Theory of The Linguistic Sign Mulder

JANUA L I N G U A R U M
STUDIA MEMORIAE
NICOLAI VAN WIJK DEDICATA
edenda curat
C. H. V A N SCHOONEVELD
Indiana University
Series Minor, 136

THEORY OF THE
LINGUISTIC SIGN
by
J.W.F. M U L D E R
and
S.G.J. H E R V E Y
1972
MOUTON
THE HAGUE • PARIS
© Copyright 1972 in The Netherlands.
Mouton & Co. N.V., Publishers, The Hague.
No part of this book may be translated or reproduced in any form by print,

photoprint, microfilm, or any other means, without written permission from the
publishers.
LIBRARY OF CONGRESS CATALOG CARD NUMBER: 72-86887
Printed in Belgium, by NICI, Printers, Ghent.

FOREWORD
In our opinion, the SIGN-CONCEPT should be considered to be the

most important issue in linguistic theories. Bloomfieldian linguis-
tics, as well as European (Saussurian and Prague) linguistics were
based on it, and all the important theories that were in the main
stream of these two directions, implicitly or explicitly have made
use of it. Although it did not employ the term 'sign', Bloomfield's
theory of the 'linguistic form' was basically a sign-theory, and
Bloomfieldian linguistics differs from European linguistics in the
first place insofar as the concept 'linguistic form' differed from
that of the Saussurian 'signe linguistique .
A theory of the linguistic sign not only determines the form and
content of 'grammar' and 'phonology', but that of 'semantics' as
well. It pervades and determines every area of linguistics, including
that of 'phonetics'. If it is a powerful theory, it may lead to a
powerful linguistics. If it harbours inconsistencies, these incon-
sistencies will be perpetuated throughout the whole of the linguistic
theory. Theories that lack some sort of a sign-theory (such as
Traditional Grammar and some recent, mainly psycholinguistic,
theories), are theories without a backbone, and the activities of the
adherents of such a theory are confined to nibbling at the surface,
while the linguists concerned are doomed to live with gratuitous
and often far-fetched assumptions, or with circularities and incon-
sistencies.
It is some time ago now that de Saussure, Bloomfield, and Hjelm-
slev, the most important theoreticians of the linguistic sign in this
century, stated their theories. These were very interesting and im-
6 FOREWORD
portant theories at the time. Looked upon as stages in the history

of linguistics, they are still very important, and Hjelmslev's theory
is even nowadays intrinsically important. From our present level
of sophistication — linguistics has gone a long way since de
Saussure and Bloomfield — it is easy to condemn Bloomfield's
theory as hopelessly inconsistent. The correctness of this judgement
will be immediately seen by anybody with a feeling for logic, who
carefully reads through Bloomfield's "A set of postulates for the
science of language", Language 2 (1926), and who compares some
statements made in that article and in his book Language (1933)
e.g. "... a form is a recurrent vocal feature which has meaning, and
a meaning is a recurrent stimulus-reaction feature which corres-
ponds to a form" (1926), "... each linguistic form has a constant
and specific meaning" (1933), "Every form is made up wholly of
phonemes" (1926), "Different forms which are alike as to pho-
nemes are homonyms" (1926). For a discussion of some of the incon-
sistencies in Bloomfield's views, see Mulder's "On the art of
definition, the double articulation of language, and some of the
consequences", Forum for Modern Language Studies, V, 2 (April
1969). The Saussurian sign-concept was not inconsistent in itself,
but it could hardly be, as it remained primitive and exemplification
was limited to easy and straightforward cases. It was mentalistic
in an unacceptable way, just as his notion 'opposition' was, but
the remnants of an outdated type of psychology, which also typified
the early days of the Prague-school, can be easily stripped off
without affecting the usefulness of the notions. For contemporary
functionalists, 'signifiant' and 'signifié', in English linguistic litera-
ture usually called 'expression' and 'content', are no longer
intrinsically psychologistic concepts, and nor is the notion 'oppo-
sition'. But functionalists have not done much to develop de
Saussure's sign-concept any further. They have rather conveniently
made use — or should we call it misuse — of its inherent vagueness.
Reference to this vagueness is not meant as a criticism of de Saussu-
re. At the moment of its introduction his sign-concept was a most
brilliant and powerful concept, and, had de Saussure lived longer,
he would almost certainly have developed and expanded it himself.
FOREWORD 7
It was Hjelmslev who presented one of the various possible

interpretations of de Saussure's sign, and who developed it into a
precise and consistent notion within his theory. This is not the
place for going into any of the details; it is enough to say that his
interpretation, or rather the theory it led to — consistent and
rigorous though it was — was not acceptable to linguists of other
schools. For some this was because the problems glossematicians
managed to solve with their system did not exactly coincide with
theirs, or there were differences of emphasis upon, and importance
attached to, different aspects of linguistic description. The non-
acceptance, or, at least, the lack of full acceptance, by functionalists
of Hjelmslev's sign-concept is mainly due to the fact that this
concept does not leave room for "the double articulation" which,
for functionalists, is THE defining property of language.
But there is at least one respect in which we believe that Hjelm-
slev should be followed, namely the notion 'sign' should be defined
in terms of 'relations' rather than solely as a certain type of entity.
For Hjelmslev, 'language' implicity, it seems to us, is a 'structure',
an abstract and theoretical 'construct', not to be 'discovered' by
the linguist, but to be 'established' by him. It is a means of
'accounting' for speech-phenomena, and not to be 'found' in the
speech-phenomena themselves. Its only relation to the speech-
phenomena is that it describes and explains them, unlike some
other structures that may be set up, which do not. It cannot, how-
ever, be demonstrated that there could not be other, quite different,
structures that can account for the same phenomena (though
perhaps not for exactly the same aspects of those phenomena)
equally well. THEREFORE, ONE MAY NOT CONCLUDE FROM THE
APPLICABILITY OF A STRUCTURE TO THE PHENOMENA THAT THIS IS
THE STRUCTURE OF THE PHENOMENA THEMSELVES. In some straight-
forward cases, for instance in the description of a motor-car, we
may perhaps reasonably assume
(a) that the motor-car has a certain structure 1 ; and
(b) that we can set up a theoretical structure that is in all relevant
respects isomorphic with that structure.
1
If only because motor-cars are intentionally CONSTRUCTED.
8 FOREWORD
With respect to speech-phenomena, we would not even go so far

as to make the former type of assumption (though we do not wish
to assume the contrary either), let alone the latter.
It is in this spirit, and along the lines of AXIOMATIC FUNCTION-
ALISM — which is based upon the two premises that only functional
features (i.e. features that are significantly opposed to their ab-
sence) are to be considered, and that 'language' (the hypothetical,
or rather theoretical, entity) has a double articulation (Mulder
1968, p. 10) — that the present Theory of the Linguistic Sign has
been evolved.
The earliest statements about some of the essentials of this
theory are to be found in Mulder's Oxford D. Phil, thesis (1966),
which was revised and published in 1968 as Sets and Relations in
Phonology; an axiomatic approach to the description of Speech. The
actual 'sign-concept' in this book was further elaborated in his
"On the art of definition, the double articulation of language, and
some of the consequences", Forum for Modern Languages Studies
V, 2 (1969). Further refinements were introduced, and some of the
consequences were drawn, in "Linguistic Sign, Word and Gramma-
teme", La Linguistique, 1 (1971). In the meantime, Hervey was
developing a Theory of Semantics based on the premises of
Mulder's axiomatic linguistic theory, in that way adding a vital
component to that theory. This work has now been completed,
and in 1970 it was submitted as an Oxford D. Phil, thesis, entitled
Functional Semantics; a linguistic theory with application to Pe-
kingese. It is hoped that a revised version of that thesis under the
title Axiomatic Semantics will soon be published. Most of Mulder's
work on Linguistic Theory, and on Grammar, since 1968, has been
influenced by the fact that he was Hervey's supervisor, which
involved him deeply in semantic theory. During that period of
close collaboration, extreme care was taken to avoid that statements
made by the one should be inconsistent with statements made
by the other. There has especially been some concern, when Hervey
had to develop a different sign-concept, because Mulder's concept,
though more suitable with respect to grammar, phonology, and
even phonetics, was not sufficiently equipped to deal with semantic
FOREWORD 9
facts. It has, however, turned out to be possible to develop the

relevant notions in such a way that they are non-identical (in fact
they are quite different) but still equivalent, because the one implies
the other, and vice versa. As Mulder's linguistic theory is firmly
embedded into SEMIOTICS, it goes without saying that also the sign-
concept must be consistent with a general semiotic theory. Its link
with semiotics is presented in a joint article by Mulder and Hervey
entitled "Index and Signum", Semiotica (1971), an article that is
itself an attempt to supplement and elaborate upon the discussion
of various types of semiotic system to be found in Mulder's Sets
and Relations in Phonology, whilst incorporating some ideas that
derive from Hervey's semantic studies.
The first chapter of the present work, bearing the same title
as the article just referred to, has partly the same content, but it is
for the rest independent from that article. Chapter II deals with
"Semiotic Systems" in a more elaborate fashion than this is done
in Sets and Relations in Phonology. Chapter III introduces Mul-
der's version of the linguistic sign, and Chapter IV Hervey's.
Chapter III pursues the consequences of this notion, via grammar
and phonology, into phonetics. Chapter IV deals, among other
things, with such notions as 'denotation', 'reference', 'synonymy',
'hyperonymy', 'hyponymy', and with the hypothetical nature of
'sign-identity' assumptions. Chapter V clarifies some controversies
that may arise because of a difference in the use of the term 'deno-
tation' by philosophers and linguists.
In our desire to apply the utmost rigour to our linguistic ventures,
we had — in the absence of a theory that could fully meet our
requirements — to develop such a theory ourselves as a sine qua
non. The essentials of this theory are presented in this work in,
as much as possible, an informal fashion. Optimistic as we are
with regard to the possibility that our readers — even if they may
not consider this theory useful for their own purposes — may find
many points in it that are of interest to them, we have added a
detailed index at the end.
St. Andrews
November 1970.
TABLE O F CONTENTS
Foreword 5
I. Index and Signum 13
II. Semi otic Systems 19
III. The Nature of the Linguistic Sign 26
IV. The Sign as a Class of Utterances 36
V. In Defence of a Denotational Theory of Semantics 53
References 64
Index 66
I
INDEX A N D SIGNUM
'Language' is said to be a communication-system, i.e. a system of

entities that convey information. It is, however, not the only
communication-system. There are innumerable others, natural,
as well as artificial, non-conventional, as well as conventional ones.
But the prerequisite for something to qualify as a communication
system is that it should contain entities, at least two, by which
information is conveyed. Those entities are not necessarily all of
them formally positive entities. In systems that contain only two
entities, one of the entities may formally be the mere absence of the
other, e.g. a red dashboard light in a car, the entities in which
system are the on- and off-state. There must be at least two entities,
otherwise there can be no information. If the red dashboard light
could only be on, the fact that it was on would not inform us about
anything. If it were always on when the ignition was switched on,
and off when the ignition was switched off, it could only inform
us about THAT state of affairs. If, however, it can be off as well as on
when the ignition is switched on — as is the case — but not on
if the ignition is switched off, it can inform us about something that
goes beyond mere information about the ignition, e.g. its off-state
may indicate that the engine is running above a certain speed. If,
moreover, the light may glow under certain conditions, even if the
engine is running above a certain speed, we may be informed about
some other conditions, provided that we know from other evidence
that the engine is running above a certain speed. Just as in language,
we may need, even in this very simple communication system,
some outside knowledge for the correct interpretation of a given
14 INDEX AND SIGNUM
item of information. This does not affect the actual information-

value, only its interpretation. The actual information-value is in this
case a purely physical one, i.e. that the amount of electrical energy
generated and the resultant magnetic field are or are not sufficient
to overcome a certain resistance in attracting a piece of metal
through which an electric circuit that includes the red dashboard
light becomes closed. Its interpretation is also dependent on further
circumstances. But neither the precise information-value, nor the
circumstances, have to be precisely known for the correct inter-
pretation; only those features in them that are pertinent to the
communication. One does not have to be a physicist to glean all
relevant information from a red dashboard light.
There are many types of communication-system. Some of them
are purely natural and for some of them even the fact that they are
communication-systems is not their raison d'être, but is only
incidental. Others are partly natural, or wholly conventional. We
can infer from lightning that there will be thunder, from heavy
clouds that it probably will rain, from the wagging of a dog's tail
that it is happy, from the whistling of a kettle that the water is
boiling, from a clock the time, from a telephone-bell ringing that
there is someone on the line, from the purple bottoms of some
male baboons that they are in love, from a girl's blushing that she
is embarrassed, and from the sentence 'get out', when uttered,
that someone emphatically wants someone to leave. We should
not confuse the information-value with the actual information
conveyed. The information conveyed depends on all the experienced
circumstances and on known facts about the source. This may
include knowledge about the information-value, but it need not.
If we hear a Hottentot (we do not know Hottentot) say something,
we receive all kinds of information, but most probably not enough
of the purported information for the Hottentot to consider the
communication to have been successful. This is mainly because
we are not acquainted with the information-value of the sentence
used in his utterance. The information-value of a given entity has
to do with the PURPORTED information to be conveyed, but, again,
it is not the same thing. We may infer from what the Hottentot says
INDEX AND SIGNUM 15
that he is there, that he is a foreigner, that he has a cold, etc., but

this is not the purport of the communication, unless the Hottentot
happens to be SAYING just that. The success of the communication
in accordance with its purport depends on knowing the informa-
tion-value of the given entity and being able to give that entity a
correct interpretation under given conditions. The 'sender' of the
message makes use of the information-value and may make use
of any number of circumstances as well, and the 'receiver' of the
message employs all these to extract information from the message.
A message may, of course, not be received, but there must be at
least a potential 'receiver', or a supposedly potential 'receiver', for
something to constitute a message. There may, however, not be a
'sender', in the usual sense, as in the case of natural systems. Still,
the 'receiver' will interpret such a message on the basis of its
information-value and, if relevant, the surrounding circumstances.
Let us call anything that has information-value an INDEX, and
the completely abstract information-value of an INDEX its DENO-
TATION. An INDEX is the DOMAIN (left-hand element) in any relation
'INDEX R DENOTATION'. It does not exist, as an index, outside such
a relation, neither does DENOTATION exist outside such a relation.
An INDEX is a 'form that has denotation', and a DENOTATION is the
'information-value of a form'. Therefore we cannot say that an
INDEX is a mere form. It is a form with certain properties that are
themselves not form. Moreover, the term 'form' indicates here
already a class of entities that are themselves 'forms', i.e. all the
variants of that form are together the 'form' we are here talking
about. We may, in fact, say that it is a 'class of forms' but not a
mere 'class of forms'; rather a class of forms each member of which
has the property of having a specific same denotation. Such member-
forms are called 'morphs', or, if there are more than one, 'allomorphs'.
The class of 'allomorphs with the same denotation' is the INDEX.
Within an index we may distinguish between EXPRESSION and
CONTENT, which are merely different aspects of the same thing.
Because INDEX, EXPRESSION and CONTENT mutually imply one
another, they are equivalent. We shall say more about this later
on (Chapter III).
16 INDEX AND SIGNUM
A natural index, as opposed to a conventional one, only needs a

