­
tation very much and it is probably responsible for the fact that Weis-
gerber’s account of the function of active and passive does not apply in
many particular instances. On the whole and especially in historical
perspective, however, that connection is very clear.
Weisgerber’s distinctions become still clearer when we view them against
the background of the language as a whole and the related world of
ideas, in contrasting our language world with others. H. Hartmann in
[54] has studied the range of passive forms in Celtic and Aryan languages,
especially Irish, in which the passive is especially pronounced, and has
placed them in relationship to the world picture and the religious ideas
of the culture that stands behind these languages. In view of his observa­
tions, which are extraordinarily informative in their very detail, he too
reaches the conclusion that the construction and use of passive forms for
expressing facts that have only an active formulation in German is con­
nected with the fact that the Indogermanic idea of the freely and autono­
mously active subject as the origin and bearer of an event withdraws
behind the idea of man as a part of the cosmos, of a total cosmic power
that works in him and through him.23 That we have to do here with two
different forms of interpretation of the event also becomes especially
clear from the rules for applying active and passive in Irish. For example,
Hartmann says this about the expression of following an order: “The
passive... has a place only when someone finds himself under some com­
pulsion and sees that he is forced to give way to it. If his obstinate rejec­
272 CHAPTER IV

tion of the demand is supposed to be expressed nevertheless, or at least

the freedom of his own decision is not wrested away from him entirely,
then the active is used for this purpose.” 24

The subject-predicate schema our sentences have is the most general and
consequently in Whorf’s sense also the most powerful interpretation
schema in our language.
By far the largest number of simple sentences in our language have
the subject-predicate-(object-) structure, i.e. they consist of proper names,
which stand for definite objects (be they concrete or abstract, things,
persons or other subjects), and a predicate, which ascribes to these objects
an attribute (be they qualities, relations, states, processes or actions).
The objects are taken to be more or less enduring givens, which maintain
their identity throughout their various momentary states, as their attri­
butes change, things which subsist beneath the phenomena in which the
event is realized and which function as bearers of the states.
We ordinarily interpret all events and facts along these lines in an
object-attribute structure. That this interpretation is not self-evidently
true, that we cannot naively project this predication schema into ontology,
already becomes clear in the cases in which we use this linguistic form
even though there can be no talk of identifiable objects. In this vein, we
say, for example ‘The wind is blowing’, ‘The sky is blue’, ‘The ocean
roars’, ‘The rain is falling’, ‘His will is inflexible’, ‘His sorrow is profound’,
etc.
Besides that, in verbs used impersonally we have a means of represent­
ing what happens in a way other than the object-attribute form: The
sentences: ‘It is raining’, ‘It is freezing’, ‘It is windy’, 6Es klopft\ 6Es wird
g eta n zt\‘Es wirdgepflugt\etc., have, to be sure, a subject-predicate struc­
ture in terms of formal grammar, but it is quite clear in these cases that
the pronoun ‘it’ [‘2&’] does not stand for a definite object, that we cannot
interpret these sentences as statements about objects. In the case of sen­
tences with impersonal verbs it makes no sense to ask after something
that supports what happens: ‘Who is raining?’. ‘What is freezing?’; and
that is equally true of the sentences ‘It is windy’, ‘It is cold’, in which the
adjectives are used impersonally. In the case of the other sentences it does
of course make sense to ask about a subject (6Wer klopftT) [‘Who
knocks?’], 6Wer tanztT [‘Who is dancing?’], ‘Wer pfliigtT [‘Who is
 LANGUAGE AND REALITY 273

plowing?’], but it is not a question of existence sentences (‘There is some­

one who knocks’ etc.); on the contrary, the addition of a subject acting
would have the character of an adverbial supplementation here: In the
sentence ‘Es klopff an event is depicted in which there is no mention of
persons or things, and that event is more closely described by citing the
agent, when one says, for example, ‘Es wird geklopft von Fritz'.25
The possibility of forming impersonal sentences shows that not every­
thing that happens has to be interpreted in terms of the object-attribute
schema. The fact that we make only relatively infrequent use of this pos­
sibility, that formally speaking even impersonal sentences have the sub-
ject-predicate-(object-) structure and that we use substantives even where
we are not concerned with identifiable objects proves how overpowering
that interpretation is.
We have seen above that Whorf assumes predication in this sense to
be unknown to the Nootka language. That would be very significant
further evidence that the object-attribute schema has to do with a form
of interpretation but not with a pre-existent ontological structure.26 Still,
since Whorf’s claim can only be tested and rendered precise by someone
who has a mastery of that language, we will reflect a bit on how a simple
language that has no subject-predicate-(object-) structure might look.
The basic units of this descriptive language - let us call it T - which
we are supposed to be able to use for communication just as well as our
own language - let us call it S - are simple declarative sentences. Now
while such sentences are analyzed in S into proper names and predicates,
and so have the logical form F(a), G(a, b), or H(a, b, c), for example, that
analysis is not supposed to be possible in T. In terms of content, then, the
sentences of T would correspond to impersonal sentences of S and so say,
e.g., ‘It is raining’, ‘Es wird gegessen\ ‘Es wird gekampft\ ‘It is cold’, etc.
But now it is not the case that a sentence constant of T can simply be
coordinated with every simple sentence of S , otherwise special basic
constants would have to be introduced for all of the facts that can be
represented in S in the form F{a), F(b),... or as G(a, b), G(b, a),..., so
that T would be much too uneconomical. Where S, for example, gets
along with five constants for expressing the 64 sentences that can be con­
structed with one three-place predicate constant and four object constants,
T would require 64 to do that.
But if we let sentence constants of T correspond to predicate constants
274 CHAPTER IV

of S, in such a way, say, that to a predicate constant F(x) there corre­

sponds in T a sentence with the content ‘There is something F-ish there’,
or ‘F-ness is here’ - so that to ‘is red’, for example, there corresponds
‘Redness occurs’, to the predicate ‘lacht9the sentence ‘Es wird gelachf -
and if we let sentence constants of T correspond to object constants of S ,
such that a sentence with the content ‘a participates’ (is present), corre­
sponds to the object constant V , then simple sentences in S of the form
‘a has the property F 9 can be translated into ‘F-ness is present and a
participates’. To be sure, conjunction must be understood in a stronger
sense than that of sentential logic, so that the sentences bound up together
by this ‘and’ express specific accounts of the same fact.27
In this fashion the simple sentences of S with n constants in the form
cited can be represented by sentences in T with the same number of con­
stants.
It becomes more difficult when we consider sentences with polyadic
predicates in S, e.g. the sentence F(a, b), which could mean the same,
say, as ‘Fritz hits Hans’. Since conjunction is commutative, one can not
simply construct in its place in T the sentence ‘There is hitting and Fritz
participates and Hans participates’. Instead, in this case we must distin­
guish between the subject and the object of the event, e.g. divide the sen­
tence ‘a participates’ into ‘a participates actively’ and ‘a participates pas­
sively’ and then say ‘There is hitting and Fritz participates actively and
Hans participates passively’.
The inferiority of language T in comparison with S is already becoming
clear at this point. It becomes still more so when instrumental participa­
tion, etc., is also introduced.
Sentences that express quantifications can then be introduced, too, for
example, ‘All participate’ or ‘Some participate’ and so analogies to simple
sentences of predicate logic can be constructed. Constructing analogues
in T to multiply quantified sentences in S, on the other hand, would
create great difficulties. S is thus considerably richer expressively than
is T, in which sentences consist of combinations of elementary sentences.
The language of set theory, on the other hand, is an example of a lan­
guage not inferior to S in wealth of expression and in which there are no
predicate constants, but only object constants and logical operators, in
which one might say, then, a different form of interpretation is expressed
than in S. The disadvantage of such linguistic constructions naturally
 LANGUAGE AND REALITY 275

lies in the fact that we still interpret them with the assistance of our own
language, so that they can not lend support to the thesis that other kinds
of grammatical forms often express a different sort of interpretation.
They may be suited, however, to make somewhat clearer the possibility
of such other kinds of interpretation forms and to get a better view of
the distinctive character of the predication schema and what it does.
In the case of the subject-predicate-(object-) structure of sentences, the
role of language in experience consists in the fact that with language we
learn to systematize and represent our experiences in the schema object-
attribute. Thus we learn, for example, to describe a fact as ‘The rose is
red’, and not as ‘The redness roses’, or ‘Redness is present and the rose
participates in it’. Such descriptions would imply an entirely different
systematization of phenomena, an entirely different ontology. By virtue
of the fact that there is available a certain stock of linguistic expressions
and forms that serve one particular interpretation and ordering of the
phenomena, that simple and stereotyped expressions and forms for other
interpretations are not available, language establishes the shape and form
of our apprehension and interpretation of what is experienced, within a
certain compass. For so long as we get along with the linguistic instru­
ments available, there is no reason for us to look around for new instru­
ments. A modification of the basic forms of our language would carry
with it such far-reaching consequences for our conceptual apparatus and
our beliefs about the world that it would be something to be realized only
after a long look and only by the linguistic community as a whole.
On this basis, the object-attribute schema, stand our logic and our
concept and theory construction in the sciences. To that extent it is not
an exaggeration to say that e.g. our mathematics and natural science and
with them our world picture that bears the imprint of natural science have
this basic linguistic foundation as their presupposition. In this vein
Whorf says: “Thus the world view of modern science arises by higher
specialization of the basic grammar of the Western Indo-European lan­
guages. Science of course was not caused by this grammar; it was simply
colored by it.” 28
To be sure this linguistic basis is itself modified: Our logical concep­
tions once again shape our understanding of our language. Logical gram­
mar is not our language’s ‘natural’ grammar, but nevertheless it rests upon
the same predication schema that is basic to it as well.
276 CHAPTER IV

