ROBERT M. W.

DIXON

Linguistic
Science |
and Logic

MOUTON & CO. - THE HAGUE

LINGUISTIC SCIENCE AND LOGIC
JANUA LINGUARUM
STUDIA MEMORIAE
NICOLAI VAN WIJK DEDICATA

edenda curat

CORNELIS H. VAN SCHOONEVELD

STANFORD UNIVERSITY

SERIES MINOR

NR. XXVIII

IN
UN
i SCIENTIAE

1963
MOUTON & CO. THE HAGUE
 a

LINGUISTIC SCIENCE
AND LOGIC

ROBERT M. W. DIXON
UNIVERSITY OF EDINBURGH

1963
MOUTON & CO. THE HAGUE
 © Copyright 1963 by Mouton & Co., Publishers, The Hague,
The Netherlands.
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.

Printed in The Netherlands

 CONTENTS

INTRODUCTORY

PART A
lie A GENERAL SCHEME OF SCIENCE . 11
1.1 Science and Observation 11
1.2 The Levels in a Science 21
1.3. Directed Action . 30
1.4 Meaning 35

A PARTICULAR CASE: LINGUISTICS. 40

2.1 Linguistics as a Science 40
2.2 Linguistics and Other Sciences 50

PART B
LINGUISTICS AND LOGIC . 59
3.1 Three kinds of ‘Logical...’ oo
3.2 Chomsky’s ‘Linguistics’ 62
3.3 Logical Linguistics? . 88
3.4 Towards a ‘Descriptive Veae 96
 INTRODUCTORY

Great advances have been made in linguistics in the last half-

century. Relative to the state of the subject fifty years ago these
advances may be as great as those in any other science. We use
the term ‘science’ purposely: one benefit of our half-century of
progress is that linguistics is now looked upon by very many of
its practitioners as a scientific discipline. But just what is meant
by ‘science’ in this context? The first part of this essay attempts
to provide a tentative answer to this question.
Just as one cannot profitably ‘do’ science without having a
theory, so must the theory fit into a more general philosophical
attitude. ‘Grammar’ loses much of its significance unless consid-
ered in relation to a complete theory of linguistics. And in the
same way linguistics can profit from a consideration of its place
relative to other sciences in a General Scheme of Science. The
General Scheme! presented here is not put forward as a particu-
larly definitive or, still less, a pragmatic outline. Nor is the De-
scriptive Linguistics which is stated as a particular case of it
believed to be the only scientific linguistics. My aim has been to
present a unified, if sketchy, picture of a particular approach
which I personally think to be worthwhile and worthy of bringing
to the attention of linguists of other schools.
The present statement owes much to the writings and teaching
of Firth: in this respect it will have much in common with the
work of other British linguists. The General Theory of Linguistics
1 When a science is referred to, in part A below, as a particular case
of the General Scheme, the implication is not necessarily that it conforms
to our criteria but rather that it is capable of doing so. It should be noted
that some subjects, and psychology appears to be an example, fail to satisfy
our requirements; in certain cases it is not clear that they are capable of
doing so without drastic changes of approach.
8 INTRODUCTORY

which is discussed in section 2.1 is essentially a statement of

the neo-Firthian linguistics which has been made definitive by
M. A. K. Halliday and others at Edinburgh. I have gained much
from discussions with Halliday, J. O. Ellis, J. McH. Sinclair,
Angus McIntosh and J. C. Catford during the preparation of this
essay.2 But despite this, and the fact that 2.1 and certain parts of
other sections are merely my restatements of accepted Edinburgh
opinions, as I understand them, the full responsibility for the
present discussion must, of course, rest squarely with me.
Part A can stand on its own as a discussion of the scientific
nature of linguistics. But it is included here solely as prolegomena
to the examination of the basic interrelations between logic and
linguistics in Part B. Section 3.2 is devoted to a discussion of the
work of the Chomsky school within the terms of reference of the
Edinburgh approach; this group has attracted much attention in
recent years and, more latterly, some pointed criticisms.? Choms-
ky’s approach differs from ours and we do not throw doubts upon
the merits of his own theory relative to its aims: the linguistics
described in section 2.1 would come off badly if discussed relative
to Chomsky’s approach. His viewpoint is quite possibly more
original than any in the last two thousand years: we examine it
with respect to the purposes of ‘linguistics’ as they were understood
before Chomsky applied the designation to his own work, and in
particular with respect to our own scientific criteria.
We consider to what extent logical methods can be employed
in doing linguistics. Finally, the revelance of linguistics to the
establishment of a descriptive, rather than prescriptive, and dis-
tributional, rather than conceptual, logic is discussed.

2 J. O. Ellis, M. A. K. Halliday, Angus McIntosh and J. McH. Sinclair

have read this essay in draft form and made extremely valuable suggestions
for improvement.
8 For example: Anton Reichling, “Principles and Method of Syntax:
Cryptanalytical Formalism”, Lingua, X, 1-17 (1961); Dwight D. Bolinger,
“Syntactic Blends and Other Matters”, Language, 37, 366-381 (1961).
PART A
 i leit aad yb
eas a si fe ee , esta % aayWw
ph! emi ee TE, Contra ah
$ mee: ey wee iy Be af ae
— that ee

4 ee Se pore tate Ful! come fib

: Lporiie
(pein mR: +oF a cy epiandy YRS His:
i= Riek & te, PSE a es raha ;
Rivera Chgirsds. Hr erenwil ee —
es ois ReKs eb the
i ocgenteeh 3, Section Pepe dist . .
cheek WS the CReraky <chate ieee he pein at
ethese tbe dara: Nis oomph ‘
toate eal, vars Saget -
aS opprencs: dil: s renee aes cen tase’oh
Bite akin oe ESS Qeeeyl Ssh ee Nanas: es. S48
ea eee cele aan etsbe oe ete
2 Riewick'y's anagteen
a } Porte reads oe sew
ic ipaiiey Seep " Billie
mi Bes eis

j Wei” <i easalee: shia: re

apity ie pane & Par oe eels Hae 5
“: We osahtor dp weg aa: Sompedi pated
a rage
ee: aeae Vee. §tealty, fhe. aemigoge bel
Tes = oF ok enapadohs 2 enue ep
Be ban a Bo matin thet 2Heen ;

Se pare “BA is , ,

tsmae: ‘Hac ey ay. ms iy ieaced

~ Ssapr vigils ah
} Pip. 2dryatelys Raya
oe Be
anaes! Fecciphata’

A) a Ras ee
 A GENERAL SCHEME OF SCIENCE

1.1 SCIENCE AND OBSERVATION!

The kind of linguistics which is described in this essay is conceived

of as a branch of science. As the term ‘science’ can cover a fairly
wide range of different methods it will probably be useful to state,
rather briefly, just what is meant by science in the present context.
We say that the ‘raw material’ of science consists of observa-
tions.2 The scientist will recognise a certain pattern which is com-
mon to a number of observations; in other words he will notice
that, in some particular way, the observations are similar to each
other. Having recognised certain patterns he will then compare
and correlate them. Pattern correlations can be of several different
sorts: the scientist can correlate different patterns all of which are
noticed in the same set of observations, or else he can correlate
patterns of a similar sort which each occur in one of a number of
distinct sets of observations. Generally speaking he will perform

1 J was able to effect considerable improvements on an earlier draft of

Section 1.1 after discussions with J. McH. Sinclair.
2 An alternative to this would be to say that the raw material consists
of ‘objects’ or ‘material things’ which have a ‘real’ and independent exist-
ence. In this case the next step would be to make observation of this ‘real’
raw material. This approach has the disadvantage of prerequiring an as-
sumption regarding the existence of ‘real things’, thus raising metaphysical
questions of considerable depth. A number of other difficulties can arise if
‘objects’ rather than ‘observations’ are taken as raw material but the extra
assumption indicated above is in itself thought to be sufficient reason for
our present approach. It should be noted that the observation method auto-
matically avoids the traditional dichotomy of ‘(object) referential’ and
‘non-referential’ meaning; ‘referential meaning’ involves correspondences
being set up between pieces of language material and ‘material objects’.
The observational approach demands a consistent and homogeneous treat-
ment of meaning: further details of this are given in section 1.4.
12 A GENERAL SCHEME OF SCIENCE

pattern correlations of both these kinds. A theory® is obtained by

generalisation upon pattern correlations. Thus a theory can be
looked upon as an abstraction from a number of observations.
What is abstracted from each is that pattern or patterns which is
seen to recur in similar form in other observations. And so the
first function of a scientific theory is that it should be descriptive.
Instead of having to refer to each recognised pattern in each ob-
servation individually we can describe these patterns in toto by
means of the theory: so that a theory is, in other words, a concise
and abbreviatory form of statement.
The term ‘observation’ requires some explanation. It is not
thought that observations are any sort of ‘entity’, either mental or
physical. Nor that they are uniquely determined or that they are
discrete things. Observations can only initially be considered in
relation to one particular person, an observer. We must assume,
I think, that an observer observes‘ continuously. For we can surely
never split up his life into a sequence of mutually exclusive, dis-
crete observations. We cannot draw some line in time and say
that one observation finishes and another starts here. The observer
is aware of what happens on either side of this line and he cannot
consider just some time-segmented memory; he may direct his
attention to a particular part of his past observation but he can-
not ever be sure that he is considering this part to the complete
exclusion of all others since some or all of the other parts are
always present, to some extent, in the background of his thoughts.
It would thus seem quite in order to say that life is one con-

3 Explicit discussion of the intermediate steps involved in theory con-

struction—such as the setting up and testing of hypotheses—has been
omitted from this brief account.
4 ‘Observe’ is here used in a very wide sense (whilst it is not necessarily
intended to imply the presence of volition): for example, it includes an
observer’s awareness of himself. It is thought that most, if not all, of our
knowledge and experience is gained by means of some sort of direct or
indirect observation. We would appear to observe in somewhat different
ways at different times. Whilst we are sleeping we must be observing in
some way since a loud noise or a blow on the head will wake us up or
promote some action; but we are obviously not observing in the same sort
of way as we may be at midday.
 SCIENCE AND OBSERVATION 13

tinuous observation. This is our most general statement and it

must always be borne in mind that such is the case. But we can
of course isolate, to some extent, some part of observation. The
‘observations’ which we referred to above as making up the raw
material of science are just such isolates. It must however be noted
that we are never able to clearly define the outer boundaries of
an isolate. An isolate from observation can be regarded as the
continuum of the observer’s observation up to that time with one
part of the continuum emphasised and brought into the most im-
mediate consideration. The rest of the continuum is then tempo-
rarily pushed into the background: but it must be present since
the isolate is merely a part of observation and can only have any
existence relative to the rest of observation. And again the isolate
is not some clearly defined part of observation such that just it is
emphasised more than the bulk of the continuum. Some part of
the isolate may be emphasised more than some other part of the
continuum but the isolate is always a part of the homogeneous
whole. The change from emphasis to non-emphasis is a gradual
and continuous one rather than being sudden and discontinuous.
Thus the isolate can be regarded as a hillock with gently sloping
sides set in the field of observation rather than as some sort of
tall tower. It will be seen that the process of isolation is to a large
extent arbitrary; indeed, the term ‘isolate’ may be something of a
misnomer. Again, since we cannot explicitly separate emphasis
from non-emphasis, we are in no position to decide whether any
two isolates are mutually exclusive. In the strictest sense, of
course, they can never be since they each contain the same foun-
dation of the observer’s complete observation up to that time. A
set of isolates can usefully be regarded as the observer’s observa-
tion with a number of different parts emphasised into the forefront
of consideration, a field containing a number of hillocks.

5 This distinguishes science from mystical experience (on which see

“Mysticism and its Language” by C. W. Morris, pp. 179-197 of Language:
An Enquiry into its Meaning and Function, edited by R. N. Anshen (New
York, 1957)-—I owe this reference to J. O. Ellis) where, if the mystics’
claims are justified, ‘observation’ is possible in which there is no distinction
of ‘fields’ and ‘hillocks’.
14 A GENERAL SCHEME OF SCIENCE

It is convenient to recognise two parameters which can be

identified within observation. The first one, length, can be used
as some sort of rough indication of the duration of a part of ob-
servation. The second, width, will measure synchronic detail of
observation. For instance, the width of a part of observation which
includes some written language material will also include such
things as the paper on which the material is written, the table on
which the paper rests, the walls of the room, the trees seen through
the window, a car-horn heard in the street, the observer’s aware-
ness that he is feeling hot and hungry or that he has a headache,
and so on. When we, in a manner of speaking, isolate from ob-
servation, the process is performed relative to both these param-
eters. Thus, when we are looking for material for linguistic
science, we may hear a conversation between two people. We
shall try to isolate a part of observation to be in our raw material
for linguistics. Although we cannot precisely delimit the isolate we
can give indications of the sort of length that we wish to emphasise
and the parts of the width that are to be given prominence. Thus
we shall want the length of the isolate to be such as to include the
whole conversation and as little more as possible. As for the width:
the main feature to be included is the actual sounds made during
the conversation; this is the part of width which is to be most
emphasised.
Often we find that not enough information is contained in some
part of our raw material. It is then, of course, an easy matter to
return to the part of observation near the isolate and extract a
larger piece than at first. In many sciences we need some parts of
observation to be the nucleus of our raw material and other re-
lated parts to be, as it were, the qualification of this nucleus. In
the particular case of linguistics we would need to know, in the
above example, which part of the conversation was being uttered
by which of the participants. Many other factors are also needed
in order to determine the full linguistic status of the conversation;
we would need to know enough to determine whether the con-
versation was of a formal or informal nature, what the social role
of the conversation was, and so on. The nucleus raw material is
 SCIENCE AND OBSERVATION 15

the basic substance of the science. Some parts of observation near

this substance* are needed in order to determine the contextual
meaning of the substance. These parts can in fact be called the
Scientific situation of the substance, the scientific situation being
an abstraction from the general situation of the piece of substance.
‘General situation’ is a technical term like ‘substance’, ‘pattern’
and ‘theory’ and it represents the whole of observation as it occurs
in relation to the position of the piece of substance. Recognising
and abstracting the substance is often fairly straightforward; the
scientific situation may well be less evident. But, as already men-
tioned, we are able to refer back to the whole situation of a piece
of substance at any time and extract as much further ‘scientific
situation’ as we need. And so scientific situation is potentially the
whole most general context, the scientist’s complete observation.
In practice only certain features of general situation will be rele-
vant in the case of each particular science.
Science is thus always carried on within the wide context of the
whole of the scientist-observer’s observation although he may
superficially appear to be dealing with just a few particular iso-
lated parts of his observation. The way in which the scientist deals
with his raw material, his recognition of patterns, is also relative
to the same wide context, his observation’? up until that time.
It will have been seen that the forming of isolates out of ob-
servation is itself an example of pattern recognition. The pattern
here is, in the case of linguistics, the presence of what we recognise
as written or spoken language material in a part of observation.
This is the defining pattern. But along with the language substance
we isolate certain external features of the observation which are
relevant to linguistic study. We do not of course merely search
through random parts of observation until we find some with the

6 In the case of linguistics, for example, some of the most relevant

scientific situation is often at the same place in length as the substance and
differs from it only in width.
7 As already mentioned a scientist’s observation will to a very large
extent define his knowledge and experience. Thus his raw materials are
extracted from observation and his methods of working with them are
derived from observation.
16 A GENERAL SCHEME OF SCIENCE

patterns we require. The gathering of raw material can be directed

by the observer; by this is meant that it is possible to focus our
attention in such a way that we will have a better than usual
chance of observing the necessary patterns. But it remains the case
that selection of the raw material is a preliminary application of
the pattern-recognition procedure.’ Then, in our scientific con-
sideration of the raw material, we will recognise and correlate
certain subtler patterns within the isolates that make up the raw
material.
It is often possible to provide stated procedures for at least
some of the correlations of patterns involved in a scientific in-
vestigation. But the actual recognition of the patterns themselves
must be, to a far greater extent, intuitive. Which patterns are
recognised as belonging to each of a number of isolates may vary
from observer to observer and, in the case of a single observer,
from time to time. The choice of patterns depends on the point
of view of the observer; and this point of view will presumably
depend, in turn, on his own experience and knowledge and also
on his particular purpose at the time the patterns are recognised.®
Thus pattern recognition is itself a subjective operation; recogni-
tion of any sort of pattern at all reflects upon the observer rather
than upon the observation.!° We are patently not suggesting that
isolates from observation ‘possess’ patterns as some sort of in-

8 One aspect of pattern recognition concerns the identification of pieces

of language material as alike or partly alike. We would say that we in-
tuitively recognise certain patterns of likeness. Bloomfield (in “A set of
postulates for the science of language”, Language, 2, 153-164) puts this in
a rather different way in his: “ASSUMPTION ONE: Within certain communi-
ties successive utterances are alike or partly alike.” Bloomfield assumes
that there is some quality of likeness between utterances. We merely re-
mark that certain like patterns can be recognised in an entirely subjective
manner.
® As already remarked his experience and knowledge are largely de-
rived from his observation. His purpose however cannot be related to ob-
servation in exactly the same way.
10 This is, in a sense, trivial inasmuch as the observation can only exist
relative to the observer in the first instance. It is intended to emphasise the
subjective nature of observation: that it cannot have any sort of independ-
ent form or existence.
 SCIENCE AND OBSERVATION 17

herent property and we cannot therefore make statements of the

form “there is some delimited number of patterns corresponding
to a set of isolates such that no other pattern can be recognised
as common to them”. Pattern recognition is thus in some ways
quite arbitrary. A theory will describe some aspects of a certain
set of isolates and the theory only exists relative to the choice of
these aspects. Thus we can only talk of ‘inventing theories’ and
hardly of ‘discovering laws’ whilst we are doing science.
We have stated that theories can only be considered in relation
to the raw material isolates out of observation and that observa-
tion only exists relative to a unique observer. This is an important
consideration and it must always be borne in mind that such is
fundamentally the case. In practice however, two scientists can
have similar points of view and have relevant parts of their ob-
servations in correspondence so that they are able to agree as to
the patterns that they recognise and the theories that they con-
struct. Again, a theory can be communicated by its instigator to
some other person and it often appears to be the case that the
second person can understand the theory within his own observa-
tion in such a way that it has the same kind of descriptive power
relative to him as it did for its inventor. This brings into focus the
problem of communication; to discuss this rather difficult topic
in any detail would be outside the scope of the present essay.
But it must again be emphasised that a theory must in some way
be changed when it is placed within the new general context of a
different observation (that is, when it is communicated to a second
person). Superficially we are usually able to work perfectly well
between ourselves; but the difference exists and should not be
forgotten inasmuch as it is always potentially a barrier to mutual
understanding of a common theory.
As already stated, the prime and original purpose of a scientific
theory is to describe. It will have succeeded if it does just this.
It may be, however, that once we have constructed a scientific

11 In fact, one aspect of a consideration of the ‘problem of communica-

tion’ will be an application of general linguistics, with particular reference
to ‘formal’ and ‘contextual’ meanings (sce below, section 1.4).
18 A GENERAL SCHEME OF SCIENCE

theory which satisfies us as to descriptive power, then we may be

able to use it to predict. Let us consider a body of isolates which
are chosen in such a way that in each is recognised the defining
pattern of the raw material, this set of isolates being chosen so as
not to include any which made up the raw material. Then we shall
be using the theory to predict when we attempt to ‘describe’ the
scientific patterns which may occur in the new body of isolates
not by an examination of this new set but rather by application of
our theory. We shall thus be employing a theory, which was de-
rived as a description of certain aspects of a particular set of
isolates, to describe a different set of isolates which are of a
similar type to the first set.
The above description of the nature of science (or at least of
the particular kind of science which is our concern in the present
essay) has necessarily been extremely general and oversimplified.
Certain other sorts of procedure can be used in order to assist and
complement the major one of pattern-correlation. One of these,
the use of statistical methods, is of particular importance since it
can be used in several different ways. If a theory is used to predict
then it is often useful to employ statistical methods at this stage
and to give probability predictions. But since the first function of
a theory is to describe, the use of statistical methods in the earlier
stages of theory construction may possibly be regarded as a more
important application.
The aim of a general theory of linguistics is to be such that a
particular theory can be propounded for each individual language
as a special case of the general theory. Thus a theory of gram-
mar,! for instance, could be looked upon as an abstraction of the
common patterns occurring in the grammars of individual lan-
guages. In practice however the theory need not be constructed
in exactly this way. The raw material can include observation of
the substance of many different languages and the theory can be
obtained directly from this raw material without the necessity of
first constructing particular, explicit grammars for individual lan-