knowledge of the natural phenomena, their laws or causal relations,
for its correct interpretation. This is obvious for the PURELY natural
indices such as lightning in lightning R thunder, but holds equally
true for artificial devices based on the exploitation of laws of
nature, such as the whistle on a kettle, the dashboard light and the
telephone bell. There may be some element of choice as to what
sort of device is used, but the RELATION between the device that
serves as an index and its denotation is a natural one. Convention
does not enter into it at all. There is an 'intrinsic' difference between
the whistling of a kettle, in its function of indicating that the water
is boiling, and a siren, in its function of indicating that an air-raid
is expected. No matter how much knowledge we have about the
workings of a siren, and the causes of its being set in motion,
without knowing the convention we shall never be able to interpret
it in the correct way. We may be able to learn the convention by
experience, it does not have to be explicitly taught to us, but unless
we know the convention, we can never arrive at interpreting the
wailing of a siren as an index for air-raid imminent.
On the basis of whether the relation between INDEX and DENO-
TATION is natural or conventional, we can distinguish between
NATURAL INDICES and SIGNA. It should be understood that for this
distinction it is only relevant whether the RELATOR, i.e. the ' R ' part
of the relation, is natural or conventional, not whether the INDEX
itself is a natural phenomenon or not. Sunrise can, for instance,
be used as a signum with time to attack as its denotation. There is
a convention necessary for sunrise to have this denotation, as there
is no natural relation between the rising sun and the time to attack.
We may further distinguish between relations based on FIXED
conventions, and relations based on OCCASIONAL conventions.
Based on fixed conventions — this has always to be considered
within a certain system — are, for instance, indices in the system
of number-writing. Their information-value is constant, and if we
know the system, and we know the relevant circumstances at the
moment of a particular use, we can give them the correct inter-
pretation, without having to look up their momentary definition
INDEX AND SIGNUM 17
first. Similarly in the Morse-code, the index ' — . ' always stands for
the letter 'a', and ' — . . ' always stands for '1'. Quite different it
is with, say, algebra, or the digital computer. In 'algebra' one has
to know the occasional value of a 'letter', which is given via a
definition for each operation, before one is able to give it an inter-
pretation; and also in the digital computer, which is based on a
system that is formally very similar to the Morse-code, one has, or
at least the computor itself has, to be separately informed about
the occasional information-value of the combinations of elements
of the binary system before a proper calculation can be made.
We call signa that are dependant on a separate (occasional)
definition for their correct interpretation: SYMBOLS; and those with
a wholly fixed conventional denotation: SIGNS. Within systems of
'symbols' there are usually still some conventions operative. For
instance in algebra, the symbols a, b, c, etc., have a different deno-
tational usage from the symbols x, y, z. We shall call these 'proper
symbols'. Also 'proper names' are PROPER SYMBOLS under this
definition, as there is, for instance, a convention that John, Peter,
Paul, etc., may only denote males, whereas Mary, Jane, Julia, etc.
may only denote females. The actual denotation, however, is not
based on fixed convention, but on a separate definition for each,
no matter how much prolonged, operation. As long as you are
operating with the former of the co-authors of this publication
(as an empirical entity), Jan, or Jan Mulder, always denotes
that particular person; but if his parents had christened him other-
wise, or if the patrilinear side of his family tree some time ago in
history had been named otherwise, matters would have been
different. What we want to say is merely that there is an intrinsic
difference between the denotations of a sign such as dog, and a
proper name such as Jan, but there is much less of a difference
between that of the latter and a symbol, such as in algebra. Symbols
whose denotation depends TOTALLY on occasional conventions we
call NONCE-SYMBOLS. An example of a nonce-symbol is whilk which,
if we wish to do so, we can define as 'a white elk with a missing eye'.
But we could equally easily have defined it as anything else and
used it in such constructions as a whilk or so ago, I whilk you,
18 INDEX AND SIGNUM
mind my whilk, etc. This does not mean that one may not borrow a
sign for use as a symbol, e.g. call one's wife 'little pigeon'; or a
'proper symbol' for use as a 'nonce-symbol', e.g. call an ashtray
'Johnny'; but in that case one has transferred a mere FORM from
one system to another, and consequently changed its identity AS
AN INDEX. The identity of an index, natural or conventional, sign
or symbol, proper symbol or nonce-symbol, depends on its dis-
tinctive function in respect of the other elements in the system, and
consequently on the system it belongs to itself.1
We call communication-systems that contain conventional
indices, i.e. signa, SEMIOTIC SYSTEMS. A 'semiotic system' is 'any
system of CONVENTIONS for communication' (Mulder, 1968, p. 10).
1
For a more detailed treatment of "Index and Signum", see the authors'
article bearing that title in Semiotica IV: 4 (1971), 324-338.
II
SEMIOTIC SYSTEMS
A semiotic system is any system of CONVENTIONS for communica-

tion, i.e. any system that contains signa. We can, on the basis of
the previous discussion, sub-divide semiotic systems into systems
that contain signs and systems that contain symbols. The latter
we can sub-divide into systems with proper symbols, and systems
with nonce-symbols. 1 A 'system' is a self-contained set of func-
tional entities, i.e. entities that are separately relevant to the purport
of the whole (Mulder, 1968, p. 10). The purport of a semiotic
system is 'communication', i.e. the conveying of information.
There is, however, also a more interesting way of sub-dividing
semiotic systems, but before we discuss this, we should like to
mention in passing yet another way of classification, i.e. one into
systems in which the elements are of a discrete, or of a non-discrete
nature. By the former a greater precision in communication can be
achieved, but by the latter the actual communication can be made
infinitely richer, because infinite gradations are possible. Systems
such as 'language' contain both types of element. This distinction
is important, but we cannot deal with it in the present work, except
very briefly. An example in 'language' is 'intonation', fully inter-
woven with 'speech' in actual communication; so much inter-
woven, in fact, that it has to be considered in many cases as an
integral part of the form of a construction. Yet, on the basis of a
possible classification into discrete and non-discrete systems 'in-
1
Such systems are not very interesting. We may reckon among those
occasional colour, letter or number codes.
20 SEMIOTIC SYSTEMS
tonation' can conveniently be treated as a separate system, an

auxiliary system to 'language'. Also sets of symbols (proper symbols
or nonce-symbols), and any other systems that can be set up on
the basis of this classification — if one wants to include them in
one's description — are best regarded as different, perhaps coor-
dinated or auxiliary systems. In the description of 'structure', as
opposed to 'systems' proper, the distinction can become again
irrelevant, and may then be ignored, e.g. John hit Paul can be
treated on a certain level as constructionally equivalent to the man
hit the ball (in spite of the fact that John and Paul are proper
symbols, not signs, and hence belong to a different system), and
even non-discrete features may be regarded as integral parts of the
structure. As we said, however, we cannot go deeper into this
matter here.
We can — and this is of great interest for the study of language
— further sub-divide semiotic systems as to whether they contain
complex elements, i.e. elements that are a combination of other
elements, and as to the nature of this complexity. This gives us first
a division into simple and complex systems, and the latter are
further subdivided into unordered and ordered complex systems.
As there are two types of ordering possible, a purely formal or
phonological one (the term phonological in a very wide sense, just
meaning 'having form', without reference to meaning) and a
simultaneously formal AND meaningful or grammatical one, we can
further sub-divide into 'phonological systems' and 'grammatical
systems', and systems that have both 'phonology' and 'grammar'.
What we call 'language' has both orderings, and this is referred to
by Martinet's dictum of the 'double articulation' of language. We
disagree slightly with Martinet on epistomological grounds, i.e.
we regard the two articulations as independent, and not the second,
or phonological, one as a continued operation performed on gram-
matical elements (i.e. on the signifiants of signs). The reasons for
this will become clear in Chapter III.
The following tree-diagram exemplifies the above classification.
A few of the points discussed in the previous chapter are also
implemented in the diagram:
SEMIOTIC SYSTEMS 21
Semiotic systems
simple complex
!
unordered ordered
I I I
only phonology only grammar both phonology
| | and grammar
wholly fixed not wholly fixed not

conventions wholly fixed conventions wholly fixed
conventions conventions
wholly partly fixed

non-fixed conventions
conventions
For easier reference we give here Table 1 (see overlaef) containing

the same information, together with examples of each type, and
some additional remarks.
By definition a semiotic system must contain signa. Let us call a
'sentence' any signum with such features that it is not an integral
part of another signum. In a simple system then, every signum is a
sentence. In an unordered, but complex, system, every signum,
whether or not a simultaneous bundle of elements, that is not itself
a simultaneous part of another signum, is a sentence. In an ordered
system any signum, complex or not, ordered or not, that cannot
be — b y virtue of certain features, e.g. 'intonation'—itself a p a r t
of another internally ordered or unordered signum is a sentence.
Consequently we may say that any semiotic system contains
SENTENCES. An animal cry (if belonging to a semiotic system) is a
sentence. A traffic-sign is a sentence. The combination ' . — ' i n the
Morse-code, signifying the letter 'a', is a sentence. Certain com-
binations of 'states' in a digital computer are (potentially) sen-
tences. So are 'syntagms' corresponding to 'statements' in algebra,
mathematical logic and arithmetic, e.g.
J
22
£ffl
Semiotic systems (systems containing signa)
complex
(combination of elements into higher level elements)
ordered
(combinations of elements between which there are
ordering-relations into higher level elements)
unordered
(simultaneous, i.e. only phonology only grammar
unordered, (single articulation) (single articulation) both phonology
simple bundles of and grammar
not wholly fixed
(no combination elements wholly fixed not wholly wholly fixed (double
conventions
of elements constituting conventions fixed conventions articulation)
into higher higher level conventions wholly partly |
level elements) elements) non-fixed fixed
conventions conventions j
non-discrete : non-discrete: 1. algebra language

SEMIOTIC SYSTEMS
1. gestures bees' dance morse-code digital arithmetic nonce-symbols 2. mathemat- Language is the
2. animal cries discrete: In language computer In language In language ical logic type of semiotic
In language also : traffic signs also: also: also: In language system with both
1. interjections In language also: phoneme- syntactic syntactic com- also: articulations, and
2. intonation 1. phonemes as complexes complexes plexes Syntactic it incorporates or
discrete: complexes of containing complexes uses as auxiliary
1. siren for air- dist. features nonce-words containing devices all the
warning 2. morphology proper names other types.
2. police-whistle
only paradigmatic relations both paradigmatic and syntagmatic relations
SEMIOTIC SYSTEMS 23
a2 — ab = (a — b)a; x u y = n 12 x 3 = 36, and

so is any syntagm when considered in a capacity of, systemically
speaking, not being coordinated with, interordinated with, or
super- or sub-ordinated to other syntagms. If we say 'systemically
speaking', we mean within that semiotic system. There may be,
in actual use, a system, semiotic or not, super-imposed over a
semiotic system, e.g. with respect to 'language' a certain literary
form, a certain narrative order, certain emotional overtones, etc.
In the case of the Morse-code, there may be a system of ordering-
relations superimposed, e.g. ' • •' means ' m a n ' but the
sentences, as far as the Morse-code is concerned, are the combi-
nations of dots and dashes that signify letters of the alphabet, etc.
I.e. there are here, as far as the Morse-code is concerned, T H R E E
sentences.
The further ordering does not belong to the Morse-code but,
in our example to the writing-conventions of English, another
semiotic system altogether.
The S E N T E N C E is the M A X I M U M S I G N U M in any semiotic system.
In SIMPLE systems it is the O N L Y type of signum. The distinction
made in the classification between purely phonological and purely
grammatical systems is not based upon whether the system contains
grammatical elements, as any system contains signa, at least
sentences, and signa are grammatical elements. Grammatical, as
opposed to phonological, elements are elements that have both
form and meaning, as mere different aspects of the same thing,
i.e. of the signum. In the case of 'symbols' this 'meaning' is of a
'potential' nature, but the sole justification for a 'symbol' is that
it M A Y be D E F I N E D . The case of 'nonce-symbol' is slightly different,
as something is not a nonce-symbol, but just a form, until it
actually H A S B E E N defined. If it might not be given a definite meaning
(by supplying a definition), it would not be a 'symbol'. 2 The sole
2
An interesting example of SYMBOLS, is a TECHNICAL VOCABULARY (e.g. in a
linguistic theory). FORMS are directly or indirectly borrowed from another
system (e.g. an 'ordinary language') or systems, and then defined for one's
purposes. This connection with the original indices, i.e. with the system from
where they are borrowed, is a DIACHRONIC matter. In fact, a TECHNICAL VOCA-
BULARY is—with regard to its domain of deployment (e.g. a particular linguistic
24 SEMIOTIC SYSTEMS
justification for a phonological element is that it can be part