NOTES
1 Whorf [56], p. 211.
2 Whorf [56], p. 221.
3 This claim of Whorf’s does not quite fit his other statements, of course, which say
that all sentential components in these languages are sentences themselves, combined by
sentential operators (as sentence suffixes).
4 See Whorf [56], p. 215f.
5 See Whorf [56], p. 141f.
6 Take account, say, of the Platonic theory of Ideas, in which what corresponds to the
shift from adjective to substantive is a shift from an attribute to an object, the Idea.
7 See also Wittgenstein’s statements about time in [53], 90, as well as Waismann’s
amusing remarks in [56], p. 348fF. - Wittgenstein’s battle against the ‘bewitchment of
our intelligence by our language’ is definitely also a struggle against the naive concep­
tion o f language that projects the surface structures of grammar onto reality without
further thought.
8 Whorf [56], p. 57.
9 Whorf [56], p. 59.
10 See also Weisgerber [62], V. II, p. 327, as well as note 30 to Chapter III.2.
11 Whorf [56], p. 253. On this point see also p. 41f. and 213f.
12 See Black [59], p. 231 and 236f.
13 See Whorf [56], p. 253f., 262.
14 See Whorf [56], p. 241 f. on this, as well as what is said about the sentence structure
of Japanese in P. Hartmann [52] and Weisgerber [62], V. II, p. 353f.
15 Whorf [56], p. 58.
16 Whorf [56], p. 59. - See also Fearing [54], p. 53fF., as well as Kohler [37].
17 See Whorf [56], p. 241f.
18 See e.g. Weisgerber [62], V. II, p. 300fF. - See also Snell [52], P. Hartmann [56] and
Brinkmann [50].
19 See Weisgerber [63b] and [63a], p. 232ff.
20 That would then correspond to another topicalisation, i.e. to another division of the
state of affairs into topic and comment, into object and something said about that
object.
21 Wackernagel [24], V. I, p. 135.
22 Along similar lines H. Hartmann says: “If the person involved in the action is un­
known or unimportant or for some special reason is kept in the background, then the
passive comes into the picture; on the other hand, if someone in the present or the past
is being described as active or as having an effect, then as a rule we make use of the
active form.” ([54], p. 12)
23 See H. Hartmann [54], pp. 33, 39 and 62.
24 H. Hartmann [54], p. 29.
25 The ‘one [man]9in impersonal sentences can also be often understood as the existen­
tial or as the universal quantifier, to be sure.
26 P. Hartmann has called attention to a language type in Japanese as in other East
Asian languages in which sentences do not have any subject-predicate structure. He
calls these languages ‘referring [referierendeY and says: “Event designation is always
the main content of the Japanese sentence. As such, it is what appears as the only
sentence component in the nominative representation. All other members o f the sen­
tence can be looked upon as attributes of this event designation.” ([52], p. 96) I.e.
mention of the subject, like mention of the object is an expendable supplement to the
 LANGUAGE AND REALITY 111

simple sentence as a pure event sentence. This ‘subjectlessness [Subjektlosigkeit]\

perhaps better: this secondary role of the subject, is a property of all East Asian
languages, according to Hartmann. The English sentence ‘The man sees the mountain’
takes, for example, the form in Japanese of ‘the man’s the-mountain-seeing’, where
‘seeing’ (in the sense of ‘there is seeing’) is the event designation, which can also stand
by itself. This event is first supplementarily described as a ‘mountain-seeing’ (‘A
mountain is seen’) and then as a mountain-seeing by the man (‘A mountain is seen by
the man’). (See [52], p. 23.) - For Hartmann a different sort of world view is expressed
by this different sort of sentence construction.
27 The English ‘and’ [German ‘w/k/ ’], as against the logical conjunction, also includes a
relationship of content or an affinity between the sentences connected.
28 Whorf [56], p. 221.
278 CHAPTER IV

4. T h e e p i s t e m o l o g i c a l p r o b l e m a t i c of
THE R E L A T IV IT Y THESIS

So far in this chapter we have discussed the thesis as to the role of language
in experience primarily in terms of the formulations and arguments laid
out for it in the general theory of language. We will now place this thesis
within our specifically philosophical discussion of the problem of meaning
and discuss its relevance to epistemology.
The relativity thesis can be divided into two claims:
(1) There is a correlation between language and world view of such a
kind that typical and profound differences in world view correspond to
typical and profound differences in linguistic forms.
(2) The world view depends on the language; the forms of interpreting
experience are imparted by language.
While the second claim says that the language we speak influences the
forms of our experience, and so implies an effect of language on experience,
the first claim says only that there are correspondences between lin­
guistic forms and forms of world view, without stating whether the former
is determined by the latter or the latter by the former. The first claim is
weaker than the second, for there can be a correlation between language
and world view not based upon any influence of language on experience; but
one can not assume that language has that sort of influence if there are no
distinct differences inexperience corresponding to even typical and profound
differences in linguistic forms. Only the second claim expresses the full
content of the relativity thesis as Humboldt, Whorf and Sapir understood it.
The empirical work of comparative studies of language and of the
history of language initially discloses nothing but correlations between
forms of language and forms of experience. In order to establish an in­
fluence above and beyond that, one can either bring to bear empirical-
psychological studies that show, for example, that certain forms of ex­
perience depend on linguistic parameters and how they do1, or one can
rely on studies in the philosophy of language. Such arguments from the
philosophy of language emerge directly out of the discussion of Wittgen­
stein’s ideas in the ‘Philosophical Investigations’ in Chapter II.4. And a
way of founding the relativity thesis on the philosophy of language is to
be expected only within the confines of a pragmatic theory of meaning
with a Wittgensteinian stamp.
 LANGUAGE AND REALITY 279

From the standpoint of realistic semantics what one would say is this:
If words and grammatical forms become meaningful by our assigning to
them as their meanings objects, concepts or propositions that are inde­
pendent of language, then the thesis of the role of language in experience
is false. It is false in a trivial way if what we understand by language are
only the vocal or written linguistic expressions, for these expressions surely
have no influence on our experience by themselves. Nor can an influence
of a purely syntactical character ever be maintained, it can never be any­
thing but an influence exerted by an interpreted meaningful language.
But on the realistic understanding a meaningful expression is an expres­
sion for an entity independent of language. So even if one does not take
concepts and propositions to be ‘objectively real’ attributes and facts
occurring in the world, and experience to be a true copy of reality in
consciousness, but rather understands them in conceptualist terms as ways
of interpreting, still the linguistic expressions do not determine those ways
of interpreting. On the contrary, they are defined as meaningful expres­
sions only in terms of them, as expressions of these ways of interpreting.
First there are the ways of interpreting and then the words and sentences
explicated through them. There is then no influence of a sign on what is
signified.
If the realistic premises of this argument are accepted, one will not be
able to say very much in the way of objection to it.
The thesis that language plays a part in experience has its place only
within the confines of pragmatic meaning theory. Indeed it has already
become clear from the quotations cited in IV. 1 from Humboldt, Sapir
and Whorf that they advocate a pragmatic theory of meaning,2 and that
with them the relativity thesis grows out of that theory of meaning. Only
if the linguistic apparatus is regarded as not just a way of expressing forms
of interpretation, but as defining those forms in the first place, if we learn
the differentiations and specifications in terms of which we articulate
what is experienced, and if language is a means of bringing order into
the world, does it make sense to say that our ways of conceiving and
apprehending are shaped by our language and that different languages
can represent different ways of interpreting [the world]. The pragmatic
theory of meaning, however, not only provides the conditions under which
the relativity thesis can be accepted, but with its doctrine of the mediation
of conceptual determinations by language it even implies that thesis. That
280 CHAPTER IV

there are languages to which distinctly or quite fundamentally different

ways of interpretation correspond, that the differences that hold among
languages are thus not all insignificant for our world view and that the
relativity thesis has something more than just theoretical significance in
principle, is to be proved separately, to be sure, and doing that requires
the empirical work of comparative linguistics.
Now the relativity thesis raises a series of epistemological problems,
which we will discuss in what follows. Since the thesis follows from the
pragmatic theory of meaning, these problems also have to do with that
theory.
Realistic semantics, as a theory of the assignment of objects, concepts
and propositions to linguistic expressions, entails no epistemological
problematic of its own. For the assignments have a conventional character;
there are no objectively correct assignments that could possibly be hidden
from us; the assignments have no influence on what is assigned and it is
presupposed of the entities to be assigned that they antecedently exist.
The question as to which ontological status these entities have, whether
they belong to a world ‘an sich\ i.e. independent of our knowledge, the
physical world of concrete things or an ideal sphere of abstract concepts,
classes and the like, such as Platonism assumes, or whether it is a matter
of things constructed by the intellect, understood conceptually, or of
ideas, and of how these entities are known, is not of great significance
for the basic ideas of realistic semantics itself.
We became acquainted in II. 1.3 with Wittgenstein’s semantics in the
Tractatus as an example of a realistic theory. But Wittgenstein also de­
veloped in the Tractatus an epistemological position which gave E. Stenius
occasion to speak of a transcendental-linguistic approach, of an analogy
in philosophy of language to Kant’s transcendental philosophy in the
‘Critique o f Pure Reason\3 We want to reflect, therefore, on whether
there is not after all some question of an epistemological problematic that
concerns realistic semantics in general.
Wittgenstein depends on the picture theory of language, which we
discussed in II. 1.3, as the support for his statements on epistemology.
We saw there, by way of a quotation from his letter to Bertrand Russell
dated August 8, 1919, that his semantic theory serves him only as a foun­
dation for his central concern in the Tractatus, his distinction between
what can be said by means of sentences and what can only be shown. What
 LANGUAGE AND REALITY 281

can be said or represented by a sentence is the fact, which a (true) sentence

pictures by virtue of its structure and the isomorphy correlation. What
on the contrary can not be represented by sentences are e.g. the corre­
spondences between the syntactic structure of the sentence and the onto­
logical structure of the facts pictured. These structures and corresponden­
ces are shown in the sentence, but they can not be represented by it, for
knowledge of these structures and correlations is the presupposition of our
understanding language at all.4 But what is a presupposition of all under­
standing of language can not be explicated by means of language. Some­
one who does not yet know English cannot learn the language from the
Oxford Dictionary, and no more can anyone say by means of a sentence,
which according to the picture theory has to be grasped as a fact itself,
and which means a fact, what a fact is. It is to no avail in this connection
to switch to a metalanguage wherein the sentences can represent the way
in which the sentences in the object language function, for the metalan­
guage, just as much as the object language and any other language, would
have to presuppose the ontology and the picturing function, but could not
communicate them. When we speak, we are always ‘inside the language’
and we can not somehow observe it from the outside.5
For Wittgenstein it follows from this that the statements in the Tractatus
that are concerned with the ontological structure of reality are also sense­
less, because they attempt to say what cannot be said. But that does not
make them valueless, for they are illuminating and help us see what can
not be said: “My propositions are illuminating by virtue of the fact that
anyone who understands me knows at the end that they are nonsensical,
when by means of them - on them - he has climbed on beyond them.
(He must throw the ladder away, as it were, after he has climbed up it) -
he must surmount these propositions, and then he will see the world
correctly.” 6
With the radicality so distinctive of him, Wittgenstein then draws the
conclusion that a philosophy as a system of theoretical propositions about
the ontological structures of reality and the basic forms of language is not
possible. Even in the Tractatus philosophy for him is not a theoretical
science, but has only maieutic and therapeutic functions.7
Philosophy, in particular the philosophy of language, is naturally not
a natural science, as behaviorism, for example, intends.8 It does not im­
part to us totally new facts about language, it does not intend to under­
282 CHAPTER IV