12 For example, that outlined in M. A. K. Halliday’s “Categories of the

Theory of Grammar”, Word, 17, 241-292.
 SCIENCE AND OBSERVATION 19

guages. In assembly of the raw material for this general theory,

or for a particular grammar, or for almost any linguistic theory,
statistical sampling techniques are employed. The whole corpus
of language material available is considered in absentia, as it were,
and a series of representative samples are chosen to be the actual
raw material for the scientific analysis.‘ Thus it is intended that
the theory should describe the whole available corpus. A tentative
theory is propounded on the basis of a first sample of the available
raw material; then more isolates are included and the theory is
changed, if need be, to accommodate them. This process is con-
tinued until several large additions to the considered corpus have
failed to necessitate any alterations to the tentative theory.
During this procedure specific sampling techniques and criteria,
built up from past experience, are used. In this way the theory is
effectively set up to describe the whole available corpus; it has
merely been found unnecessary to consider explicitly every pos-
sible relevant isolate.
Thus a distinction is to be made between a theory which is
built out of all the raw material which it is required to describe
and a theory which is built out of a certain set of raw material
and which is required to predict the patternings in another set."
Which kind of theory is constructed for each science depends en-
tirely on the nature of the raw material and the patterns that are
looked for in it. Some sciences, such as linguistics, are essentially
‘descriptive sciences’; others, and physics would appear to be an

13 The sampling is a two-part operation. A representative sample of the

world’s languages are chosen and within each selected language a repre-
sentative sample of language material is considered. Different criteria are
in operation in the distinct sampling operations.
14 Tt will be noticed that when a theory has been constructed out of
some sample which is statistically selected from a corpus we do not say
that the theory will predict the patternings of the extra-sample part of the
corpus. Since we have employed statistical techniques in assembling the
sample and in checking whether the sample was large enough and repre-
sentative enough for the theory to be a description valid for the complete
corpus we could only use the term ‘predict’ in a quite trivial sense here.
In fact we restrict our use of the word to application of a theory to some
material which was not considered either directly or indirectly during its
construction.
20 A GENERAL SCHEME OF SCIENCE

example here, are ‘predictive sciences’.1* Two factors can be ad-

vanced as affecting the predictiveness of a science. It appears to
be the case that the raw material of descriptive disciplines, such
as linguistics, is pretty well delimited (although it is tremendously
large) and if necessary a surprisingly large fraction of it could be
directly considered. Whereas physics can choose its substance in
many diverse ways and it is not obvious here that the corpus is
delimited to anything like the same extent. Because of this the
predictive sciences usually propound some quite simple descriptive
theory and use it to predict; as the techniques employed in such
sciences improve" their theories will be embellished and extended.
Descriptive sciences however are able to consider the complete set
of possible raw material, to sample from it, and to put forward
definitive descriptive theories at a fairly early stage; in this case
there is nothing left to predict17! The second factor is that pre-

15 We are not trying to effect a dichotomous division of sciences into

‘descriptive’ and ‘predictive’. Rather we establish a continuous scale, whose
ends points are called ‘descriptive’ and ‘predictive’, along which sciences
can be arranged; this scale is more validly used for purposes of compari-
son than for explicit pinpointing of individual sciences.
16 Reference should be made to section 1.3 for a description of ‘di-
rected action’ methods: these may be partly responsible for such improved
techniques. One of the many factors involved is the use of assisted ob-
servation (by means of scientific instruments) instead of the direct variety.
Our present discussion of science does not take account of the sort of
change which can occur in a particular sort of patterning over a period
of time. If this were considered then linguistic theories, for instance, might
be used to ‘predict’ in a sense quite different from that mentioned above.
But here we are only concerned with examination of patterns observed
more or less synchronically —or at least treated as if they had been ob-
served synchronically. Compare Saussure’s remarks: “In practice a lan-
guage-state is not a point but rather a certain span of time during which
the sum of the modifications that have supervened is minimal.” (Course
in General Linguistics, p. 101).
17 In the whole of this section we have implicitly rejected the idea that
there are, for example, any phenomena which we would classify as ‘lan-
guages’ and which are of extra-terrestrial origin. This exclusion may be
misguided: but in our present state of knowledge it remains very much a
matter of personal opinion as to whether it is a justifiable assumption.
If extra-terrestrial ‘languages’ are held to be likely then there is indeed
some predictive work for our linguistic theories to embark on. On this
general topic it is relevant to note the work of Hans Freudenthal (Lincos:
 SCIENCE AND OBSERVATION Dal

dictive sciences appear often able to formulate simple theories,

which are abstracted from a small number of observations and
are then used to predict, whereas the nature of the patterning in
descriptive sciences does not yield any obvious simple theory and
so the whole corpus has to be considered in the construction of a
suitable, and necessarily rather more complicated, theory.

1.2 THE LEVELS IN A SCIENCE

We have already specified that a scientific theory is set up to

describe certain patterns which are recognised in certain isolates
from observation, these comprising the raw material of the sci-
ence. It will have been seen that the raw material has existence
relative to observation, independently of whether any science is
later built around it — it is what we are ‘given’. A formal theory,
on the other hand, is a construct, purposive and arbitrary in many
ways: it is put forward as a model'8 for the pertinent isolate pat-

Design of a Language for Cosmic Intercourse, Part One, Amsterdam,

1960). He has put forward the hypothesis that any people who possessed
short-wave radio communication would have at least some ideas similar
to our arithmetic, mathematics and logic. Thus he constructs a ‘language’
which starts with formulae involving numbers and gradually introduces
new ideas into itself until it can carry more complicated, and more every-
day, messages. This is put forward as the ‘artifical language’ which extra-
terrestrials would be most likely to understand. This is an interesting
hypothesis: but it might be possible to advance on it by bringing in Gener-
al Linguistic theories to a greater extent than Freudenthal has done.
18 The terms ‘model’ and ‘theory’ are in a sense interchangeable. Quite
generally we can say that a scientific theory is a model for the parts of
observation which it describes. That is, it provides a ‘representation’ of the
structuring of patterns which the scientist recognises in the raw material.
It is probably the case that the word ‘model’ is normally employed in this
sense. But it should be noted that this use is very different from that of
some mathematicians and logicians who will rather present some parts of
observation as a ‘model’ for a theory. Here the theory is virtually dis-
sociated from its original raw material and considered as having an in-
dependent existence. Then parts of observation are considered and patterns
looked for in them such that correspondences can be set up between these
patterns and the theory. If such a set of isolates is distinct from the
22 A GENERAL SCHEME OF SCIENCE

ternings. It is not a ‘way of dealing with’ the raw material but is

rather a fully explicit mirror in which will be reflected certain
scientific patterns (the ‘form’) of the raw material. All theoretic
operations and descriptions take place entirely within the model,
on the plane of the mirror: we cannot say that theoretic categories
directly describe the raw material; a theory will be complete in
itself and will describe certain items which are part of itself. Then
correspondences can be established between these theoretic items
and the parts of substance of which they are reflections: it is
through such correspondences that the theory can be said to
‘describe’ the patternings of the raw material.
The essentially different statuses of the defining patterns and
subpatterns, and the formal!® patterns must be made explicit in
our description of the general nature of sciences. We do this by
saying that there are a number of different Jevels within any sci-
ence; the defining patterns and subpatterns are dealt with at the
level of substance and the theories which account for the formal
patterns are at the level of form. Substance is the more basic level
since here we are directly concerned with the raw material. The
raw material exhibits formal patterns, but these are only to be
described, interrelated and compared by reference to the theoretic
level, the level of form. Form can thus be regarded as the ‘inner’

original raw material (to give a very rough and ready interpretation of a
complex theoretic point) it is called a ‘non-standard model’ for the theory.
This procedure is an accepted part of abstract mathematics; but in science,
as already stated, we are concerned with making meaningful statements
about observed patterns. We are interested in describing parts of observa-
tion rather than in obtaining deductive ‘results’ which are of interest for
their own sake instead of for the light they can throw upon observed
patterns. And so we can only assign any meaningful existence to a formal
theory in the context of the raw material from which it was abstracted
and we only employ the term ‘model’ to describe a theory in relation to
its raw material.
10 Formal patterns are those which are internal to the scientific raw
material and which can be shown to be significant (by commutation, etc.)
with respect to the occurrence of the raw material relative to its general
situation (the rest of observation). ‘Formal meaning’ is given fuller discus-
sion in section 1.4.
 THE LEVELS IN A SCIENCE 23

level; the theories set up to account for internally meaningful

patternings in the raw material will belong to this level. The sci-
entific formal patterns which have been recognised in the raw
material and have been correlated and generalised upon are re-
presented by theoretical classifications and categories at the sec-
ond level of form.
Thus theories are derived, in the first place, from the substance
of the science. Intermediate stages of correlation, generalisation,
and hypothesis setting up and testing are gone through and even-
tually a theory is devised. The theory itself is an abstraction from
the substance and as such is part of a different scientific level.
But a theory is set up to describe raw material patterns, the raw
material is only considered per se at the level of substance and so
we must, as already implied, have some means of relating the two
levels of form and substance. We can do this by providing an
interlevel, as it were, between them. This interlevel is an essential
component of the science: as necessary a component, in fact, as
the levels which it relates. It should be noted that an interlevel is
in no sense a ‘meta-level’; it is not ‘about’ two levels but should
rather be regarded as a smooth slope with top and bottom touch-
ing the two levels which it is to ‘join’. An interlevel will itself
contain theories which will perform the necessary level-relating,
in this respect only an interlevel is similar to the inner level of
form.
But despite the theoretic nature of both the level of form and
the interlevel which relates form to substance there is a very large
difference between them. Formal theories describe the formally
significant patterns within the raw material. These patterns are
significant by virtue of the nature and mode of their occurrence
in the scientific substance, both as such and in relation to the
significant correspondent patterns in the scientific situation. If,
for example, two patterns appear to be in free variation in the raw
material and are not distinguishable by recognised and relevant
situational patterns then they will not be treated as two distinct
entities by the formal theories; their free variation makes any dif-
ference between them non-significant and they can be said to have
24 A GENERAL SCHEME OF SCIENCE

the same ‘scientific meaning’.2° Thus formal theories are set up to

describe patterns which are significantly contrastive within a sci-
ence. Since a formal theory must be completely contained within
the level of form it will only directly describe items which exist
within this level—_we can call these ‘formal items’. And so a
formal theory consists of a set of formal items and of an apparatus
for describing the formal properties of these items.
Interlevel theories will describe and relate the components of
two levels; since formal items are the basic components of the
level of form, theories of the interlevel between form and sub-
stance will establish correspondences between them and defining
subpatterns considered at the level of substance. Generally, the
formally meaningful patterns of a science may not be identical
with the pattern or patterns which define an isolate from observa-
tion as a part of scientific substance. And again the formal pat-
terns may not be clearly distinguishable as such within the raw
material: it may be that they will be recognisable only as the sum
of a number of simpler patterns. It is the function of the substance/
form interlevel to provide an inventory of distinctions such that
correspondences between formal items and substance patterns can
be stated in terms of these distinctions or combinations of them.
This interlevel does not appear to have been given any generally
applicable name; particular names are given to it or to parts of it
in specific sciences; since its function is to relate formal items to
substance items the title interlevel of item-relation might be con-
sidered suitable. It may be, of course, that in some sciences the
meaningful patterns can be directly recognised as such in sub-
stance. Here there is no need for an item-relation interlevel with
its complex of theories; in this case a direct correspondence can
be set up between the formal items and the raw material. And so
an intrinsic difference between levels and interlevels has been
demonstrated: levels are common to all sciences and must always
be present whereas interlevels are, in the general statement, only

20 ‘Scientific meaning’ is here intended to represent the sum of formal

meaning and contextual meaning for each particular science.
 THE LEVELS IN A SCIENCE 25

to be considered as potentially necessary. Whether specific inter-

levels need to be set up in particular instances, and the relative
complexities of corresponding interlevels in different sciences, is
a function of the various types of patterning of those sciences.
So far we have described two levels—form and substance —
which must be integral parts of a science, and an interlevel of
item-relation which may be needed to provide the necessary cor-
respondences between form and substance. These two levels are
not in themselves enough, however, to account for all the neces-
sary types of phenomena that have to be considered in a science.
For instance, we have already mentioned that the position of an
isolate relative to the rest of observation must be relevant to our
scientific consideration of the isolate. The place of an isolate in
observation is called its general situation— this suggests that we
should establish a third scientific level and call it the level of
situation.21 At the level of substance we consider the actual sci-
entific isolate from observation, chosen because it exhibits the
defining patterns of that science; the level of situation deals with
those parts of observation near to the scientific isolate with respect
to both the parameter of width and that of length. Thus situation
will initially indicate the location of the substance isolate; in addi-
tion to this it will describe the particular parts of observation
which are near to this location.
It must, of course, be realised that the whole process of the
setting up of levels is quite abstract and subjective; this process
is itself part of the general scheme. Since the scheme has to do
with the description of the components and internal structuring
of any given science it can be looked upon as a theory of the
philosophy of science. Our levels do not imply any inherent divi-
sions within observation: they are artificial devices to segregate
21. The names of the levels —form, substance and situation —and of the
interlevel of context as used here coincide with the names for linguistic
levels put forward by M. A. K. Halliday (“Categories of the Theory of
Grammar”, Word, 17, 241-292). The present general theory of science was
more or less worked out before I became familiar with Halliday’s terms;
but his names would appear to be widely applicable and it has seemed
quite natural to use them for levels in general.
26 A GENERAL SCHEME OF SCIENCE

phenomena according to their particular modes of existence rela-

tive to the science that we are concerned with.
We have now postulated three levels, and these three do con-
stitute the full complement of levels for a science. The isolates
from observation which display the defining patterns of the science
under consideration are dealt with at the level of substance; the
parts of observation which are, according to either parameter,
near to a defining isolate, are considered at the level of situation;
and the theories which are set up to describe the formal patterns
belong exclusively to the level of form. In addition to these levels
we require a number of interlevels which will relate the levels to
each other; they also contain theories, not to describe internally
meaningful scientific patterns as do the formal theories but in-
stead to establish correspondences between the pattern carriers of
each level. Thus far we have given detailed consideration to the
levels of form and substance and discussed the interlevel of item-
relation which may be needed to relate them in the absence of any
absolutely direct correspondences between the items of the two
levels. It remains to consider the types of correspondences which
must be established between the levels of form and situation and
of substance and situation. The second of these is simple: since
scientific substance (raw material) and scientific situation are both
extracted from observation by the same general type of process
there automatically exists a direct correspondence between the
components of the two levels. So that in this case we will never
have need of an interlevel; situation and substance must always
be directly relatable to each other by virtue of their method of
derivation. But this general lack of need for a situation/substance
interlevel does not imply that these two levels are any less distinct
than, for example, substance and form. Substance is chosen by
recognition of the defining pattern or patterns of the science, as
such it is the basic object of study for the science. Situation con-
tains peripheral parts of observation which are not in themselves
the raw material of the science and are important only inasmuch
as they occur near to and sympathetically patterned with the raw
material.
 THE LEVELS IN A SCIENCE a7)

Frequently, the significance of formal patterns in some part of

observation will depend in part on the composition of observation
near? those patterns. In other words, the occurrence of certain
patterns in our raw material can often be correlated with certain
general patterns occurring in the relevant parts of the general
situation; we have previously called these parts the ‘scientific
situation’. In order to incorporate this important relationship into
our scheme of science we will usually require to set up a second
interlevel — namely, one which will link form to the level of situa-
tion, which deals with this scientific situation. This second inter-
level can be conveniently called the interlevel of context; its role,
having extracted the scientific situation from the general situation,
will be to recognise patterns in it and to establish correspondences
between these situational patterns and patterns at the level of
form. This interlevel will contain theories (to be called theories
of context) by means of which the required correspondences can
be effected. It is quite conceivable that direct relationships might
be establishable between form and situation such that no proper
interlevel of context was required. Most generally we must allow
that context, like item-relation, is potentially dispensable; but I
cannot at present think of an example in which some at least
slight interlevel of context would not be an essential component.
The three levels,?? as described above, are the necessary com-

22 The statement that the scientific situation will be amongst those

parts of general situation ‘near’ to an isolate exhibiting the defining pat-
terns of the science under consideration must be a considerable simplifica-
tion if ‘near’ is accorded an everyday sense. Quite complex pattern cor-
relations may be necessary within general situation in order to account for
all relevant features; these will always be correlations emanating from a
part of observation close to the raw material isolate, however, and so the
initially relevant situation may justifiably be described as being in proxim-
ity, relative to width and length, to a defining pattern. ‘Near’ is thus used
as a technical term implying adjacency relative to the two parameters and
through the application of pattern correlations.
23 This representation of ‘three levels’ is, of course, only one of a num-
ber of possible approaches. We could instead make a theoretical distinction
into two contours (calling them form and substance) and then say that the
form contour consisted of one level (our level of form above) whilst the
substance contour consisted of two levels: immediate-substance (our level
28 A GENERAL SCHEME OF SCIENCE

ponents of a science.*4 Within this general framework we can

construct particular sciences: we should need to provide criteria
for identifying patterns in observation and thus extracting the raw

of substance above) and situational- or peripheral-substance (our level of

situation above). This might better stress the universal lack of an interlevel
between ‘substance’ and ‘situation’ (to revert to our original names): but
the advantage seems unworthy of the extra complication that would have
to be introduced.—An impressionistic representation of our levels and
interlevels (including also the contours) may be of interest:

ce aaa
~ ea

contour of substance

24 J. O. Ellis has suggested, in conversation, the alternative view: that

most sciences only need two levels for adequate description and that it is
a measure of the uniqueness of linguistics that it requires three levels. The
tendency here is to conflate substance and the most immediate scientific
situation into a single level; the important thing to realise is that a piece
of substance, an isolate in which is recognised the defining pattern of the
science, is, by nature of its occurrence, effectively delimited in extent. The
scientific situation, however, cannot be thought of as delimited in any-
thing like the same way — we have already mentioned that it is potentially
the whole of the general situation, of the rest of observation. If we extract
some scientific situation and consider it together with substance (as of
limited extent, as substance must always be) then we are effectively and
finally rejecting the rest of the situation. This is an artificial and dangerous
step: one should not be misled by the fact that it is slightly less wounding
in physics than in linguistics, for example. For we have already shown
that an isolate can only have existence relative to the whole of observation
and should thus always be considered in relation to its complete situation.
In a chemistry experiment in which two gases are being subjected to
abnormal temperature and pressure in order to study their reaction: the
gases are the substance and the temperature and pressure are the most
important scientific situation; to say that the temperature and pressure are
 THE LEVELS IN A SCIENCE 29

material of the science, and we should need to recognise the sub-

tler scientific patterns within the substance and correspondent
patterns in situation and to construct theories at the level of form
and the interlevels of context and item-relation (if these are re-
quired in the particular case) in order to describe and relate these
patterns.” The roles of the three levels and the type of interrela-
tions between them will naturally vary in different sciences. The
distinct natures of different sciences can in fact be stated partly in
terms of the roles of the levels in these sciences and the type of
interrelations between these levels; these properties can be looked
upon as one aspect of the defining characteristics of a science. It
is relevant to note that, although every science must have a level
of situation and (normally) an interlevel of context, these may
be more easily recognisable as such in one science than in another.
It is essential however, that they should be recognised and that
they should be clearly differentiated from substance, item-relation
and form; any conflation of levels will expose the possibility of
serious confusion within that science and can often lead to scien-
tifically unsound conclusions.