(perhaps the only part) of an INSTANCE OF THE FORM OF A SIGNUM.
By 'form' we mean purely systemic, i.e. functional, form. It does
not have any reference to its substance, nor to its particular realiza-
tion at a given moment. Yet, it is important to talk about an
instance (not to be confused with realization) of a given form, as,
especially in language, the form of a given sign may be rather a
class of different phonological forms than a phonological form
itself. Only the 'members' of such a class, i.e. instances, may be
said to have phonological form proper. If we say that a 'sign' in
language has phonological form, something we may say for con-
venience sake, we do not actually mean this literally. What is, for
instance, THE phonological form of the 'plural' moneme in English ?
If we say that this moneme has phonological form, we can only
mean that not all the members of that class are 'zero', not that
the form of the 'plural' moneme in English consists of a phoneme
or a sequence of phonemes, or has any other single phonological
feature. Some of the instances of the form of that moneme consist
of a phoneme, phonemes, or other phonological features. In fact
all instances, except 'zero' (as in the plural of 'sheep'), have phono-
logical form in that sense.
As Martinet has pointed out 3 — and this is the beauty of his
moneme-theory — it is immaterial whether phonological form can
be described in terms of phonemes. It is sufficient that there are
certain phonological features by which a form can be distinguished
from all other forms. This holds also for classes of forms, i.e. the
'phonological forms' of signs. As long as they differ in respect to
one element (e.g. the 'plural' and the 'genitive' in English have
not exactly the same class of phonological forms corresponding
to them), they are formally different classes. We come back to this
below.
theory) — a system of PROPER SYMBOLS. It is a system of proper symbols, not

of nonce-symbols, as there are generally some conventions operative, and those
conventions — it has to be admitted — have, at least partly, to do with the
DIACHRONIC origin of the symbols (technical terms).
3
A. Martinet, A Functional View of Language (Oxford, 1962).
SEMIOTIC SYSTEMS 25
In those systems where the form of a signum has only one mem-
ber, as, for instance, in the Morse-code, we can ignore the distinc-
tion between the form of a sign, and an instance of that form. In
systems where also the converse is true, i.e. where the same form
always belongs to the same sign, i.e. where there are no HOMONYMS,
such as, again, in the Morse-code, we can also ignore the distinc-
tion between 'phonological form' (in both senses) and 'expression',
a vital distinction in language, as we shall see.
As we were saying, the distinction made in the classification
between purely phonological, and purely grammatical systems, is
not based on whether the system contains signa. It is based upon
whether it contains complex ordered elements that are articulated
into signa (in which case those complex ordered elements must be
signa themselves) or into phonological entities. In the latter case,
the complex ordered elements must, strictly speaking, be phono-
logical entities themselves, but in such cases as in the Morse-code
those phonological entities completely coincide with the signum,
and therefore we may ignore the distinction.
It must also be understood that the classification is based on
what is intuitively TYPICAL for the system in question, except that
there is an inverse hierarchy from simple, via unordered, complex,
ordered, to doubly ordered. That is, as soon as elements belonging
to a higher step in the hierarchy are found in terms of complexity
or ordering, the system is reckoned to be of the 'higher' type. But
in systems of 'symbols', such as algebra and mathematical logic,
one finds 'signs' as well, i.e. signa the meaning of which is deter-
mined by completely fixed conventions, e.g. in algebra ' + ' , '—',
' = ' , etc., and signs of that kind used in mathematical logic.
Typical for algebra and mathematical logic, however, is their use
o f SYMBOLS.
If, in a semiotic system, we find sub-systems that are of the same
kind as the full system of another semiotic system, we treat it, in
analysis, just as we would treat that other semiotic system, e.g. we
describe the morphology of a language in a similar way as we
would describe the system of traffic-signs, mutatis mutandis of
course.
Ill
THE N A T U R E OF THE LINGUISTIC SIGN
The two aspects under which we can view the linguistic sign, i.e.
its 'formal' aspect and its 'meaning-bearing' aspect, called 'sig-
nifiant' and 'signifié' respectively by de Saussure, are in English
usually referred to as EXPRESSION and CONTENT respectively. It is
generally agreed that the two are inseparably united and that the
one implies the other, and vice versa, i.e. they are in a one-to-one
relation of mutual implication for each instance of a sign. It is not
possible to set the notion 'sign' up in such a way that it is 'meaning'
alone that determines the identity of the sign. If one did that,
synonyms would be identical signs. A sign would then be a certain
'meaning' attached to no matter what form, i.e. there must be
'form', but what that form is, would be totally irrelevant. Also to
let 'form' exclusively determine the identity of the sign would have
undesirable consequences. In that case, there could be no homo-
nyms, i.e. hair and hare in spoken English would be identical signs,
and, moreover, there would be as many signs as there are phonolo-
gical forms of signs, e.g. there would be several signs of the plural
(plural-monemes) in English. The notion 'sign' would refer to a
certain phonological form to which meaning is attached, but what
kind of meaning was attached would be irrelevant for the identity
of that sign.
So, either one must deny the possibility of synonyms, or that of
homonyms, or one must agree that the identity of signs is as much
determined by their form, as by their meaning. But this can only
be done if we regard 'form' and 'meaning' with regard to signs
merely as different aspects of the same thing. Otherwise the situation
T H E NATURE O F T H E L I N G U I S T I C SIGN 27
would be even worse than in the case of solely depending on 'form',

or solely depending on 'meaning'. Not only would all different
'forms' belong to different signs, but all similar 'forms' with
different 'meanings' would be different signs as well, and in the
end there would be as many signs as there are utterances, as we
may assume that no two utterances ever mean exactly the same,
nor are they formally (we would not be able to arrive at 'distinctive
function', i.e. at 'phonology') ever exactly the same; we would,
in other words, have no criteria to establish 'formal' or 'semantic'
equivalence between utterances. In order to limit the number of
signs in the inventory we must establish certain forms as being
equivalent, and this can only be done on the basis of 'meaning'
correlations, or we must establish certain 'meanings' as equivalent,
and this can only be done on the basis of 'form' correlations.
Therefore we may say that the meaning of a sign implies its form,
and the form of a sign implies its meaning. From this follows that
(a) different signs have neither the same form nor the same meaning,
and (b) form and meaning of signs are not just form and just
meaning, but both are, in a sense, form as well as meaning.
Consequently the terms 'form' and 'meaning' used of signs do
not, and cannot, have the same value as that which those terms
normally and intuitively have. They are entirely different things.
Therefore, they are given different names: in French 'signifianf
and 'signifie', in English 'expression' and 'content'.
We may, then, say that a sign is the CONJUNCTION of an EX-
PRESSION and a CONTENT, and that a particular 'expression' and
a particular 'content' mutually imply each other. Each of them
implies a specific sign, and vice versa. Therefore, in a mathematical
sense, the notions 'sign', 'expression' and 'content' are equivalent.
They represent three ways of looking at the same thing. Using the
term 'sign' implies looking at the 'sign' in its totality, using
'expression' implies looking at the 'sign' from a formal angle, and
using the term 'content' implies looking at the 'sign' from the side
of meaning. Expression and content are each other's converse,
nothing more, and moreover, because they mutually imply one
another, they are equivalent. Except for that which is connected
28 THE NATURE OF THE LINGUISTIC SIGN
with the difference in 'aspect', everything we say about one of the

three can be equally said about any of the three. One would
normally be inclined to say that an 'expression' is a class of
allomorphs, or that the 'sign' is a class of allomorphs, because when
we speak about allomorphs we are usually either primarily concerned
with form or with the totality of 'form and meaning'. We could,
however, equally well say that the 'content' is a class of allomorphs,
as will become clear below. If we do that, the allomorph referred
to is, as a relational structure, the converse of the allomorph as
a member of the expression, but, as there is mutual implication
between the relation and its converse in this case, the converse
is equivalent to the relation it is the converse of. Similarly, if we
say that a 'sign' is a class of allomorphs, we imply that an allomorph
is a conjunction of the previous two relations, relations which are
equivalent to each other, and each equivalent to the conjunction
of the two. We can symbolize this as follows:
sign = e &c, i.e. a particular 'sign' is the conjunction of a particular

'expression' and a particular 'content'.
e <-» c, i.e. a particular 'expression' and a particular 'content'
mutually imply one another, and are therefore equiva-
lent to one another.
e sign and c <->• sign, i.e. both a particular 'expression', and a
particular 'content' mutually imply a particular 'sign',
hence e, c and sign are equivalent.
We may, then, say that:
sign *-> (e <-> c), i.e. a particular sign is equivalent to a particular
expression and a particular content in an equivalence
relation.
Let us now determine what e, c and sign are. If we determine what

one of them is, e.g. e, we know what the others are, as, c is the
CONVERSE of e, and sign is the conjunction of e and c.
If we assume, i.e. a priori accept, that, say, the different forms
of the plural in English belong to one and the same sign, then we
must accept that the so-called 'phonological' form of a sign is
THE NATURE OF THE LINGUISTIC SIGN 29
actually a CLASS of phonological forms. I shall symbolize a phono-
logical form by p, and a class of phonological forms by {/>}.
{p} = pi u pi ... KJpn, e.g. the class of phonological forms that
constitutes the phonological form of the plural in English equals
/iz/ u /z/ u /Z/ u /rn/ u / 0 / u /men ~ man/ u /mais ~ maus/
u /uaivZ/ ~ /uaif/ u ... etc. 1 We cannot go into matters of pho-
nology here, but the phonological representations (written between
slant lines) refer to the forms of the plural-allomorphs in such
words as houses, days, huts (or heads) (i.e. the archiphoneme /Z/,
symbolizing a suspension of opposition between /s/ and /z/), oxen,
sheep, men, mice, and wives respectively (see: Mulder, 1968;
Mulder and Hurren, 1968). The symbol (in fact this is a 'sign') ' <J'
signifies the 'union' of terms. The whole is called a logical sum,
which means that the terms stand in an 'either ..., or ..., or both'
relation, i.e. x u y means either x, or y, or both x and y. A class
is here, for all practical purposes, the 'logical sum' of its members.
It is clear that, if the members of a class stand in a ' u ' relation,
each member can REPRESENT the class. In the case under considera-
tion, i.e. that of the form of a certain 'sign', the members (individual
phonological forms) of the class are equivalent in respect to the
sign, and each of them can therefore 'represent' the form of the
sign in question.
A class {p}, if it is the phonological form of a sign, stands in a
certain relation R with (i.e. it 'has') a certain distinctive function,
distinctive that is within the domain to which that sign belongs,
i.e. grammar. The distinctive function of an entity is determined
by the class of entities with which it commutes (Mulder, 1968).
The distinctive function of a 'sign' is determined by the set of signs
with which it commutes, i.e. those 'signs' to which it is opposed in
equivalent contexts. The allomorphs, i.e. variants with regard to
their phonological forms, of signs are always COMBINATORY or
CONTEXTUAL variants. Therefore allomorphs do not commute
with one another, i.e. they are not opposed to one another. In the
case of FREE variants, e.g. perhaps /ai<3r/ and /iRdr/ 'either' for some
1
For the phonemic notation of English used in this work, see Mulder and
Hurren, 1968.
speakers in some dialects of English, or /felOu/ 'fellow' and

/blOuk/ 'bloke', these are considered to be SYNONYMS, i.e. different
signs, SYNONYMS are different signs, but with all their SEMANTIC
FEATURES (Hervey, 1970) in common. As SIGN-IDENTITY depends
in the first place on the distinctive function of a sign, and as distinc-
tive function depends on the item being opposed to other signs
(or to 'zero') in a certain position, COMBINATORY VARIANTS OF
SIGNS, (e.g. / and me) if having the same semantic features, cannot
be regarded as different signs, but free variants have to be regarded
as different signs. They are FORMALLY DIFFERENT, and they are
opposed to, i.e. they have a different distinctive function from, each
other.
If two tentative signs are both FORMALLY DIFFERENT and OPPOSED
to one another, they constitute different signs: SYNONYMS, if they
are denotationally the SAME; JUST DIFFERENT SIGNS, if they are
denotationally DIFFERENT.
If two tentative signs are FORMALLY DIFFERENT but NOT OPPOSED
to one another, NOR DENOTATIONALLY DIFFERENT, they are one and
the same sign, i.e. COMBINATORY VARIANTS.
If two tentative signs are FORMALLY THE SAME but DENOTA-
TIONALLY DIFFERENT, e.g. hair and hare, they are DIFFERENT SIGNS,
whether opposed to one another (in a strict paradigmatic sense;
of course in the overall inventory different items are a priori
distinctive, i.e. opposed to one another) or not. In that case we
call them HOMONYMS.
If two tentative signs are both FORMALLY and DENOTATIONALLY
THE SAME — which, incidentally, implies that they are NOT OPPOSED
to one another — they belong to one and the same sign. So, the
same phonological forms may either pertain to the same sign, or to
different signs (i.e. HOMONYMS), and different phonological forms
may either pertain to the same sign, in which case the formally
different items are called combinatory variants (i.e. allomorphs),
or they may pertain to different signs. In the latter case the signs
are called SYNONYMS, if they are denotationally the same, i.e. if
they have identical sets of semantic features (Hervey, 1970).
As just an example of what difficulties we would let ourselves in
for if we used formal SIMILARITY (e.g. /aidr/ ~ /iRJr/) rather than

formal IDENTITY as a criterion to distinguish between SYNONYMS
and other types of free variants, we should like to say the following.
If we cannot give a precise criterion for what is still similar and
what is not, the whole procedure would remain an arbitrary one,
and nothing would prevent us from regarding fellow and bloke as
just free variants, but identical signs, as they have the phonological
feature / l O u / in common. Also, without Hervey's precise criteria
for semantic equivalence, or, in fact, semantic identity, we would
be unable to demonstrate that hair and hare are different signs, as
both may refer to soft, hairy objects. We cannot deal with those
criteria here in extenso, but it is, for the purpose of this discussion,
enough to say that if two terms d o not imply the same set of MORE
GENERAL TERMS (called 'hyperonyms' by Hervey), they are seman-
tically non-equivalent. There is a hare, implies, among other things,
'there is an animal', but there is a hair does not. Therefore, hare
and hair are semantically non-equivalent, and this implies that if
they are signs (which they are), they are different signs. The reverse
(i.e. the condition of having two terms that do imply the same set
of more general terms) implies semantic equivalence, but not
necessarily sign-identity.
Mulder symbolizes distinctive function of a sign b y ' s ' , in order
to distinguish it from distinctive function in phonology, which
latter he symbolizes b y ' d ' (Mulder 1968, 1969, 1971). In earlier
publications (Mulder, 1968, 1969), he has equated the distinctive
function of signs with their SEMANTIC FUNCTION. Since Hervey's
refinement of the latter notion, we recognise that there is a difference
between the two, but only with a view to cases of 'synonymy'.
SYNONYMS are different signs, and therefore their distinctive func-
tion differs. Their semantic function is, however, by definition,
the same, as they share the same semantic features.
We can now symbolize 'expression' by {/?}Rs, i.e. a certain class
of phonological forms standing in a relation with a certain gram-
matically distinctive function, or, in plain language, a certain class
of phonological forms with a certain distinctive function which
allows it to carry a certain non-phonological information. The
content is the CONVERSE of this, i.e. JR{/>}, where R indicates the