stand language for the first time by means of language, but it does intend
to analyze systematically the antecedent understanding of language (that
aids us in understanding its own statements as well), so that we can better
comprehend the way in which the language functions. But that its limits
are drawn so narrowly as Wittgenstein says, with his reference point the
semantic picture theory of the Tractatus cannot be maintained in the face
of the fact that this picture theory is a much too restricted theory and
even when limited to simple descriptive sentences a thoroughly dubious
one.
Wittgenstein’s theses are statements about the limits of language’s
capacity for successful representation. That the world is as we judge it
to be in language, or as it turns out to be in our statements, Wittgenstein
does not doubt. When he says: “The limits of my language indicate the
limits of my world,” 9 he does not mean to suggest that the world is always
mediated through language and that for that reason it would make no
sense to speak of the world as it already is prior to any conceptual-
linguistic specification. Just as for Kant the a priori concepts [Verstandes-
begriffe] refer exclusively to possible experience and cannot be applied
beyond experience (in rational metaphysics, for example: in rational
psychology, cosmology or theology as it is represented in the transcen­
dental dialectic) so for Wittgenstein the representational function of lan­
guage relates only to empirical facts. To that extent the analogy holds.
But it does not go so far that language also has a role in the limits of its
applicability, as Kant assumes there is a part played by reason in expe­
rience. At the limits of its applicability language has according to Witt­
genstein a purely pictorial function, so that the title ‘transcendental lin-
gualism’ is not quite adequate for his position in the Tractatus.
The question as to the extent in which we can describe the way language
functions with linguistic means is not, as we have seen in section 2.4.3,
just a problem that applies to realistic semantics in particular, however.
Beyond that, Wittgenstein’s statements about the limits of language can
be separated from the realistic semantics of the Tractatus as well. What
Wittgenstein intends by the proposition cited above, “The limits of my
language indicate the limits of my world”, as the explanation that follows
in the Tractatus shows, is that in a language one can never speak of the
things that belong to the ontology of that language, that the only ques­
tions about existence it makes sense to formulate in it are, as Carnap
 LANGUAGE AND REALITY 283

formulated it later, internal, but not external ones.10 But that is equally
true, as we shall see below, of the pragmatic conception of language, in­
deed it is true in a still stronger sense because of the correlation between
language and world view.
That pragmatic theories of meaning, as compared with realistic, imply
a distinctive and inseparable epistemological component has already be­
come clear in the discussions in the second chapter. There are already
epistemological considerations with which pragmatic criticism attacks the
presuppositions of the realistic semantics, when it urges that concepts can
not be indicated, differentiated or defined without using predicates and
that there is no ‘knowledge’ of concepts not mediated by language. But
the relativity thesis points out the most important problematic for these
theories of meaning: If language is not only a means for the expression
of properties, distinctions and facts discovered beforehand, but defines
them, then one cannot say that the world, i.e. the concrete things, states,
processes, etc., are determinate in and of themselves [an sick], and so
prior to having been established by language in any way. Nor can we say
that we express their antecedent determinateness more or less correctly
in our statements about them. On the contrary, nothing like a determinate
world ever comes to be except by way of linguistic interpretation, the
world is always mediated by language.
Thus the thesis of the part played by language in experience parallels
Kant’s transcendental-philosophical thesis of the role played by reason in
experience: As for Kant, experience arises out of an indefinite ‘welter of
sensations’ only through the application of the categories of the under­
standing and the world as it is in itself remains unknowable in principle,
so experience here comes to be only through the application of language
and its categories and statements about the ‘objective’ world in itself, i.e.
not mediated through linguistic interpretation have no sense.
But while for Kant the concepts of the understanding pertain to a rea­
son as organized universally among all human beings, so that experiential
knowledge is at least intersubjectively valid, according to the relativity
thesis there are fundamentally different, but in principle equally correct
forms of experiential knowledge.
The thesis of the role played by language in experience implies, then,
a still more radical relativization of our knowledge than Kant’s thesis of
the role of reason in experience.
284 CHAPTER IV

The relativity thesis immediately takes on the color of the remarkable

and unusual: The real world on which our knowledge is always directed so
far as its intention is concerned does not exist at all; our knowledge is
under the dominion of our language and never shows us anything except
what the forms of that language permit; there is no correct world picture,
but only various equally warranted world pictures belonging to various
languages. So the relativity thesis appears to be a fundamentally skeptical
thesis.
But those are all false or misleading formulations:
(1) In the first place, we urged in II.4.5 that even from the pragmatic
point of view language cannot be conceived as a game played with ex­
pressions, but must be understood as a system of speech acts. That a game
pla>ed with expressions should determine our knowledge would in fact
be remarkable and unusual. Not at all remarkable and unusual, on the
other hand, is the close connection between speech acts and experiences.
Speech activities encompass very much more than utterances of words.
Declaratives are speech acts, because we can only make them within lin­
guistic forms. But making a declaration does not consist just in the ut­
terance of words, but encompasses all of the features we take on when we
know and judge. So it does not make sense to look upon speech acts and
acts of determining something of a cognitive character as two acts inde­
pendent of each other; they form a unity within which determining some­
thing cognitively, i.e. judging, is always realized in language and within
which when we are asserting declarative sentences we are always judging.
There is then nothing astonishing in saying that we can never make judg­
ments except by means of (linguistic) declaratives and that therefore the
possibilities of making judgments (in language) define our possibilities of
knowledge. Only when language is looked upon as merely an instrument
of expression, when language is freed from its close interrelationship with
our whole behavior and knowledge, does the relativity thesis become
dubious.
The non-trivial character of this thesis resides solely in how closely, (in
contrast to the so very natural semantic realism that is stressed in the
ordinary understanding of language) speaking and apprehending, de­
scribing, proving are bound up with each other; that we learn these activ­
ities, as speech acts, only along with language; that there is a cultural
tradition hidden within what we learn there and how we learn it; that
 LANGUAGE AND REALITY 285

knowledge is not a passive acceptance of distinctions antecedently given,

but an active exercise of an art that is learned; that the picture we make of
the world is the product of a cultural accomplishment in which language
plays an important part.
Along these lines, Johann Gottfried Herder in his Abhandlung iiber den
Ursprung der Sprache [91] explicitly conceives language and reason to be
a unity and the condition of the possibility of all specifically human cultur­
al accomplishments, and he adopts therefore the Aristotelian definition
of man as a £coov A,oyov £%ov.n It is on the horizon of precisely this under­
standing of language that the formula of the world mediated by language
has its place.
(2) Further, we have already urged in Chapter IV.2 that two aspects
must be distinguished in speaking of a ‘dominion language has over our
knowledge’: A culture forms its language together with its understanding
of the world in the course of its encounter with its environment. There is
surely no dominion of language over knowledge here, then, but rather,
on the large view, a unity of language formation and the cognitive process.
The individual, however, receives the languages he uses from the culture
he belongs to. When he makes a picture of the world for himself, he is in
every respect subject to a determinative influence. Again, however, this
influence is not language as an isolated system of signs but the whole cul­
tural tradition, into which language is intimately woven. Furthermore, this
influence, so far as knowledge is concerned, consists not so much in a
system of restrictions as in a system of contrivances placed at the individ­
ual’s disposal to help him to know. Nothing hinders him from adding new
devices or modifying the old. The limits he confronts in acquiring a lan­
guage are not so very much restraints that are forced upon him, but lie
rather in the limited number of possibilities that are offered him and -
this must of course be added - in the difficulty of creating new possibilities
himself.
(3) If there is talk of language determining experience, that is not to be
understood as meaning that language determines what we experience. To
begin with, what language places at our disposal is only the concepts, as
descriptive resources for empirical observations, the predicate ‘red’, for
example, for distinguishing between red and non-red things. Whether a
specific object is red or not is then a matter of observation and is not
prejudiced by language. In a way, then, language is only a net of coordi­
286 CHAPTER IV

nates for specifying the position of a point, a net which does not yet fix
the point’s coordinates, but first makes it possible to specify them.
One can also say that language does settle the validity of some proposi­
tions, the analytic propositions. Even if the boundary between analytic
and synthetic is not a sharp one, as we discussed that point in Chapter
II.3.2, there is nevertheless a large set of unambiguously synthetic sen­
tences, and language says nothing at all about their truth.
Furthermore, language is made for the analysis and systematization
of experiences and the natural languages are such as have proved them­
selves equal to that task. The languages are somewhat like systems of
tools that have proved useful in working some material. Thus they were
not free inventions, but were tested, and to that extent experience also
determines language, as the material determines the choice of tools. For
example, not every conceptual system is equally suited for the formulation
of simple law statements with a high degree of generality. So the intro­
duction of fruitful concepts into the sciences signifies an important ac­
complishment and that accomplishment is confirmed empirically by the
fact that we are successful in giving a simple formulation to a large num­
ber of lawlike connections by means of those concepts.12
Of course it is not only the material that determines the choice of tools,
but also the purpose of working on the material. And in the same way,
what has been called somewhat vaguely, but impressively ‘the will of the
culture [KulturwollenY, in particular the cognitive intent and interest, also
determines the choice of language form, and so there are even entirely
different languages for working on the same ‘materials of experience’.
(4) The relativity thesis is not a negative existential thesis with the
content that there is no one true reality or reality an sich.
Every ‘There is’ - or ‘There is no’ - statement makes sense only within
the boundaries of the language to which it belongs, and its meaning is
defined within the whole system of that language. For every descriptive
language, as a language with which we speak about something, there is an
ontology, i.e. a totality of objects, properties, descriptions, and facts we
use the language to speak about. Whether this ontology is viewed as in­
dependent of language within the realistic understanding of language or
along pragmatic lines as mediated by language plays no role in this
connection. The understanding of this language always relates to this
ontology. It does not make sense, therefore, to deny the existence of the
 LANGUAGE AND REALITY 287

objects presupposed within the language, or to assert the existence of

objects that do not belong to this ontology.
For purposes of illustration and clarification, two examples:
(a) The declarative sentence ‘a does not exist’ - in symbols ~i \Jx(x=a)
- is trivially false or meaningless. For if ‘a’ is a name, then there is an
object a that it designates, and the statement is then (analytically) false;
or if ‘0’ is not a name, then it is meaningless.13
(b) In constructing a language of predicate logic, one grounds the
interpretation of it in a non-empty domain of objects y as the set of all
objects one can talk about in this language. A sentence of the form There
is an F 9is then interpreted to mean that there is an object in y with the
property F. The sentences There are no objects in y9 and There is an
object that is not contained in y9are then trivially false as sentences in this
language. If you are convinced that y is empty or if you wish to speak of
objects that are not contained in y, then you must just choose a new lan­
guage.14 The only existential sentences that are informative concern the
distinctions within the ontology of the language, i.e. sentences of the form
There is an F 9, where *F9 is a predicate defined on y, but not one which
defines y.15 In Wittgenstein’s words: The language defines the limits of
the range of objects of which we can make meaningful statements in it.
J. L. Austin’s statements in [62b], Chapter 7, are aimed in the same
direction. According to them, ‘real’ is not a predicate with which we can
sketch out an ontology but serves to make descriptive distinctions that
differ from case to case within the confines of an ontology. Thus we dis­
tinguish ‘real’ friends from pretended or untrustworthy friends, ‘real’
illnesses from imaginary, and so on. A use of this word in the context
‘real world’, ‘really existing object’, on the other hand, is not defined.16
There stands behind these arguments the simple fact that we can never
make statements with the content that something has such and such a
character except in a particular language, and that we are thereby pre­
supposing in using the language an ontology that we cannot meaning­
fully call into question within that language. Of what lies beyond the
limits of a language, what cannot be said in it, we cannot speak in it. We
cannot even say in it that there is something beyond those limits. It does
not follow from our understanding that for every language there are
limits of what can be said in it that there is something that cannot be said
in it. That can only be shown, by saying something in another language
288 CHAPTER IV