‘part of substance’ is plainly misleading! It may be that some non-predicted

but unfortunately relevant situational feature occurs: a violent thunder-
storm or an earthquake would conceivably affect the result of the experi-
ment. And these are therefore very relevant situational features: the
science is necessarily conducted within the framework of ordinary life and
although we may try to isolate our experiments from unwanted external
happenings we cannot always be sure of doing so. Scientific situation is
generally much more easily extractable from general situation in the case
of physical sciences than in the case of human sciences, and so human
sciences will require more complex interlevels of context: this is a measure
of the distinct natures of the different types of science. Conflation of
substance and situation levels would, in addition, make it virtually im-
possible to discern the amount of admittable ‘direct action’ which is em-
ployed in a science (see the discussion in section 1.3).
25 We will also, of course, have theories at the level of substance — pho-
netic theories for example; but it seems unlikely that ‘theories of situation’
are possible.
30 A GENERAL SCHEME OF SCIENCE

1.3 DIRECTED ACTION

In sciences such as linguistics and economics our observation can,

of course, be directed in the sense that we shall try to observe
more linguistic or economic patterns than might randomly come
our way. We are interested in setting up theories within one of
these sciences and so we deliberately search out the relevant types
of isolate: by this means we can assemble our necessary raw ma-
terial in a quite short time. But although, when doing linguistics
or economics, we may direct our observation in this way we would
rarely attempt to direct our actions in order to provide observable
patterns of the required kind. In other words an economist would
not establish a new industry in order to have more trade figures
to study, a linguist would not write a book himself in order to have
more material to analyse; more than that: a linguist or economist
would not even, in the normal course of events, often try to es-
tablish an artificial situation in which a particular sort of pattern
might be likely to occur. In each of these cases the scientist is a
— he does not try to alter or change what he ob-
pure observer
serves but merely to describe it adequately, to set up theories
which will reflect the way in which his scientific isolates from
observation appear, to him, to be structured. Linguistics and eco-
nomics are alike in that they study what might be loosely referred
to as natural happenings; the above remarks will apply to every
science which just takes its raw material from ordinary day-by-day
observation — whether it considers some aspect of human behav-
iour, such as linguistics does, or some more general physical or
even non-terrestrial patternings, as do archaeology or astronomy.
Practitioners of such sciences are detached and impartial ob-
servers — they can direct their observation but they will not nor-
mally direct their actions in order to provide themselves with
constructed patterns. In these cases the scientist stands apart from
the normal flow of life: his observation is of the ordinary and
natural happenings which occur as the world progresses rather
than of happenings which were directed to occur simply so that
they could be observed and included in a set of raw material.
 DIRECTED ACTION 31

Observation is only made of phenomena which occur as purpose-

ful units within the normal matrix of behaviour.?¢
But in subjects such as physics and chemistry, for example, the
scientist will not only direct his observation but he will also make
definitely directed actions in order to obtain observation of pat-
terns which interest him. If a chemist merely maintained a sharp
look-out for interesting phenomena he would gather only a very
small amount of raw material during any delimited time. In the
first instance chemistry was born out of purely random observa-
tion being seen to display a certain type of patterning. But as the
science progressed more directed observation was made and at
the same time the scientist would arrange things so that he was
able to make the right type of observation for his purpose. After
a while some of the scientist’s directed actions produced condi-
tions which had a negligible chance of occurring in the normal
course of events??; so that sciences like physics and chemistry
could never have progressed to their present states unless directed
action had been taken by their exponents. The methods of such
sciences require that observation be made which will include, for
instance, abnormal temperatures and pressures. In the absence of
directed action some subsciences, and nuclear physics is an ex-
ample, would never have come into existence.
So far we have talked about ‘directed action’ in a quite general
way; but we can, and indeed must, differentiate between two
distinct kinds of directed action. These can be exemplified within
linguistics as follows: firstly a linguist sits in an armchair in front
of his fire and tries to think up new and interestingly analysable
collective nouns; secondly the linguist devises a rather hideous

26 A linguist will himself use language in his daily life and in working
on his scientific studies — but he will use it in the same way as non-linguists.
He will never purposely talk in jabberwocky (considering every item-part
of his utterance in relation to a potential later analysis of it) just so that
he can analyse this artificial utterance: he will not construct a synthetic
language in order to have something to describe. For such description
would be of little account —it would be trivial inasmuch as he would be
scientifically analysing what he had just scientifically synthesised.
27. The implication is: in conditions suitable for human life.
32 A GENERAL SCHEME OF SCIENCE

practical joke — such as causing a bucket of water to fall on some

person’s head. In the second case the linguist makes careful note
of the utterances of this unfortunate immediately after the joke or
whatever has had its effect. In both these examples the scientist
has directed his action in some way: in his armchair he directed
action onto himself in order to directly affect (synthesise) language
raw material; the second direction was only to the situation part
of observation, he hoped to be able to observe a particular type
of utterance patterning (exclamatory, or abusive) in the quite
artificial situation he had created. In both cases the linguist was
directing action to some specific linguistic purpose — but the dif-
ference lies in action directly affecting the substance, and action
directly affecting just the situation and the substance only (if at
all) through the patterns inflicted on situation.
It is clear that when a chemist or physicist directs his personal
actions in the way described in the last paragraph but one he is,
in effect, directing situation. It must be the case that no scientist
can directly create or affect scientific substance or else he will
run into the vicious circle encountered by the linguist who writes
(purposely constructing) a book merely to be able to analyse it.?8
The chemist will purposely arrange phenomena which will belong
not to the level of substance but to that of situation: his aim in
doing this is to study the substance patterns which are likely to
occur in the environment of the type of situation that he has
created. Whereas the economist or linguist merely looks for the
sort of patterns he is interested in and examines them, taking into
consideration the related contextual patterns, the physicist or
chemist increases his scope by also seeing what kind of patterns
will occur in particular situations. Thus he either accelerates the
rate at which he would normally expect to observe certain types
of situation or else he creates situations which would never be
likely to happen otherwise.
We are now able to effect a rough division of sciences into two

8 That is, who constructs each item making up the book in a purpose-
ful manner, with the later analysis very much in mind.
 DIRECTED ACTION 33

types: those that employ ‘pure observation’ and those that might
rather be described as ‘observation and action’ sciences. In actual
fact, of course, no really dichotomous classification can be ef-
fected. Each science will tend to employ some, perhaps very small,
amount of directed action; the most we can do is to arrange the
sciences along a continuous linear scale whose endpoints are ‘a
negligible amount of directed action’ and ‘a very large amount of
directed action’. For purposes of reference only we may rather
loosely refer to sciences, such as linguistics, which are near to the
first endpoint as ‘pure observation’ sciences. Some of the examples
of the last few pages have rather oversimplified the category of
‘directed action’. Like our previous division into ‘descriptive’/
‘predictive’, the ‘directed action’ criterion will be more useful for
comparing the amount of directed action involved in two sciences
than for classifying them individually. Every science is primarily
descriptive; some sciences must, by nature of the patternings in
their raw material and the total possible amount of raw material,
be almost entirely descriptive, but others can also be, to a certain
extent, predictive. It would appear, for instance, that astronomy
has a rather more predictive nature than does linguistics and that
physics is more predictive than is botany. Again, every science
will have directed observation; the properties of being fully de-
scriptive and of making directed observation might indeed be
looked upon as two of the major defining features of a science.
We have seen that in addition to this universal habit of directing
observation some particular sciences will direct action to create
particular situations. It seems reasonable to state that botany is
likely to involve more directed action than linguistics and that
physics will certainly involve more than will astronomy.”® In ap-
plying both the pure-description/description-and-prediction and

29 The examples quoted throughout section 1.3 are of a rather tentative

nature: more detailed investigation of particular sciences than has been
possible so far might not yield the same comparisons as those given here.
The whole concept of ‘directed action’ is at a pre-scientific stage of devel-
opment inasmuch as no advice is given here as to how the place of a
science on the undirected/directed action scale could be determined.
34 A GENERAL SCHEME OF SCIENCE

pure-observation/observation-and-action categories and in giving

the above examples we have had to rely upon extremely subjective
and speculative ideas. Individual opinions may vary as to where
a science should be placed on each of the above scales; the first
important step is to realise that these scales do have some signifi-
cance — that sciences do vary in these two ways. At a later stage
we can make the scales definitive by statement of formal criteria
for placement on them.
We discovered certain reasons which explained or partially ex-
plained why some sciences were restricted to a purely descriptive
role whilst others could also have a predictive side. In the same
way we can indicate certain reasons for the presence or absence
of directed action in a science. It would seem that in principle
directed action should be included amongst any scientist’s tools
unless there is some specific reason for its exclusion. In other
words: a scientist can usually benefit to a quite considerable extent
by using directed action; there need be no positive criteria for the
use of this technique — there can only be some negative reason for
its rejection. It does appear that in every ‘pure observation’ sci-
ence there is some peculiar facet of that science which proscribes
directed action methods. In astronomy, for instance, directed ac-
tion is not used simply because the relevant parts of the general
situation— that is, the scientific situation — of astronomical pat-
nernings is beyond the effects of any direct action by the observer;
if an astronomer could change the topology of the space-time
continuum no doubt he would in order to see, in true scientific
tradition, ‘what would happen’. In sciences such as linguistics
which deal with manifestations of human behaviour, directed ac-
tion would be likely to intrude upon the normal manner of con-
struction of the parts of observation which carry scientific pat-
terns. For directed action can only be allowed to affect those parts
of observation which belong to the level of situation — it can affect
scientific substance through its situation but it must not itself have
any direct effect on potential raw material. And it is due to the
particular nature of linguistic substance, for instance, that it can
only exist (in a significant sense) under more-or-less normal situa-
 DIRECTED ACTION 35

tional conditions. There does not appear to be any analogy in

linguistics to the abnormal situations which a chemist can create
when, for example, he reduces the temperature of some substance
to —272°C or so. It is usually because of the nature of the raw
material of a science that directed action cannot be usefully em-
ployed (or cannot be employed to any great extent; as already
stressed there is doubtless a certain amount of directed action in-
volved in every science, our point is that the technique is em-
ployed more widely and is more important in some sciences than in
others) — and often these reasons are to an extent dependent upon
whether or not the raw material is some direct offshoot of human
behaviour. So that, as in a consideration of ‘degree of predictive-
ness’, the type and extent of directed action employed in a science
can be looked upon as a further defining characteristic of that
science.*1

1.4 MEANING

A science, of the general form sketched above, is a particular sort

of academic discipline with certain stated aims and inclinations.
A scientist will be interested in certain types of patternings which
he recognises in observation and he will want to investigate and
correlate these patterns —in short, to see how they are put to-
gether and how they interrelate. He might be said to have a desire
to find out how things work; in order to fulfil this desire as satis-

30 Quite false descriptive categories may be set up if language is studied

not in the course of its normal social functioning but with reference to a
set of reactions from isolated informants. Within our terms the ‘informant
response’ method, so important a factor in the ‘procedure’ of many lin-
guists, is thus hardly scientific. Quirk remarks that work with informants
will show “what is barely possible rather than what is actual and normal
in linguistic behaviour” (TPS, 1960, p. 54). We should prefer to put it
somewhat more strongly than that.
31 To some extent the amount of directed action employed in a par-
ticular science will be an indication of the maturity of the science; a dis-
cipline is likely to begin with observation only, and, if it can, gradually
make more use of directed action methods.
36 A GENERAL SCHEME OF SCIENCE

factorily as possible he must have as few as possible preconceived

ideas about the arrangements of patterns that he is likely to en-
counter. He must argue by analogy and from previous experience
in many instances but he should always refer to actual observation
whenever he can. If his theories are properly to describe the pat-
ternings of a certain type occurring in observation then he must
be certain that he has built them out of the patterns which do
occur and not just from some preconceived idea of which patterns
he thinks ought to occur.
But when we construct a particular science it is not sufficient
merely to specify the levels and state their theorems: a science is
not just a framework for us to admire, it is essentially a tool. A
science is constructed with a particular purpose in mind, to pro-
vide a critical and selective description. In order to employ a
science in this way we must have some means of interpreting its
descriptions; and we can do this by appeal to meaning.®? Meaning
is therefore to be an essential part of each science: it is the final
link in the application of a constructed science to the description
of those parts of observation from which the theories of the sci-
ence were originally abstracted. Because of this importance it is
quite vital that we should have clear ideas concerning ‘what mean-
ing is’ within the general context of scientific work and exactly
how and where it is applicable. To rely entirely on ‘every-day’
ideas concerning meaning would be insufficient especially inas-
much as meaning appears to be one thing concerning which there
is hardly any general unanimity. We shall not pretend that our
scientific meaning is the only sort of meaning but instead that it

32 Firth expressed this point very well in reference to the science of

linguistics: “The object of linguistic analysis as here understood is to make
statements of meaning so that we may see how we use language to live.”
(Synopsis, p. 23). If we were to insist that meaning did not concern us in
our pursuit of some particular science we should not be doing science at
all (at least in the sense of this essay) but merely ‘playing a game’ (further
discussion of the science/game scale is presented in section 2.2); in this
case we should be unable to throw any light on how things worked, our
scheme would not have any provision for actual description of observation.
For ‘meaning’ is our descriptive tool and can be our only such.
 MEANING 37
is the only type of meaning relevant to our present discussion;
‘meaning’ is here used as a technical term.
When we use ‘meaning’ we are making significant statements
about some patterns in observation; but we have already distin-
guished several different types of patterning and set up levels in
our science to handle them. And so it is quite natural that we
should find it necessary to distinguish between at least two%* dis-
tinct types of scientific meaning. ‘Formal meaning’ will refer to
the significant patterns within the raw material, these being sig-
nificant with respect to the way in which the raw material occurs
relative to the rest of observation; they are obtained by abstraction
from substance and exclusively handled by theories at the level
of form: thus formal meaning is determined by reference to this
‘inner level’.8¢ The second type, ‘contextual meaning’, refers to the
patterns observed in the scientific situation and in their correlation
with formal patterns: it is obtained from theories at the interlevel
of context.
Formal theories will, as we have already seen, describe formal
items and thus these items are the meaningful units at this level.
If two pieces of raw material are related through the interlevel of
item-relation to the same formal item then they are said to have
the same formal meaning; but if correspondences can be set up
between two pieces of raw material and two distinct formal items
then those pieces must have different formal meanings. So that
formal meaning is completely dependent on the structuring and
composition of the various theories at the level of form: as Halli-
day has remarked ®> it is analogous to the ‘information’ of ‘infor-
mation theory’.
Contextual meaning, on the other hand, depends on the situa-

83 Other types: ‘item-relations meaning’ and ‘substantial meaning’ are

not discussed in this present brief outline.
84 Sweet appears to have used ‘meaning’ in senses including the formal
one: “The whole of syntax is nothing else but an investigation of the
meanings of grammatical forms” (“Words, Logic and Grammar”, in Col-
lected Papers, p. 31; quoted and commented upon by Firth in “Structural
Linguistics”, TPS, 1955, p. 85).
35 “Categories of the Theory of Grammar”, Word, 17, 244.
38 A GENERAL SCHEME OF SCIENCE

tion in which a piece of substance occurs. Most strictly, two dis-

tinct items in substance will always be different in contextual
meaning; for they must always occur at different places in ob-
servation and thus their widest situations must eventually differ.
But in practice we can always recognise some degree of likeness
between two isolates from observation. We have already men-
tioned that the significant part of situation relative to a particular
science is the ‘scientific situation’; we can thus compare the sci-
entific situations of two isolates — such a comparison being made
by means of the theories of the interlevel of context. In fact we
can say, as a general statement, that the contextual meaning?* of
a piece of scientific substance is the situation of that substance
—its place in observation relative to both the parameter of length
and that of width — with special reference to the particular con-
textual patterning, according to the theories of the interlevel of
context for that science, of the relevant parts of the situation.
To phrase it more explicitly, the formal meaning of an item is
its description in terms of the theories at the level of form, its
contextual meaning is its description in terms of the theories at
the interlevel of context. The formal meaning is a statement of the
significant patterns within the raw material and the contextual
meaning is a statement of the significant patterns of correspond-
ences which can be set up between formal patterns and situational
patterns; the significance of each of these two types of patterns is
measured relative to the other type.37

36 There may be a tendency to state that this contextual meaning of

science is quite similar to every-day notions of meaning. Such an equation
can be very dangerous: ‘contextual meaning’ is a technical term, clearly
defined within the general scheme of science. In certain particular cases
it may be similar to a particular instantial abstract from the range of the
every-day term ‘meaning’: but this does not permit us to adopt any general
hypothesis concerning the correspondence between a term within our
exposition and an ordinary unrestricted language item.
87 Jtem-relational patterns are set up to provide correlations between
defining subpatterns and formal patterns—no direct commutational cor-
respondences can be set up between them and situational or contextual
patterns. Thus ‘item-relational meaning’ is not directly relevant to the way
in which the raw material occurs relative to the rest of observation.
 MEANING 39

This completes our necessarily brief description of what will

be meant by science in this discussion. The scheme presented is
put forward as a quite general framework: many more details
would have to be supplied before it could be considered a really
efficient theory — but the outline presented above does, I think,
include the main and important points. The next section considers
linguistics as a particular case of the general scheme.
 A PARTICULAR CASE: LINGUISTICS

2.1 LINGUISTICS AS A SCIENCE

The status of linguistics with respect to the two general categories

of ‘prediction’ and ‘directed action’ has already been discussed.
We have seen that the science is placed very near the ‘pure de-
scription’ end of the non-predictive/predictive scale; it may well
be that it is the least predictive of all sciences. If this isso then
it is because of the particular sort of patternings observable in
linguistic isolates and of the potentially limited nature of a general
linguistic corpus. Linguistics was again seen to be close to one
end of the second, non-directed action/directed action scale. But
we should be rather more hesitant to say, here, that it is the sci-
ence which is nearest to the ‘pure observation’ end; the other
human sciences will occupy about the same position on the scale
and astronomy, for example, is obviously closer than linguistics
to its first endpoint.?

1 Or at least this is the position of the subject at present. 4t should be

noted that there is more scope for directed action in General Phonetics,
considered as an immanent discipline in its own right, than in General
Linguistics (including phonetics in a transcendent sense).
2 It has often been the fashion to grade sciences according to whether
they are ‘classificatory’ (for example, botany) or rather ‘explanatory’ (such
as physics); it would appear that a high degree of explanatoriness is depend-
ent on fairly powerful descriptive theories. We can see that this category
is to some extent independent of the others we have mentioned: a descrip-
tive science can be either classificatory or explanatory but a predictive
one can probably only be explanatory. Linguistics appears to be both
‘descriptive’ and fairly ‘explanatory’; physics is ‘predictive’ and ‘explana-
tory’, botany is ‘descriptive’ and ‘classificatory’.
 LINGUISTICS AS A SCIENCE 41

Linguistic science will recognize several different defining pat-

terns in observation. The most important two are that an isolate
should contain what is recognized as spoken language material
(particular sorts of ‘disturbances of the air’) or else written lan-
guage material (particular sorts of ‘marks on paper’).? It is a
commonplace that the defining patterns of linguistics bear some
similarity to those in physical sciences whilst the patterns at the
level of linguistic form are rather to be compared with formal
patterns in other human sciences. Certainly it is a large measure
of the individual character of linguistics that quite complicated
interlevels of item-relation and of context‘ are necessary in order
to interrelate the various recognisable defining subpatterns and the
very different formally meaningful patterns which are handled by
theories at the level of form, and to correlate formal patterns with
situational ones.
Because of the two main and distinct types of linguistic defining
patterns the linguistic interlevel of item-relation requires two quite
different types of theory to perform a relating of these substance
patterns with the significant, formal patterns. The discipline of
phonetics provides the means whereby we can perform the actual
recognition of patterns in phonic substance; ‘phonology’ is the
name given to that part of the interlevel of item-relation which
deals with these phonic patterns. And so the alter-interlevel of
phonology will contain theories which describe and compare the
patterns initially extracted from observation by phonetics. Phonol-

3 In addition a number of secondary ones may be allowed in the case

of gesture languages, such as a deaf-dumb one (meaningful placement of
parts of the body) or a tactile one (meaningful physical distortion of a
paper or other surface —recognized by appeal to the sense of touch). It
should be noted that Braille is, like the Morse code, a mere coding of
written language and thus only indirectly a tactile mode for the occurrence
of language patterns. Our discussion below is limited to phonic and graphic
defining patterns.
4 These two complex interlevels supply linguistics with a large portion
of its unique character. If a nickname were ever needed for linguistics
‘the interlevel science’ would be as apt as any.
42 A PARTICULAR CASE: LINGUISTICS

ogy® deals in abstractions from the phonic raw material, its cate-
gories are set up with regard being paid to phonetic (featural)
distinctions, general distributional criteria and formal considera-
tions. The significant patterns which are thus to be described in
a phonology are found to be of varying extent, and to be such
that a larger pattern can be regarded as ‘made up’, in a certain
sense, of a sequence of smaller ones. Thus we are led to postulate
a hierarchy of units. The phonological items which are the de-
scribendum of this theory each belong to one of the units; each
item which belongs to a unit above the lowest has a structure
— selected from a category of possible structures. Each structure
has a sequence of places and with each place is associated an
element of that structure: to each element of a structure of an
item associated with a certain unit corresponds a term of a system;
the system is a set of items which are associated with the unit next
below the given one in the unit hierarchy. Thus the category of
system details the choice of one term, rather than the others, from
its set, to act at a given element of a given structure. The smallest
unit (the lowest on the hierarchy) is generally called the phoneme:
the phoneme is thus the smallest (phonologically) meaningful
(phonological) unit. In addition to these three phonological cate-
gories certain features can be recognized which need not be as-