converse of R , and, together with the permutation of domain and
counter-domain, it indicates the converse of the whole relation.
The sign, being the conjunction of expression and content, i.e.
sign = e & c, is therefore {/)}RJ & JR{/?}. The intuitive inter-
pretation of {p}Rs is: {p} in its capacity of having the specific
function s. The intuitive interpretation of ¿ R { p } is: the function
j of {/>}.
Because {p} may consist of more than one member (one of the
members may even be 'zero', or less than 'zero', as in the case of
the plural of French 'bceuf') one can say that {p}Rs is a set,
containing the members / n R i , p^Rs, etc. That is to say =
/nRy u / ^ R i u ... u /}nR.i. The members />iR.s, etc. are called
MORPHS. If there are more than one i.e. if there are different morphs,
they are called ALLOMORPHS. Similarly JR{/>} = sR/>i U sR/72
u ... u jR/7n, and sign = .piRs & sR/?i u /?2Rj & sRp2 u ...
u pnRs & sRpn. A s these are equivalent (i.e. sign, expression and
content are equivalent) we can call them all allomorphs, i.e. the
sign is a class or set of allomorphs, so is the expression, so is the
content. F o r convenience's sake, because we concentrate on form
here, we shall use the EXPRESSION for our further discussion, but
we could have used equally well the content, or the sign in its
totality, though the latter would be clumsy, because of the longer
formula. So {p} = pi u p2 u ... u pn, i.e. a set of phonological
forms, and {/>}Rr = p i R s u />2Rj u ... u pnRs, i.e. a set of
allomorphs. One sees that not only is there a distinction between
'expression' (signifiant) and 'morph' (or 'allomorph'), but also
between 'morph' (or 'allomorph') and 'phonological f o r m ' . Only
the latter can be directly described in terms of phonological features,
e.g. phonemes. This should be absolutely clear, for h o w could an
allomorph, let alone a class of allomorphs, be analyzed into
phonemes ? Unless one is prepared to say that, for instance, "/iz/
R 'plural' " can be analyzed into "/i/ R 'plural' & /z/ R 'plural' ",
i.e. unless one assigns to phonemes a grammatically distinctive
function as well as a phonological one, saying that an allomorph
CONSISTS of phonemes is logically contradictory. Yet, most linguists
say this not only of allomorphs, but even of the 'expression', i.e.
a CLASS of allomorphs.
We have said that a p, member of a class {p}, is a phonological
form, i.e. a phonological feature which often, in its turn, is a
complex of phonological features. This is, of course, not entirely
correct, as also 'zero' may be a member of such a class, and this is,
of course, not a phonological feature. One should actually say
that p is a phonological form or feature, or — in the case that p
is a member of {/>}, and one of the members of {p} is 'zero' —•
p may be 'zero'. This is how it should be understood, but it would
be clumsy to say this every time we want to say what {/»}stands for.
In fact, perhaps, even the term 'phonological f o r m ' may be mis-
leading, i.e. 'phonological feature' may be better. In the case of
French la grande montagne blanche we have a discontinuous phono-
logical feature /a ... d ... môtan ... s/, in the case of men we have
/e/ ~ /a/ (i.e. /men/ as phonologically distinct from /man/) and
in the case of French bœufs we have /bô/ ~ /bôf/ or even 0 ~ /f/
(i.e., as it were, 'subtraction' of /f/)). All these we can call 'phono-
logical features', no matter whether we can describe them ex-
clusively in terms of phonemes or sequences of phonemes. One of
Martinet's examples (Martinet, 1968) is the Latin case-endings,
e.g. orum in dominorum, where there is amalgamation of case,
number, and a, grammatically speaking, non-functional gender-
feature. Still we can abstract the phonological form of each allo-
morph of the monemes present, e.g. the allomorph of the 'plural',
by, keeping everything else the same, opposing it to its commutant,
here a particular allomorph of the 'singular'. The phonological
form of this particular allomorph of the plural-moneme is, then,
orum ~ i. Other allomorphs of the plural have the forms arum ~ ae,
ae~a, etc. Unlike under the Bloomfieldian 'morpheme-theories',
under Martinet's 'moneme-theory', there is no difficulty at all in
recognizing and dealing with any type of functionally differential
feature, and we should exhaust all the possibilities that are inherent
in this approach. Therefore, if we are talking about features, we
mean any type of differential feature. Something is a phonological
feature (or form), if it is functionally (though not grammatically)
differential. Something is a grammatical feature, if it is grammat-

ically differential. Something is a phonetic feature, if it is phonetic-
ally (i.e. acoustically or articulatorily) differential.
Members p of a set {p} which constitutes the domain of {/>}RJ
stand in a relation to s, which simply means that they are GRAM-
MATICALLY differential. As such, i.e. as a p, however, they are just
phonological forms or features, i.e. they are only PHONOLOGICALLY
differential. Any phonologically permissible (i.e. allowed by the
rules of the phonological system in question) feature or combination
of features constitutes a p of the language, and any set of p con-
stitutes a {p} (Mulder, 1969). If, and only if, such a {p} is, and is
being considered as, the domain of a particular {/>}R.s, then such
a p is the phonological form of a MORPH (or allomorph), and such
a {p} is the phonological form (but 'phonological f o r m ' has here
a quite different meaning, i.e. it means 'a CLASS of phonological
forms') of an EXPRESSION.
Let us now see what a p in itself actually is. As we have said
already, it is a phonologically distinctive (i.e. differential) form or
feature. It can be regarded as a class, i.e. a class of ALLOPHONES.
But an allophone too is not just a phonetic form or feature. A
phonetic form or feature is an ALLOPHONE if, and only if, when,
and only when, it has phonologically differential function, just as
a phonological form or feature is an ALLOMORPH, if, and only if,
when, and only when, it has grammatically differential function.
That is, also allophones ARE not phonetic features or forms, but
they HAVE phonetic features or forms. Now, the phonetic features
or forms of allophones do not, in actual fact, belong to a finite
and discrete set, as n o two productions of sounds are ever exactly
identical, and an infinite number of 'in-between' realizations are
possible. However, for convenience's sake, we may bring in here,
a bit of distortion of the DATA, and act as if they were a set of
discrete entities. This is, for instance, done if we use, no matter
how narrow, a transcription by the aid of the symbols (in fact
within their system, i.e. the 'alphabet', they are 'signs') of the
International Phonetic Alphabet.
If we symbolize such a rather imaginary phonetic feature or
form by / , and a phonologically distinctive function (assessed by

commutation, as in the case of grammatically distinctive function)
by d, we may represent a PHONOLOGICAL form or feature by { f}Rd.
Analogous to = /?IRJ U / ^ R s U ••• U p n R s ' we may say
{f}Rd ==/iRrf uf2Rd u ... u fnRd. That is, a phonological fea-
ture or form is a class or set of allophones, just as a sign or an
expression is a class or set of allomorphs. Analogously to the fact
of {p} in {p} Ri, if we say that a phoneme HAS PHONETIC FORM, we
use the term 'phonetic form' in a quite different sense from if we
say that an allophone has phonetic form, i.e. the phonetic form of a
phoneme is a logical sum of phonetic forms, i.e. it is a CLASS of
phonetic forms. As long as we are aware of this, we may use the
same term for the two different things.
We may, however, never say that an allophone, let alone a
phoneme, is phonetic form, and consequently we cannot analyze
an allophone, let alone a phoneme, into phonetic features or forms,
any more than we can analyze a morph (or allomorph), let alone
an 'expression', into phonological features.
In a way analogous to the definition of a particular EXPRESSION
as 'a specific CLASS of phonological forms {p} in its capacity of
having a specific function s in the grammatical system in question', 2
and a corresponding ALLOMORPH as 'a specific phonological form
in its capacity of having that specific function s\ we may say that a
particular PHONEME is 'a specific CLASS of phonetic forms {/} in
its capacity of having a specific function d in the phonological
system in question', and a corresponding ALLOPHONE is 'a specific
phonetic form in its capacity of having such a specific function d\
The symbol s stands, as we have said, for grammatically (i.e. qua
sign) distinctive function, and the symbol d stands for phonolog-
ically (i.e. qua form) distinctive function.
2
CONTENT is defined as 'the function s of {/>}' (see the forgoing, a n d Mulder
1971).
IV
THE SIGN AS A CLASS OF UTTERANCES
The notion sign defined as Expression ^ sufficiently

Content
refined concept to satisfy the purposes of grammar. As it has been
shown in the previous chapter, with the notion as it stands, 'sign'
can be linked up with phonological form, or, more appropriately,
with phonological features, and ultimately, through the phono-
logical features, with phonetic form or features.
When one comes to semantics, however, the need becomes felt
for providing a link between the sign and its 'meaning'. This
'meaning', roughly speaking, concerns the capacity to provide
information (which can be inferred from realizations of the sign)
about some entity or entities other than the sign itself or its reali-
zations. This can be symbolized as follows :
realization of sign 'a' provides information b (where b ^ 'a')
sign 'a' has information value b (where b ^ 'a')
Consequently, we may say that, whereas the sign appears as a

member of a semiotic set, for instance the set of signs in a 'language',
the information value is a priori, an entity outside of the semiotic
system in question.1 What we mean by 'linking the sign to its
1
The elements of semiotic systems can of course be projected, when we
consider them as denotables, into the extra-linguistic world. Both the sign
man in Man is an animal, and the sign man in man is a noun, have extra-
linguistic reference, except that the denotable corresponding to an utterance of
THE SIGN AS A CLASS OF UTTERANCES 37
meaning', is the bridging of the gap, by using certain relations,

between signs, which are within semiotic systems, and their in-
formation value, which lies outside the semiotic systems. For the
very reason that these entities belong to the complement of the
given semiotic domain, e.g. language, and this complement can be
considered to be the whole of the universe excluding the given
semiotic domain, we consider the relation of linguistic sign to
information value as a relation between a linguistic element and
an extra-linguistic entity.
We propose to brave the disapprobation of philosophers, (see
also the subsequent chapter), and use the term DENOTABLE for any
entity (entity in the widest possible sense) when considered outside
of a realization of an INDEX R INFORMATION relation. A DENOTABLE
considered as the right hand term of such a relation we shall then
call a DENOTATUM, for, once we make the assumption that an entity
is the information inferred from the realization of an index, that
entity is not potentially DENOTABLE, but is actually BEING DENOTED.
We do not think that it is counter-intuitive to use the term DENO-
TATUM in our sense, for denotata are usually conceived of as extra-
linguistic entities to which signs refer.
Here we must mention the fact that an entity which may else-
where constitute the left hand term of a 'realization of sign R
information' relation, may in turn be considered as a DENOTABLE
itself, and may therefore itself become, in turn, a DENOTATUM
within such a relation. We shall, however, want to distinguish,
for instance, between an utterance of the sign horse and the
DENOTATUM 'the sign horse', for the two are clearly kept apart
according to their occurrence as left hand and right hand terms,
respectively, in relations of 'realization of sign R information'.
Apparent difficulties arise — and the need for dealing with them
will be shown to lead to a definition of 'sign' as an equivalence-
class — when we consider the relation of SIGN to DENOTATUM.
There is an apparent paradox which can be exemplified by taking
the latter is, in turn, a model in a linguistic description of English, whereas the
denotable corresponding to an utterance of the former is the species homo
sapiens.
38 THE SIGN AS A CLASS OF UTTERANCES
the sign 'table' and its information value, or DENOTATION. If we

were to say that the sign 'table' denotes a single DENOTATUM (i.e.
one empirical entity) we should doubtless be wrong, for the whole
purpose of having a sign table in a language, it seems to us, is to
be able to talk about any given TABLE in the universe. From which
it would appear that the sign table denotes, not a single TABLE, but
a CLASS OF TABLES. On the other hand this alternative is also wrong,
for that would make the sign table synonymous with the sign the
class of tables. This is manifestly not so, but rather it is the case
that, every time the sign table is USED (i.e. realized), it denotes a
SINGLE DENOTATUM.
The above paradox can be resolved by the simple device of
distinguishing between the sign table and realizations of the sign
table, for which distinction the alternative paradox itself serves as
a justification. Once we make this distinction, then we can say that,
when the sign is realized, each of its realizations denotes a single
DENOTATUM, and that this single DENOTATUM, when considered
out of context of its relation to the realization by which it is denot-
ed, can be said to be a DENOTABLE.
The link between the linguistic sign and the extra-linguistic
complement of language within the Universe (i.e. the class of all
DENOTABLES) will be visualized as follows:
SIGN
is realized
as an
• UTTERANCE denotes
*• DENOTATUM
which under a different
aspect, can be considered
AS A DENOTABLE
An important fact is that denotables may belong to the field of

study of various non-linguistic sciences, from which criteria,
empirical in the case of denotables studied by empirical sciences,
can be brought to bear upon DENOTATA. The fact that a denota-
tional theory of (linguistic) semantics does not shed light on the
nature of DENOTABLES as empirical entities, cannot be considered
to be a shortcoming of that theory, for here, there is a case of one-

way traffic. We have suggested that criteria from the various sciences
studying what we call DENOTABLES, might usefully be brought to
bear upon DENOTATA, and therefore play a part in semantic des-
cription. We are NOT suggesting that criteria from semantics can
necessarily be usefully applied in non-linguistic sciences.
In order to build the link between SIGN and DENOTATUM (and
ultimately DENOTABLE) into the theory itself, some changes have
to be made in the notion of the sign as outlined in the previous
chapter.
If an utterance is a single realization of a sign (in fact it is a
MODEL for a single realization, as we shall see), then by adding
together all the realizations of the sign into a potentially infinite
class
{ u i U U2 U U 3 U ... Un}
we should arrive at the definition of a given sign as an equivalence-

class, 'equivalent' used in the sense that each utterance 'represents',
as it were, the same sign. Now, if each utterance of the sign denotes
a single DENOTATUM, then the equivalence-class of utterances that
is the sign, will correspond to a whole class of the individual
denotata of its member utterances. Thus, to return briefly to the
example of table:
corresponds to a class of denotata

table (sign) whose members are
single tables
1
is instanced as is instanced as
an utterance denotes a single denotatum

"/teibl/" table
In defining the sign as an equivalence-class of utterances, we must

be careful to avoid a contradiction which we shall outline below.
The avoidance of this contradiction together with the necessity

of incorporating the notion DENOTATUM in the semantic theory,
has led us to include UTTERANCE (or utterance-model) as a model
in the theory itself.
Usually when Mulder speaks of the distinction between sign
and its realizations, he equates it with a clearcut distinction drawn
between ABSTRACT MODEL and CONCRETE DATA. Sign is a notion
in the model; its realizations are in the realm of the data. The
relation between the model and the data described is a certain
isomorphism, whereby the model, say the sign, is said to ACCOUNT
FOR a certain number of realizations. Thus, to say that the sign
both ACCOUNTS FOR a set of realizations, and is a class of those
realizations, would be to imply an absurdity. Firstly, this statement
would imply the false conclusion that the relation of isomorphism,
that is of 'descriptive application' (e.g. sign x applies to a certain
number of concrete acoustic realizations) is EQUIVALENT to the
relation of CLASS TO MEMBER. Secondly, an abstract model cannot
directly CONSIST of members which are concrete entities. The
alternatives are then EITHER that SIGN ceases to be an abstract
model OR that the UTTERANCES belonging as members to the sign
as a class are no longer to be considered as concrete realizations,
i.e. as speech-events. The first alternative would be tantamount to a
rejection of SIGN as a theoretical concept in linguistics, and will
therefore be ignored. The second alternative, the one which we
have adopted, leads to the setting up of UTTERANCE as a model in
the theory, as distinct from concrete REALIZATIONS of a sign in the
data. An UTTERANCE can be said to be the following things: It is
a MODEL FOR A SINGLE CONCRETE REALIZATION OF A SIGN. I t i s a
MEMBER OF A CLASS OF EQUIVALENT UTTERANCES, i.e. of a sign. It
has a formal component called a FORM, to account for the phonetic
features occurring in the concrete realization for which it is a model.
It has a referential component called REFERENT, in order to account
for the information conveyed by the concrete realization for which
it is a model. It is a ONE-TO-ONE RELATION BETWEEN A FORM AND A
REFERENT, accounting for the fact that no two concrete realizations
that ARE empirically different may ever be assumed to have (and
probably never do have) exactly the same phonetic features and to