that cannot be translated into the first. That there is something that lies
beyond all bounds of linguistic possibility, something that cannot be said
at all, cannot be shown, then.
All that does not mean that there is for us only the ‘bespoken’ world,
as represented in our language, but no ‘real’ world. ‘Real’ and ‘bespoken’
are superfluous epithets: The world in which we live and act is for us the
‘real’ world, if that word is supposed to have any meaning at all - there
is none more real. And this is the world we talk about, for language is part
of our life and action.
Again the relativity thesis reduces to a very simple statement, then:
Our knowledge, in the stronger sense in which, for example, it is different
in type from the experience of animals, is a matter of conceiving, grasping
in language. For that reason, there is no knowledge beyond the limits of
language. As far as my knowledge will reach, so far will my language
reach, too; wherever there is something to be known, it is also to be
grasped in language. The significance of this simple statement here again
lies only in the fact that we are not permitted to distinguish the world
antecedently given and language as an instrument for describing it. In­
stead we must always see that the world conceived, known and spoken of
is always a world disclosed to us with the cognitive instruments of know­
ledge, and represents the outcome of a cultural process and accomplish­
ment.
(5) The relativity thesis is no skeptical thesis. If it is said that for us
‘reality’ always coincides with the ontology of our language, it does not
follow that we are shut up in our language as in a cage and are incapable
of knowing what can not be said in it, or that ‘reality in itself’ is hidden
from our view.
We have already insisted: Language is an open system, so far as our
knowledge will reach, our language will also reach. And the statement
that there is a reality we cannot comprehend in terms of language is not
only just as unprovable as the old principle of skepticism ‘There is no
certain knowledge’ (in order to prove that proposition, one would definitely
have to know for certain that it is true), but downright meaningless. It
was already shown why a sentence in language S: ‘There are objects of
which one can not speak in S ’ is meaningless. Wittgenstein has this to say
along these lines in the Tractatus: ‘Skepticism is not irrefutable, but ob­
viously nonsensical, when it chooses to doubt where no question can be
 LANGUAGE AND REALITY 289

raised. For there can be doubt only where there is a question; a question,
only where there is an answer; and that, only where something can be
said.'17
If there are different languages with different ontologies, what does
emerge from that fact is a relativism, but no skepticism. One cannot say:
There can be only one correct ontology, but which one is correct we are
incapable of deciding, for all of them represent interpretations that are
possible, i.e. compatible with experience. Only sentences are true or false
in this sense, not conceptual systems. We must add to the conceptual sys­
tems, then, the basic assumptions that are formulated in terms of those
concepts, and so look on different world views as theories. Theories
formulated in quite different languages can stand all alongside each other
without our being able to say that the correctness of the one excludes the
other’s being correct. In this sense, then, relativism does not imply any
skeptical component with the content: only one world view can be cor­
rect, but we do not know which.

NOTES

1 See e.g. the work by Brown and Lenneberg [54] discussed in Chapter IV.2.
2 The appelation ‘pragmatic theory of meaning’ is of course not quite suitable espe­
cially for Humboldt’s ideas, since he never identified meaning and use - for him mean­
ings belong to an ideal mental sphere - but what decides the matter in this case is that
linguistic expressions are not viewed as signs for communicating pre-existent objects,
but the objects are mediated only by way of language.
3 On the following, see the discussions in Stenius [60], Chapter XI, as well as the
presentation in Stegmiiller [65], Chapter XI, I, 4.
4 See Wittgenstein [22], 4.12ff.
5 According to Wittgenstein the like is also true of the meanings of expressions. Thus
Wittgenstein says: “The identity o f meaning of two expressions is not something about
which a claim can be made. For in order to make some claim about their meaning, I
must be cognizant of [kennen] their meaning; and in that I am cognizant of their mean­
ing, I know that they mean the same or different things.” ([22], 6.2322). See also [22],
3.262.
6 Wittgenstein [22], 6.54.
7 See Wittgenstein [22], 4.111, 4.112 and 4.003.
8 See Wittgenstein [22], 4.111 and 4.112.
9 Wittgenstein [22], 5.6.
10 See Carnap [50].
11 See Aristotle, Politics A, 1253al0.
12 On this point see what Hempel has to say about the systematic import of concepts
in [65a].
13 We have pointed out in Chapter III.2.1.2 how it is possible in a language like L to
290 CHAPTER IV

speak of things that are regarded as non-existent. If a distinction is made between the
set U of possible objects and the set U' of existing objects, then the sentence ‘a does not
exist’ can be formulated either as ‘n \/ x ( x = d y in the sense of ‘a is not a possible
object’ or as ‘ 1 V .x(x=aY in the sense of ‘a is not a real object’. In either case the object
a has to belong to the ontology of L , i.e. to the set £/, or U'.
14 ‘Language’ always means ‘interpreted language’ in this connection. Naturally it is
also possible to interpret the expressions of a language as a syntactic system concerning
a different domain o f objects.
15 On this point see R. Carnap’s remarks in [50].
16 See also Savigny [69], p. 233ff.
17 Wittgenstein [22], 6.5.1.
 B IB L IO G R A P H Y

Abraham, S. and Kiefer, F. [66]: A Theory o f Structural Semantics, The Hague 1966.
Alston, W. P. [63a]: ‘Meaning and Use’, Philosophical Quarterly 13 (1963), 107-124;
reprinted in Parkinson [68].
Alston, W. P. [63b]: ‘The Quest for Meanings’, Mind 12 (1963), 79-87.
Alston, W. P. [67]: Philosophy o f Language, in Edwards [67], Vol. IV, pp. 386-390.
Antal, L. [61]: ‘Sign, Meaning, Context’, Lingua 10 (1961), 211-219.
Arnauld, A. and Lancelot, C. [60]: Grammaire generate et raisonnee (The Grammar
of Port Royal), Paris 1660. Edited by H. Brekle, 2 Vols., Stuttgart 1966.
Austin, J. L. [61]: Philosophical Papers, Oxford 1961.
Austin, J. L. [62a]: How to Do Things with Words, Cambridge/Mass. 1962.
Austin, J. L. [62b]: Sense and Sensibilia (ed. by G. J. Warnock), Oxford 1962.
Ayer, A. J. [63]: The Concept o f a Person and Other Essays, London 1963.
Ayer, A. J. (ed.) [59]: Logical Positivism, New York 1959.
Bach, E. and Harms, R. T. (eds.) [68a]: Universals in Linguistic Theory, New York 1968.
Bach, E. [68b]: ‘Nouns and Noun Phrases’, in Bach [68a], pp. 91-122.
Bacon, R. [57]: The Works o f Francis Bacon, edited by Spedding, Ellis and Heath,
London 1857-1874.
Bacon, F. [69]: Grammatica Graeca, Oxford Manuscripts (ed. by E. Charles), 1869.
Bambrough, R. [60]: ‘Universals and Family Resemblances’, Proceedings o f the
Aristotelian Society 61 (1960/61), 2-22; reprinted in Pitcher [66].
Bar-Hillel, Y. [50]: ‘On Syntactical Categories’, Journal o f Symbolic Logic 15 (1950),
1-16.
Bar-Hillel, Y. [53]: ‘A Quasi-Arithmetical Notation for Syntactic Description’,
Language 29 (1953), 47-58.
Bar-Hillel, Y. [70]: Aspects o f Language, Jerusalem 1970.
Bar-Hillel, Y. and Shamir, E. [60a]: ‘Finite-State Languages: Formal Representation
and Adequacy Problems’, The Bulletin o f the Research Council o f Israel, 8F, No. 3
(1960), 155-166.
Bar-Hillel, Y., Gaifman, C., and Shamir, E. [60b]: ‘On Categorical and Phrase Struc­
ture Grammars’, The Bulletin o f the Research Council o f Israel, 9F, No. 3 (1960),
1-16.
Basilius, H. [52]: ‘Neo-Humboldtian Ethnolinguistics’, Word 8 (1952), 95-105.
Becker, K. F. [42]: Ausfuhrliche deutsche Grammatik, Vol. I, Prague 1842.
Beckmann, G. A. [63]: ‘Die Nachfolgekonstruktionen des lateinischen Instrumentalis
in den romanischen Sprachen’, Beihefte zur Zeitschrift fur romanische Philologie,
Tubingen 1963.
Bierwisch, M. [66]: ‘Strukturalismus, Geschichte, Probleme und Methoden’, Kursbuch
5 (1966), 77-152.
Black, M. [59]: ‘Linguistic Relativity’, Philosophical Review 68 (1959), 228-238.
Bohnert, H. G. [63]: ‘Carnap’s Theory of Definition and Analyticity’, in Schilpp [63],
pp. 407-430.
292 BIBLIOGRAPHY

Brekle, H. E. [69a]: ‘Generative Semantics vs. Deep Syntax’, in Kiefer [69], pp. 80-90.
Brekle, H. E. [69b]: ‘Review of N . Chomsky “Cartesian Linguistics’” , Linguistics 49
(1969), 74-91.
Brekle, H. E. [70a]: Generative Satzsemantik und transformationelle Syntax im System
der englischen Nominalkomposition, Munich 1970.
Brekle, H. E. [70b]: ‘Generative Satzsemantik versus generative Syntax als Komponen-
ten eines Grammatikmodells’, Linguistik und Didaktik 1 (1970), 129-136.
Brentano, F. [30]: Wahrheit und Evidenz (ed. by O. Kraus), Leipzig 1930; English edi­
tion: The True and the Evident (transl. by Roderick M. Chisholm e ta l.\ London 1966.
Brinkmann, H. [50]: ‘Die Wortarten im Deutschen’, Wirkendes Wort 1 (1950), 65-79.
Brown, R. W. and Lenneberg, H. E. [54]: ‘A Study in Language and Cognition’,
Journal o f Abnormal and Social Psychology 49 (1954), 454-462.
Btihler, K. [34]: Sprachtheorie, Stuttgart 1934, 21965.
Carnap, R. [28]: Der logische Aufbau der Welt, 1Leipzig 1928, 2Hamburg 1961.
English edition: The Logical Structure o f the World (transl. by Rolf A. George),
Berkeley 1967.
Carnap, R. [32]: ‘Ober Protokollsatze’, Erkenntnis 3 (1932/33), 215-228.
Carnap, R. [37]: The Logical Syntax o f Language, London 1937.
Carnap, R. [49]: ‘A Reply to L. Linsky’, Philosophy o f Science 16 (1949), 347-350.
Carnap, R. [50]: ‘Empiricism, Semantics, and Ontology’, Revue Internationale de
Philosophie 11 (1950), 20-40; reprinted in Linsky [52].
Carnap, R. [52]: ‘Meaning Postulates’, Philosophical Studies 3 (1952), 65-80; reprinted
in Carnap [56].
Carnap, R. [55]: ‘Meaning and Synonymy in Natural Languages’, Philosophical Studies
7 (1955), 33-47; reprinted in Carnap [56].
Carnap, R. [56]: Meaning and Necessity, Chicago 21956.
Carnap, R. [63]: ‘Reply to Quine’, in Schilpp [63], pp. 915-922.
Chappell, V. C. (ed.) [64]: Ordinary Language, Englewood ClifTs/N.J. 1964.
Charlesworth, M. J. [59]: Philosophy and Linguistic Analysis, Pittsburgh, Louvain
1959.
Chomsky, N. [56]: ‘Three Models for the Description of Language’, I.R.E. Transactions
o f Information Theory, Vol. IT 2 (1956), 113-124; reprinted in Luce [65].
Chomsky, N. [57]: Syntactic Structures, The Hague 1957.
Chomsky, N. [59a]: ‘On Certain Formal Properties of Grammars’, Information and
Control 2 (1959), 137-167; reprinted in Luce [65].
Chomsky, N. [59b]: ‘Review of Skinner’s “Verbal Behavior” ’, Language 35 (1959),
26-58; reprinted in Fodor [64].
Chomsky, N. [61a]: ‘On the Notion “Rule of Grammar” ’, in Jakobson [61],pp. 6-24;
reprinted in Fodor [64].
Chomsky, N. [61b]: ‘Some Methodological Remarks on Generative Grammar’,
Word 17 (1961), 219-239.
Chomsky, N. [62a]: ‘A Transformational Approach to Syntax’, in A .A . Hill (ed.):
Proceedings o f the Third Texas Conference o f Linguistic Analysis in English, Austin
1962, pp. 124-158; reprinted in Fodor [64].
Chomsky, N. [62b]: ‘Explanatory Models in Linguistics’, in Nagel [62], pp. 528-550.
Chomsky, N. [64]: ‘Current Issues in Linguistic Theory’, in Fodor [64], pp. 50-118.
Chomsky, N. [65]: Aspects o f the Theory o f Syntax, Cambridge/Mass. 1965.
Chomsky, N. [66a]: ‘Topics in the Theory of Generative Grammar’, in Sebeok [66],
pp. 1-60.
 BIBLIOGRAPHY 293