5 This description of the components of linguistic theory is, as men-

tioned earlier, pretty well in accordance with the views of M. A. K. Halli-
day and others at Edinburgh. For detailed treatment of the topics men-
tioned here reference should be made to, for example: M. A. K. Halliday,
“Categories of the Theory of Grammar”, Word, 17, 241-292; “Gram-
matical Categories in Modern Chinese”, TPS, 1956, 177-224; The Language
of the Chinese “Secret History of the Mongols’ (Oxford, 1959). — Angus
McIntosh, “Patterns and Ranges”, Lg, 37, 325-337; “‘Graphology’ and
Meaning”, ArchL, 13, 107-120.—_M. A. K. Halliday and J. McH. Sinclair,
An Outline Grammar of Spoken English, to be published shortly by
Longmans, Green & Co.— Much of this work owes a good deal to Firth’s
example, particularly his phonological ideas and his views on meaning and
what he called ‘context of situation’. A detailed bibliography of Firth’s
work (by R. H. Robins) is given in Lg, 37, 199-200, and BSOAS, XXIV,
417-418. The work of Firth’s pupils and colleagues at the School of Ori-
ental and African Studies, London, is also highly relevant to the present
discussion: see the bibliography to G. L. Bursill-Hall, “Levels Analysis”,
JCLA/RACL, 6, 186-191.
 LINGUISTICS AS A SCIENCE 43

signed to any unit but which can be of variable length — these

features are called prosodies. Prosodies need not be delimited in
statement, that is either their beginning or their end or both need
not be explicitly noted; they can be associated with any extent of
phonological ‘text’ (the minimum extent being a single phoneme).®
The alter-interlevel which is directly concerned with graphic
substance is called graphology.In actual fact the two disciplines
— phonology and graphology— are intimately connected and ref-
erence may be made to both in performing an item-relational
theoretic treatment on some raw material.? Graphological pattern-
ings may be of a simpler sort, easier to handle and recognise than
phonological ones. But graphology has yet to receive the explicit
statement that has been (and is being) accorded phonology.
At the level of linguistic form we can notice two initially dif-
ferent types of pattern — thus we must use rather different tech-
niques in order to describe them to best advantage. Firstly, we
may have lexical groupings of linguistic items: a (lexical) item is
described by giving the probabilities that various other items will
occur next to it in a piece of language substance, next item but
one to it, and so on. Thus an item is described in terms of the
items with which it can collocate— that is, near to which it is
likely to occur; for each item a scale is set up ranging from items
most likely to occur next to the given item to items least likely to
do so. This is the category of collocation; the second lexical cate-
gory is that of lexical set’: a lexical set is a collection of lexical

6 This use of the term ‘prosody’ is probably consistent with Firth’s use
of the term although it differs somewhat from R. H. Robins’s definition
in “Aspects of Prosodic Analysis”, Proceedings of the University of Dur-
ham Philosophical Society, Series B, 1, 1-12 (1957). The British use of the
term ‘prosody’, as exemplified in Firth’s “Sounds and Prosodies”, is com-
pletely misrepresented by A. A. Hill in “Suprasegmentals, Prosodies,
Prosodemes: Comparison and Discussion”, Lg, 37, 457-464.
7 In the case of a phonological script, that of English for example,
reference will have to be made to both the alter-interlevel of phonology
and that of graphology; but for a formal script, for example the Chinese
script, the linguist will only need to refer to graphology.
8 Called set by Halliday (“Categories”, Word, 17, 276) and Firth
(“Synopsis”, p. 26).
44 A PARTICULAR CASE: LINGUISTICS

items which can, roughly, occur in the same linguistic environ-

ments. Here again the results should be expressed in terms of
probabilities — there are patently no two lexical items which can
occur each in exactly the same linguistic environments; this is to
say that there is no absolute synonomy.
Theoretically, the whole of language substance could be divided
up into ‘lexical items’; but it can soon be seen that some of the
commonest parts of substance (for example, the items ‘the’, ‘me’
and so on in English) can be given only an insignificant descrip-
tion by application of the lexical categories. Thus lexis can ac-
count perfectly well for some of the formal patterns within lin-
guistics but is totally inadequate to deal with all patterns. And so
a second approach is to construct a grammar which will describe
the complex mutual interrelations of the large number of sets of
linguistic items which each contain relatively few members; lexis,
on the other hand, described the quite simple interrelations of the
smaller number of sets which each contained a large number of
items. No sharp dividing line between grammatical and lexical
patternings can be set up: we might rather consider language pat-
terns as arranged along a scale with the complex, completely
grammatical patternings at one end and the simpler lexical pat-
terns at the other. Language patterns will occur, continuously,
beween these two endpoints. We can now see in just what way
our two theories are mutually complementary. The theory of lexis
is most efficient when applied to patterns at the lexical extreme
of the scale, the theory of grammar at the grammatical end. Each
theory becomes progressively less efficient as it is used to describe
patterns further from its endpoint. As the theory of lexis is ap-
plied nearer to the grammatical extreme it becomes virtually use-
less, the descriptions it gives are so weak as to be descriptively
insignificant. Likewise it seems probable that grammatical cate-
gories could not distinguish between (lexically) distinct patternings
towards the lexical end; but this question remains, for the time
being, fairly open: more detailed investigation would be needed
before we could say that, and at what point, grammatical criteria
cease to be of use.
 LINGUISTICS AS A SCIENCE 45

There will be some patterns, near neither end of the scale,

which can be described by both grammatical and lexical theories.
Because neither theory will be as powerful in this region as it
would be at its own extreme it will, in fact, be useful to describe
these ‘middle patterns’ in terms of both theories. Theories of
grammar and lexis are both needed in order to describe the mean-
ingful patternings at the level of linguistic form. But the theories
themselves are, in internal construction, not at all analogous. A
theory of grammar must be (by virtue of the patterns out of which
it is abstracted) a very powerful tool. Halliday has described as
the “grammarian’s dream” a state of affairs in which all formal
patternings could be accounted for grammatically.? But, as we
have seen, a theory of grammar gets less efficient as it is applied
nearer to the ‘large set’ end of the scale. A lexical theory is, on
its home ground, weak by comparison with a grammatical theory
describing grammatical patternings; it works in terms of proba-
bilities rather than tightly woven systemic relationships. But even
so it does appear unlikely that a lexical theory could ever be
dispensed with. Poor relation as it is it will do a more adequate
job at its own end of the scale than a theory of grammar appears
likely to be capable of at that end.
Since theories of grammar and lexis approach the description
of linguistic formal patternings in two quite distinct ways it may
be misleading to merely say that they are both theories at the level
of form. Better that we should initially divide form into two demi-
levels which we can call ‘grammar’ and ‘lexis’. Then we can put
forward theories within these demi-levels; by the nature of this
existence these theories will not properly be comparable one to
another. Because of the essential grammar/lexis distinction it is
quite natural that each demi-level should contain its own set of
formal items: patternings from the middle of the before-mentioned
scale may be described at both demi-levels, but, since they are
described in completely different ways, they will not need the
same sort of ‘pattern carrier’. What is grammatically significant
may not be so in a lexical sense. So that formal grammatical items
9 “Categories”, Word, 17, 267.
46 A PARTICULAR CASE: LINGUISTICS

will not generally coincide with formal lexical items even when
both correspond to the same piece of language substance. In fact
formal lexical items will not need to correspond to all parts of
substance: the most grammatical patternings cannot usefully be
treated within lexis and so the parts of substance in which they
occur have no lexical item correspondents. Grammar, because of
its different approach, will have formal items corresponding to
all parts of language substance even though some of the most
lexical patternings may not be distinguishable by application of
the theory of grammar.
The theory of grammar bears certain resemblances to the pho-
nological theory we described above. It has three rather analo-
gous categories which can be called by the same names of unit,
structure and system. The units are arranged in a hierarchy —
which we may refer to as the rank scale: they represent the dif-
ferent sorts of stretches of language substance which carry gram-
matical patterns. Again each formal item is associated with some
structure: structures are set up to represent the sequences of
similar patterns which make up other patterns. Systems account
for the sets of patterns which are identified by their similarity in
making up other patterns: it details the choice of one event rather
than another from among a set of possible events. A scale of
delicacy operates within the categories of structure and system:
as one goes from structure to more delicate structure, for example,
the breakdown of patterns becomes more detailed and finer cri-
teria are brought into play. Delicacy in system and structure go
hand-in-hand: correspondences are set up between ‘elements’ of
structures and ‘terms’ in systems but these always involve systems
and structures at the same degree of delicacy. One of the points
of dissimilarity relative to the theory of phonology is that gram-
matical items can ‘jump’ downwards on the rank-scale and act as
terms in systems which have correspondences with structures
which are associated with rank not just one higher than are the
items but can be one higher, the same or lower. The scale of depth
describes this phenomenon of ‘rank-shift’ by detailing the extent
to which a pattern within an item is made up of other patterns
 LINGUISTICS AS A SCIENCE 47

carried by items which have greater extent, pattern-carrying-wise,

than the original item.
We have yet to add a fourth member to the categories and a
fourth to the scales of the theory of grammar. We have already
stated that, at present at least, grammatical criteria cannot be used
to distinguish all contrastive formal patterns. So that, in our gram-
matical description of some formal item, we may be left not with
some unique description but instead with a description which can
serve equally well for several other formal items: recourse will
have to be made to the theory of lexis if we are to distinguish
between these items within the level of form. More explicitly, we
may obtain, as end product, a system of several terms such that
we are unable to evolve grammatical reasons for the selection of
one term rather than of another. The fourth category of class
deals with just such situations. Class is a fully grammatical cate-
gory but it might be said to act as a liaison between grammar and
lexis: if — vain hope — the whole of formal language activity could
be described grammatically we would have no further need of a
separate category of class. Finally a scale — of a different general
type, perhaps, but of equal status with ‘rank’, ‘delicacy’ and
‘depth’ — of exponence is needed in order to link formal items to
the categories, to join the categories to each other and to inter-
relate parts of categories. A formal item can be linked, quite
directly, to any one of the four categories by an exponence rela-
tion. It is through exponence that we are able to effect the cate-
gorical description of formal items and thus, by means of inter-
level correspondences, of language raw material.
The demi-level of grammar and the alter-interlevel of phono-
logy can now be said to possess really adequate and powerful
descriptive theories. Lexis, on the other hand, has received less
attention and should be considered as being in a rather less defi-
nitive state: the actual construction of a particular case of the gen-
eral theory of lexis would probably involve far more detailed
study of actual language raw material than would the construction
of a particular tightly-structured grammar. Such textual examina-
tion could be profitably aided by the use of mechanical devices
48 A PARTICULAR CASE: LINGUISTICS

such as computers, although the actual pattern-recognition and

methodological steps would always have to be carefully supervised
by linguists.
Theories at the interlevel of context can also be said to be at
an earlier stage of development when compared to grammatical
and phonological theories. Preliminary suggestions of the type of
situational factors which need to be considered have been made
by Firth’:
A context of situation for linguistic work brings into relation the
following categories:
A. The relevant features of participants: persons, personalities.
(i) The verbal action of the participants
(ii) The non-verbal action of the participants
B. The relevant objects
C. The effect of the verbal action
J. O. Ellis! has developed and extended this outline to propose
tentative contextual categories of, amongst others, participants
(general relevant features of them); immediate situation (relevant
parts of observation ‘near’ to the substance isolate); wider situa-
tion (everything more generally relevant from the rest of observa-
tion); register range (register being a division of idiolect according
to formal considerations and correlatable with immediate situation-
types: register range is the system of registers available to, for
example, the speaker); and register choice (selection of a register
from this system). The categories are to be interrelated to each
other and correspondences set up between them and patternings
at the level of form. Much work remains to be done before a
definitive theory (or theories) of context can be formulated; it is
probably quite natural that this interlevel should be one of the last
to receive full consideration: phonology, grammar and lexis are

10 Firth, Papers in Linguistics (London, 1957), p. 182; “A Synopsis of

Linguistic Theory, 1930-1955” in Studies in Linguistic Analysis (Oxford,
1957), p. 9. See also, amongst others, M. A. K. Halliday, The Language
of the Chinese “Secret History of the Mongols’ (Oxford, 1959), p. 13-14.
J. O. Ellis, “On Contextual Meaning” (to appear in the forthcoming
volume of essays in commemoration of J. R. Firth).
 LINGUISTICS AS A SCIENCE 49

both more concrete and perhaps more basic, in the sense of lin-
guistic science construction, than context. But only once each
level and interlevel has its full complement of descriptive theories
can the science function as an integral whole, each part of it being
essential to the complete framework and as important as every
other part.
Linguistics, like every other science, has formal meaning and
contextual meaning. Formal meaning will here have two compo-
nents corresponding to the demi-levels within the level of form:
formal grammatical meaning and formal lexical meaning. Con-
textual meaning will be determinable by consideration of the inter-
level of context once that interlevel has been provided with a set
of descriptively adequate theories.
The scheme of linguistic science presented above must be un-
derstood as an outline of general linguistics: each theory in it is a
general theory. Particular cases of the general scheme can be
constructed for each specific language. Thus each language can
be supplied with particular cases of the theories of graphology,
phonology, grammar, lexis and context. The general categories
will be the same in every grammar: but the way in which these
categories are made up and their components interrelated will vary
with the language under consideration. Context-descriptions, in
the cases of different languages, may be rather more similar than,
say, phonologies: situational features which occur in observation
near and relevant to a raw material isolate may not vary to the
same extent as the substance patternings if they correspond to
different languages. To summarise, a grammar, for instance, of a
given language will be a particular case of a general theory of
grammar at the level of form in linguistic science, this being itself
a particular case of our general scheme of science.
It remains to emphasise that none of the ‘theories’ mentioned
in this section will furnish heuristic (discovery) procedures for
‘automatic’ analysis. A general theory of grammar will certainly
not provide a procedure for the construction of particular gram-
mars for specific languages. And the grammar of a language is,
in turn, not a procedure for mechanical analysis of the sentences
50 A PARTICULAR CASE: LINGUISTICS

or whatever of that language. Instead, a general theory is merely

a framework within which particular grammars can be formulated;
and a grammar is again a categorical framework within which
descriptions of formal grammatical items can be stated. These
remarks apply equally to lexis, graphology, phonology and con-
text: in the same way, of course, our general scheme was far from
providing a procedure for the construction of any particular sci-
ence: what it did detail was a necessary, most general, structuring
for each science.

2.2 LINGUISTICS AND OTHER SCIENCES

Having described a general scheme of science, such that any par-

ticular science can be given explicit statement within its general
framework, we can usefully consider the interrelations between
these various particular sciences. The framework of the general
scheme — its levels and possible interlevels — will be a common
property of every science; if we consider a restricted set of sci-
ences we may be able to find more delicate similarities between
them. In order to discover all such similarities we should first
consider every possible pair of sciences and work out the features
common to the pairs. Then it would be a simple matter to deduce,
from these pair similarities, the common properties of a larger set
of sciences. If we have three sciences — A, B and C — their com-
mon features would be those that were common to the pairs A
and B, and A and C (and, of course, B and C). But this is not
suggested as a straightforward procedure — rather as a line of at-
tack which should only be followed if constant reference be made
to the detailed structures of the considered sciences. For any
analogic correspondences between parts of different sciences can
only be put forward in a fairly subjective and necessarily tenta-
tive manner. Distinct sciences will search for different types of
defining patterns in observation and have contrastive approaches
to the treatment of these patterns, as the purposes of their analyses
vary; sciences are recognizable as distinct, in the first instance, by
 LINGUISTICS AND OTHER SCIENCES 51

virtue of these different defining patterns and overlying ap-

proaches. So that, although we may discover analogous bits in
different disciplines we should not, strictly, go any further: anal-
ogies are in order, statements of ‘overlapping’ or ‘identity’ should
be treated with caution.
We are here mainly concerned with linguistics and thus in this
section with its interrelations with other sciences. Within our gen-
eral context we can examine whether one science can ever be said
to be, in some way, more basic than another. Most generally we
might tend to say that this is impossible: that each science has its
own approach, its own defining patterns, and is as important as
every other. When we do physics we must use language to write
down our observations and work out our results and predictions;
in the same way the physical properties of things are exploited
whenever we examine linguistic raw material. It might be objected
that linguistics, for example, is more basic than a science which
makes all its observations under conditions which are induced by
directed action and which would never be likely to occur without
such directed action application. For here the scientist uses lan-
guage as a tool under quite normal conditions (when writing down
his results, for instance) whereas the patterns he himself recog-
nizes would never occur in the same environment as linguistic
patterns. But this is, for two reasons, a considerably misleading
objection. There is no science which employs directed action to
the extent that it only considers patterns occurring in situations
which are completely abnormal: parts of sciences may do this but
any science, with its distinctive viewpoint, purpose and defining
patterns, is only to be significantly considered as a unified whole.
Secondly, this ‘more basic’ relation would be of little direct com-
parative interest since it merely means that a basic science would
need to be used to investigate the way in which the operations of
a less basic science are performed (a meta-investigation) whereas
the reverse would not be the case. So that we must allow our
statement to stand: no science can ever be said to be more basic
than another.
We have emphasised that each science is an entity in itself. It
52) A PARTICULAR CASE: LINGUISTICS

does not share parts with other sciences — two different sciences
do not have common scientific theories. If two disciplines do over-
lap, not just in the types of patterns they deal with but also in the
way they deal with them, then they must be part of the same sci-
ence, as we are here using the term. But it does appear that all or
most of what are commonly regarded as separate sciences would
be treated as separate within our general scheme. And so it is
vitally important that the methods and theories of one science
should not be transferred wholesale to another: this quite frequent-
ly is done, often on slender analogic grounds, under the misap-
prehension that one discipline, despite its distinctive purpose and
defining patterns, can be a particular case of another! Such an
apotheosis of one science with respect to another would appear
to be one of the most common errors made by scientists who have
insufficiant awareness of the different natures and points of view
of disciplines other than their own.??
12 Mathematics is also frequently apotheosised in this way. C. F.
Hockett (“Two Models of Grammatical Description”, Word, 10, 210-231)
establishes very dubious analogies between mathematics and linguistics and
then proceeds to generalise, analogically, within linguistics! He also states
in a footnote (14): “The inventors of mathematics, as speaking humans,
distill mathematical notions out of the raw material of everyday language.
Their notions are derived by leaving something out of the nearest every-
day-language analogs. A linguist analysing English must assume that two
plus three and three plus two are different forms. The mathematician
chooses to ignore everything which differentiates the meanings of these
two forms, and by so doing, he renders his ‘addition’ commutative. It may
seem strange that we who are concerned with the total complexity of lan-
guage should turn to mathematics for help, considering the ultimate source
of mathematics. But the circle is not in fact, closed: the mathematician
derives his notions by abstraction from language whereas we are deriving
not language itself, but a way of handling language, from mathematics.”
It may be misleading to say that mathematics ‘developed out’ of ‘everyday
language’ (although in origin they must have developed together) and the
observation that every mathematical formula can have some sort of rough
language write-out tells us nothing of the origin of either ‘mathematics’ or
‘language’. Most mathematics, today, is in no way abstracted from ‘lan-
guage’ and cannot be efficiently expressed in ‘ordinary language’. Thus
the mathematician does not “choose to ignore everything which differ-
entiates the meanings of two linguistic forms”. He will make direct ob-
servation not of language patterns but of general mathematical patterns
which occur in the world—or, most usually, he will work by correlating
 LINGUISTICS AND OTHER SCIENCES 53

As they stand at present, some parts of some ‘sciences’ would

hardly satisfy our stated criteria for a science. They may be based
on ideas of how things ‘ought to be’ instead of actual observation
of occurring patterns. And again the orientation of an observer
will depend upon his cultural background — the way in which he
recognises patterns and particularly his purpose in correlating
them will vary as he belongs, for instance, to the mediaeval Eng-
lish society or a twentieth century Russian community. A science
with a long history must be careful to examine its traditional foun-
dations and to ensure that they are acceptable within the context
of its present-day purpose: ‘theories’ which were motivated by
religious or mystical passion may be, coincidentally, valid scien-
tific theories; but we must make detailed examination of them
relative to their raw material to make sure that such is the case.
A particularly blatant scientific error is to take some part of a
science which is itself notional rather than observation-orientated
and to use this unsound thing in another science; even if the first
‘discipline’ were really scientific such uncritical transference would
be inexcusable! 18

and generalising patterns out of his previous patterns. Mathematics, like

every other subject, has some original correspondences to observation, but
it does not limit itself to describing observation. Only trivially can parts
of mathematics be called a science—it is better regarded as an abstract
game, a synthesising rather than an analysing discipline. It is also relevant
to note that Hockett believes mathematics to be a ‘way of handling’ parts
of language and, in this respect, his circle is indeed closed. He has gener-
alised from a particular case (which ‘ignores everything which differ-
entiates the meanings ...’) to generality.
18 The use of psychological ‘mental concepts’ (and so on) in linguistics
is particularly to be deplored. Although de Saussure was, after Sweet, one
of the first linguists to insist on the prior importance of a synchronic
approach his work was inevitably marred, to some extent, by the insistence
that language elements (signifiers) could only exist relative to mental con-
cepts (signifieds). In the present approach everything is necessarily relative
to the observer (or his mind, if you wish). Thus ‘mind’ is the vantage point
from which we make observation—the linguistic patterns in substance
then have existence relative to the observer’s ‘mind’: we have no idea what
happens in a mind when linguistic patterns are observed and, in any case,
this is not relevant to our science. Saussure’s ‘mind’ (as observer) estab-
lishes correspondences between ‘concepts’ in a speaker’s ‘mind’ and lin-
54 A PARTICULAR CASE: LINGUISTICS