convey exactly the same information. 2
Just as no two concrete realizations of a sign are ever identical
in substance or in information, so no two utterances, which are
models each for separate concrete realizations, may ever be iden-
tical, by definition, in FORM or REFERENT. An utterance can therefore
be said to be the conjunction at a given, or presumed, point in
time and space of a given FORM and a given REFERENT.
While no two utterances may ever be IDENTICAL to one another,
there are, however, certain other relations which may hold be-
tween utterances.
Utterances may be SIMILAR as to FORM, when they have the same
phonological properties (for determining this, criteria from pho-
nology must be applied). These are so called FORMALLY SIMILAR
UTTERANCES. Utterances may be SIMILAR as to REFERENT, when their
DENOTATA are identical on the level of DENOTABLES (the condition
for this is demonstrated or assumed empirical identity). These are
so called REFERENTIALLY SIMILAR UTTERANCES. Utterances may also
b e BOTH FORMALLY a n d REFERENTIALLY SIMILAR.
The only other relations that we need consider as holding
between a number of utterances are that of 'belonging to one and
the same sign' and, of course, that of 'not belonging to the same
sign', which is simply the negation of the former relation. For the
sake of convenience we use the term EQUIVALENCE tautologously
with 'the relation of belonging to one and the same sign'. This is
a special restricted use of 'equivalence', for, strictly speaking,
utterances may be equivalent with respect to various other classes
to which they belong. In our sense here EQUIVALENCE (when applied
to utterances) is to be understood as EQUIVALENCE WITH RESPECT
TO A SIGN. A S we have said, the only purpose of introducing the term
EQUIVALENCE is one of convenience, in order to enable us to refer
to the sign, when it is considered as a class of a certain number of
utterances, tautologously, as a CLASS OF EQUIVALENT UTTERANCES.
Furthermore, by tautology, instead of having to say each time that
2
If for no other reason than the fact that physically distinct events occupy
different positons from one another in time and space.
a number of utterances are all members of one and the same sign,
we can say that these utterances are EQUIVALENT.
Ultimately we can only arrive at the statement that two or more
utterances are equivalent, or that a certain set of utterances is co-
extensive with a class of equivalent utterances (i.e. a sign) via
statements of sign-identity in the description of a language. This
procedure is not circular, for criteria are set up in the semantic
theory (Hervey, 1970) whereby statements of sign-identity, made
as hypotheses, can be tested (i.e. refuted or corroborated, i.e. if
refutation has been attempted, but without succes).
It is our contention, that sign-identity, i.e. equivalence of utter-
ances, cannot, in any description of a language, be proved in a
positive sense. It can only be set up as a hypothesis in the descrip-
tion, and at best, demonstrated via the refutation of all possible
alternative hypotheses, or refuted by confrontation with the data
to which it is meant to apply.
The argument for this is that sign-identity, or equivalence of
utterances could only be subject to positive (empirical) proof, if a
both sufficient and necessary condition for the notion EQUIVALENCE
could be devised from external, empirical criteria applicable to
utterances. To our knowledge there are only two external, empirical
approaches to utterances
a) via their phonological features

b) via their denotata
Consequently, any conditions for positive proof of equivalence

between utterances would have to be based
either on identity of phonological features, that is to say FORMAL

SIMILARITY between utterances,
or on identity of denotata on the level of denotables, that is
to say REFERENTIAL SIMILARITY between utterances
or on identity both of phonological features and of denotata
on the level of denotables, that is to say FORMAL-REFEREN-
TIAL SIMILARITY between utterances.
That neither of these three alternatives will do as a sufficient and

necessary condition for equivalence between utterances, i.e. for
sign-identity, can be demonstrated if we consider the conclusions
to which accepting them respectively as sufficient and necessary
conditions, would lead.
If all formally similar utterances must also be equivalent, then
all utterances having the phonological form /her/ must belong to
one and the same sign. Therefore, homonyms do not exist, and
there is only one sign "hair/hare", which is unacceptable, since
this would lead to an INADEQUATE model for linguistic descriptions,
e.g. certain types of 'ambiguity' would remain unaccounted for,
such as the 'ambiguity' of an utterance "/3er iz r her in mai sup/"
("there is a hair/hare in my soup"). Therefore FORMAL similarity is
not a sufficient — and thus not a both sufficient and necessary —
condition. As a matter of fact, it is not a necessary condition either,
but we need not demonstrate this here.
If all referentially similar utterances must be equivalent, we
obtain the following absurdity. Let us take a denotable whose
empirical identity is premised, say the entity Elizabeth II, Queen
of England. Any two utterances denoting this denotable must by
definition be referentially similar, and consequently, by our first
premise, also equivalent. Therefore an utterance "/sRi/" ("she")
denoting Elizabeth II must be equivalent to an utterance "/5r
prezrnt kuRin of irilrnd/" ("The present Queen of England").
Which is absurd, since the utterance "/sRi/" is an utterance of a
simple sign, whereas the utterance "/5r prezrnt kuRin of irilrnd/"
is that of a complex sign, and since one and the same sign cannot
be both simple and complex in grammatical structure, there must
be two separate signs she and the present Queen of England to
which the above two utterances belong respectively. These two
utterances are therefore not equivalent. Consequently, REFEREN-
TIAL similarity is not a sufficient (and so not a both sufficient and
necessary) condition for equivalence. It is, as a matter of fact,
not a necessary one either.
If only FORMAL-REFERENTIALLY similar utterances may be equiv-
alent then no utterance "/haus/i" denoting a house in, say, London,
44 THE S I G N AS A CLASS OF UTTERANCES
may ever belong to the same sign as any other utterance "/haus/2",
denoting a house in, say, San Francisco. Furthermore this would
imply that the denotation class of EVERY sign consisted of a single
member when considered on the level of denotables. This would
be tantamount to denying the ability of realizations of one and
the same sign to communicate more than one (empirically the
same) information, and therefore in direct contradiction to one
of the tenets of this theory of the sign. Consequently FORMAL-
REFERENTIAL similarity is not a necessary condition (and so not a
both sufficient and necessary condition) for equivalence. It is a
sufficient condition (for two or more utterances to be equivalent),
but this is trivial for the problem of sign-identity.
As the above approaches do not lead to a criterion which is both
sufficient and necessary for equivalence, we must conclude that
equivalence cannot be determined through positive empirical
criteria, or indeed, be solved on the utterance-level at all. It must
therefore, in each description of a language, be decided on the
level of the sign, and this is done by making and testing hypothetical
assumptions.
In the previous chapter a brief mention has already been made
of the criteria that operate on "tentative signs" (i.e. on hypothetical
assumptions of sign-identity). We cannot here go into these proce-
dures, and the theorems that underlie these procedures, as these
do not directly concern our theory of the sign so much as have to
do with the setting up of signs in linguistic descriptions.
A consequence of treating the sign as a class of utterances, and
of making UTTERANCE a model in the theory, is that we can make
use of the relations of set-theory in various operations with classes
of utterances. 3
For instance, by specifying a certain phonological feature or
set of phonological features, we can class together all the utterances
which have that feature or those features, into what Hervey (1970)
calls a FORM-CLASS (not to be confused with a class of phonological
3
There are all kinds of secondary advantages as well, e.g. 'puns' can be
a c c o u n t e d f o r a s ONE REALIZATION t o w h i c h c o r r e s p o n d TWO OR MORE UTTER
ANCES (models, members of different signs).
forms). Thus for instance, we can talk of the FORM-CLASS {"/iz/"},

namely the class of all utterances which have the phonological
feature /iz/. Since the members of such a class are all of the same
order as those of a sign (i.e. in both cases the members are utter-
ances), we may use a set-theoretical approach in relating form-
classes to signs. We may take for instance the form-class {"/iz/"}
and the sign to be (assuming here that its identity has been estab-
lished previously). Their relation will be a relation of PARTIAL
INTERSECTION, as shown below:
sign to be
A ~ B = members of {"/iz/"} which do not belong to to be

e.g. members of "plural" (/horsiz/ 'horses')
B ~ A = members of to be which do not belong to {"/iz/"},
e.g. utterances of am as in / AM here.
AB = members of both to be and {"/iz/"},
e.g. utterances of is as in he is there.
The possible relations between any given form-class and any given
sign are as laid down by set-theory :
I. Partial Intersection
II. Proper Inclusion

III. Disjunction
form
sign
class
IV. Total Overlap
foim sign
class
We can now come to a different definition of morph and allomorph,

which is nevertheless equivalent to that given in the previous
chapter. A morph is the intersection of a form-class with a sign
(and is therefore, itself, a class of utterances). A sign has allomorphs
if and only if it intersects with more than one form class, that is to
say, if it has more than one morph. When we say that this notion
allomorph is equivalent to the notion allomorph discussed in the
previous chapter, we can substantiate this claim. Just as in the
earlier approach a sign can be regarded as a class of allomorphs,
so it can under the latter definition, with the only difference, that
in the former the allomorphs of which a sign is a class are themselves
regarded as single entities, whereas in the latter approach they
themselves are, in turn, classes. However, there is no contradiction
implied in saying that a sign is a class of classes (allomorphs) whose
members are utterances. The two notions ALLOMORPH are equiv-
alent, by virtue of having the same relation (of member to class)
to SIGN. They differ only in internal structure, in a way analogous
with the distinction between the treatment of SIGN in the previous
chapter, and that in the present. Namely, in the foregoing we did
not treat either of them as a class of utterances, simply because
we can dispense with the notion UTTERANCE in grammar. A s stated
earlier, the treatment of sign as a class of utterances comes into
semantics in order to give access to denotation.
In what follows we propose to show that
(a) the notion 'class of equivalent forms' is equivalent to the

notion Expression;
(b) the notion 'class of equivalent referents' is equivalent to the
notion Content;
(c) thenotion'sign'asaclassofequivalentutterances,is equivalent
to the notion 'sign' as given in the previous chapter.
Each utterance in a class of equivalent utterances, i.e. in a sign,

is a one-one relation between a form and a referent. When viewed
from a formal aspect, all the forms of the utterances within one
sign, constitute a class of forms, which are equivalent with respect
to that sign. This class of forms we can call a CLASS OF EQUIVALENT
FORMS.
Similarly, when viewed from a referential aspect, all the referents
of the utterances within one sign constitute a class of referents,
which are equivalent with respect to the sign. This class of referents
we can call a CLASS OF EQUIVALENT REFERENTS.
A class of equivalent forms, as a class, implies, and is implied by
a sign. Furthermore, it is a class whose members, individually
imply (due to the one-one relation of form and referent in utter-
ances), and are implied by, single referents belonging to the class
of equivalent referents, proper to the same sign.
Conversely, a class of equivalent referents, which, as a class,
also implies a sign, and vice versa, is a class whose individual
members imply, and are implied by, single forms belonging to the
class of equivalent forms proper to the same sign. From this it can
be seen that a class of equivalent forms is a certain CLASS OF FORMS
in a relation with a certain CLASS OF REFERENTS, and a CLASS OF
EQUIVALENT REFERENTS is a certain class of referents standing in a
relation (which is the converse of the former relation) with a
certain CLASS OF EQUIVALENT FORMS. Therefore it would appear:
(a) that the class of equivalent forms and the class of equivalent
referents proper to one and the same sign, are the converse
of one another;
(b) that they both imply, and are implied by, the sign in question;
(c) that they are mutually equivalent aspects of the sign viewed
from the point of view of 'form' and of 'meaning' respec-
tively;
(d) that the sign is the conjunction of its class of equivalent
forms and its class of equivalent referents.
If we now compare this with the terms EXPRESSION and CONTENT,

we can conclude that these are equivalent to CLASS OF EQUIVALENT
FORMS and CLASS OF EQUIVALENT REFERENTS respectively, since we
find that:
(a) expression and content are the converse of one another
({/>}RJ and respectively);
(b) they both imply the sign in question, and vice versa;
(c) they are mutually equivalent aspects of the sign viewed from
the point of view of 'form' and of 'meaning' respectively;
(d) the sign is the conjunction of its expression and its content.
On all these points CLASS OF EQUIVALENT FORMS is equivalent to

EXPRESSION and CLASS OF EQUIVALENT REFERENTS is equivalent to
CONTENT. By demonstrating this, we have already demonstrated
also, that the two notions SIGN are equivalent, but we shall outline
below the extent of this equivalence:
(a) the sign is a class of allomorphs (in both views);

(b) the sign is the conjunction of EXPRESSION (<-> class of equiv-
alent forms) and CONTENT (<-* class of equivalent referents);
(c) the sign implies its expression
(<-> class of equivalent forms), and vice versa;
(d) the sign implies its content
(«-» class of equivalent referents), and vice versa.
The isomorphism stops only at 'the sign implies a class of equiv-
alent utterances, and vice versa\ but this is a trivial consequence
of the fact that the former view (presented in the previous chapter)
does not include the notion UTTERANCE in the model.
For a discussion of the descriptive potential of a model in which
the sign can be regarded as a class of utterances, we must turn to
semantics proper. As it has already been intimated, a sign, whose

identity has been established, will have as its members a potentially
infinite, but nevertheless determined set of utterances. To each of
these utterances corresponds the denotatum which it denotes, and
therefore to the whole class of utterances corresponds the class
of denotata whose members are denoted by the single utterances.
Any set of utterances (e.g. a set of {"/teibl/", "/kat/", etc.}) can
be said to have a class of denotata corresponding to it; but when,
and only when, the set of utterances in question is co-extensive
with a sign, the corresponding class of denotata shall be called
DENOTATION CLASS.
Once the identity of the sign has been determined, that sign can
be said to govern a specific denotation class; we may say that
governing this denotation class is a property, the semantic property,
of the sign in question. The members of denotation classes are
denotata, and denotata can be considered on another level as
DENOTABLES, to which outside criteria may be applied. Therefore,
we can compare the semantic properties of signs; that is to say, we
can set up semantic relations between signs by asking ourselves
whether one and the same denotable may ever be a denotatum
belonging to the denotation classes of all those signs, or, alter-
natively, whether one and the same denotable may ever be a
denotatum belonging to the denotation class of one of those signs,
but not to the denotation class of the other sign, or signs.
Here again, any of the relations possible in set theory may hold
between the respective denotation classes of two signs. See diagrams
overleaf.
When the denotation class of the sign x properly includes the
denotation class of the sign y, we say that the sign x is a HYPERO-
NYM of the sign y, and that the sign y, in turn, is a HYPONYM of the
sign x 4 An instance of such a relation from English would be that
of horse and stallion, where the sign horse is a hyperonym of the
sign stallion, and the sign stallion a hyponym of the sign horse.
When the denotation class of the sign x totally overlaps with
4
Lyons uses the terms "superordinate term" and "hyponym" in a rather
similar, though less formal, way (Lyons, 1968).
50 THE SIGN AS A 3S OF UTTERANCES
Den Y
Den X 5 properly includes Den Y 8
Den Y properly includes Den X
Den X partially intersects with Den Y,

and vice versa
Den X is disjunct with Den Y Den Y
Den X totally overlaps with Den Y
that of the sign y, we say that signs x and y are SYNONYMS. An

instance of such a relation in English would be that of the sign
stallion and the sign adult male horse. The validity of the statement
that stallion and adult male horse are synonyms in the sense that
synonymy is used here, depends on whether or not we can refute
the proposition that 'any denotable which may be a denotatum
belonging to the denotation class of stallion, may also be a deno-
tatum belonging to the denotation class of adult male horse, and
vice versa'. A single contradictory instance is sufficient to refute
the statement that stallion and adult male horse are synonyms, but
in the absence of such refutation, they will be described as synony-
mous signs.
N o special semantic relations are set up between signs whose
denotation classes intersect partially, or are disjunct. Partial inter-
section becomes interesting for semantics in cases where the inter-
section of two or more denotation classes is co-extensive with
the denotation class of yet another sign. For instance the denotation
5
Den X refers to the denotation class of the sign x.
9
Den Y refers to the denotation class of the sign y.
classes of the signs to refer to (class A in the diagram below) and to

allude to (class B in the diagram below) intersect in such a way that
their intersection is co-extensive with the denotation class of to
mention.
A ~ B: denotation class of to refer to something at length (not briefly)-,