Chomsky, N. [66b]: Cartesian Linguistics, New York 1966.

Chomsky, N. [67]: ‘Recent Contributions to the Theory of Innate Ideas’, Synthese 17
(1967), 2-11.
Chomsky, N. [68]: ‘Quine’s Empirical Assumptions’, Synthese 19 (1968/69), 53-68.
Chomsky, N. and Miller, G. A. [59]: ‘Finite State Languages’, Information and Control
1 (1959), 91-112; reprinted in Luce [65].
Church, A. [43a]: ‘Carnap’s “Introduction to Semantics” ’, Philosophical Review 52
(1943), 298-304.
Church, A. [43b]: ‘Review o f Quine: Notes on Existence and Necessity’, Journal o f
Symbolic Logic 8 (1943), 45-47.
Church, A. [46]: ‘Reviews of M. G. White and M. Black’, Journal o f Symbolic Logic 11
(1946), 132-134.
Church, A. [50]: ‘On Carnap’s Analysis of Statements of Assertion and Belief’,
Analysis 10 (1950), 97-99; reprinted in Macdonald [54].
Church, A. [51a]: ‘The Need for Abstract Entities in Semantic Analysis’, Proceedings
o f the American Academy o f Arts and Sciences 80 (1951), 100-112.
Church, A. [51b]: ‘A Formulation of the Logic of Sense and Denotation’, in Henle
[51], pp. 3-24.
Church, A. [54]: ‘Intensional Isomorphism and Identity of Belief’, Philosophical Studies
5 (1954), 65-73.
Cohen, L. J. [64]: ‘D o illocutionary forces exist?’, Philosophical Quarterly 14 (1964),
118-137.
Copi, I. M. and Beard, R. W. (eds.) [66]: Essays on Wittgenstein*s Tractatus, London
1966.
Cordemoy, G. de [77]: ‘Discours Physique de la Parole’, facsimile reprint o f the 1677
edition with a commentary by H. E. Brekle, Stuttgart 1970.
Cresswell, M. J. [73]: Logics and Languages, London 1973.
Curry, H. B. [61]: ‘Some Logical Aspects o f Grammatical Structure’, in Jakobson
[61], pp. 56-68.
Davidson, D . [63]: ‘The Method o f Extension and Intension’, in Schilpp [63], pp.
311-349.
Davidson, D. [68]: ‘On Saying That’, Synthese 19 (1968/69), 130-146.
Davis, M. [58]: Computability and Unsolvability, New York 1958.
Duhem, P. [06]: La theorie physique, son objet et sa structure, Paris 1906; English
edition: The Aim and Structure o f Physical Theory (transl. by Philip P. Wiener),
Princeton 1954.
Dummett, Michael [73]: Frege Philosophy o f Language, Londen 1973. Bibliography
includes a list of English translations of Frege’s works.
Edwards, P. (ed.) [67]: Encyclopedia o f Philosophy, 8 Vols., New York 1967.
Fearing, F. [54]: ‘An Examination o f the Conceptions of B. Whorf in the Light of
Theories of Perception and Cognition’, in Hoijer [54a], pp. 47-81.
Feigl, H. and Sellars, W. (eds.) [49]: Readings in Philosophical Analysis, New York 1949.
Fillmore, Ch. [68]: ‘The Case for Case’, in Bach [68a], pp. 1-88.
Findlay, J. N. [61]: ‘Use, Usage, and Meaning’, Proceedings o f the Aristotelian Society
Suppl. 35 (1961), 231-242.
Fodor, J. A. and Katz, J. J. [63]: ‘The Structure of a Semantic Theory’, Language 39
(1963), 170-210; reprinted in Fodor [64].
Fodor, J. A. and Katz, J. J. (eds.) [64]: The Structure o f Language, Readings in the
Philosophy o f Language, Englewood Cliffs/N.J. 1964.
294 BIBLIOGRAPHY

Frege, G. [79]: Begriffsschrift. Eine der arithmetischen nachgebildete Formelsprache

des reinen Denkens, Halle 18792 (ed. by I. Angelelli), Darmstadt 1964.
Frege, G. [92a]: ‘Ober Sinn und Bedeutung’, Zeitschrift fiir Philosophic und philo-
sophische Kritik, N.F. 100 (1892), 25-50; reprinted in Frege [67].
Frege, G. [92b]: ‘Ober Begriff und Gegenstand’, Vierteljahresschrift fur wissenschaft-
liche Philosophic 16 (1892), 192-205; reprinted in Frege [67].
Frege, G. [18]: ‘Logische Untersuchungen, I: Der Gedanke’, Beitrage zur Philosophic
des deutschen Idealismus 1 (1918/19), 58-77, ‘II: Die Verneinung’, ibid., 143-157,
‘III: Gedankengefuge’, ibid. 3 (1923/26), 36-51; reprinted in Frege [67].
Frege, G. [67]: Kleine Schriften (ed. by I. Angelelli), Darmstadt 1967.
Frege, G. [69]: Nachgelassene Schriften (ed. by H. Hermes, F. Kambartel and F.
Kaulbach), Hamburg 1969.
Frege, G. [English translations]: See Bibliography in Dummett [73].
Gewirth, A. [53]: ‘The Distinction Between Analytic and Synthetic Truths’, Journal o f
Philosophy 50 (1953), 397-425.
Gipper, H. [69]: Bausteine zur Sprachinhaltsforschung, Diisseldorf 21969.
Goodman, N. [49]: ‘On Likeness o f Meaning’, Analysis 10 (1949/50), 1-7; revised draft
reprinted in Linsky [52].
Goodman, N. [51]: The Structure o f Appearance, Cambridge/Mass. 1951,2Indianapolis
1966.
Goodman, N. [52]: ‘On Some Differences about Meaning’, Analysis 13 (1952/53)
90-96; reprinted in Macdonald [54].
Goodman, N. [61]: ‘About’, M indlto (1961), 1-24.
Goodman, N. [67]: ‘The Epistemological Argument’, Synthese 17 (1967), 23-28.
Grebe, P. (ed.) [66]: ‘Grammatik der deutschen Gegenwartssprache’, Der grosse
Duden, Bd. 4, Mannheim 21966.
Grice, H. P. [57]: ‘Meaning’, Philosophical Review 66 (1957), 377-388.
Grice, H. P. and Strawson, P. F. [56]: ‘In Defense o f a Dogma’, Philosophical Review
65 (1956), 141-158.
Griffin, J. [64]: Wittgenstein's Logical Atomism, Oxford 1964.
Groot, A. W. de [56]: ‘Classification o f the Uses of a Case Illustrated on the Genitive
in Latin’, Lingua 6 (1956/57), 8-66.
Hallet, G. [67]: Wittgenstein's Definition o f Meaning as Use, New York 1967.
Harman, G. H. [63]: ‘Generative Grammars Without Transformation Rules: a Defense
of Phrase Structure’, Language 39 (1963), 597-616.
Harman, G. H. [67]: ‘Quine on Meaning and Existence’, The Review o f Metaphysics
21 (1967/68), 124-151, 343-364.
Harman, G. H. [68]: ‘An Introduction to Translation and Meaning, chapter two of
Word and Object’, Synthese 19 (1968/69), 14-26.
Harman, G. H. and Davidson, D . [72]: Semantics o f Natural Language, Dordrecht 1972.
Hartmann, H. [54]: Das Passiv. Eine Studie zur Geistesgeschichte der Kelten, Italiker,
Arier, Heidelberg 1954.
Hartmann, P. [52]: Einige Grundzuge des japanischen Satzbaus - gezeigt an den Aus-
driicken fiir das Sehen, Heidelberg 1952.
Hartmann, P. [56]: Wortart und Aussageform, Heidelberg 1956.
Heger, K. [63]: ‘Homographie, Homonymie und Polysemie’, Zeitschrift fiir Romanische
Philologie 79 (1963), 471-491.
Heger, K. [69]: ‘Die Semantik und die Dichotomie von Langue und Parole’, Zeitschrift
fiir Romanische Philologie 85 (1969), 144-215.
 BIBLIOGRAPHY 295