If one science is, in some respects, at an earlier stage of devel-

opment than another it may very well be able to gain useful hints
from consideration of the methods employed in the second disci-
pline. Procedures must not be transferred in a wholesale fashion
—they are unlikely to be ideal for the new needs of a different
science; but they can often be usefully employed as a tentative
framework in the construction of the new science, to be altered
and adapted to properly fulfill their purpose. As an example we
can cite: statistics.!4 Statistics, as the term is normally employed,
deals with raw material in which can be recognised sociological
patterns: the name has an implication of sociological application.
But other sciences need to employ statistical methods and these
have and are being set up for each particular science as adapta-
tions of the sociological methods. Statistical mechanics is a well-
developed subject which gained immeasurably from consideration
of the methods employed in (sociological) statistics. At present
work is being done at Edinburgh (amongst other places) on the
use of statistical methods in linguistics —for example, sampling
techniques in raw material assembly and correlation procedures
in grammatical and lexical analysis. Such methods are being de-
veloped out of standard sociological procedures and within the
framework of general linguistic theories.
We have seen, then, that every science is basically independent
of every other. Care must be taken not to over-emphasise any one

guistic patterns. He mentions aspects of situation only as a part of the

rag-bag of ‘external linguistics’, Again, many sciences are wont to take
over non-linguistic (or pre-scientific-linguistic) ideas on language and then
imagine that they have actually drawn on a scientific linguistics; all they
have done is to undermine the scientific nature of their own discipline.
14 Statistics, as an empirical discipline, is not a part of mathematics.
It draws on mathematical methods, particularly those of probability
theory, but it is orientated towards description and explanation of observed
patterns whereas mathematics (which is, of course, not conceived of as a
descriptive science) is an abstract discipline, theory-orientated. Mathematics
is, it is true, sometimes loosely referred to as including statistics—a mis-
leading state of affairs— and this may account for the tendency to use the
name ‘mathematical linguistics’ for the new body of techniques which are
properly referred to as ‘statistical linguistic methods’.
 LINGUISTICS AND OTHER SCIENCES 55

discipline in such a way that results and techniques are thought-

lessly and dangerously transferred. But each discipline is likely to
benefit to a very considerable extent from a consideration of the
methods used in other sciences — new ideas can be gained which
can then be developed within the theoretical framework of the
new discipline. Linguistics, in particular, is not, in the most gen-
eral sense, any more important than any other subject; but it is
certainly not less important, either, and it is not any sort of par-
ticular case of any other science or of mathematics.

15 ‘Subject’ here refers only to such disciplines that are sciences ac-
cording to our General Scheme. Most of what is called ‘philosophy’ is,
for us, highly unscientific: the relative importance of scientific linguistics
with respect to any philosophical investigation is an important question
but one which is not discussed in the present essay.
 es eats
pan [-14 Pr isp
spllgacthion élaoatsricieal cemnaiel
‘
ae onHee Raat, ee S Hy cerqaltey MaENEical sneha TEE. Sent
eer Tag Totalhate backiy Set cap Tae alk avertee ABET ales ate aA
ae it ed Maaiel\ joie! Trkisane: Statics:Peal tnacharus ite well
Serratia! abbas witiek gues) Agaric rBbiy tases
sone tetation: :
ee the Paaihicet: Hhhd
« ‘yea tr roped) Stes tates et erann’
0: % Dey duarai. Edindorgh; famevign othe ghee) mh the
pike of pga methods iy Hpigbistics -~ omanpley SN:
ec Halips Pave. matecig | assembly aux! codidatih Daxdause |
, SS ROE oa and iexical yy natpwinsSech. wyeteaie ee, Deng:He
“Weloaat tut of etandacd soctohinion! Prreedages art.
oo
~ troonerek ofgeo in stn 5athetenis 3
cA Rhee denier, 3 ese, hat -owery wears ‘>
Tay tale a
oatsini take. e aa: said et =RAR, *
ie caaas, an¥one z ;

Gees, Sit Pen | Bier deny- army's 2Aeris pe

i’
SEO OT Seah cents, Aga, aauie Weep: ber
See menses eh, de> GYSRE iiae) mend on tie
ia eva Rae an: aly See es 4 S03
Ree ERR io. ee meannchiisie te polenta: ‘pation: ri
a: eet E:
oh RR, aN HME, 5 Be. chin eel bi‘llc: yo
oe ae +atm £ its ae
neMY aie
PART B
 LINGUISTICS AND LOGIC

3.1 THREE KINDS OF ‘LOGICAL...

The fifth of the six entries under logical in the New English Dic-
tionary on Historical Principles (1908) reads:
“S. [nonce-uses, after Gr.doy.x6¢] Characterised by reason;
rational, reasonable”
This is the wide, every-day sense of the word. Any action which
seems relevant to the situation, or for which some reason is given,
may be referred to as ‘logical’; it has been suggested that man
differs from other animals in that he acts purposively — in that he
is a ‘logical’ or ‘rational’ being. Needless to say there are many
instances of the use of logical, which a professional logician!
would find impossible to account for in terms of his theories and
rules. So that logical; refers not to some fixed way of reasoning
but rather to the general intuitive impression that some sort of
reasoning is present.
The second entry under logical and the first under logic in the
NED are:
“2. That is in accordance with the principles of logic; con-
formable to the laws of correct reasoning”
and
“1. The branch of philosophy that treats of the forms of think-
ing in general, and more especially of inference and of sci-
entific method”
In other words, the branch of philosophy that grew out of the
work of Aristotle; his syllogisms still make up the nucleus of logi-

1 Either a logicalg or a logical investigator.

60 LINGUISTICS AND LOGIC

calg philosophy. If a chain of reasoning can be expressed or re-

expressed in one of a set number of forms then it is logicale. Here
we have left behind the intuitive impression that something is
logical; and moved on to a formal, scientific analysis of whether
or not it is logicalg. Intuition will still be needed, however, in order
to attempt to fit the state of affairs into one of the logical, frame-
works: we have a method rather than an explicit procedure.
Whether the ‘laws of correct reasoning’ do reflect the way in
which people ‘reason’ or whether they are insufficient or prescript-
ive or whatever is an important question: we shall return to it
in section 3.4.
Little real addition was made to Aristotle’s logic during the
first 2,150 years following his death. When his theories were
eventually extended and improved upon in a significant manner
the motivation behind such alteration was rather different from
Aristotle’s original purpose. Begun by such as Boole, de Morgan,
Frege and Russell and continued by an ever-increasing number
of mathematico-logicians, the aim of the new logic was to be such
that predominantly mathematical work could be stated in terms
of its ideas. The aim of Hilbert and his followers? was to express
all of mathematics as rigidly formulated logical; systems: such a
system contains a number of basic symbols, any sequence of which
is a formula; a number of formulae postulated to be ‘true’, these
called axioms; and a number of rules of inference whereby the
‘truth’ of a formula could be deduced from the given ‘truth’ of a
number of other formulae. Then a sequence of formulae such that
the first is an axiom and each succeeding one is either an axiom
or else deducable as ‘true’ from some of the formulae preceeding
it in the sequence is called a proof of the last member of the
sequence, which is a theorem. Thus their ambition was to be able
to deduce all ‘true’ mathematical theorems by application of a
number of specified and carefully chosen rules of inference on

2 Gé6del’s famous result, unanticipated by Hilbert, has led to even more

abstract work in Mathematical logic, totally unrelated to the patterns
recognisable in any sort of natural observation.
 THREE KINDS OF ‘LOGICAL...’ 61

a given set of axioms; the rules of inference employed were similar

in form to Aristotle’s types of syllogisms.
So that logical; is a general term with a wide range of applica-
tion. Logicalg is a technical term in philosophy
— it can be used
to describe a wide range of phenomena. Logicals, on the other
hand, has become a technical term in the general field of Mathe-
matics — it implies that some complex of ‘results’ can be expressed
as the end-product of a logicals system. More accurately, the name
‘mathematical logic’ is used instead of the ‘logic’ referred to in
logical; but such a large amount of attention has been focussed
on logicals work in recent years that the ‘mathematical’ is often
dropped: the implication that non-mathematical work can be de-
scribed as logical; is, as we shall see in more detail later, mis-
leading. Logics) was developed within the framework of mathe-
matics and attempts at random and uncritical application of it
would be a definitive apotheosis of this subject. Its possible ap-
plicability to descriptive sciences should be the subject of detailed
investigation.?
The risk of confusion between the three distinct uses of ‘logical’
that we have chosen to distinguish is great. In the remainder of
this essay we will not only be concerned with a discussion of some
of the kinds of ‘logical’ linguistics that have so far been put for-
ward: we will, in addition, suggest a new ‘logical, linguistics’ which
will have a place in our scientific scheme.

3 Discussion of this applicability is provided in section 3.3.

4 J. M. Bochenski discusses varieties of formal logic in his A History
of Formal Logic (English Translation; Notre Dame University Press, 1961,
p. 12-13): “There are in essence, as far as we can determine, four such
forms: 1. The Ancient Variety of Logic. [ie. Greek and Stoic]; 2. The
Scholastic Variety of Logic; 3. The Mathematical Variety of Logic; 4. The
Indian Variety of Logic.” We here use logical to refer to both the an-
cient variety and also the scholastic variety. Indian logic is not, at present,
relevant to our limited discussion.
62 LINGUISTICS AND LOGIC

3.2 CHOMSKY’S ‘LINGUISTICS’

When Firth wrote: “Our studies of speech and language, and in-
deed a good deal of our educational methodology, have been
dominated far too much by logic and psychology” he undoubtedly
included de Saussure as one who placed uncritical reliance upon
psychological tenets. But Saussure himself was very much aware
that “logic has given us bad grammar and taken the heart out of
language”.> His Course in General Linguistics begins*:

The science that has been developed around the facts of language
passed through three stages before finding its true and unique object.
First something called ‘grammar’ was studied. This study, initiated
by the Greeks and continued mainly by the French, was based on
logic.7 It lacked a scientific approach and was detached from lan-
guage itself. Its only aim was to give rules for distinguishing between
correct and incorrect forms; it was a normative discipline, far re-
moved from actual observation, and its scope was limited.

During the last few years the work of Chomsky and his colleagues
has been the focus of much attention. Chomsky’s most compre-
hensive work is called The Logical Structure of Linguistic Theory:
we shall examine this new application of logic to linguistics in
this section.®

Papers, p. 186.
English Translation, London, 1961, p. 1.
The logical, discipline.
oar
Oa
We shall talk of ‘Chomsky’ when actually referring to Noam Choms-
ky, Morris Halle, R. B. Lees and other members of the ‘transformational’
school: this for brevity. The main works consulted (and referred to in
future by initials only) are: by CHomsky: The Logical Structure of Lin-
guistic Theory, mimeographed, Cambridge, Mass., 1955 (LSLT) (there are,
I believe, two distinct versions of this in existence — the one consulted con-
tains ten chapters and textual pages numbered up to 752); Syntactic
Structures, The Hague, 1957 (SS); “The Transformational Basis of Syntax”,
to appear in Proceedings of the fourth Univ. of Texas Symposium on
English and Syntax (1959) (TBS) (page references are to a mimeographed
version of the paper); “Explanatory Models in Linguistics”, to appear in
the Proceedings of the International Congress on Logic, Methodology and
Philosophy of Science, Stanford (1960) (EMIL) (page references are to a
 CHOMSKY’S ‘LINGUISTICS’ 63

Chomsky links linguistics with psychology: “A reasonable,

though still remote goal for linguistics and psychology would be
to construct a device capable of duplicating this performance [of
a child learning a language], or certain aspects of it. The lin-
guistic abilities of the mature speaker can in part be characterized
by what we might call a ‘formalized grammar’ of his language.
I will, for the remainder of this discussion, consider only these
aspects of linguistic competence.” He then suggests a device which
“represents a function that maps a set of observed utterances into
the formalized grammar of the language of which they are a sam-
ple. ... A description of this device would therefore represent a
hypothesis about the innate intellectual equipment that a child
brings to bear in language learning.”®
It should be noted that Chomsky states that a ‘formalized gram-
mar’ will only represent a part of our ‘linguistic abilities’. His
‘formalized’ indicates the presence of something analogous to a
logicals system?°: so that his hypothesis is that our minds employ
logical rules in part of their linguistic functioning. Obviously, as
a speculative hypothesis, this is as good as any other. But it is so
speculative and, in a positivist sense, unverifiable, that it seems
hardly worthy of inclusion in an account of linguistics. We might
equally well and equally uselessly suppose that our ‘minds’ work
in terms of grammatical categories such as were described in sec-
tion 2.1. But Chomsky’s dialectic later (in the same paper) ceases
to consider his device as ‘representing a hypothesis’ about a child’s
abilities and he appears convinced that his ‘linguistic’ theory is
psychologically valid: “A collection of ad hoc rules may succeed
in enumerating a great mass of facts without thereby giving any

mimeographed version of the paper).— By LExEs: “What are Transforma-

tions?”, a Russian translation of which appeared in Voprosy Yazykozna-
niya, 1961, 3, 69-77 (WAT) (page references are to a mimeographed ver-
sion of the paper in English).—I also had the benefit of a number of
conversations with R. B. Lees on ‘Chomsky linguistics’ whilst he was in
Edinburgh in January, 1962.
® EMIL, p. 3-4.
10 This seems to follow, in this instance, from the context of its men-
tion — but see footnote 16 below.
64 LINGUISTICS AND LOGIC

insight into the formal properties that distinguish the correct set
of structural descriptions from alternative sets that would be given
by totally different grammars; without, in short, contributing seri-
ously to the study of general features of linguistic structure or the
study of the nature of a universal language-learning device that
exhibits the intellectual abilities of the child’.
So that Chomsky’s main motivation appears to be to provide a
model for our (postulated) intuitive grammatical (logicals) rules.1?
The subject-matter of his ‘grammar’ is not identical with that of
the demi-level of grammar presented in section 2.1, although it is
intended to deal with roughly the same type of tightly-structured
small-set patternings. R. B. Lees has assured me (in private con-
versation) that all the formal patternings of a language can be
handled by a Chomsky ‘grammar’. Chomsky does, in fact, have
no equivalent to our lexis; but it will be seen below that his gram-
mar is not so much more powerful than ours that it can adequately
fulfill the ‘grammarian’s dream’.
In agreement with our statements in section 2.1 Chomsky does
not intend his general theory to be a heuristic procedure for dis-
covering the grammars of particular languages. He rather says:
“One function of this [general] theory is to provide a general
method for selecting a grammar for each language, given a corpus
of sentences of this language”’.18 He realises that there is need of
some criterion for distinguishing the ‘better’ of several grammars
for a specific language which are all obtained as particular in-

11 EMIL, p. 8.
12 His position is put even more unequivocably later in the same
paper: “This general theory can therefore be regarded as a definition of
the notion ‘natural language’ (insofar as we are concerned with its formal
properties). Its goal should be to exhibit the built-in information process-
ing capacities of the child that lead him to develop the specific linguistic
competence characterized in a fully explicit grammar. In this sense, it
should aim to express just what is ‘essential’ to natural language.” (EMIL,
p. 16.)
18 §S, p. 11 (my italics).
 CHOMSKY’S ‘LINGUISTICS’ 65

stances of the same general theory.‘ He mentions that “in careful

descriptive and methodological work, we almost always find that
one of the considerations involved in choosing between analyses
is the ‘simplicity’ of the grammar in which these analyses are
presented”.'5 Simplicity is thought to be of prime importance and
the word is used in a technical sense when he states: “Suppose
that we use the word ‘simplicity’ to refer to the set of formal 1é
properties of grammars that we shall consider in choosing among
them’”,7 and “in other words, we define simplicity so that, in cer-
tain clear cases, simplest grammars are in fact the correct ones”.'8
He discusses such a ‘definition’ of simplicity*® and concludes “we
have seen that as a consequence of the decision to consider the
simplicity of grammar to be a factor in the choice among gram-
mars, hence a factor in the determination of linguistic structure,
we find it necessary to make a detailed study of the form of gram-

14 Chomsky does not, of course, countenance the comparison of gram-

mars which are particular cases of different general theories. In such a
case a different initial approach may well have been made in each case
and it would be impossible significantly to compare either the particular
grammars or else the distinct general theories in any formal way. He
further specifies (LSLT, p. 7-8) that a general theory can be abstracted
from particular grammars only if the grammars are constructed within
the same general approach.
18 LSLT, p. 72.
16 The word ‘formal’ can be used in at least two quite different senses.
It can be used to describe the linguistic ‘form’ of language substance as
dealt with and described by our linguistic level of form. Or else it can
refer to ‘formal logic’: ‘formalized’ means that a loosely worded descrip-
tion is restated (without self-distortion) as a system of formal logic (if this
is possible: it is often not). But ‘formalized’ can also refer to linguistic
form and mean that a theory is explicitly stated within the level of form.
Thus ‘formal methods’ or a ‘formalized treatment’ may refer either to a
logical, or logical, statement or else to an explicit and scientific linguistic
statement. At many places in Chomsky’s work one is unsure which of
these two senses is intended. Even in Reichling’s telling criticism of
Chomsky’s ‘Cryptanalytic Formalism’ it is not clear which sense of ‘for-
malization’ is intended when he remarks “I am convinced that the future
lies with formalization, but only if developed along lines different from
those traced in ‘Syntactic Structures’” (Lingua, X, 17).
tO SSaips, 53%
18 LSLT, p. 81.
i9 LSLT, Chapter II.
66 LINGUISTICS AND LOGIC

mars and the methods by which we derive structural information

from a grammar. We have outlined a conception of grammar as
an ordered set of rules of the form a>, and have discussed the
question of how such a grammar might be evaluated in terms of
its simplicity. ... This conception is incomplete in many respects.
For one thing we have not really stated an evaluation procedure,
but only indicated how one might be stated. . . . It would be quite
easy to fill in these gaps one way or another, but it would also be
quite pointless at this stage of our knowledge.” ?° It is difficult to
disagree with Chomsky’s contention that a criterion of ‘simplicity’
must involve consideration of the ‘form’ of grammars. He suggests
that a grammar be a completely heuristic procedure (‘an ordered
set of rules...’) and can thus outline a ‘definition’ of simplicity
in terms of this procedure. But he later insists that grammars can
only be properly compared if they are mechanical procedures:
“It is important to note that despite the very great amount of
important grammatical work done in the last few decades, there
are unfortunately no published grammars which we can use as
they stand as data for determining the adequacy of these theoreti-
cal constructions concerning simplicity and the form of grammars.
The reason is that none of the existing grammars provides a liter-
ally mechanical method for generating sentences. Even though it
may be intuitively evident to every reader how to use these gram-
mars to generate utterances, we cannot know how much the for-
mulation of the missing steps will add to the complexity of the
grammar. The difficulty of presenting a really mechanical gram-
mar can be easily underestimated. But it is absolutely crucial to
develop such grammars if we wish to investigate seriously the
adequacy of a proposed linguistic theory in which such notions
of ‘simplicity of grammar’ play a part”.21 Chomsky’s simplicity is
thus so narrow that it can only be applied to ‘mechanical proce-
dures’ and not to ‘general methods’ such as grammars which are
particular cases of the general theory of grammar presented in
section 2.1; these ‘procedures’ are, in part, justified on the grounds