B ~ A: denotation class of to allude to something without being explicit;
AB : to mention (i.e. to refer to something briefly and explicitly).
We cannot, without exceeding the scope of the present work, do

more than to mention the use of relations between denotation
classes in the setting up of SEMANTIC FEATURES for signs. Here we
need a notion 'direct hyperonym', which, without going into
details, can be defined as 'a hyperonym z of the sign x, such that,
in the language in question, there is no sign y which is both a
hyponym of the sign z and a hyperonym of the sign x\ In brief
exemplification of this complex statement we note here that, in a
case where the denotation classes of signs x, y and z (Den X, Den Y
and Den Z respectively) are in a relation as shown in the diagram
below, the sign z is a hyperonym, but not a DIRECT hyperonym of
the sign x, whereas the sign y is a DIRECT hyperonym of sign x.
A further use of the HYPERONYM-HYPONYM relation and of the

relation of SYNONYMY, is that signs can be ordered into a HIER-
ARCHICAL NETWORK7 o n t h e b a s i s o f t h e s e r e l a t i o n s . A sketchy

r e p r e s e n t a t i o n o f s u c h o r d e r i n g , i n w h i c h a c o n n e c t i n g line r e p r e s e n t s
a h y p e r o n y m - h y p o n y m r e l a t i o n , is g i v e n in f i g u r e 1.
Fig. 1.
7
It may be interesting to note that the 'semantic structure' of languages is
most faithfully represented as a hierarchical network, and not as either just
a hierarchy, or just a network.
8
Synonyms of stallion are, for instance, adult male horse, adult male
mammal of the species 'horse', etc. These synonyms have qua signs (by require-
ment) exactly the same semantic features. In order not to complicate the
discussion, we have left out of consideration the sign gelding (castrated adult
male^horse), which may refute the hypothesis that adult male horse, etc., and
stallion are^synonyms.
V
IN DEFENCE O F A DENOTATIONAL
THEORY O F SEMANTICS
Most of the arguments against denotational theories known to us

come not from linguists, but from philosophers. This places us in
a paradoxical situation, where, although we are neither competent,
nor, indeed, particularly inclined to take our stand in the domain
of philosophical theory, we feel nevertheless obliged to present a
philosophically slanted defence for our semantic approach. It is
our purpose here to try to indicate the particular use of 'denota-
tum' in our theory, and to show that, some of the absurdities
normally incurred by denotational theories, are avoided by our
use of that term. We shall also allude to the scope of denotational
semantics as compared to 'theories (general) of meaning', and the
advantage of restricting this scope, as compared to other possible
semantic theories.
We shall start by re-iterating a part of our discussion from the
previous chapter. In that chapter we have discussed the possibility
of resolving a paradox concerning denotation, by distinguishing
between SIGN and UTTERANCE, as well as, in a parallel way, between
DENOTATION CLASS (class of denotata corresponding to a SIGN) and
DENOTATUM (corresponding to, or denoted by, an UTTERANCE).
It has also been stated in the previous chapter that denotata,
when considered outside of their relation to utterances, are being
viewed, on another level, as DENOTABLES. That empirical criteria,
from the sciences studying entities in the empirical world, CAN be
brought to bear on DENOTABLES, need not imply that positive
characterization of all denotables (and therefore of all denotata)
is always possible. What is more, this statement is not to be inter-
54 IN DEFENCE OF A DENOTATIONAL THEORY OF SEMANTICS
preted in a logical positivist sense as implying that all denotables

are empirically existing entities (directly experienceable through
the senses or through an extension of the senses).
At least some philosophers will agree that there are entities other
than physical objects 'existing' in some sense of the word. It is our
contention that at least one of the functions of human communica-
tion (e.g. speech) is to DENOTE entities, whether physical objects
or not.
It follows from Mulder's axioms and definitions, 1 that this is
the sole raison d'être of grammatical entities in some communica-
tion systems (e.g. the sentences ['letters'] of the Morse-code, which
cannot, as such, be used to 'connote', only to denote) and specif-
ically of linguistic signs (qua signs) as defined in our theory (this
will be discussed further on).
The term 'denotable', and even the term 'entity', is not entirely
devoid of 'substantive' implications. In our theory they are, how-
ever, to be understood in a sense entirely deprived of these 'sub-
stantive' overtones. By this is meant not only that entity and deno-
table are not to be interpreted as: 'empirical (physical) object',
but that they are not to be interpreted as: 'object', in any sense
at all. Some denotata, of course, may be 'objects' (e.g. table,
unicorn), some may even be 'concretely existing objects' (e.g.
table), but apart from these, there are denotables which are neither.
Attributes, actions, processes, relations and all sorts of complex
facts and circumstances are also denotables, and may in turn
become the denotata of utterances.
Consequently, when considering denotables, it is immaterial
for our purposes, whether they are entities whose existence has
been instanced in a concrete sense, entities whose existence has
not been instanced, but may yet be instanced, entities whose
existence can never in fact be instanced, or entities whose existence
can never in principle be concretely instanced. Of the first kind,
entities such as table, red, hit, underneath (object, quality, action
and relation respectively) would serve as examples. The case of the
1
See Mulder, Sets and Relations in Phonology (Oxford, 1968), 10-12.
IN DEFENCE OF A DENOTATIONAL THEORY OF SEMANTICS 55
Abominable Snowman can furnish an example of a entity which,

although it has not been instanced to date, may at some future
time be empirically observed. A n interesting instance of an entity
(a complex circumstance) belonging at some time to the latter type,
but becoming relegated to the type of entities whose existence can
never IN FACT be instanced, is The moon is made of green cheese.
Before any actual empirical observation of the substance of which
the moon consists was possible, the fact (entity) that The moon is
made of green cheese, was potentially instanceable. Since empirical
observation of the surface of the moon, this complex entity has
changed its status to 'entity that can never IN FACT be observed'.
This change in the status of the denotable entity may perhaps effect
the truth value of the proposition "The moon is made of green
cheese", but we see no reason to suppose that it has in any way
affected the information value of the sentence (sign) The moon is
made of green cheese.
Our use of the term 'denotatum' is free of restrictions of scope
on the basis of any possible non-linguistic classification of deno-
tables. It is probably in this regard that our use of 'denotatum' is
most unlike the use philosophers make of that term. Philosophers
start by specifying what conditions a denotatum as an ENTITY
(denotable) must fulfil before one can say that it is 'denoted' by a
linguistic expression. A t the most restricted, only concrete objects
can be said to be 'denoted', but even at best, there will still be a
specification as to what type of entities may be said to be denoted.
Where this line is drawn may vary considerably, but drawn it
usually is, rather arbitrarily, it seems to us, but whether it is drawn
arbitrarily or not is irrelevant to the present discussion. On the
other hand, in our approach to denotation, which is via the sign
TOWARDS denotables (and not vice versa, as seems to be the case
with philosophers), we merely note the fact that denotables may be
classified in various ways, but insist on disregarding such classifica-
tion as a possible criterion in a definition of 'denotatum'. Since
we are using 'denotable' in the widest sense, and 'denotatum' in
the sense of 'denotable being denoted', it would be absurd for us
to say that 'SOME denotables, such that they are denoted, and such
that they fulfil certain empirical conditions as well, are DENOTATA'.2

By this we would imply both that 'all entities are denotables' and
that 'entities are denotables if an only if they fulfil certain em-
pirical conditions'. Either one or the other of these contradictory
statements must be rejected. Which one is rejected, that is to say
how the term 'denotable' is to be understood, depends, of course,
on the purpose to which the notion is to be put. In semantics it is
not our purpose to construct a model for denotables, but rather
a model for speech phenomena. This in itself is a good reason for
choosing 'all entities are denotables' irrespective of subclassifica-
tion of entities, as a definition in semantics (entity is left as a
primitive term, but its use has been explained to be the widest
possible). Furthermore, philosophers who have opted for the
restricted use of 'denotatum' (note that they do not usually distin-
guish between the level of denotata and the level of denotables)
are fond themselves of demonstrating that their notion 'denota-
tion' (under their own definition) is a useless one. Therefore it
seems to us justified on at least two counts to adopt the wider
definition of 'denotatum'. The following chain of definitions will
clarify our view:
'denotatum' for 'denotable being denoted';

'denotable' for 'any entity' (i.e. anything that can be, directly
or indirectly, referred to);
'being denoted' for 'occurring as the right hand term in a
relation a R b, where a is the realization of an index and b is
the information conveyed by a, provided that b ^ a\
From the above chain of definitions it might seem that we are

indulging in a trivial tautology to the effect that 'everything that is
the information conveyed by the realization of an index is a deno-
tatum' and 'everything that conveys information is the realization
2
We are thinking here specifically of the view according to which, say, table
(in red table) DENOTES (because it 'refers to' something classified as an OBJECT),
whereas red CONNOTES (because it 'refers to' something classified as a QUALITY
of the object denoted).
of an index'. This, however is not the case, for it is always in-

dependently decidable, that is without recourse to circularity,
whether a given element conveys information or not. If from the
given element some entity other than its own identity is being
inferred, then, in that instance, the given element is a realization of
an index, and the entity inferred from it is its DENOTATUM. Whether
or not we can say any more about that denotatum, say on the level
of DENOTABLES, is not necessarily material to semantics. We might
of course be interested in comparing two or more denotata on the
level of denotables, where proof of their identity or non-identity
may be used for setting up relations (e.g. of referential similarity)
between the realizations (of indices) to which they correspond
respectively. But for proof of such identity or non-identity positive
characterization of denotables is not necessary. Unicorns do not
have to exist, nor do we have to know their 'defining qualities'
(this has to do, in any case, with devising paraphrases which are
ultimately trivial), before we can say with a high degree of certainty
that an entity unicorn is not identical to other entities such as table
or atom or force or fairy. In the case of two realizations of the sign
unicorn it may be undecidable whether their denotata are identical
or not on the level of denotables, but for all our purposes in
semantics, it is sufficient that they MAY be identical.
In the same way, whatever realizations of and denote (and we do
not wish to be drawn into giving a paraphrase, since paraphrasa-
bility is a comparatively trivial, and certainly entirely language-
bound property), we can say with certainty that they denote some
denotatum, for they indubitably convey information (other than
their own, presumably acoustic, identity). That they do so, follows
inevitably from the fact that they differ in information value from
realizations for instance of or (a book and a pen as opposed to the
constructionally (i.e. syntactically) equivalent a book or a pen).
Furthermore, one needs no concrete specification of the denotata
of realizations of and in order to be able to determine that these
denotata are, on the level of denotables, non-identical to the deno-
tata of realizations of signs such as table, unicorn and or. Whether
two realizations of and have or have not denotata identical on
the level of denotables may not be decidable. But we may assume

for instance that this relation (denotable) is the same, provided it
holds between the same empirical objects in identical conditions
of time and space. This 'same' relation can be denoted by more
than one realization of and, and therefore two realizations of and
MAY have denotata which are identical on the level of denotables.
The same DENOTABLE (relation) may be denoted by a realization
of together with, implying that the denotation class of and inter-
sects with that of together with. A s long as intersections between
denotation classes, and the type of intersection, are decidable in
our view of denotation, we feel that our use of the terms 'denota-
ble', 'denotatum' and 'denotes' are justified by their usefulness
in leading to a consistent semantic description of speech.
Another line of attack must be covered. The example has been
used by philosophers of the evening star and the morning star as being
both denotationally equivalent to the planet Venus and to one
another. Clearly, before the identification of the planet Venus, the
two were not synonymous in even the narrowest sense, in fact
they would have had to be described as having non-intersecting
denotation classes. If they have become synonymous, this would
seem to be purely a result of the change in the status of their
respective denotata from being regarded as non-identical on the
level of denotables to being regarded as identical on that level.
Here non-linguistic facts would, then, appear to affect linguistic
descriptions. This is true, but only in a sense, as will be seen in
what follows. The conventional information values of the two
expressions, that is to say their denotation classes, were originally
demonstrably disjunct only on the assumption that these classes
were limited (because of the way these expressions were USED by
speakers) to 'particular star visible in the evening but not in the
morning' and 'particular star visible in the morning but not in the
evening' (two non-intersecting classes), respectively. N o w it could
be argued that this limitation is still extant and that therefore we
have two different approaches to the denotation of the expressions
in question, leading to two diametrically opposed conclusions,
namely that
a) realizations of the morning star and the evening star denote

one and the same denotatum (the planet Venus) and therefore there
is total overlap between their denotation classes
b) the denotation classes of the same two expressions are
disjunct.
Such a paradox can, however, only be obtained by falsifying the
LINGUISTIC data. The denotata of realizations of these signs must
of course be entities AS SPEAKERS KNOW THEM, and not as they are
qua entities, but individual speakers do not have to be aware of
the synonymy of expressions when they realize them. If we were
accounting for 'the native speaker's intuitive knowledge of his
language' we should have to say that some speakers are aware of
the fact that the morning star and the evening star are both denota-
tionally identical to the planet Venus, while other speakers are not.
But what we are describing is SPEECH DATA, in other words, roughly,
what people SAY. NOW we shall clarify what we meant at the be-
ginning of this paragraph by "falsifying the LINGUISTIC data". If
on the one hand we accept the data equating (denotationally) the
two expressions above, with the planet Venus, then we must ask
the same informants from whom this data was gathered, as to
whether it is true that 'the morning star is visible both in the
morning and in the evening'. The informant, KNOWING about the
planet Venus (and this fact was GIVEN) cannot negate this outright,
but will say "Yes, but in the evening we call it the evening star"
(or some such). In other words, it will no longer be the case, as it
will become clear from the data, that the DENOTATION (as opposed
to 'connotation') of the morning star will be limited (according
to the speaker's usage) to 'star visible in the morning but not in
the evening'.
Consequently, we can say that the discovery of the planet Venus
and the knowledge of the fact that the same entity is denoted by
utterances of all three expressions the morning star, the evening
star and the planet Venus, will be seen to have affected the SPEECH-
DATA. By this we mean simply that, whereas previously, speakers
would have said categorically that "the morning star can be seen
in the morning, but not in the evening", now at least some speakers
will say that "the morning star can be seen in the morning AND
in the evening". This alteration in the speech-data, provided by
informants, will lead to setting up a new sign the morning star
in such a way that it is semantically (and therefore also qua sign)
non-identical to the sign the morning star as it was set up previous
to the discovery of the planet Venus. The empirical fact of this
discovery affects sign identity only because it affects the speech-
data; if it did not do so, it would have no linguistic repercussions.
Therefore, it is not true to say that empirical discoveries, as such,
affect linguistic descriptions, only that changes in the speech-data
may cause changes in the linguistic description (which is as it
should be), and empirical discoveries, of course, may occasion
changes in the speech-phenomena.
Yet another point which can be demonstrated from the example
of the morning star and the evening star is one that cannot be
sufficiently stressed. When we say that, denotationally speaking,
these two expressions are SYNONYMOUS both with the planet Venus
(sign) and with one another, we most emphatically DO NOT imply
that there is no significant difference of 'meaning' (in the wide
sense) between these expressions. Indeed we firmly believe that
there are other (by tautology NON-DENOTATIONAL) meaning-
differences between them, or at least between respective utterances
of them. Most denotata of utterances of the evening star TEND TO
belong to the class of 'star seen in the evening', and most denotata
of utterances of the morning star TEND to belong to the class of
'star seen in the morning'. This however, is only a TENDENCY in
the utterances, and since it is not a general rule, it may not be made
into a characteristic (semantic) feature of the SIGNS in question.
Rather than making this tendency a feature of the sign on a deno-
tational level, where it would lead to contradictions, we could
specify on another (perhaps stylistic, but certainly non-denota-
tional) level that the expression (sign) the morning star is CORRECTLY
USED when it correctly implies (not in the logical sense) that its
denotatum actually belongs to the class of 'star seen in the mor-
ning'. In a similar manner, one could specify the CORRECT USE of
the evening star. But this would be definitely on a level other than
that of the DENOTATION of SIGNS, for the latter is concerned only