Hempel, C. G. [34]: ‘On the Logical Positivists’ Theory of Truth’, Analysis 2 (1934/35),
49-59.
Hempel, C. G. [65]: Aspects o f Scientific Explanation and Other Essays in the Philosophy
o f Science, New York 1965.
Hempel, C. G. and Oppenheim, P. [36]: Der Typusbegriff im Lichte der neuen Logik,
Leiden 1936.
Henle, P. (ed.) [65]: Language, Thought and Culture, Chicago 21965.
Henle, P., Kallen, H. M ., and Langer, S. K. (eds.) [51]: Structure, Method, and Mean­
ing: Essays in Honor o f Henry M . Sheffer, New York 1951.
Herder, J. G. [91]: ‘Abhandlung uber den Ursprung der Sprache’, in J. G. Herder:
Samtliche Werke (ed. by B. Suphan), Vol. V, Berlin 1891.
Hermes, H. [61]: Aufzahlbarkeit,Entscheidbarkeit,Berechenbarkeit, Berlin 1961; English
edition: Enumerability, Decidability, Computability (transl. by G. T. Herman and
O. Plassmann), New-York 1965.
Hintikka, J. J. [68]: ‘Behavioral Criteria of Radical Translation’, Synthese 19 (1968/69),
69-81.
Hjelmslev, L. [43]: Prolegomena to a Theory o f Language (from the Danish (1943),
transl. by F. J. Whitefield), Bloomington/Ind. 1953.
Hormann, H. [67]: Psychologie der Sprache, Berlin 1967.
Hoijer, H. (ed.) [54a]: Language in Culture, Chicago 1954.
Hoijer, H. [54b]: ‘The Sapir-Whorf Hypothesis’, in Hoijer [54a], pp. 92-105.
Holdcroft, D . [64]: ‘Meaning and Illocutionary Acts’, Ratio 6 (1964), 128-143;
reprinted in Parkinson [68].
Humboldt, W. v. [03]: Gesammelte Schriften, Berlin Academy Edition, Berlin 1903.
Husserl, E. [00]: Logische Untersuchungen, Vol. 1, Halle a.S. 1900.
Jakobson, R. (ed.) [61]: ‘Structure of Language and Its Mathematical Aspects’,
Proceedings o f the 12th Symposium in Applied Mathematics, Providence 1961.
Kamlah, W. and Lorenzen, P. [67]: Logische Propadeutik, Mannheim 1967.
Katz, J. J. [64a]: ‘Analyticity and Contradiction in Natural Language’, in Fodor [64],
pp. 519-543.
Katz, J J. [66]: The Philosophy o f Language, New York 1966.
Katz, J. J. and Postal, P. [64b]: An Integrated Theory o f Linguistic Descriptions,
Cambridge/Mass. 1964.
Keenan, E. [72]: ‘On Semantically Based Grammar’, Linguistic Inquiry 7/74 (1972).
Kemeny, J. G. [63]: ‘Analyticity vs. Fuzziness’, Synthese 15 (1963), 57-80.
Kiefer, F. (ed.) [69]: Studies in Syntax and Semantics, Dordrecht 1969.
Kohler, W. [37]: ‘Psychological Remarks on Some Questions of Anthropology’,
American Journal o f Psychology 50 (1937), 271-288.
Kripke, S. [72]: ‘Naming and Necessity’; in Harman and Davidson [72], pp. 253-355.
Kutschera, F. v. [64]: Die Antinomien der Logik, Freiburg 1964.
Kutschera, F. v. [67]: Elementare Logik, Vienna 1967.
Kutschera, F. v. [72]: Wissenschaftstheorie. Grundziige der allgemeinen Methodologie
der empirischen Wissenschaften, Munich 1972.
Kutschera, F. v. [72a]: ‘Eine logische Analyse des sprachwissenschaftlichen Feld-
begriffs, Studia Leibnitiana, Sonderheft 3 (1973), 71-84.
Kutschera, F. v. [73]: Einfuhrung in die Logik der Normen, Werte und Entscheidungen,
Freiburg 1973.
Kutschera F. v. and Breitkopf, A. [71]: Einfuhrung in die moderne Logik, Freiburg
1971.
296 BIBLIOGRAPHY

Kutschera, F. v. [74]: ‘Partial Interpretations’, in Keenan (ed.): Formal Semantics o f

Natural Language, Cambridge 1974.
Lakoff, G. [70]: ‘Linguistics and Natural Logic’, Ann Arbor 1970 (mimeographed).
Lambek, J. [61]: ‘On the Calculus o f Syntactic Types’, in Jakobson [61], pp. 166-178.
Lambert, K. (ed.) [70]: Philosophical Problems in Logic, Dordrecht 1970.
Langford, C. H. [42]: ‘The Notion o f Analysis in Moore’s Philosophy’, in Schilpp
[42], pp. 321-342.
Lashley, K. S. [51]: ‘The Problem o f Serial Order in Behavior’, in: Jeffress (ed.):
Hixon Symposium on Cerebral Mechanism in Behavior, New York 1951.
Lees, R. B. [57]: ‘Review o f Chomsky’s Syntactic Structures’, Language 33 (1957),
375-408.
Lenk, H. [67]: ‘Zu Wittgensteins Theorie der Sprachspiele’, Kant-Studien 58 (1967),
458-480.
Lenneberg, E. H. and Roberts, J. M. [53]: ‘The Denotata of Color Terms’, Paper read
at Linguistic Society of America, Bloomington/Ind., August 1953.
Lewis, C. I. [43]: ‘The Modes of Meaning’, Philosophy and Phenomenological Research
4 (1943/44), 236-249; reprinted in Linsky [52].
Lewis, D . [68]: ‘Counterpart Theory and Quantified Modal Logic’, The Journal o f
Philosophy 65 (1968), 113-126.
Lewis, D . [69]: Convention, A Philosophical Study, Cambridge/Mass. 1969.
Lewis, D . [70]: ‘General Semantics’, Synthese 22 (1970), 18-67.
Lewis, D . [73]: Counterfactuals, Oxford 1973.
Lewis, H. D. (ed.) [56]: Contemporary British Philosophy, London 1956.
Linsky, L. [49]: ‘Some Notes on Carnap’s Concept of Intensional Isomorphism and the
Paradox of Analysis’, Philosophy o f Science 16 (1949), 343-347.
Linsky, L. [52]: Semantics and the Philosophy o f Language, Urbana/Ill. 1952.
Linsky, L. [67]: Referring, London 1967.
Lorenz, K. [70]: Elemente der Sprachkritik. Eine Alternative zum Dogmatismus und
Skeptizismus in der analytischen Philosophic, Frankfurt a.M. 1970.
Luce, R. D ., Bush, R. R., and Galanter, E. (eds.) [65]: Readings in Mathematical
Psychology, Vol. II, New York 1965.
Lyons, J. [69]: Introduction to Theoretical Linguistics, Cambridge 1969.
Lyons, J. [66]: ‘Towards a ‘Notional’ Theory of the ‘Parts of Speech’ ’, Journal o f
Linguistics 2 (1966), 209-236.
Macdonald, M. (ed.) [54]: Philosophy and Analysis, Oxford 1954.
Mace, C. A. (ed.) [66]: British Philosophy in the Mid-Century, A Cambridge Sympo­
sium, London 21966.
Malcolm, N. (ed.) [58]: Ludwig Wittgenstein - A Memoir, London 1958.
Marchand, H. [69]: The Categories and Types o f Present-Day English Word-Formation:
A Synchronic-Diachronic Approach, 2 Munich 1969.
Martin, R. M. [52]: ‘On “Analytic” ’, Philosophical Studies 3 (1952), 42-47.
Martin, R. M. [59]: The Notion o f Analytic Truth, Philadelphia 1959.
Mates, B. [50]: ‘Synonymity’, University o f California Publications in Philosophy 25
(1950); reprinted in Linsky [52].
Mates, B. [51]: ‘Analytic Sentences’, Philosophical Review 60 (1951), 525-534.
Mates, B. [58]: ‘On the Verification of Statements about Ordinary Language’, Inquiry 1
(1958); reprinted in Chappell [64], pp. 64-74.
McCawley, J. D . [68]: ‘The Role of Semantics in a Grammar’, in Bach [68a], pp. 124-
169.
 BIBLIOGRAPHY 297

Mittelstrass, J. [68]: ‘Die Pradikation und die Wiederkehr des Gleichen’, Ratio 10
(1968), 53-61.
Montague, R. [70]: ‘Universal Grammar’, Theoria36 (1970), 373-398. German transla­
tion with a commentary by H. Schnelle, Braunschweig 1972.
Montague, R. [70a]: ‘The Proper Treatment of Quantification in Ordinary English’,
manuscript o f a lecture at Stanford 1970.
Morris, Ch. W. [38]: Foundations o f the Theory o f Signs {International Encyclopedia o f
Unified Science, Vol. 1, No. 2), Chicago 1938.
Morris, Ch. [46]: Signs, Language and Behavior, New York 1946, 21955.
Naess, A. [49]: ‘Toward a Theory o f Interpretation and Preciseness’, Theoria 15 (1949),
220-241; reprinted in Linsky [52].
Naess, A. [53]: ‘Interpretation and Preciseness: A Contribution to the Theory of
Communication’, Skrifter Norske Videnskaps Akademi, Oslo II. Hist-Filos. Klasse
(1953), No. 1.
Nagel, E., Suppes, P., and Tarski, A. (eds.) [62]: Logic, Methodology, and Philosophy
o f Science, Stanford 1962.
Neurath, O. [32]: ‘Protokollsatze’, Erkenntnis 3 (1932/33), 204-214.
Ohmann, S. [51]: Wortinhalt und Weltbild, Stockholm 1951.
Pap, A. [55]: ‘Belief, Synonymity and Analysis’, Philosophical Studies 6 (1955), 11-
15.
Pap, A. [57]: ‘Belief and Propositions’, Philosophy o f Science 24 (1957), 123-136.
Pap, A. [58]: Semantics and Necessary Truth, New Haven/Conn. 1958.
Parkinson, G. H. R. (ed.) [68]: The Theory o f Meaning, Oxford 1968.
Patzig, G. [70]: Sprache und Logik, Gottingen 1970.
Peach, B. [52]: ‘A Nondescriptive Theory o f the Analytic’, Philosophical Review 61
(1952), 349-367.
Pitcher, G. [64]: The Philosophy o f Wittgenstein, Englewood Cliffs/N.J. 1964.
Pitcher, G. (ed.) [66]: Wittgenstein - The Philosophical Investigations, New York 1966.
Postal, P. [64a]: Constituent Structure, Bloomington 1964.
Postal, P. [64b]: ‘Limitations o f Phrase Structure Grammars’, in Fodor [64], pp.
137-151.
Postal, P. [64c]: ‘Underlying and Superficial Linguistic Structure’, Harvard Educa­
tional Review 34 (1964), 246-266.
Putnam, H. [54]: ‘Synonymity and the Analysis of Belief Sentences’, Analysis 14
(1953/54), 114-122.
Putnam, H. [61]: ‘Some Issues in the Theory of Grammar’, in Jakobson [61], pp. 25-42.
Putnam, [62a]: ‘The Analytic and the Synthetic’, in H. Feigl and G. Maxwell (eds.):
Minnesota Studies in the Philosophy o f Science, Vol. Ill, Minneapolis 1962, pp.
358-397.
Putnam, H. [62b]: ‘It Ain’t Necessarily So’, Journal o f Philosophy 59 (1962), 658-671.
Putnam, H. [67]: ‘The ‘Innateness Hypothesis’ and Explanatory Models in Linguistics’,
Synthese 17 (1967), 12-22.
Quine, W. V. [48]: ‘On What There Is’, The Review o f Metaphysics 2 (1948/49), 21-38;
reprinted in Quine [64a].
Quine, W. V. [51a]: ‘Two Dogmas of Empiricism’, The Philosophical Review 60 (1951),
20-43; reprinted in Quine [64a].
Quine, W. V. [51b]: Mathematical Logic, Cambridge/Mass. 21951.
Quine, W. V. [56]: ‘Quantifiers and Propositional Attitudes’, The Journal o f Philosophy
53 (1956), 177-187; reprinted in Quine [66].
298 BIBLIOGRAPHY