20 LSLT, p. 105-106.
2 OESED puntods
 CHOMSKY’S ‘LINGUISTICS’ 67

that they are necessary for simplicity evaluations. We too must

consider simplicity as an important factor — but as less important
than our prime purpose: to effect a faithful and adequate descrip-
tion of language patterning. Overemphasis on ‘simplicity’ to the
detriment of descriptive power can result in a conflation of pat-
terns which should properly be considered as at different degrees
of delicacy. Thus although the grammar itself may be made rather
simpler (relative to itself) it will give a less clear picture of ob-
served patternings and be more difficult to apply, to actually use
to describe language.?2
Chomsky thus departs from our scheme in that, although his
general theory is a method, particular cases of it are very much
discovery procedures; the concessions which have to be made in
order to achieve this indeed highly desirable state (but normally
unattainable in the face of other, more stringent, requirements)
will appear below. His viewpoint leads him to “regard the theory
of linguistic structure as being, essentially, the abstract study of
levels of representation”’.2* Following Rosenbloom” he states that
“a linguistic level L has the basic form of a concatenation alge-
bra’”’.2> Then “the best schema for grammatical description (. . .)
is the following: a grammar will consist of three components, a
constituent structure, transformational and morphophonemic com-
ponent. The constituent structure consists of a finite number of
rules of the form XAY—XZY, where A is a single symbol and
X and Y (but not Z) may be null. It contains a designated ‘initial
symbol’ S (standing for sentence) and a set of terminal symbols
which never play the role of A in rules of the form cited above. ...
Given such a system we can construct derivations of terminal
strings. ... The transformational component consists of a set of

22 This presupposes that a theory, or a particular case of it, only has

meaningful existence relative to its application to a previously chosen set
of raw material. As will emerge below, this differs widely from Chomsky’s
view.
23 LSLT, p. 60.
24 Paul Rosenbloom, The Elements of Mathematical Logic (New York,
1950), Appendix 2 (p. 189-193).
25 LSLT, p. 61-62.
68 LINGUISTICS AND LOGIC

complex operations, each of which converts a string with a given

constituent structure (...) into a new string with new constituent
structure determined by that of the given strings and the character
of the transformation. ...Some of the transformations are op-
tional; others are obligatory. ...The morphophonemic compo-
nent of the grammar is a sequence of rules, almost all obligatory,
of the form @-y, each interpreted again as an instruction of the
form ‘rewrite » as y’. Optional aspects of this component involve
only free variation or stylistic variation.”6
Historically “the notion of grammatical transformation arose
...in the course of Zellig Harris’ work on discourse analysis”’.27
“Harris himself viewed the entire study as an extension of de-
scriptive linguistics, not a part of it.” 28 But transformations were
used in a very different sense when “Chomsky attempted to for-
mulate a theory for grammatical description which would correct-
ly account for all or most of what we know of linguistic structure,
and he chose to think of this theory in the form which is often
employed in the study of logic and the foundations of mathe-
matics, namely as a set of abstract rules to generate from a finite
vocabulary of symbols the infinite set of grammatical sentences,
assigning automatically to each generated string its correct gram-
matical analysis”.2® So that transformations themselves can be
used within a descriptive linguistics as an optional additional tool,
or even as an integral part of a theory of descriptive linguistics,
be it procedural or methodological. Harris’ model was essentially
an IC one; transformations could equally well be employed within
the theories of grammar and phonology described in section 2.1.
In this sense they do not affect the general theoretical approach
(IC analysis in the case of Zellig Harris; ‘grammatical categories’
in our kind of linguistics) but merely provide a possible extra tool
within this approach.
Chomsky has retained Harris’ IC (heuristic) procedure in gen-

26 TBS, p. 8-10.
27 WAT, p. 1.
28 WAT, p. 2.
22 WAT, p. 4.
 CHOMSKY’S ‘LINGUISTICS’ 69

eral outline but he has subjugated the importance of the constitu-

ent structure component and introduced the transformational com-
ponent as the other integral part of his grammatical theory (we
can temporarily restrict his ‘grammar’ so as to exclude the mor-
phophonemic component — which in any case presupposes spoken
language material — and thus bring it more into line with our use
of this term). But he has effectively reversed the traditional IC
procedure in that he is not interested in analysing observed sen-
tences but rather in generating all possible sentences. The con-
stituent structure component provides the first type of generation
(morpheme to phrase, etc.) and the transformational component
the second type (active phrase to passive phrase, etc.).3° More
recently the original IC procedure has been improved upon: for
example, in place of the traditional binary divisions (or syntheses,
in the reversed model) ternary divisions (or even larger ones, in
exceptional cases) are allowed where appropriate*!; the constituent
structure component has been arranged in such a way that it
forms, with the transformational and morphophonemic compo-
nents, an integral whole. Chomsky does not, in fact, separate
grammar (meaning constituent structure and tranformational com-

30 Jn SS (Chap. 5) Chomsky demonstrates the inadequacy of an IC

model (by itself) and uses this to justify the introduction of a transforma-
tional component. But this throws light only on the general insufficiency
of IC—not on the general necessity for transformations. Within our own
approach we would merely suggest replacing the IC model by a grammar
and lexis as outlined in section 2.1. Moreover, Chomsky has a number of
‘kernel sentences’, those that are generated by the constituent structure
component. This leads to a peculiarly one-sided scheme of description.
A sentence which is in the kernel can be described directly in terms of the
constituent structure component whereas a non-kernel sentence has to be
described in terms of, first, the transformational component and, secondly,
the constituent structure component. This contrasts with our completely
homogeneous description of a formal grammatical item in terms of the
four categories of our theory of grammar. But it must be remarked that
there has recently arisen a tendency to play down the importance of the
extremely subjective ‘kernel’ or even to do away with it altogether (R. B.
Lees assured me, in private conversation, that he at least tends to this
opinion).
31. See WAT, for example.
70 LINGUISTICS AND LOGIC

ponents) and phonology (morphophonemic component) in the way

we do.
There is, as we have mentioned above,*? a very strong analogy
between Chomsky’s generative grammar and mathematical logic
—so much so that his grammar can be looked upon as a particular
type of a logical; system. So that The Logical Structure of Lin-
guistic Theory is, by its terseness, a rather misleading title. Here
logicals is intended and could be replaced by ‘mathematical logical’
for greater explicitness. There is nothing particularly new in
Chomsky’s way of looking at language (he himself mentions how
similar his viewpoint is to that of ‘traditional grammar’**) and
parts of his theoretical apparatus are highly derivative.*4 But the
way all these are put together to make a ‘Linguistic Theory’ is
completely original — the first true (logical) ‘generative’ grammar.
Chomsky has not restated any earlier theory but has instead in-
vented a new ‘linguistics’. Since this theory can be regarded as a
logical; system it must seem reasonable to expand the title of his
large exposition to: The Mathematical Logical Structure of a
(Mathematical) Logical Linguistic Theory; in this way we can
bring out the full nature of his examination (logicals) and of his
‘new’ theory (the first logicals linguistics).3> Inasmuch as Choms-
ky’s logical; theory differs in so many respects from all previous
linguistic theories we are extending the meaning of the term lin-
guistics,®* to a significant extent, when we refer to his theory as a
linguistics. It is thus particularly to be deplored that the first ac-
32 WAT, p. 4: quoted above.
33 “The Logical Basis of Linguistic Theory”, to appear in the Pro-
ceedings of the Ninth International Congress of Linguists.
34 His constituent structure component is very similar in detail to
standard American immediate constituent procedures; his transformation-
al component grew out of Zellig Harris’ work on transformations in de-
scriptive linguistics.
35 The title of his work, when thus expanded, is seen to be extremely
misleading. The Logical Structure of Linguistic Theory implies that the
logical structure (logicaly or logical, presumably) of some already existing
linguistic theory is being examined. Whereas Chomsky actually injects logic
into linguistics and then the ‘logical structure’ is just this injection. A
Logically Structured Linguistic Theory would be a more appropriate title.
36 Its formal lexical meaning.
 CHOMSKY’S ‘LINGUISTICS’ 71

count of this new theory should merely refer to it as a ‘linguistic

theory’.
Chomsky’s treatment of ‘meaning’ has aroused a large amount
of comment and criticism. In section 1.4 we gave an explicit dis-
cussion of the types of meaning which would be employed in ob-
servational sciences — the word ‘meaning’ can be used in a large
number of different, every-day senses most of which need not
concern us in our pursuit of descriptive linguistics, or of other
sciences. The glossematicians*’ do introduce an additional type of
‘meaning’ in their discussions of content-substance (and content-
form) when they say?* that the content-entity ‘boy’ is equal to the
syntagmatic association of the content-entities ‘young’, ‘male’ and
‘human’; in this it seems that they presuppose some kind of refer-
ential meaning, although this does not appear to be explicitly
stated in any of the works of Hjelmslev and Uldall. Chomsky
mentions that “there are two distinct sets of semantic notions.
Following Quine*® we may distinguish between the theory of refer-
ence and the theory of menaning. ... The distinction between
meaning and reference becomes clear when we recognize that not
all words have reference (e.g., ‘and’, ‘of’), even though they are
presumably not without meaning, and that even among referring
expressions, certain expressions with the same reference may not
have the same meaning. Thus ‘centaur’ and ‘unicorn’ have the
same (null) reference, but are not synonyms.”4° 41 Chomsky’s

37 It must be remarked that glossematics is not a proper observational

science according to the scheme of section 1, since it fails to emphasise
the necessary descriptive (faithful representational) nature of any science.
38 Outline of Glossematics: Part One, General Theory by H. J. Uldall
(Copenhagen, 1957), p. 46.
39 See Quine, From a Logical Point of View (Cambridge, Mass., 1953),
p. 47 et seq.
40 Chomsky does not appear to use the term ‘synonym’ to refer ex-
clusively to items which have the same grammatical descriptions (we would
insist on the same lexical descriptions as well) but rather in a notional
sense e.g. (SS, p. 95) he states that ‘bachelor’ and ‘unmarried man’ are
synonyms. Throughout his work Chomsky insists on formal criteria but
frequently disregards his own advice and employs psychological and no-
tional (not to mention prescriptive) criteria.
41 LSLT, p. 23.
T2, LINGUISTICS AND LOGIC

standpoint on the whole question of ‘meaning’ is summarised

when he says “that semantic notions are quite irrelevant to prob-
lems of formal structure,*? that only their unclarity disguises their
irrelevance, and that when the claim is put forward that linguistic
analysis can not be carried out without the use of meaning, what
is really expressed is that it can not be carried out without intui-
tion”.48 And more explicitly “I believe that the inadequacy of sug-
gestions about the use of meaning in grammatical analysis fails to
be apparent only because of their vagueness and because of an un-
fortunate tendency to confuse ‘intuition about linguistic form’ with
‘intuition about meaning’, two terms that have in common only
their vagueness and their undesirability in linguistic theory”.44 The
main obstacle to an understanding of Chomsky’s attitude to mean-
ing is, however, that his own idea of ‘meaning’ appears, to me at
least, to be extremely vague and unclear. We will mention below
some of the things that Chomsky appears to have subsumed under
‘meaning’. It must be remarked that his acknowledged opinions
concerning the unclarity of ‘meaning’ and his failure to explain
his own use of the term do not preclude him from discussing
‘meanings’ at great length, often in an extremely bewildering man-
ner. If one wishes to employ a tool in any scientific or mathe-
matical investigation then it must be clearly presented and de-
scribed in advance of its application. This is what we have done
in section 1.4 with our particular ‘contextual’ and ‘formal’ mean-

42 Chomsky’s opinion that ‘meaning’ has no place in the study of ‘form’

is or has been shared by many other American linguists and is the prime
doctrine of logical, formalists (notably the ‘logical positivists’ such as
Schlick and Carnap): “In daily speech no word has a perfectly fixed mean-
ing; there is always some amount of free play, the greater, the more ab-
stract the notion is. This makes people miss each other’s point, also in non-
formalized mathematical reasonings. The only way to achieve absolute
rigour is to abstract all meaning from the mathematical statements and to
consider them for their own sake, as sequences of signs, neglecting the
sense they may convey. Then it is possible to formulate definite rules for
deducing new statements from those already known and to avoid the un-
certainty resulting from the ambiguity of language” (speech by “FORM” in
A. Heyting, Intuitionism, Amsterdam, 1956, p. 3).
43. DESLT, p. 23.
44 68S, p. 94.
 CHOMSKY’S ‘LINGUISTICS’ 13

ings. Chomsky, on the other hand, appears to be searching for

‘God’s Truth’ in the welter of everyday uses of meaning. Instead
of specifying some particular types of meaning he avoids the issue
by saying that “its vagueness and unclarity make the notion of
meaning an ineffective criterion of analysis and a dubious foun-
dation for linguistic theory” *>; an investigator has only himself to
blame if a potential tool is ‘vague and unclear’.
He then goes on to discuss meaning: “If we take ‘meaningful-
ness’ or ‘significance’ seriously I think we must admit that
I noticed a round square
or
Colorless green ideas sleep furiously
are thoroughly meaningless and non-significant, but it seems to
me that as a speaker of English, I would regard these as in some
way “grammatical sentences’, and it can certainly be argued that
the establishment of their non-significance lies outside grammar.” 46
This touches on a very deep and vital question concerning intui-
tive ideas on ‘meaningfulness’ of utterances which cannot be gone
into in any detail here. But a distinction must be made between
the reaction of logicians (and logical linguists) to a question re-
garding the ‘meaningfulness’ of sentences and that of ordinary
language users. I have carried out an experiment which consisted
of dropping utterances, purposely constructed so as to be mean-
ingless in a certain sense, into everyday conversation. Such is the
way in which language is normally used that if an utterance can
be meaningless in a certain sense then it will always be understood
in a different sense, a more meaningful one, if such a sense can
be found. In other words an ordinary language user will tend to
read as much meaning as possible into any utterance; this is, in
fact, one of the most important ways in which language can fulfill
its everyday job of communication. Items in ordinary language
will normally have a quite wide range of potential meaning‘? and
45. ESET p. 29:
46 LSLT, p. 37.
47 Cf. Angus McIntosh, “‘Graphology’ and Meaning”, ArchL, 13,
107-120.
74 LINGUISTICS AND LOGIC

it is thus almost always possible to read some meaning into any

utterance. In a more restricted technical language linguistic items
will have far more restricted potential meanings and it is thus
easier to construct truly ‘meaningless’ sentences in such a restrict-
ed language. The wide potential meanings inherent in normal dis-
course virtually proscribe any attempt at constructing completely
meaningless everyday sentences. The restricted language of logic
is particularly rich in meaningless sentences since logicians are
continually on the look out for ‘paradoxes’ and ‘contradictions’;
a normal language user will only label anything as ‘meaningless’
as a very last resort. Thus Chomsky’s identification of ‘I noticed
a round square’ as meaningless must definitely be regarded as a
minority opinion. ‘I noticed a round square’ was understood by
the subjects of my experiment (all English) to refer to a grass-
covered ‘square’ in a city such as London which was circular in
shape,‘8 or else to be a statement in the restricted but well-known
‘jazz language’. Similar, if more complex, results were obtained
with the equally meaningful (and by now quite memorable) ‘color-
less green ideas sleep furiously’.
The danger of any exemplificatory, as against observational,
technique is that the investigator’s own uses of and ideas con-
cerning, in this case, language may be (perhaps for the very rea-
son that he is that sort of investigator) quite abnormal. This is
certainly so in the case of Chomsky and his remarks concerning
meaningfulness can only be regarded as having any validity with
respect to the particular restricted language used by logicians,
mathematicians, some philosophers and a small set of logical
linguists.
Chomsky categorises sentences as meaningful or meaningless
and also as grammatical or ungrammatical (although more re-

48 Not being familiar with American ‘city terminology’, before writing

this essay I rang up J. C. Catford to ask if there were any ‘squares’ in
New York as there are in some English cities. Without knowing the reason
for my question his first reply was: “Well, there are round squares, called
circles or circuses or something”, showing that the collocation ‘round
square’ can occur in a completely meaningful and yet spontaneous man-
ner in ordinary conversation.
 CHOMSKY’S ‘LINGUISTICS’ aq

cently the name ‘well-formed’ appears to have been used in pref-

erence to ‘grammatical’). Thus he says:
“Sentences (1) and (2) are equally nonsensical [not meaning-
ful], but any speaker of English will recognize that only the for-
mer is grammatical
(1) Colorless green ideas sleep furiously
(2) Furiously sleep ideas green colorless”4%
Chomsky’s ‘grammaticalness’ is, to some extent, covered by our
formal grammatical meaning — here we would account for the non-
occurrence of (2) in terms of firstly its formal grammatical mean-
ing and, secondly, its formal lexical meaning.5° We could stretch
our terminology and say that (2) is ungrammatical and unlexical
whereas (1) is grammatical but unlexical. That is to say: the col-
location of ‘colorless’, ‘green’, ‘ideas’, ‘sleep’ and ‘furiously’ does
not occur (or rather, has only a very small probability of occur-
ring) in English. Chomsky does not consider the situation of any
linguistic item (he considers only formal patternings looked at
solely in relation to themselves) and thus has nothing correspond-
ing to our contextual meaning.
So that on the basis of the above we can tentatively and partial-
ly identify Chomsky’s ‘grammatical’ with our ‘grammatical’ and
his ‘meaningful’ with our ‘lexical’. His statement that meaning is
not part of linguistics proscribes his formal study of what we call
lexis. And although, as stated earlier, Chomsky and his associates
believe that any sort of formal patterning can be described within
their grammar (the fulfillment of the grammarian’s dream), in the
examples above Chomsky himself shows that this is not so, that
some patterns are of a lexical nature and can only adequately be
described in a non-grammatical way.
But, although their fields of application are overlapping,
Chomsky’s ‘grammatical’ is used in quite a different way from
ours. Our formal grammatical meaning is part of our theoretical

49 SS, p. 15.
50 Our formal lexical meaning must not be confused with Chomsky’s
use of the term ‘lexical meaning’ (SS, p. 104, 108); the two are quite dis-
similar.
76 LINGUISTICS AND LOGIC

framework and is only employed in theoretical description. We

make no claim that it will correspond, in any sort of direct way,
to any of our intuitive ‘ideas’. Chomsky, on the other hand, talks
of the ‘intuitive grammatical sense’ of speakers. He mentions that
‘meaningful’ cannot be identified with ‘grammatical’ * — this is to
say that the two criteria are not identical, and from this he infers
that they are completely independent of each other.®* This again
emphasises the need for two distinct approaches to the study of
‘linguistic form’ — grammar and lexis. Since to Chomsky ‘linguistic
form’ consists solely of grammar he can say, when considering the
field of operation of his theory, “it appears then that in a certain
sense the ultimate criterion remains the speaker’s intuition about
linguistic form. . .”’.58
Chomsky exemplifies this intuitive idea of ‘grammaticalness’ or
‘well-formedness’: “given the sequences:
(a) the dog looks terrifying
(b) the dog looks barking
(c) the dog looks lamb
any speaker of English knows that (a) is well-formed, and that (b)
and (c) are formally deficient (although, if they were sentences,
their meaning would be clear and fairly unambiguous). He also
knows that (b) is somewhat closer to well-formedness than (c).
In this and inumerable other perfectly clear cases, we should re-
quire that the formalized grammar make the same discrimination,
assigning to each sequence of words or morphemes what we can
call a ‘degree of grammaticalness’.”’*4 Purely for simplicity Choms-
ky has reduced this scale of grammaticalness to a dichotomy of
grammatical/ungrammatical in much of his work — he is aware of
this simplification and of the need for later generalisation. It seems
unfortunate that a speaker’s intuition about ‘grammaticalness’
should have to be employed in such an investigation — the reason
for this is Chomsky’s neglect of lexis and lexical patternings. In a

st §S, p. 15.
82 §S, p. 17.
53 SLT, p. 58.
54 EMIL, p. 6.
 CHOMSKY’S ‘LINGUISTICS’ Wi)

fully comprehensive linguistic investigation every sort of formal

patterning will be observed — grammar and lexis employ indepen-
dent approaches but each is pursued in consultation with the ma-
terial, methods and categories of the other; in this way grammar
and lexis serve, to a very large extent, to delimit and define each
other. If grammar is studied and lexis is ignored then some exter-
nal means of selecting grammatical and rejecting lexical patterns
has to be employed.
Thus we must examine in greater detail the ‘intuitive gram-
matical senses’ of our speakers which are so necessary for initial
selection of Chomsky’s raw material. He implies that ‘Sincerity
admires Bill’ and ‘Golf admires Bill’ are more grammatical than
‘The admires Bill’, since ‘Jones admires Bill’ is perfectly gram-
matical and ‘Golf’ and ‘Sincerity’ are, like ‘Jones’, nouns whereas
‘the’ is not a noun.® It is certainly the case that many English-
speaking people will refer to sentences as ‘grammatical’ or ‘un-
grammatical’ in the ordinary course of their lives. From where do
they get their criterion of ‘grammaticalness’?>* The answer’ ap-
pears to be that it is largely based upon the ‘grammar’ they were
taught in school or college. The majority of school grammar
courses are, of course, very traditional in outlook and can fairly
be described as largely derived from the original Greek ideas of
grammar as ‘denounced’ in our passage from Saussure at the be-
ginning of this section: great stress is laid upon the parts of speech
— nouns as ‘naming words’, verbs as ‘doing words’, and so on; the
traditional notions of ‘subject’ and ‘predicate’, so misleadingly
borrowed by Aristotle from his logic, are still widely emphasised.
Prescriptive teaching in schools will inevitably affect the language
to a certain extent — but the small extent of this influence is shown
by the continued inadequacy of traditional grammars as descrip-