with WHAT utterances (any and all) of a sign may denote, and does not
extend to exploring under which circumstances it is a more or less
'correct usage' for those utterances to denote what they denote.
In other words, all the information conveyed by utterances, if it
is of a WHOLLY FIXED CONVENTIONAL nature (and may not vary
between members of a class of utterances) is accounted for as the
DENOTATION of the sign of which the utterances are members.
Denotation classes are set up, semantic features determined, on
the basis of WHOLLY FIXED CONVENTIONAL information values.
This is neither more nor less than what is demanded by the defini-
tion of SIGN.3 Consequently, if a realization conveys information
other than a member of the denotation class of the sign of which
it is a realization, we must conclude that this information is NOT
the denotatum or part of the denotatum of the UTTERANCE which
is a model for the realization, as well as being a member of the sign
in question. We may then either say that that information is
'connoted' by the utterance (and TENDS to be 'connoted' by the
sign) or that it is DENOTED not by the UTTERANCE, but by some
other model (e.g. stylistic, but not a strictly linguistic one), set up
on a different level, for the given realization. We believe, in other
words, that the strictly LINGUISTIC 'meaning' of realizations [DA
prezidant kikt bAkit] ("the President kicked the bucket") is no
more and no less than that of realizations [<?e prszidant dajd] ("the
President died"), and the additional overtones of 'disrespect or
dislike' in the former, are NOT strictly a matter of LINGUISTIC
'meaning'.
We have already indicated, both in the first chapter of the present
work and in our article on indices 4 , that elements are often present
in the speech phenomena (such as tone of voice and the 'meanings'
conveyed by this, or the conveying of other than the 'purported'
information, e.g. the fact that the speaker is present, etc.) which,
although they are functions of 'meaning' in the realizations, are
3
See also especially the authors' article, "Index and signum", Semiotica, IV,
1971,4.
4
Ibid.
not accounted for on the level of signs. Such elements do not form
a part of denotational sign-semantics. Strictly speaking therefore,
we should not say that 'a linguistic UTTERANCE (member of a
linguistic sign) connotes a certain information value', for by
definition, whatever information value the actual UTTERANCE as a
model conveys is its DENOTATUM or PART OF THAT DENOTATUM, a
member, or part of a member, of the denotation class of the sign
to which the given utterance belongs. It must be clearly understood
that, wherever the realization of a sign is regarded as conveying
information other than what is determined by the FIXED CONVEN-
TIONS governing the information value of the sign whose realization
it is, we are really dealing with a model for that realization on
some level other than sign-semantics (such as psychology, or
stylistics, etc.). If this is understood, then there is no harm in
defining 'connotes' in such a way as to make it a function of the
UTTERANCE (albeit an indirect one).
'utterance connotes x' for 'the realization corresponding to the

utterance conveys information x, such
that x is not the denotatum (or part of
the denotatum) of the utterance'.
We may, furthermore, characterize signs (on a non-denotational,
and therefore in our sense non-semantic level) by the tendency of
their utterances to connote certain types of information,
'sign tends to connote x' for 'a conspicuously large number of
(but by requirement NOT ALL) utter-
ances of the sign connote .x'.
In the case, therefore, of gobble, we can say that this sign TENDS
to CONNOTE that 'the performer of the action is an animal' (N.B.
it MAY be a human) and a given UTTERANCE of this sign may connote
that 'the action of eating is performed by an animal'. This in-
formation cannot be said to be DENOTED by the UTTERANCE, for it
is not WHOLLY FIXED CONVENTIONAL, but variable between members
(utterances) of the sign gobble. Such information value may well
be a part of the 'meaning' of a given realization, but not of the
denotatum of an utterance "/gobl/". Therefore we say that denota-
tion is only an 'aspect of meaning'.
We do not believe that our view of 'denotation' implies internal

inconsistencies within the theory of the sign as presented in this
work. Consequently, any arguments against this view have to be
external ones. The objection that we allow utterances (e.g. of and)
to DENOTE, which others might maintain have no denotation, has
been answered, simply by pointing out that we use the term
'denote' in such a way that all utterances of signs do denote.
Other serious objections would arise if
(a) we were to claim that all questions of 'meaning' in speech
can be resolved in denotational sign-semantics;
(b) it could be demonstrated that denotational sign-semantics
blocked the way to the study of other aspects of 'linguistic
meaning'.
The former point has been discussed above, where we made
clear our belief that 'meaning', in the wide sense, has to be studied
on AT LEAST one level other than that of denotational sign-se-
mantics.
The latter point deserves to be mentioned in conclusion of our
discussion. Not only is it not the case that denotational sign se-
mantics impedes access to the remainder of 'linguistic meaning',
but it in fact could be of valuable assistance in defining the scope
(by elimination) of other disciplines studying such 'meaning'. If a
realization conveys information other than what is accountable for
in terms of the denotation of the sign whose realization it is, then
we can be sure of having isolated a feature of the 'meaning' of that
realization on some other level. Contradictions which might arise
on a single-level approach to 'linguistic meaning', can, as we saw
in the case of the morning star and the evening star, be resolved if
a separation is made between levels.
REFERENCES
Alston, W. P.
1964 Philosophy of Language (Englewood Cliffs).
Bloomfield, L.
1926 "A set of postulates for the science of language", Language 2.
1933 Language (New York).
Ebeling, C. L.
1960 Linguistic Units (= Janua Linguarum, series minor 12) (The Hague:
Mouton).
Gardiner, A. H .
1932 The Theory of Speech and Language (Oxford: Clarendon Press).
Harré, R.
1970 The Principles of Scientific Thinking (Macmillan).
Harris, Z. S.
1942 "Morpheme alternants in linguistic analysis", Language 18.
Hervey, S. G. J.
1970 Fnnctional Semantics. A linguistic theorie with application to
Pekingese, (Oxford D. Phil, thesis).
1971 "Notions in the manipulation of non-denotational meaning in
speech", La Linguistique.
Hjelmslev, L.
1953 Prolegomena to a Theory of Language (Bloomington : Indiana
University Press).
Lyons, J.
1963 Structural Semantics (Oxford: Blackwell).
1968 Introduction to Theoretical Linguistics (Cambridge University Press).
Martinet, A.
1960 Éléments de linguistique générale (Paris).
1962 A Functional View of Language (Oxford: Clarendon Press).
Morris, C.
1947a Signs, Language and Behavior (New York). Reprinted in Writings
on the General Theory of Signs (The Hague: Mouton, 1971).
1947b Foundations of the Theory of Signs (Chicago). Reprinted in Writings
on the General Theory of Signs (The Hague: Mouton, 1971).
Mulder, J. W. F.
1968 Sets and Relations in Phonology (Oxford: Clarendon Press).
REFERENCES 65
1969 " O n the art of definition, the double articulation of language and
some of the consequences", Forum for Modern Language Studies V, 2
(April).
1971 "Linguistic Sign, Word, and Grammateme", La Linguistique.
Mulder, J. W. F., and S. G. J. Hervey
1971 "Index and Signum", Semiotica IV, 4.
Mulder, J. W. F., and H . A. Hurren
1968 "The English vowel phonemes from a functional point of view and a
statement of their distribution", La Linguistique.
Ogden, C., and I. Richards
1927 The Meaning of Meaning (New York).
Parkinson, G. H . R., ed.
1968 The Theory of Meaning (Oxford University Press).
Prieto, L.
1964 Principes de noologie ( Janua Linguarum, series minor 35) (The Hague :
Mouton).
1966 Messages et signaux (Paris).
Saussure, F. de
1955 Cours de linguistique générale (Paris: Payot).
Ullmann, S.
1964 Semantics: An introduction to the science of meaning (Oxford:
Blackwell).
Verbürg, P. A.
n.d. "Delosis and clarity", Philosophy and Christianity : Essays dedicated
to Professor Dr. Herman Dooyeweerd (Kampen: J. H . Kok).
INDEX
account for 40 class(es) 15, 24, 29, 32-34, 38, 40,

acoustical 34 44-47, 49, 61
actions 54 class of allomorphs (see also {/>}
adequacy 43 R s) 28, 32, 33, 35, 48
algebra 17, 21, 22, 25 class of allophones (see also { / }
allomorph(s) (see also pi R s, etc.) R d) 34, 35
15, 28, 29, 32-35, 46 class of equivalent forms 47, 48
allophone(s) (sec also fi R d, etc.) class of equivalent referents 47, 48
34, 35 class of equivalent utterances 40-42,
alphabet 23 47, 48
amalgamation 33 class of phonetic forms (see also
ambiguity 43 { / } ) 35
animal cries 22 class of phonological forms (see
archiphoneme 29 also p) 24, 29, 31-35, 44, 45
arithmetic 21, 22 combinatory variants 29, 30
articulation(s) 20, 22, 25 communication 14, 15, 18, 54
articulatory 34 communication system(s) 13, 14, 18.
aspect(s) 26 54
attributes 54 commutant 33
auxiliary system(s) 20, 22 commutation 29, 35
axiomatic 8 complex 20-22, 25
axiomatic functionalism 8 complex circumstance 55
axiomatic semantics 8 complex facts 54
axioms 54 complex sign 43
complex system(s) 20
bees' dance 22 complexity 20, 25
being denoted 55, 56, 58, 62 computer, see digital computer
Bloomfield 5, 6 concrete objects 55
Bloomfieldian 5, 33 conjunction 27, 28, 32, 41, 48
borrowing 18 connotation 59
connote(s) 54, 56, 61, 62
c, see content construction 19
case 33 constructionally equivalent 20, 57
case-endings 33 content (see also s R {/?} and c)
causal 16 6, 15, 26-28, 32, 35, 36, 47, 48
chain of definitions 56 contextual variants, see combina-
circumstances 54 tory variants
INDEX 67
convention(s) 16-19, 21, 22, 24, 61 double articulation 7, 8, 20, 22

conventional 13, 14, 16, 17 doubly ordered 25
conventional index 16, 18
converse 27, 28, 32, 47, 48 e, see expression
converse of R 32 e & c, see sign
coordination 23 emotional overtones 23
correct usage 60, 61 empirical 55, 60
correspond(s) to 39, 49, 53, 57 empirical criteria 42, 53, 56
c o r r o b o r a t i o n 42 empirical entity 38, 54
counter-domain 32 empirical identity 41, 43
empirical object(s) 54, 58
d, see distinctive function in phono- empirical sciences 38
logy English 23, 24, 26, 28, 29, 45, 49,
de Saussure, see Saussure 51
definition(s) 16, 17, 23, 46, 54-56, entity 38, 54-57, 59
61 equivalence 9, 15, 27-29, 39-44, 46-
denotablc(s) 36-39, 41-44, 49, 50, 48
53-58 equivalence class(es) 37, 39
denotation 9, 15-17, 38, 46, 53, 55, equivalent contexts 29
58, 59, 61-63 existence 54
denotation class(es) 44, 49-51, 53, expressions (see also { p j R i a n d
58, 59, 61, 62 e) 6, 15, 25-28, 31-36, 47, 48
denotational semantics 53 external criteria 42, 44
denotational sign-semantics 62, 63 extra-linguistic 36, 38
denotational theory 38, 53
denotationally 30, 60 /, see phonetic form
denotationally different 30 { / } , see class of phonetic forms
denotationally equivalent 58 facts 54
denotationally the s a m e 30, 59 fixed conventions 16, 62
d e n o t a t u m 37-42, 49, 50, 53-62 f o r m 15, 19, 20, 23-29, 48
denote(s) 17, 37-39, 43, 44, 49, 53- f o r m (as opposed to referent) 40,
57, 59 41, 47
descriptive application 40 f o r m a l 20, 26, 30, 40, 47
diachronic 23, 24 f o r m a l equivalence 27
differential f e a t u r e 33 f o r m a l identity 31
digital c o m p u t e r 17, 21, 22 f o r m a l similarity 31, 41-43
direct h y p e r o n y m 51 f o r m a l l y different 30
discontinuous 33 formally similar utterances 41, 43
discrete 19, 22, 34 f o r m a l l y the same 30
disjunction 46, 50, 58 formal-referential similarity 42-44
distinctive features 22 formal-referentially similar utter-
distinctive function 18, 27, 29, 30, ances 43
35 form-class 44-46
distinctive function of a sign (see f o r m s 23
also s) 31, 32, 34, 35 { / } R d, see class of allophones
distinctive function in phonology fi R d, etc., see alloplione(s)
(see also d) 31, 35 f r e e variants 29-31
d o m a i n 15, 32 F r e n c h 32, 33
68 INDEX
functional 8, 19, 24 language 7, 13, 20, 22-25, 36-38,