Quine, W. V. [57]: ‘The Scope and Language of Science’, The British Journal for the
Philosophy o f Science 8 (1957/58), 1-17; reprinted in Quine [66].
Quine, W. V. [58]: ‘Speaking of Objects’, Proceedings and Addresses o f the American
Philosophical Association, Eastern Division, 1958; reprinted in Quine [69a].
Quine, W. V. [59]: Methods o f Logic, New York 21959.
Quine, W. V. [60]: Word and Object, Cambridge/Mass. 1960.
Quine, W. V. [63]: ‘Carnap and Logical Truth’, in Schilpp [63], pp. 385-406.
Quine, W. V. [64a]: From a Logical Point o f View, Cambridge/Mass. 21964.
Quine, W. V. [64b]: ‘Reference and Modality’, in Quine [64a], pp. 139-159.
Quine, W. V. [64c]: ‘Notes on the Theory o f Reference’, in Quine [64a], pp. 130-138.
Quine, W. V. [66]: The Ways o f Paradox and Other Essays, New York 1966.
Quine, W. V. [69a]: Ontological Relativity and Other Essays, New York 1969.
Quine, W. V. [69b]: ‘Propositional Objects’, in Quine [69a], pp. 139-160.
Quine, W. V. [69c]: ‘Natural Kinds’, in Quine [69a], pp. 114-138.
Reichenbach, H. [47]: Elements o f Symbolic Logic, New York 1947.
Rescher, N. [68]: Topics in Philosophical Logic, Dordrecht 1968.
Robins, R. H. [51]: Ancient and Medieval Grammatical Theory in Europe with Particular
Reference to Modern Linguistic Doctrine, London 1951.
Russell, B. [05]: ‘On Denoting’, M ind 14 (1905), 479-493; reprinted in Feigl [49].
Russell, B. [18]: ‘The Philosophy o f Logical Atomism’, The Monist 28 (1918), 495-527;
29 (1919), 32-63, 190-222, 345-380.
Russell, B. [40]: An Inquiry into Meaning and Truth, London 1940.
Russell, B. [56a]: ‘Logic and Knowledge’, Essays 1901-1950 (ed. by R. C. Marsh),
London 1956.
Russell, B. [56]: ‘Logical Atomism’, in Russell [56a], pp. 323-343.
Ryle, G. [49]: The Concept o f Mind, London 1949.
Ryle, G. [53]: ‘Ordinary Language’, The Philosophical Review 62 (1953), 167-186;
reprinted in Chappell [64].
Ryle, G. [60]: ‘Use, Usage and Meaning’, Proceedings o f the Aristotelian Society, Suppl.
55 (1960), 223-230; reprinted in Parkinson [68].
Ryle, G. [66]: ‘The Theory o f Meaning’, in Mace [66], pp. 237-264.
Sapir, E. [12]: ‘Language and Environment’, American Anthropologist 14 (1912),
226-242; reprinted in Sapir [49].
Sapir, E. [29]: ‘The Status of Linguistics as a Science’, Language 5 (1929), 207-214.
Sapir, E. [49]: SelectedWritings o f Eduard Sapir (ed. by D . G. Mandelbaum), Berkeley
1949.
Saussure, F. de [16]: Corns de Linguistique Gendrale, Paris 11916. German: Grundfragen
der allgemeinen Sprachwissenschaft, 2Berlin 1967; English edition: Course in General
Linguistics (transl. by Wade Baskin), New York 1959.
Savigny, E. v. [69]: Die Philosophic der normalen Sprache, Frankfurt a.M. 1969.
Savigny, E. v. [70]: Analytische Philosophic, Freiburg 1970.
Scheffler, J. [54]: ‘An Inscriptional Approach to Indirect Quotation’, Analysis 14
(1953/54), 83-90.
Scheffler, J. [55]: ‘On Synonymy and Indirect Discourse’, Philosophy o f Science 22
(1955), 39-44.
Schilpp, P. (ed.) [42]: The Philosophy o f G. E. Moore, Evanston/Ill. 1942.
Schilpp, P. (ed.) [63]: The Philosophy o f Rudolf Carnap, La Salle/Ill. 1963.
Schlick, M. [36]: ‘Meaning and Verification’, The Philosophical Review 45 (1936),
339-369.
 BIBLIOGRAPHY 299

Schmidt, F. [61]: Logik der Syntax, Berlin 31961.

Schnelle, H. [73]: Sprachphilosophie und Linguistik, Reinbek b. Hamburg 1973.
Schwyzer, H. R. G. [62]: ‘Wittgenstein’s Picture Theory of Language’, Inquiry 5 (1962),
46-64; reprinted in Copi [66].
Searle, J. R. [58]: ‘Proper Names’, Mind 67 (1958), 166-173.
Searle, J. R. [68]: ‘Austin on Locutionary and Illocutionary Acts’, The Philosophicla
Review 77 (1968), 405-424.
Searle, J. R. [69]: Speech Acts, Cambridge 1969.
Sebeok, T. A. (ed.) [66]: Current Trends in Linguistics, Vol. 3, Theoretical Foundations,
The Hague 1966.
Shwayder, D . S. [60]: ‘Uses o f Language and Uses o f Words’, Theoria 26 (1960),
31-43; reprinted in Parkinson [68].
Shwayder, D . S. [63]: ‘Critical notice on E. Stenius “Wittgenstein’s Tractatus” ’,
M ind 12 (1963), 275-288; shortened version under the title ‘On the Picture Theory of
Language’ in Copi [66].
Skinner, B. F. [57]: Verbal Behavior, New York 1957.
Snell, B. [52]: Der Aufbau der Sprache, Hamburg 1952.
Stegmuller, W. [56]: ‘Das Universalienproblem einst und jetzt’, Archiv fur Philosophic 6
(1956), 192-225 and 7 (1957), 45-81.
Stegmuller, W. [57]: Das Wahrheitsproblem und die Idee der Semantik, Vienna 1957.
Stegmuller, W. [65]: Hauptstromungen der Gegenwartsphilosophie, 3Stuttgart 1965.
English edition: Main Currents o f Contemporary German, British and American
Philosophy (transl. by Albert E. Blumberg), Bloomington/Ind. 1970.
Stegmuller, W. [70]: Theorie und Erfahrung, Berlin 1970.
Stenius, E. [60]: Wittgenstein’s Tractatus, A Critical Exposition o f its Main Lines o f
Thought, Oxford 1960.
Stenius, E. [67]: ‘Mood and Language-Game’, Synthese 17 (1967), 254-279.
Stenzel, J. [34]: Philosophic der Sprache, Berlin 1934.
Stevenson, Ch. [44]: Ethics and Language, New Haven 1944.
Strawson, P. F. [49]: ‘Truth’, Analysis 9 (1948/49), 83-97; reprinted in Macdonald
[54], pp. 260-277.
Strawson, P. F. [50]: ‘Truth’, Proceedings o f the Aristotelian Society, Suppl. 24 (1950),
129-156.
Strawson, P. F. [50a]: ‘On Referring’, M ind 59 (1950), 320-344.
Strawson, P. F. [54]: ‘Review of Wittgenstein’s Philosophical Investigations’, M ind 63
(1954), 70-99; reprinted in Pitcher [66].
Strawson, P. F. [64]: ‘Intention and Convention in Speech Acts’, The Philosophical
Review 73 (1964), 439-460.
Tarski, A. [23]: ‘O wyrazie pierwotnym logistyki’, Przeglad filozoficzny 26 (1923),
68-89. English translation ‘On the Primitive Term of Logistic’, in Tarski [56].
Tarski A. [35]: ‘Der WahrheitsbegrifF in den formalisierten Sprachen’, Studia Philo-
sophica 1 (1935/36), 261-405, English translation in Tarski [56].
Tarski, A. [44]: ‘The Semantic Conception o f Truth and the Foundations o f Semantics’,
Journal o f Philosophy and Phenomenological Research 4 (1943/44), 341-375; reprinted
in Feigl [49].
Tarski, A. [56]: Logic, Semantics, Metamathematics, Oxford 1956.
Thiel, Ch. [65]: Sinn und Bedeutung in der Logik Gottlob Freges, Meisenheim a.G.
1965.
Tinbergen, N . [51]: The Study o f Instinct, Toronto 1951.
3 00 BIBLIOGRAPHY

Trier, J. [31]: Der deutsche Wortschatz im Sinnbezirk des Verstandes - Die Geschichte
eines sprachlichen Feldes, Heidelberg 1931.
Trier, J. [32]: ‘Sprachliche Felder’, Zeitschrift fur deutsche Bildung 8 (1932), A ll-A ll.
Trier, J. [34a]: ‘Deutsche Bedeutungsforschung’, in: Germanische Philologie, Ergebnisse
und Aufgaben, Festschrift fur Otto Behaghel (ed. by Alfred Goetze and Wilhelm
Horn), Heidelberg 1934, pp. 173-200.
Trier, J. [34b]: ‘Das sprachliche Feld’, Neue Jahrbiicher fur Wissenschaft und Jugend-
bildung 10 (1934), 428-449.
Uexkiill, J. v. [21]: Umwelt und Innenwelt der Tiere, Berlin 21921.
Uexkiill, J. v. [28]: Theoretische Biologie, Berlin 21928.
Wackernagel, J. [24]: Vorlesungen uber Syntax, 2 Vols., Basel 1924/26.
Waismann, F. [56]: ‘How I See Philosophy’, in Lewis [56]; reprinted in Ayer [59].
Wang, Hao [55]: ‘Notes on the Analytic-Synthetic Distinction’, Theoria 21 (1955),
158-178.
Weinrich, H. [64]: Tempus. Besprochene und erzahlte Welt, Stuttgart 1964.
Weisgerber, L. [58]: ‘Der Mensch im Akkusativ’, Wirkendes Wort 8 (1958), 193-205.
Weisgerber, L. [62]: Von den Kraften der deutschen Sprache, Vol. 1: ‘Grundziige der
inhaltsbezogenen Grammatik’, Dusseldorf 31962. Vol. II: ‘Die sprachliche Gestal-
tung der Welt’, Diisseldorf 31962. Vol. Ill: ‘Die Muttersprache im Aufbau unserer
Kultur’, Dusseldorf 11950, 21964. Vol. IV: ‘Die geschichtliche Kraft der deutschen
Sprache’, Dusseldorf 21959.
Weisgerber, L. [63a]: Die vier Stufen in der Erforschung der Sprachen, Dusseldorf 1963.
Weisgerber, L. [63b]: ‘Die Welt im “Passiv” ’, in: Die Wissenschaft von deutscher
Sprache und Dichtung, Festschrift fur F. Maurer (ed. by Siegfried Gutenbrunner),
Stuttgart 1963.
Wheatley, J. [63]: ‘Some Aspects o f Meaning and Use (abstract)’, Journal o f Philosophy
60 (1963), 643-644.
White, M. G. [50]: ‘The Analytic and the Synthetic: an Untenable Dualism’, in S. Hook
(ed.): John Dewey: Philosopher o f Science and Freedom, New York 1950, pp. 316—
330; reprinted in Linsky [52].
Whorf, E. L. [56]: Language, Thought and Reality (ed. by John B. Carroll), New York
1956.
Wittgenstein, L. [61]: Notebooks 1914-1918 (ed. by G. H. von Wright and G. E. M.
Anscombe), Oxford 1961.
Wittgenstein, L. [22]: ‘Tractatus logico-philosophicus’, in Ostwalds Annalen der
Naturphilosophie (1921); German-English edition, London 1922.
Wittgenstein, L. [53]: Philosophical Investigations (ed. by G. Anscombe and R. Rhees),
Oxford 1953, 21965, 31967.
Wittgenstein, L. [56]: Remarks on the Foundations o f Mathematics (ed. by G. H.
v. Wright, R. Rhees, and G. E. M. Anscombe), Oxford 1956.
Wittgenstein, L. [58]: The Blue and Brown Books, Oxford 1958.
Woodruff, P. W. [70]: ‘Logic and Truth Value Gaps’, in Lambert [70], pp. 121-142.
Wright, G. H. von [55]: ‘Ludwig Wittgenstein: A Biographical Sketch’, The Philosophi­
cal Review 64 (1955); reprinted in Malcolm [58], pp. 1-22.
Xenakis, J. [54]: ‘Meaning’, Methodos 6 (1954), 299-327.
Ziff, P. [60]: Semantic Analysis, Ithaca/N.Y. 1960.
 IN D E X OF SUBJECTS