55 DLSLT, p. 110-111; it is in just such cases that we find the scale of

delicacy most valuable.
56 Or, more fully: From where do they get the term ‘grammatical’ for
part of their intuitive linguistic sense, and some of their criteria for the
linguistically acceptable (which they, like Chomsky, over-extend the term
‘grammatical’ to)?
78 LINGUISTICS AND LOGIC

tive vehicles. And the many people who use language completely
instinctively and without any school analysis or prescription —
there are a considerable number of English speakers who fall into
this category and in the cases of other languages the vast majority
of speakers can be thus described — will certainly have no intuitive
‘srammatical sense’. So that Chomsky delimits his field of investi-
gation through application of ideas which are based on the worst
type of traditional and prescriptive teaching and which are only
held by those (relatively) few people who have been subjected to
this type of teaching. That virtually all the people with whom a
linguist comes into contact and those who read his work will nec-
essarily have learnt ‘traditional grammar’ at school should not
blind him to the artificiality of their ‘intuitive grammatical ideas’
and the fact that in building on these he is presupposing the va-
lidity of a type of grammar long condemned by professional lin-
guists and unfortunately (and necessarily misleadingly) cast into
the mould of Aristotelian logic.
A. A. Hill5’ has indicated, by experiment, that ideas concern-
ing ‘grammaticalness’ will vary from individual to individual. This
had previously been realised by Chomsky and in itself it does not
destroy the possible usefulness of his criterion. It is the general
origin of ‘grammaticalness’ that makes it unsuitable for use as a
procedure in a new type of linguistics. I must confess to being
personally bewildered by Chomsky’s divisions into grammatical
and ungrammatical. I may have a ‘weak’ grammatical sense. But
I am unable to uphold the IC analysis of ‘my friend enjoyed the
book’ and ‘my family liked the book’ into ‘my friend/enjoyed the
book’ and ‘my family/liked the book’ rather than ‘my friend en-
joyed/the book’ and ‘my family liked/the book’ on the grounds
that ‘my friend enjoyed and my family liked the book’ is ungram-
matical whereas, for instance, ‘my friend enjoyed the play and
liked the book’ is perfectly grammatical.58 I am quite unable to

57 Archibald A. Hill, “Grammatically”, Word, 17, 1-10.

58 ESET, p. 272-273.
 CHOMSKY’S ‘LINGUISTICS’ 719

distinguish, grammaticalnesswise, between these two sentences.°9

Chomsky continually insists upon ‘formal grounds’ being used
throughout a linguistic investigation; yet his procedures depend
upon a weak, prescriptive and entirely notional criterion of ac-
ceptable/non-acceptable, correct/incorrect or good/bad instead of
upon the entirely consistent and scientifically valid observational
criterion.®* As Quirk remarked * in quoting from Latham’s Com-
parative Philology (1862): “What is called ‘bad grammar’, says
Latham, is a detail in which a speaker differs from someone else
who calls his form of speech ‘good grammar’, but imperfect and
misguided writings on this subject have confused the issue and
obscured the quite separate ‘philological truth that whatever is,
is right’.” So that the major difference between the grammar de-
scribed in section 2.1 and Chomsky’s grammar will now emerge:
our linguistics is an observational science, uninfluenced by previ-
ous linguistics insofar as is possible, which looks for linguistic
patternings in language raw material. ‘Grammar’ is our name for
that demi-level at the level of form which deals with the intricately
interrelated, closed set patterns; it only has existence relative to
lexis and the other levels of linguistic theory. Chomsky’s grammar
has independent existence apart from lexis, situation or substance
and it is constructed with attention given to intuitive, school-
prescribed ideas. So that our grammar and that of Chomsky,
whilst overlapping to some extent in applicability, are definitely
not directly correspondent or comparable. The exemplificatory
approach is, of course, more valid if a speaker’s intuition about
complete linguistic form is used—then he merely distinguishes

59 One is tempted here to apply Chomsky’s own remarks (SS, p. 23):

“It is difficult to conceive of any possible motivation for excluding them
from the set of grammatical English sentences” to ‘my friend enjoyed and
my family liked the book’ and ‘the little tug chugged up and the great liner
sailed down the river’ (LSLT, p. 271).
60 Our criticisms are based upon Chomsky’s dichotomy of grammatical/
ungrammatical as employed throughout his work. He admits that this is
a simplification and that several degrees of grammaticalness would be
preferable — our criticisms, suitably modified, would still apply in this case.
61 Randolph Quirk, The Study of the Mother-Tongue (Inaugural lecture
delivered at University College, London, 1961), p. 15.
80 LINGUISTICS AND LOGIC

between what is in his language and what is excluded. The division

into grammar and lexis in our approach or the isolation of gram-
mar in the traditional one are purely theoretic and artificial de-
vices which in no way mirror any inherent dichotomy within lan-
guage. If observational grammar and lexis were taught in schools
(and thus, trivially, prescribed) then it would still be unwise to
employ ‘intuitive grammatical (or lexical) senses’ since a vicious
circle would be set up which would prevent us from noting
changes in the language or improving upon our descriptive lin-
guistic theories.
In a normal scientific investigation the raw material will be
more or less delimited, by the recognition of the defining patterns
of the science in question, before explicit theories are constructed.
Or rather this is our ideal — the difficulty of getting a fully descrip-
tive set of theories will mean that a certain, small adjustment to
this delimitation may have to be made at a post-theory-construc-
tion stage. So that when Chomsky says: “But a field of investiga-
tion cannot be clearly delimited in advance of a theory dealing
with this subject matter. In fact, it is one of the functions of a
theory to give a precise delimitation of a field of investigation”
,®
we are bound to agree, in essence, with his first sentence. But as
a descriptive science is more fully developed the aim must always
be to construct theories which will more and more nearly describe
the exact set of raw material which was delimited in the first in-
stance — and only that set of raw material; so that our goal must
always be to construct theories such that the field of investigation
as previously delimited can be perfectly and adequately decribed.
For our theories only exist relative to the raw material and their
sole purpose is to provide succinct and accurate descriptions of it:
the theory is to be a model for the raw material.
Chomsky uses ‘model’ in something approximating to this sense:
“Such a grammar could properly be called an explanatory model,
a theory of the linguistic intuition of the native speaker.” ® But

ESET alAs
63 EMIL, p. 9.
 CHOMSKY’S ‘LINGUISTICS’ 81

he also employs the idea* in the entirely different ‘mathematical

sense’ ® in which a set of raw material is a model for a theory:
“. , . linguistic theory presents an abstract construction of a system
of levels of representation related to each other in certain fixed
ways, and formulated in terms that are operationally interpretable
in a certain specified manner. The actual process of linguistic
analysis, then, is the process of finding an interpretation for this
system in a given corpus of linguistic material.”®* So that for
Chomsky a grammar can exist apart from any raw material,
“...Wwe were studying language as an instrument or a tool, at-
tempting to describe its structure®? with no explicit reference to
the way in which this instrument is put to use”,®* just as, for
Hjelmslev, a system (language) can exist without there being any
process (text) belonging to it.*® In fact Chomsky will construct
his grammar out of a set of raw material (those utterances which
are ‘grammatical’ or ‘well-formed’) but the theory then appears
to be assigned an independent existence, the original raw material
is ignored, and an interpretation is looked for for the theory — this
interpretation will then become ‘that which is described by the
theory’. So that in the final resort a grammatical sentence is to be
defined not in terms of our intuitive grammatical senses but rather
as ‘that which can be generated by the linguistic theory’. Formal
criteria are substituted for the intuitive ones but the formal ones
are only correlated out of the intuitive ones and it appears even
less likely that they will be adequate in any descriptive sense. And
the theory does not limit itself to the description of occurring sen-
tences which are grammatical —it can generate many sentences
which will be formally grammatical although they would never be
encountered in practice. For instance R. B. Lees assured me (in

64 He does not use the actual word ‘model’ here but the implication
of ‘finding an interpretation for’ is clear enough.
65 See above, p. 21, footnote 18.
66 TSLT, p. 54.
67 Meaning by this just its grammatical structure.
68 §S, p. 103.
69 Prolegomena to a Theory of Language, English Translation (Balti-
more, 1953), p. 25.
82 LINGUISTICS AND LOGIC

private conversation) that since (in traditional terms) ‘take off

occurs both as a transitive and an intransitive verb and ‘take
down’ as an intransitive verb then a sentence such as ‘the plane
took down’ would most probably be considered grammatical al-
though it was “not at present in use”. If lexis were studied as well
the dual criteria of ‘grammaticalness’ and ‘lexicalness’ would nec-
essarily restrict an investigator to ‘actual’ rather then ‘potential’
language material. The examples quoted above will clearly indi-
cate how meaningless an isolated study of ‘grammar’ can become.
Any actual language is clearly delimited in corpus at any given
time. It can be imagined that Chomsky’s generative grammar
transcends this limitation and is capable of generating an ‘infinite’
number of sentences, all of which must, however, be of finite
length. A descriptive theory will, as we have already mentioned,
consider a statistically valid and representative sample of actual
language material. The vast difference between this approach and
that of Chomsky is clearly seen when one encounters Chomsky’s
amazing statement that “in general, the assumption that languages
are infinite is made in order to simplify the description of these
languages”’.7° Surely no better reason could be put forward for
using a descriptive approach as outlined in Part A in preference
to a ‘mathematical-generative’ viewpoint; Chomsky’s ‘description’
is of language not as it is but as it might be, the ‘might be’ being
deducible from Chomsky’s own theory! Some of the blame for the
‘infinite’ aspect of his work lies with his insistence on definite and
fixed (often dichotomous) distinctions in place of ranges of proba-
bilities or continuum-type scales. As Halliday has stated: “As the
description increases in delicacy the network of grammatical rela-
tions becomes more complex. The interaction of criteria makes
the relation between categories, and between category and expo-
nent, increasingly one of ‘more/less’ rather than ‘either/or’. It be-
comes necessary to weight criteria and to make statements in terms
of probabilities... but the ‘more/less’ relation itself, far from
being an unexpected complication in grammar, is in fact a basic

TOP YSIS De nz
 CHOMSKY’S ‘LINGUISTICS’ 83

feature of language and is treated as such by the theory”.7! So that

it is necessary not only to give the ‘grammatical rules’ of a lan-
guage but also to detail the extent of their applicability: Chomsky’s
supporters frequently cite the example of a sentence which is the
conjunction of a certain number of clauses: it can always be made
one clause larger by the conjunction of another clause, thus prov-
ing that there are an infinite number of sentences in a language.
The actual state of affairs is that if a sentence, S, has proba-
bility, p, of occurring and if S’ is formed from S by the conjunc-
tion of another clause, S’ having probability p’ of occurring, then
p’ need not be identical with p. In fact, if S is a conjunction of,
say, more than five clauses then p’ is very likely to be much less
than p. To decide upon the size of a language at a particular time
it is necessary not just to count the number of sentences allowable,
but to sum the probabilities assigned to each allowable sentence.
In the case of sentences which each consist of a conjunction of
clauses we are clearly unable to say that there is any definite
number, N, such that no sentence contains more than N clauses.72
But we can say that, as n gets larger, then p,, the probability”
of a sentence of n clauses occurring, will get smaller (once n is
larger than two or three). So that the sum of the infinite series:
Pit pe+tps+pi+...
will be a finite number, that is to say the series will be convergent.
From this point of view, that of language as it is, every language
is synchronically finite. It is only when one considers a language
as it might be, according to a generative theory like Chomsky’s,
that one is led to regard it as ‘infinite’. Chomsky mentions: “The
grammatical approach, with its establishment of a sharp dividing

71 “Categories of the Theory of Grammar”, Word, 17, 259; my italics.

72 In case any reader disagrees with this and puts forward some defi-
nite N we can remind him of occurrences such as the 1943 report of the
President of Columbia University which “filled 11 pages, contained 4,284
words and was punctuated as a single sentence” (Fries, The Structure of
English, London, 1957, p. 11).
73 This ‘total probability’ can be deduced from separate ‘grammatical’
and ‘lexical’ probabilities, in turn deduced from the theories at the two
demi-levels within the level of form.
84 LINGUISTICS AND LOGIC

line between the two classes G and g [grammatical and ungram-

matical] contrasts with a statistical approach that assigns a proba-
bility of occurrence to each sequence... .” He further remarks
that whilst “it would be easy to picture the statistical approach
just described as being a generalisation of the grammatical ap-
proach” this would be, in fact, “a gross misconception” since
Chomsky’s theory will account for statements such as ‘colorless
green ideas sleep furiously’ which is grammatical but (to him)
meaningless but not ‘furiously sleep ideas green colorless’ which
is both ungrammatical and meaningless.74 We have already seen
that this arises from his failure to contemplate a formal study of
lexis. The probability of occurrence (meaning the ‘total proba-
bility’) will obviously not be a simple generalisation of a purely
grammatical dichotomy. If, however, the ‘grammatical probabili-
ty’ were separated off from the ‘lexical probability’ then Chomsky’s
dichotomy should indeed be a particular case of the former as far
as actual occurring sentences are concerned (excluding such as
‘the plane took down’). But such an isolated ‘grammatical proba-
bility’ would be a purely theoretical artifice shedding little light
on the actual probabilities of language occurrence; ‘lexical proba-
bility’ must be considered together with it for any meaningful in-
sight to be obtained into the way language works. Then Chomsky
requests us to “notice that there is no question being raised here
as to the legitimacy of the probabilistic approach, just as the
legitimacy of the study of meaning was in no way brought into
question when we pointed out (...) that projection cannot be
defined in semantic terms. Whether or not the statistical study of
language can contribute to grammar, it surely can be justified on
quite independent grounds. These three approaches to language
(grammatical, semantic, statistical) are independently impor-
tant. ... Nevertheless, these three approaches are in some way
related. The object of investigation is ultimately the same, and
ultimately, we might expect them to fall into place in some larger
semiotic theory”.7> We would put this rather more strongly and
i4° ESLT, p..145.
% LSLT, p. 151, footnote 27.
 CHOMSKY’S ‘LINGUISTICS’ 85

say that not ultimately but at all times they must be considered
as part of a general theory. We use the term linguistics to cover
Chomsky’s grammatical, semantic (corresponding to our lexis to
some extent) and statistical (subsumed under grammar, lexis,
phonology etc. in our discipline) approaches as well as context
and graphology. All of these components are very definitely
interrelated and they can only be properly studied in conjunction
with one another. To isolate (non-statistical) grammar, for in-
stance, can give most misleading and often quite meaningless
results.
We have already noted that Chomsky’s approach is largely
exemplificatory. He appears to have made little use of textual
analysis, whether written or spoken, but has mainly relied upon
what, to him, would seem most reasonable. Occasionally he will
refer to observations he has made as when, in discussing ‘he has
been being taken’, ‘he will be being taken’ and ‘he will have been
being taken’ he considers whether they should be called ‘gram-
matical’ or not. He mentions that “they certainly have a clear
meaning, which can be expressed in no other way, and I have
noted several instances of these forms in normal conversational
speech. If they are excluded [from being called grammatical] on
some grounds, then no doubt such forms as ‘he has been being
very cooperative today’ will be excluded on the same grounds”’.76
If Chomsky had carried his observations a little further he would
have found that combinations of up to three tense-elements occur
regularly in written English, whilst up to five-element combina-
tions can occur in spoken English.77 In fact Chomsky’s work does
not appear to take account of the very great difference, in formal
patterning, between written and spoken languages. His inclusion
of a morphophonemic component in his grammar implies a con-

7% LSLT, p. 279-280.
77 These statements are due to M. A. K. Halliday and are based on
observation of British English; for example a reply to the statement “He'll
have been working full strength for quite some time soon” might be “He’s
already been going to have been working full strength for some time soon
for the last twenty-five years”, exhibiting a five-element combination
(example due to Halliday).
86 LINGUISTICS AND LOGIC

sideration of spoken language. Yet he quotes a rule for converting

‘him’ to ‘whom’ (rather than to ‘who’, viz: wh he = who; wh him
— whom) in the transformation ‘He saw him yesterday > Whom
did he see yesterday’ with the implication that “Whom did he see
yesterday’ occurs in spoken English.78 ‘Whom’ is, in actual prac-
tice, almost completely restricted to the written mode although it
may often (and vainly) be prescribed as ‘correct speech’. One
wonders if Chomsky would regard the actually occurring “Who
did he see yesterday’ as ungrammatical. The implication is that
he would.
The discussions above must not be allowed to obscure the main
difference between Chomsky’s work and that of almost every other
present-day linguist. Instead of attempting to construct a descrip-
tive theory to account for observed language patternings, within
a scientific context such as that described in Part A, he put for-
ward a generative system, more-or-less a system of mathematical
logic. As such it should be considered as a special type of logic |
and only incidentally as a linguistics. Chomsky’s methods might
be usefully employed within a descriptive discipline (for example,
that of section 2.1); by themselves they are of little scientific
worth. He is, in fact, concerned not so much with describing ob-
servation but rather with obtaining ‘results’, of a mathematical
nature,” within his logical linguistics. A ‘result’ which is logically
quite powerful may be useless in a descriptive sense. Chomsky
thus has a new way of looking at language —he is interested
in finding analogies within language to the patternings in such
synthetic-systemic disciplines as mathematics. We are interested in
linguistic, rather than logical, patterns in language and in an
analytic-systemic® description of them. Within our approach the

78 LSLT, p. 528-529.
79 So that it is of continual interest to Chomsky whether a feature is
‘interesting’ (LSLT, p. 15, 16 etc.); we should only consider whether it was
‘relevant’ to the description.
80 We must be careful to distinguish between at least three different
senses of ‘system’: 1) a system of mathematical logic (this is the sense used
in ‘synthetic-systemic’ here); 2) a system which is just a set of terms — the
implication being that one term is chosen at any one time out of this
 CHOMSKY’S ‘LINGUISTICS’ 87

criterion of ‘faithful description’ is of paramount importance and

no set procedure can, as yet, satisfy this criterion or adequately
account for the correlations between formal and contextual pat-
terns, and so on. So that we are forced to a methodological ap-
proach. Chomsky’s theory demands a procedural approach®! and
this leads to dogmatism and neglect of relevant features which
cannot be easily fitted into a simply procedural (generative)
treatment.
Chomsky appears peculiarly dogmatic about the ‘nature of lin-
guistic theory’. It is clear that he foresees no other formal lin-
guistics than his own. This is shown in particular by his comments
upon the ‘hocus pocus’/‘God’s truth’ distinction: “In this connec-
tion we can return to the question of whether the linguist ‘plays
mathematical games’ or ‘describes reality’. ... The danger in the
‘God’s truth’ approach (FOOTNOTE: As the alternative to ‘playing
mathematical games’ has been styled by Householder) is that it
sometimes verges on mysticism ... .”82 Here Chomsky identifies
‘hocus pocus’ with ‘playing mathematical games’. Householder,
himself a God’s truth man, put it this way: “The hocus-pocus
linguist believes (.. .) that a language (. . .) is a mass of incoher-
ent, formless data, and the job of the linguist is somehow to ar-
range and organize this mass, imposing on it some sort of struc-
ture (. . .).” 88 Bazell has commented on this approach, saying that
linguists “have also come to see that this structure is not some-
thing given in the language, but that it has to be ‘analysed out’ of
the material by the linguist, and that if one linguist differs in his
analysis from the other it does not follow that either he or the
other linguist is wrong”.84 The alternative to the ‘God’s Truth’

system in preference to the other terms (as used in section 2.1 and in
‘analytic-systemic’ here); 3) system in a completely everyday sense, with a
vague implication of ‘rational’, ‘planned’ or just ‘reasonable’.
81 He also falls into the error of putting ‘prediction’ above ‘description’
(SS, p. 49).
82) CSTD p59
88 JJAL, 18, 260.
84 CC. E. Bazell, Linguistic Typology (Inaugural Lecture, University of
London, 1958), p. 1.
88 LINGUISTICS AND LOGIC

attitude is a subjective approach of which Chomsky’s ‘playing

mathematical games’ is but one of many instances. The majority
of hocus-pocus linguists employ a descriptive, not a generative,
approach and could rather be described as ‘playing scientific
games’.
If we wished to bring Saussure’s survey (with which this section
began) up to date we would need to include a new paragraph:
“Fourthly, something, again called ‘grammar’, was studied. This
study, initiated by Chomsky and others, was based on mathemati-
cal logic and, in part, constructed out of intuitive ideas derived
from a knowledge of the first, Greek initiated, logicale grammar:
ideas which were thus prescriptive and concerned with distinguish-
ing between grammatical and ungrammatical forms. It lacked a
scientific approach and was detached from language itself. It was
a mathematical and largely normative discipline, far removed from
actual observation, and its scope was both too limited in some
respects and too wide in others.”