functionalism 6-8 51, 52
functionally differential 33, 34 Latin 33
Latin case-endings 33
gender 33 letter(s) 23, 54
genitive 24 linguistic description(s) 43, 44, 58,
gestures 22 60
glossematics 7 linguistic expression(s) 55, 58-60
grammar 5, 8, 9, 20-22, 29, 46 linguistic form 5
grammatical 20 linguistic sign(s) 26, 54, 62
grammatical elements 23 linguistic theory 8
grammatical entities 54 linguists 53
grammatical feature 34 literary form 23
grammatical structure 43 logical positivist 54
grammatical system(s) 20, 23, 25 logical sum 35
grammatically differential 33, 34 Lyons 49
grammatically distinctive function,
see distinctive function of a sign Martinet 20, 24, 33
mathematical logic 21, 22, 25
Hervey 8, 9, 30, 31, 42, 44 maximum Signum 23
hierarchical network 51, 52 meaning 20, 23, 26-28, 36, 37, 48,
hierarchy 25, 52 60-63
Hjelmslev 5-7 meaning-bearing 26
homonym(s) 25, 26, 30 member(s) 29, 33, 40, 44, 45, 47,
homonymy 43 49, 62
Hurren 29 mentalistic 6
hyperonymy 9, 31, 49, 51, 52 message 15
hyponymy 9, 49, 51, 52 model 39, 40, 43, 44, 48, 56, 61, 62
hypotheses 42, 52 moneme(s) 24, 26, 33
hypothetical 9, 44 moneme-theory 24, 33
morph(s) (see also p< R s, etc.) 15,
identity 18, 26, 41, 57
32, 34, 35, 46
implication 47, 48
morpheme-theory 33
index 9, 15-18, 23, 37, 56, 57, 61
morphology 22, 25
inference 57
Morse-code 17, 21-23, 25, 54
information 13-15, 19, 31, 36, 37,
Mulder 6, 8, 9, 18, 19, 29, 31, 34,
40, 41, 44, 56, 57, 61, 62
35, 40, 54
information-value 14-17, 37, 38, 55,
mutual implication 26-28, 47
57, 58, 61, 62
instance(s) 24-26
instanced 39, 54, 55 narrative order 23
interjections 22 natural 16
International Phonetic Alphabet 34 natural index 16, 18
interordination 23 necessary condition 42-44
interpretation 13 network 52
intersection 46, 50, 51, 58 non-denotational 60, 62
intonation 19, 21, 22 non-discrete 19, 22
isomorphism 40, 48 non-functional 33
INDEX 69
non-intersecting (see also disjunc- phonologically distinctive function,
tion 58 see distinctive function in phono-
non-linguistic classification 55 logy
nonce-symbol(s) 17-20, 22-24 phonology 5, 9, 20-22, 27
nonce-words 22 physical object(s) 54
not wholly fixed conventions 21, 22 plural 24, 26, 28, 29, 32, 33, 45
number 33 plural allomorphs 29
number-writing 16 plural moneme 24, 26, 33
police whistle 22
object 54-56 position 30
observation 55 Prague 5
occasional 17 primitive term 56
occasional conventions 16, 17 processes 54
one-to-one relation 26, 40, 47 proof 42
opposition 6, 8, 29, 30, 33 proper inclusion 45, 49, 50
ordered 20-22, 25 proper name(s) 17, 22
ordering 20, 25 proper symbol(s) 17-20, 24
ordering relations 22, 23 { p } R J, see expression
overlap 46 pi R s, see allomorph(s)
Pi R s & s S. pi (see also allo-
p, see phonological form morph(s) 32
{/>}, see class of phonological psycholinguistic 5
forms psychology 62
paradigmatic 22, 30 puns 44
paradox 37, 38, 53, 59 purport 15
paraphrases 57 purported information 14, 61
partial intersection 45, 50
partly fixed conventions 21, 22
permutation 32 quality 54, 56
philosophers 53-56, 58
philosophical theory 53 R, see relator
phoneme(s) 22, 24, 32-35 ft, see converse of R
phoneme-complexes 22 realization(s) 24, 34, 36-41, 44, 56-
phonetic feature(s) 34, 36, 40, 41 59, 61, 63
phonetic form(s) 34-36 red dashboard light 13
phonetically differential 34 reference 9, 36, 37, 47, 56
phonetics 5, 9 referent(s) 40, 41, 47
phonological 20, 24, 25, 29, 32, 33 referential 47
phonological elements 23 referential similarity 41-43, 57
phonological features 24, 31-36, 42, referentially similar utterances 41,
44, 45 43
phonological form(s) (see also { / } refutation 42, 50, 52
R d and p) 24-26, 28-30, 32-36, relation(s) 7, 15, 16, 28, 29, 32,
43 34, 37, 38, 41, 44, 45, 47, 49-54,
phonological properties 41 56-58
phonological system(s) 20, 23, 25, relator (see also R & È) 15, 16, 29,
34, 35 37, 56
phonologically differential 34 represent(s) 29, 39
70 INDEX
s, see distinctive function of a sign statements 21

de Saussure 5-7, 26 structure 7, 20, 46
Saussurean 5, 6 stylistic 60, 61
scope 63 stylistics 62
self-contained 19 sub-ordination 23
semantic description 39, 58 substance 24, 41, 55
semantic equivalence 27, 31 sufficient condition 42-44
semantic features 30, 31, 51, 52, 61 superordinate term 49
semantic function 31 super-ordination 23
semantic identity 31 suspension of opposition 29
semantic properties 49 symbol(s) 17-20, 23-25, 29
semantic relations 49, 50 synonym(s) 9, 26, 30, 31, 58-60
semantic structure 52 synonymy 31, 38, 49-52
semantic theory 40, 42, 53 syntactic complexes 22
semantically non-equivalent 31 syntagm(s) 21, 23
semantically non-identical 60 syntagmatic 22
semantics 5, 8, 9, 36, 38, 39, 46, system(s) 13, 18-25
49, 50, 53, 56, 57
semiotic system(s) 9, 18-23, 25, 36, technical vocabulary 23
37 tendency 60-62
semiotics 9 tentative signs 44
sentence(s) 21, 23, 54, 55 testing 42, 44
separation between levels 63 theorems 44
set, see class theories of meaning 53
set-theory 44, 45, 49 theory 8, 40, 53, 54, 63
sign(s) 5, 7, 17-20, 24, 26-28, 30-32, tone of voice 61
34-48, 50-55, 57, 59-63 total overlap 49, 50, 59
sign-concept 5-9 traditional grammar 5
sign-identity 9, 30, 31, 42-44, 49, 60 traffic signs 22, 25
signe linguistique 5 transcription 34
signifiant 6, 20, 26, 27, 32 truth value 55
signifié 6, 26, 27
signum 9, 16-19, 21-25, 61 union 29
similarity 41 unordered 20-22, 25
simple 21, 22, 25 utterance(s) (see also utterance-
simple sign 43 model) 27, 37, 39-44, 46-49, 53.
simple system(s) 20, 23 54, 59-63
simultaneity 22 utterance-model, see utterance(s)
simultaneous bundle 21, 22 Venn-diagrams 45, 46, 50, 51
single articulation 22
single level approach 63 wholly conventional 14
singular 33 wholly fixed conventional, 17, 61,
siren 22 62
speech 19, 54, 58-60, 63 wholly fixed conventions 21, 22, 25
speech-phenomena 7, 56, 60, 61 wholly non-fixed conventions 21, 22
s R /?,, etc. (see also allomorph(s)) writing conventions 23
32
s ft {/>}, see content zero 24, 30, 32, 33
JANUA LINGUARUM
STUDIA MEMORIAE NICOLAI VAN WIJK DEDICATA

Edited by C. H. van Schooneveld
SERIES MINOR
42. MILKA ivid: Trends in Linguistics. Translated by Muriel Heppell.

1965. 260 pp. Gld. 28.—
44. THEODORE M. DRANGE: Type Crossings: Sentential Meaninglessness
in the Border Area of Linguistics and Philosophy. 1966. 218 pp.
Gld. 29.—
45. WARREN H. FAY: Temporal Sequence in the Perception of Speech.
1966. 126 pp., 29 figs. Gld. 23.—
47. BOWMAN CLARKE: Language and Natural Theology. 1966. 181 pp.
Gld. 30 —
49. SAMUEL ABRAHAM and FERENC KIEFER: A Theory of Structural
Semantics. 1966. 98 pp., 20 figs. Gld. 16.—
50. ROBERT j. SCHOLES: Phonotactic Grammatically. 1966. 117 pp.,
many figs. Gld. 20.—
51. HOWARD R. POLLIO: The Structural Basis of Word Association Beha-
vior. 1966. 96 pp., 4 folding tables, 8 pp. graphs, figs. Gld. 18.—
52. JEFFREY ELLIS: Towards and General Comparative Linguistics. 1966.
170 pp. Gld. 26.—
54. RANDOLPH QUIRK and JAN SVARTVIK: Investigating Linguistic Ac-
ceptability. 1966. 118 pp., 14 figs., 4 tables. Gld. 20.—
55. THOMAS A. SEBEOK (ED.): Selected Writings of Gyula Laziczius. 1966.
226 pp. Gld. 33.—
56. NOAM CHOMSKY: Topics in the Theory of Generative Grammar.
1966. 96 pp. Gld. 12.—
58. LOUIS G. HELLER and JAMES MACRIS: Parametric Linguistics. 1967.
80 pp., 23 tables. Gld. 14.—
59. JOSEPH H. GREENBERG: Language Universals: With Special Reference
to Feature Hierarchies. 1966. 89 pp. Gld. 14.—
60. CHARLES F. HOCKETT: Language, Mathematics, and Linguistics. 1967.
244 pp., some figs. Gld. 28.—
62. B. USPENSKY: Principles of Structural Typology. 1968. 80 pp.
Gld. 16.—
63. v. z. PANFILOV: Grammar and Logic. 1968. 160 pp. Gld. 18.—
64. JAMES c. MORRISON : Meaning and Truth in Wittgenstein's Tractatus.
1968. 148 pp. Gld. 20 —
65. ROGER L. BROWN: Wilhelm von Humboldt's Conception of Linguistic
Relativity. 1967. 132 pp. Gld. 20 —
66. EUGENE j. BRIERE: A Psycholinguistic Study of Phonological Inter-
ference. 1968. 84 pp. Gld. 14.—
67. ROBERT L. MILLER: The Linguistic Relativity Principle and New
Humboldtian Ethnolinguistics: A History and Appraisal. 1968.
127 pp. Gld. 20.—
69. i. M. SCHLESINGER: Sentence Structure and the Reading Process.
1968. 172 pp. Gld. 22.—
70. A. ORTIZ and E. ZIERER: Set Theory and Linguistics. 1968. 64 pp.
Gld. 12.—
71. HANS-HEINRICH LIEB: Communication Complexes and Their Stages.
1968. 140 pp. Gld. 20.—
72. ROMAN JAKOBSON: Child Language, Aphasia and Phonological
Universals. 1968. 104 pp. Gld. 12.—
73. CHARLES F. HOCKETT: The State of the Art. 1968. 124 pp. Gld. 18.—
74. A. JUILLAND and HANS-HEINRICH LIEB : ' 'Klasse" und ' 'Klassifikation"
in der Sprachwissenschaft. 1968. 75 pp. Gld. 14.—
76. URSULA OOMEN: Automatische Syntaktische Analyse. 1968. 84 pp.
Gld. 16.—
77. ALDO D. SCAGLIONE: Ars Grammatica. 1970. 151 pp. Gld. 18.—
106. HENRIK BIRNBAUM: Problems of Typological and Genetic Linguistics
Viewed in a Generative Framework. 1971.132 pp. Gld. 16.—
107. NOAM CHOMSKY: Studies on Semantics in Generative Grammar. 1972.
207 pp. Gld. 24.—
110. MANFRED BIERWISCH : Modern Linguistics. Its Development, Methods
and Problems. 1971.105 pp. Gld. 12.—
113. ERHARD AGRICOLA: Semantische Relationen im Text und im System.
1972. 127 pp. Gld. 26.—
114. ROMAN JAKOBSON: Studies on Child Language and Aphasia. 1971.
132 pp. Gld. 16.—
117. D. L. OLMSTED: Out of the Mouth of Babes. 1971. 260 pp. Gld. 36 —
119. HERMAN PARRET: Language and Discourse. 1971. 292 pp. Gld. 32.—
123. JOHN w. OLLER: Coding Information in Natural Languages. 1971.
120 pp. Gld. 20 —
134. ROMAN JAKOBSON: A Bibliography of His Writings. With a Foreword
by C. H. Van Schooneveld. 1971. 60 pp. Gld. 10.—
MOUTON • PUBLISHERS • THE HAGUE

Theory of The Linguistic Sign Mulder

Uploaded by

Copyright:

Available Formats

Theory of The Linguistic Sign Mulder

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Theory of The Linguistic Sign Mulder

Uploaded by

Copyright:

Available Formats

JANUA L I N G U A R U M

Series Minor, 136

No part of this book may be translated or reproduced in any form by print,

LIBRARY OF CONGRESS CATALOG CARD NUMBER: 72-86887

Printed in Belgium, by NICI, Printers, Ghent.

In our opinion, the SIGN-CONCEPT should be considered to be the

portant theories at the time. Looked upon as stages in the history

It was Hjelmslev who presented one of the various possible

With respect to speech-phenomena, we would not even go so far

facts. It has, however, turned out to be possible to develop the

I. Index and Signum 13

II. Semi otic Systems 19

III. The Nature of the Linguistic Sign 26

IV. The Sign as a Class of Utterances 36

V. In Defence of a Denotational Theory of Semantics 53

'Language' is said to be a communication-system, i.e. a system of

item of information. This does not affect the actual information-

that he is there, that he is a foreigner, that he has a cold, etc., but

A natural index, as opposed to a conventional one, only needs a

A semiotic system is any system of CONVENTIONS for communica-

tonation' can conveniently be treated as a separate system, an

wholly fixed not wholly fixed not

wholly partly fixed

For easier reference we give here Table 1 (see overlaef) containing

non-discrete : non-discrete: 1. algebra language

a2 — ab = (a — b)a; x u y = n 12 x 3 = 36, and

justification for a phonological element is that it can be part

theory) — a system of PROPER SYMBOLS. It is a system of proper symbols, not

THE N A T U R E OF THE LINGUISTIC SIGN

would be even worse than in the case of solely depending on 'form',

with the difference in 'aspect', everything we say about one of the

sign = e &c, i.e. a particular 'sign' is the conjunction of a particular

Let us now determine what e, c and sign are. If we determine what

speakers in some dialects of English, or /felOu/ 'fellow' and

for if we used formal SIMILARITY (e.g. /aidr/ ~ /iRJr/) rather than

content is the CONVERSE of this, i.e. JR{/>}, where R indicates the

differential. Something is a grammatical feature, if it is grammat-

form by / , and a phonologically distinctive function (assessed by

THE SIGN AS A CLASS OF UTTERANCES

The notion sign defined as Expression ^ sufficiently

realization of sign 'a' provides information b (where b ^ 'a')

sign 'a' has information value b (where b ^ 'a')

Consequently, we may say that, whereas the sign appears as a

meaning', is the bridging of the gap, by using certain relations,

the sign 'table' and its information value, or DENOTATION. If we

An important fact is that denotables may belong to the field of

to be a shortcoming of that theory, for here, there is a case of one-

we should arrive at the definition of a given sign as an equivalence-

corresponds to a class of denotata

an utterance denotes a single denotatum

In defining the sign as an equivalence-class of utterances, we must

The avoidance of this contradiction together with the necessity

probably never do have) exactly the same phonetic features and to

a) via their phonological features

Consequently, any conditions for positive proof of equivalence

either on identity of phonological features, that is to say FORMAL

That neither of these three alternatives will do as a sufficient and

forms). Thus for instance, we can talk of the FORM-CLASS {"/iz/"},

A ~ B = members of {"/iz/"} which do not belong to to be