ablative 173 category symbols 221

absolutism 34, 106 characteristic (Merkmal) 232ff.
accusative 173 class 41
achievement verbs 132f. class terms 169, 193, 262
action-free sentence 269f. classificatory concepts 152
action sentence 269f. codability 259
advertive 211 coherence theory 50
active passive interpretation schema collective terms 169, 193
268ff. comitative function 173
adequacy, of language 22 comment 181 n. 28
- , semantic, syntactic 166 comparative 173, 203f.
adjective 170, 194, 261f. comparative concept 153
adverb 170f., 198fF. complex sentence 34
adverbial modification 175f., 207f. concept 25f.
agentative function 173 conditioned response 67
allomorphy 85 conjunctions 171, 200f.
ambiguity 85, 232 context, dependence on 86ff., 105f.
- , grammatical 86 context, extensional 45f., 56 n. 64
- , semantic 210, 217 n. 43 - , indirect (oblique) 40, 43, 75
- , structural 210, 217 n. 43 - , intensional 45f., 56 n. 64
ambivalence (of meaning) 85 - , non-intensional 46, 56 n. 64
analysis, paradox of 41 ff. - , pragmatic 8, 15, 113
analytical philosophy 1 convention theory (of truth) 50
antinomies, semantic 18 conventionalism (linguistic) 20ff., 53
Appell 14 n. 14
articles 170, 185 copula 175
atomic sentence 34 copy theory (semantic) 19, 24ff.
assertive sentence 13 see also picture theory
atomism 34, 106 correspondence theory 49ff.
attribute 24ff., 35, 173, 175, 207 counterpart relation 188
attributive combination 233
automata, finite, deterministic and dative 173
non-deterministic 226 defining relation 218ff.
auxiliary verbs 175 definite description 31f., 48, 213
deontic contexts 190
background phenomena, linguistic 261 deep structure 209ff., 225
belief sentences 46f. derivation 219
descriptive component 14
case 173, 203 descriptive meaning 13, 135ff.
categories, formal, substantial 240 descriptor 63
logical 183fF. distinctor 228
302 INDEX OF SUBJECTS

equivalence, definitional 101 n. 18 indeterminacy of translation 90fF.

equivalence, logical 45 index expressions 8, 15
essential predicates 188 indirect context 40, 43, 75
evaluator 64 individual concept 48
evidence theory (of truth) 50 inflection, theory of 173ff., 201ff.
evocative component 14f. innateness hypothesis 234ff.
evocative meaning 139 instrumental behavior 67
example-classes 153f. instrumental function 173
existence, internal and external intension 48, 79, 187f.
questions 282f. - , identity of 44ff.
expressions (linguistic) 6 intensional context 45f., 56 n. 64, 76
expressive component 14 intensional isomorphism 45ff.
expressive meaning 139 intensional logic 183
extension 44ff., 185f. intensional semantics 48, 79f., 212
extensional concept 40f., 79 interjections 171
extensional context 45f., 56 n. 64 interpretant 62
interpretation 182
fact 25 -exten sion al 185, 191
family resemblance 148ff. - , intensional 189
field concept 156f. - , partial 192,213
form (of behavior) 6 - , pragmatic 191
formation rules 166 - , rule of 35f.
formator 64 - , semantic 165f., 220, 224
forms of life 108 interpretive hypotheses 124
functors, incompletely defined 212 intersubjective agreement 125ff., 146f.
isogeny 226f.
games (language and) 9 isomorphy 34ff.
see also language games - , intensional 45ff.
gender 173, 201 isomorphy-correlator 35
generalisations, empty 213
Kundgabe 14
genitive 173
grammar 179, 182
language, animal 5
- , a priori 217 n. 39
- , artificial 5
- , logical 182ff.
- , gesture 5
- , rational, universal 217 n. 39
- , ideal 36, 106f.
grammars, categorical 225f.
- , logical 82, 165f., 182
- , natural 3, 82, 182, 209ff.
hearer (speaker, referent) 14 - , ordinary 5, 106f.
homography 84 - , signal 5
homonymy 86 - , standard 5
homophony 84 - , potency of 244
Language Acquisition Device 235
ideas, innate 234ff. language games 108ff., 140
identification, nominative of 181 n. 26 langue (vs parole) 11
identificator 63
lingualism, transcendental 282
identity, trans-world 188 locutionary act 161 n. 61
illocutionary role 113, 130ff.
impersonal verbs 264, 272f. mass terms 170
implicit definition 33, 83 meaning (in Frege’s sense) 37ff.
 INDEX OF SUBJECTS 303

meaning (vs reference) 29f. phonology 10

meaninglessness 23If. phonological component 165
mention 17 phrase-marker 221
metalanguage 17f. picture theory 33ff.
modalities 187 see also copy theory
modal operators 190 plasticity of meaning 179f.
mood 173, 205 polysemy 86
morpheme 166 positive (form) 173
possessive 173
name-relation 25ff., 36 possible objects 187
naturalism, semantic 20f. possible worlds 47, 187
nominalism 27f., 149 pragmatic theory of meaning 59fF.,
nominative 173 278ff.
number 173, 201f. pragmatics 18
number words (numeralia) 170, 198 pragmatism 59, 150
predicates, meaning of 24ff., 39ff.
object, direct 173ff., 207 prepositions 171, 200
- , indirect 173ff., 207 prescriptor 64
object-attribute structure 272ff. presuppositions, invalid 213
object language 17 private expression 142ff.
objects, abstract 10, 27 private, expression for the 142ff.
- , concrete 10, 27 private language 122ff., 142ff.
- , empirical 26 product (utterance as) 6f.
- , non-existent projection rule 228lf.
occurrence (of a word) 6 pronouns 170, 195ff.
- , extensional 45f. proper names 24ff., 169, 193
- , intensional 45f. - , ostensive 32f., 48
- , non-intensional 46 property (vs characteristic) 232ff.
onomatopoeia 20 proposition 25f.
ontology 23, 2 8 ,105ff., 243, 261, 281ff. provable 219
openness o f concepts 150, 159f. n. 46 psychologism 53 n. 26
operant 68f.
operators, logical 186 readings (of a sentence) 228ff.
ordinary-language philosophy If., 107 realization (of forms of behavior) 6
realization (of a word) 9fF., I l l
parole (vs langue) 11 reduction, method of 177
particles 174 reference 29f., 37, 73f.
parts of speech 169fF., 193ff. reference system 89f.
performance utterance as 6f., 11 referent 14, 62
performative description 137ff. reinforcement 68f.
performative meaning 135ff. - , secondary 69
performative mode 132ff. relativity, linguistic principle of 25Ilf.,
performative operator 138 278ff.
performative utterance 116 resemblance relation 152
performative verb 134 rhema 181 n. 28
performative version 137
perlocutionary act 13If. satisfaction 52
person 173 scientific discourse 13
phonetic aspect 7 SD-grammars, general 222
3 04 INDEX OF SUBJECTS

SD-grammars, simple 222 surface structure 209ff., 225

selector 228 syncategorematic expressions 27
sema 181 n. 28 synonymity 42f., 75ff., 151, 232
semantic aspect 7 synsemantic expressions 27
semantic component 165 syntactic component 165
semantic function 59 syntactic relations 35
semantic region 156f., 255ff. syntax 18, 175ff., 182, 206ff.
semantics 18, 37, 182 synthetic (vs analytic) 79ff.
semiotic 18
semi-Thue-systems 218ff. tense 173, 204f.
sense 38f. theoretic sentences 99
- , identity of 47 thought [iGedanke] 39
sentence design 177 time concept 262f.
sentence model 177 topic 181 n. 28
sentence sequence 176 transformable 218f.
sentences, theory o f : see syntax transformation rules 34, 224f.
sign, general concept 60f. transformational grammar 224f.
speaker (hearer, referent) 14 truth value 15, 30
speech acts 6, 129fF. Turing machines 218
standard names 33, 48f., 188 type 6
states-of-affairs [Sachverhalten] 25 f., type concepts 155ff.
34
state systems, finite 226 universals, linguistic 235
stimulus, preparatory 61 use, criteria of 119ff., 143
- , substitute 61 use (manner, instances of) 112
stimulus synonymy 99f. use (vs mention) 17
structural description 22Iff. utterance, linguistic 6ff.
structure, rules of 222
subject (of a sentence) 173, 175, 206f. value-distribution [Wertverlauf] 186
subject-predicate schema (of verbs 170, 193f., 261ff.
interpretation) 268ff. verification theory 52 n. 2
subject-predicate structure 263f., vocabulary, non-terminal 221
268ff. - , terminal 220f.
substance terms 262 voice 173, 206
substantives 169, 193, 261ff.
substitution 78 weak ordering 153
substitution categories 226ff. word fields 156, 255ff.
substitution principle 39 word formation 174
substitution, rules o f 222, 228 word types: see parts o f speech
superlative 173 world view [Weltansicht] 243ff.
 IN D E X O F L O G IC A L SY M B O LS

” 1 not
A and
V or
z> implies (if - then)
= is equivalent to (if and only if)
A for every
V there is (at least) one
V —n there are precisely n
= is identical with
:= is by definition identical with
—► entails (as a logical consequence)
A null set
kx class of x such that -
class with elements x l9...9x n
e is a member (element) of
n intersection
u union
c= is a subset of
nl intersection of all classes with indices i

ui union of all classes with indices i