3.3 LOGICAL LINGUISTICS?

As with every discipline we can have two distinct approaches

to logic: namely the ‘God’s truth’ one and the alternative, ‘hocus
pocus’ attitude. The extreme flank of the ‘God’s truth’ brigade
will work towards the ‘discovery’ of our ‘laws of thought’, these
being looked upon as a basic, universal and immutable charac-
teristic of our ‘minds’. Whereas a hocus pocus adherent will treat
logic as an empirical science. We obviously cannot make any
direct observation of our mental ‘reasoning’ and so observations
are made of the relevant manifestations of this reasoning, be they
manifested through language, or in some other way. The only
evidence we have for the existence of such ‘reasoning’ is these
manifestations — to be quite scientific we must regard them not as
manifestations at all but as our sole and basic raw material. The
hypothesis that some mental working lies behind this raw mate-
rial, reasonable as it may seem, is quite notional and cannot thus
 LOGICAL LINGUISTICS? 89

be considered during the course of our scientific investigation.

The study of logic, excluding for the moment mathematical
logic, is still, to a quite large extent, set in the pattern originally
created by Aristotle. It would be difficult to find anyone who
would maintain that Aristotle’s logic is completely exhaustive
— that it presents an explicit theory which will account for all the
different types of phenomena that can be recognised as ‘logical;
reasoning’; but it does serve to throw considerable light on a cer-
tain and particular type of reasoning. The general shapes of syl-
logisms make any everyday ‘application’ of them into a definite
kind of ‘reasoning procedure’. Aristotelian logic (as it has been
enlarged and expanded since his time) is thus more suitable for
application to certain academic disciplines than to others. In
particular it is ideally suited for use, in a reworked but quite
analogous form, in mathematics. Frege and others can be said to
have initiated a logicale study of mathematics when they general-
ised the single inference of an Aristotelian syllogism to ‘rules of
inference’ and were able to restate many mathematical proofs in
the form of a procedure, starting from certain axioms (postulated
as ‘true’ — as in a conditional syllogism) and proceeding, by means
of the rules of inference, to ‘theorems’. This application of logic
is restricted and particular in scope: although it was constructed
out of the logical discipline its form is, in a certain sense, more
comprehensive, and its subject-matter rather different so that we
have, in section 3.1, used ‘logicals’ to describe an application of
this new ‘mathematical logic’.
There may, of course, be historical reasons for the neat appli-
cability of logicale methods to mathematics: in the same general
vein we shall, in the next section, mention the trivial applicability
of logicals-influenced grammar to logicals syllogisms (in the work
of Latham). But an explanation of the suitability of mathematics
to be ‘logicalized’ need not concern us here. Suffice to say that
there was this suitability and that it is not universal but that the
extent of its existence for any particular discipline is very much
a feature of the general nature of that subject. Mathematics, con-
sisting of largely synthetic ‘built-up’ proofs, shows just that par-
90 LINGUISTICS AND LOGIC

ticular type of systematic®> nature which can be thus ‘procedural-

ized’; whether the whole of mathematics can be faithfully stated
as a series of logical; calculi is, it must be noted, still a subject
of some debate.
It will easily be seen that a procedural approach, either of a
logicals or of a logicals type, is not generally suitable for the state-
ment of a theory in a descriptive science. A purely descriptive
theory will normally provide a framework within which isolates
from observation can be categorically described, use being made
of the observer’s scientific intuition in assigning this description.
In this case the theory will be stated as economically as possible
(compatible with its function of providing a full description): its
components will, of course, be interrelated but they will not gen-
erally be definable®’* one in terms of another. Thus a descriptive
theory will normally be a sum of parts, each part depending on
the others to a limited extent and having its existence relative to
the other parts and the whole; each part is, however, in a sense
autonomous within the theory and completely necessary to it. If
a theory is restated as a logical; (axiom) system then it is said to
be ‘axiomatized’ or, perhaps, ‘formalized’. Carnap describes this:
“By an axiom system we understand the representation of a theory
in such a way that certain sentences of the theory (the axioms)
are placed at the beginning and from them further sentences (the
theorems) are derived by means of logical deduction”.8? So that
the statement of a descriptive theory is plainly not, in general,
‘axiomatizable’ and, further, its application to description will
only be (potentially) axiomatizable if it is a heuristic procedure.
This state of affairs is quite in keeping with the essential differ-
ence between mathematics (a synthetic discipline) and the em-
pirical sciences (analytic disciplines). What may be more amenable
to axiomatic treatment are the predictive parts of those sciences
85 ‘Systematic’ referring to ‘system,’ of footnote 80, p. 86.
86 Definable in a logical sense. Even it they were, the linguistic validity
of making such ‘definitions’ (employing all the resources of ‘logical mean-
ing’ etc.) is dubious.
87 R. Carnap, Introduction to Symbolic Logic and Its Applications
(New York, 1958), p. 171.
 LOGICAL LINGUISTICS? 91

which fall towards the far end of the descriptive/predictive scale.

But linguistics itself is almost wholly descriptive and so any con-
sideration of this point would be outside our present scope.
To stress it again, then, the vital point here is that only set
procedures are capable of axiomatization — general methods can-
not be so restated.
Although he went into no details Firth saw very clearly the
inapplicability of essentially mathematical methods of inference
to an empirical science: “In my opinion, any fully axiomatized
mathematical linguistics will not be found workable in a truly
empirical science....”88 It is significant that very soon after
Firth’s remarks Chomsky should have published an account of a
‘fully axiomatized’ linguistics which, as we have seen in section
3.2, is indeed ‘not workable’ within the context of linguistics as a
‘truly empirical science’. Firth’s paper “Structural Linguistics”
dis-
cusses the ‘axiomatizations’ of Bloomfield, Bloch and others and
the unnecessary and linguistically invalid ‘logical definitions’ of
terms which had to be included in their statements. The straight-
jacket of ‘propositions’ is another handicap in any axiomatization
of a scientific theory. A linguist cannot validly use the notion of
‘formal logical meaning’ — whereby statements can be rewritten
into a set pattern and keep their (logical) meaning — in his work.*®
Without this idea of identity of meaning we cannot employ the
standard propositional terminology. Moreover we have already
noted that statements in an empirical science will always eventu-
ally be of a ‘more/less’ rather than of a definitive ‘either/or’
nature: we will thus require to talk not about some actual ‘event’
but rather of a certain probability of the occurrence of the event.
To fit such statements into a propositional framework would in-
volve propositions of ‘one type higher’ and, as well as being quite
unnecessary, would be very unwieldy.
Quite apart from the general question of the use of logical

88 “Structural Linguistics”, TPS, 1955, p. 99.

89 A more detailed discussion of this point is given in section 3.4.
90 The point concerning probabilities must again be taken into account
if logical, methods are used in the predictive part of a science.
92 LINGUISTICS AND LOGIC

procedures in linguistics it is in order to comment upon some

more particular ‘borrowing’ from the theories of mathematics and
logic. It is recognised that any multi-valued relation can be stated
as a set of binary relations: this is, of course, a major factor in the
preparation of data for analysis by digital computers. In a syn-
thetic discipline, in which simplicity is one over-riding considera-
tion, this principle can often be used to great advantage. But in
observational sciences we have a ‘paramount principle’ which is
of greater general moment than considerations of simplicity: that
the theory should provide a full, adequate and faithful description.
Consider a general scientific example® in which an item, W, has
a structure which clearly consists of three elements, X, Y and Z.
X and Z cannot be allocated any structure (excluding the trivial
case of one-element structures) but more detailed examination
reveals that Y can be represented as consisting of two elements
Y;, and Ye. The structure XYZ is the important and obvious one;
XY,Y2Z is apparent as such only at a far later stage of analysis
and reflects less significant natural patternings than XYZ. So that
the whole situation is summed up most neatly when we say that
W has a three element structure XYZ and that Y has a two ele-
ment structure Y;Y2. No doubt the simplest,* in a certain sense,
approach would be to say that W has structure ST and that S and
T have structures XY; and YeZ respectively: but this would lead
to a quite useless description since it would completely obscure
the essential difference in status between X, Z and Y;, Ye. Such
uncalled for binarizations are particularly to be guarded against:
specific linguistic examples (which are not necessarily similar to
the above) are the universal binary cuts of traditional IC analysis

91 This example is quite arbitrary and is not intended to apply to any

particular scientific instance, linguistic or otherwise. ‘Structure’ is not used
in either a grammatical or a phonological sense.
% ‘Simplicity’ is another term which is used by different people in
widely different senses. Some people would insist that a ternary division
of, for example, P into Q, R and S was ‘simpler’ than an initial binary
division into Q and RS, and a further binary division of RS into R and S;
others would maintain that the second is ‘simpler’ than the first.
 LOGICAL LINGUISTICS? 93

and the ‘binary functions’ of the glossematicians.® It is of interest

here to notice Lees’ ternary- or larger-cuts in his revised and im-
proved version of the IC-component of a tranformational gram-
mar. In any scientific investigation we must begin with an open
mind concerning the size and extent of patterns which we except
to find.
Within our European civilisation it is important to recall that,
as Saussure has already reminded us, the first step towards a
“linguistics” was an application of Aristotelian logic to language
analysis. That particular restricted language which is used in doing
logic is indeed one part of the raw material of linguistics: it is
important to recall that it is just one of many such parts. In the
work of the Greek grammarians a wholesale transference of logi-
cal categories into linguistics was effected — to say the least, this
produces an ineffective and inadequate descriptive framework.
Much of our present-day prescriptive thinking may be traceable
to the ancient logicale grammar. It is only within the last few
years that a really non-logicale, descriptive theory of linguistics
has been put forward. It has been recognised that much more
comprehensive categories and methods are needed than could be
supplied by traditional logic. In addition to this the whole ap-
proach is different from that of a logical, investigator — not least
through the emphasis on descriptive rather than normative judge-
ments. Whether this is due to the different intrinsic natures of
logic and linguistics or to some still-persisting failing of logicians
will be examined in section 3.4.

93 “Tt can be foreseen for all three kinds of functions that they may be
contracted by more than two functives; but these multilateral functions
can be considered as functions between bilateral functions” (Prolegomena,
p. 22). But it must be admitted that in the present dearth of examples of
the application of glossematic methods to specific texts the full significance
of this statement is unclear.
94 See WAT. See also Bazell’s comment (Linguistic Form, Istanbul,
1953, p. 5): “The idea of an overwhelmingly predominating binary struc-
ture of immediate constituents is again largely illusory. ...The success of
this method is a tribute, not to the importance of the binary character of
the syntagm, but simply to the acharnement with which the linguists seek
to split up utterance-segments into two.”
94 LINGUISTICS AND LOGIC

The logical positivist approach to language was quite ‘formal’

and extremely narrow in perspective. Carnap’s Logical Syntax of
Language begins by considering two ‘particular’ languages — both
of them logical; systems. The surprising thing is that Carnap
regards his ‘languages’ as special cases of ordinary language, and
implies that his results could be generalised to apply to more
comprehensive languages.®* Whereas, in fact, his languages are
structurally such simplified versions of a spoken language that no
such generalisations are possible.°* A secondary implication, and
one which has cropped up in many guises during the last two or
three thousand years, is the normative opinion that only ‘logical
languages’ are ‘exact’ enough to be capable of expressing any
definite and worthwhile results. It is commonly stated that ‘Eng-

% Carnap’s conception of ‘language’ is, not unnaturally quite similar

to Chomsky’s: “By a language we mean here in general any sort of cal-
culus, that is to say, a system of formation and transformation rules con-
cerning what we called expressions i.e., finite, ordered, series of elements
of any kind, namely what are called symbols. In pure syntax, only syn-
tactical properties of expressions, in other words, those that are dependent
only upon the kind and order of the symbols of the expression, are dealt
with” (Logical Syntax of Language, London, 1937, p. 167-168).
96 Carnap explains that “although here, for the reasons indicated above,
we shall be dealing with symbolic languages, the syntactical concepts and
rules — not in detail but in their general character —may also be applied to
the analysis of the incredibly complicated word-languages” (Logical Syntax
of Language, p. 8; my italics). That this is not so is demonstrated by the
failure of Chomsky’s valiant attempt at the construction of a ‘logicals lin-
guistics’. Considered from a logical point of view (which is meant to deal
with synthetic, constructed ‘languages’) ‘word-languages’ are indeed ‘in-
credibly complicated’ — and more— for their patterns are not simply logical
ones. Ordinary language items will have ranges of potential meanings,
unlike logical predicates, and so on. The whole point here is that the study
of these ‘incredibly complicated’ languages is a subject for scientific, not
mathematical, study. We can formulate definitive, complete and final
procedures for describing synthetic ‘symbolic’ languages. But we can only
give methodological, analytic descriptions of natural languages — descrip-
tions which we can carry on improving but which may never be absolutely
complete or final. Carnap’s ‘logical syntax’ can be stated as a particular
case of our ‘categories of the theory of grammar’ —the reverse is not the
case.
 LOGICAL LINGUISTICS? 95

lish is not adequate as a meta-language’ (or similar),°7 but it must

be realised that any ‘logical language’ is but a particular and
restricted part of a given natural language and that the ‘faults’
which may have been recognised in the larger language are most
often still present in the isolate. Ambiguity is largely a function
of the reader or listener relative to the writer or speaker rather
than of the language as such: restricted languages are certainly to
be preferred for certain purposes but not because of the reasons
usually given.*® Opinions such as those mentioned above are rem-
iniscent of the situation some three hundred years ago when Latin
was held in such esteem that its particular grammatical categories
were thought to be suitable for description of any ‘lesser lan-
guages’ such as English, and it was maintained that only Latin
was capable of being a vehicle for the expression of valued
academic opinions.
We are thus drawn to the conclusion that, as the various dif-
ferent types of logic stand at present, we can have no logical lin-
guistics, except in the trivial sense of logical,. The aim of scien-
tific logicians should surely be to construct a new ‘logical theory’
which would account for every observed pattern commonly refer-
red to as logical. If this ideal were ever realised we would be able
to state every theory in every science in terms of this ‘logic’. But
this is a considerably misleading simplification, as will appear in
section 3.4.
Mathematical logic has given us, in addition to its general
method, a considerable array of symbolism. By means of this
equipment one is able to make neat, abbreviatory symbolizations
of ordinary language statements if these are written in a consist-
ent, explicit and simple manner. The great advantage of such
symbolization is that it enables a trained investigator to get a
theoretical framework into better perspective and it will readily
97 “Since the discussion will take place in English, English is our meta-
language. We have already noted that this is an unsuitable language from
the logical point of view, but it is used for the sake of ease of presenta-
tion” (Encyclopaedia Britannica, 1959, Volume 20, p. 313A).
98 A more detailed discussion on these lines, following Firth’s general
theories of meaning, would be outside the scnpe of the present discussion.
96 LINGUISTICS AND LOGIC

expose any ‘gaps’ or ‘inconsistencies’ in the theory. Such an ap-

plication of logicals techniques (but not of mathematical logic
itself) appears to supply as much as linguistics can at present
safely accept from contemporary logics. My “Logical Statement
of Grammatical Theory’ is just such an application: here the
categories are stated as finite sets and their interrelations in the
form of 91 formulae. A new method of writing formulae was
devised to fit the needs of such a statement: but the techniques
used in modern symbolic logic were heavily drawn upon. Such a
symbolic treatment has proved its worth by allowing a complete,
consistent and explicit statement to be made of the very complex
interrelations between the components of the theory (whilst com-
pletely preserving the descriptive power of the theory); the simple
symbolic structuring of the formulae renders transparent the im-
plicit similarities between the relationships. Such a statement is
nothing more than a descriptive device — it is far from making the
theory into a logical; system: we have seen that this is impossible
in the case of an empirical, methodological theory. The formulae
could not be called axioms or else there would be an implication
of ‘logical deduction’ and ‘theorems’ 1° — we require no ‘theorems’
from such a theory but rather ‘grammatical descriptions’ of for-
mal items. It remains to explain the title: ‘Logical Statement’ can-
not be reconciled with any of the three senses of ‘logical’ discussed
in secion 3.1. This ‘logical,’ refers to a utilisation of the tech-
niques, but not the systematic framework, of mathematical logic.
It is a scientific-symbolic statement, not a mathematical one.

3.4 TOWARDS A ‘DESCRIPTIVE LOGIC’

In the last section we gave some discussion of the applicability of

various types of contemporary ‘logical’ methods to linguistics. One
aim of general logicians must be to construct a fully descriptive

9 This being the theory at the demi-level of grammar which was out-
lined in section 2.1. To be published in Language 39 (September 1963).
100 Compare with the quotation from Carnap on page 90 above.
 TOWARDS A ‘DESCRIPTIVE LOGIC’ 97

theory which will account for our everyday (non-mathematical)

logical, patterns. Aristotle’s logic is very far from being adequate
for this purpose — and opinions vary as to whether it was ever
intended to satisfy such an aim. So that, with respect to this
ideal, logic today must be considered very far from definitive or
final. The ‘aim’ we have mentioned is in fact a scientific one: a
logic fulfilling it might be such that it could be called a science
within the terms of our general scheme. Can such a ‘logic’ be
developed? And, if so, would it be a full science in its own
right? 11
It is now in order to enquire into the relationship between this
hypothetical ‘ideal logic’ and our science of linguistics. Both ap-
pear to be very vital disciplines with definite interrelations: we
use language in ‘doing’ both logic and linguistics and we also
‘reason’ in our pursuit of both subjects. We have already remark-
ed upon the scientific invalidity of considering logical raw mate-
rial as ‘manifestations’ of a mental process. Whereas previously
it might have been said that our ‘laws of thought’ could be mani-
fested either in language or else in some non-linguistic observable
action, we would now prefer to say that we can recognise ‘pat-
ternings of reasoning’ in many different types of situation. This is
an important consideration — for logical; can be used to describe
many types of action and we must take account of all of them:
our ideal logic is not to be restricted to patterns observed in lan-
guage material. The observer may recognise a set of actions as an
instance of reasoning when no language whatever is in the im-
mediate situation: for example, if a man tries to open a door by
pulling it towards him and finds that he cannot move it, and then
tries to push it instead of attempting to pull it again, we might
well say that he employed his power of ‘reasoning’ and that this
incident should be amongst the raw material of our scientific
logic. Whether his action would be described by psychologists as
‘thought out’ or as ‘intuitive’ is of no concern to us; nor is the

101 See page 50 above: “sciences are recognisable as distinct, in the

first instance, by virtue of these different defining patterns and overlying
approaches”.
 98 LINGUISTICS AND LOGIC

remark that his action was based on prediction by pattern cor-

relation out of past experience our immediate concern, although
such considerations will be likely to interest us at a later stage in
the construction of a ‘theory of logic’.
And in cases when a recognised instance of ‘reason’ does in-
volve language material, this material is, of course, only to be
considered in relation to the relevant contextual patterns in its
situation. The logical patterns will be partly in the language isolate
and partly in the situation: but here we err in referring to a
‘language isolate’. In linguistics language is our object of study
and the defining pattern was the presence of language material
in observation. Our raw material isolate was, effectively, language.
In the case of logic the defining patterns of ‘reason’ will not be
completely contained in language and so here our raw material
isolate must include all the relevant parts of the width and length
of the part of observation that we are considering. Logical raw
material may contain some language substance but will never be
entirely composed of language. This contrasts with previous con-
siderations of ‘logic’ which have invariably looked upon logical
patternings as being entirely expressed in language and, as such,
completely autonomous.
We could classify pieces of logical raw material according to
their amount of language content. No definite binary division
would be possible, since, as we stressed in section 1.1, a so-called
isolate is not a definitely delimited part of observation. But we
can arrange pieces of logical raw material along a scale according
to, for example, whether they have a large language element, a
smaller but still appreciable language element, a minor language
element or effectively no language element at all. We will restrict
our discussion, for the time being, to those which contain a fairly
important language component. Later we will relate these par-
ticular considerations to a complete ‘ideal logic’. Insofar as Aris-
totle’s logic aimed to describe logical; patterns his results will be
relatable to ours: for he only considered logic in language, and
that independent of non-verbal situation.
Initially, our ideal logic is to be a descriptive science. We re-
 TOWARDS A ‘DESCRIPTIVE LOGIC’ 99

quire to discover how and where logical patterns do occur, not to

formalize our ideas concerning where we think they ought to
occur. It may, of course, be possible to formulate a predictive
component after our description is well advanced. Linguistics,
itself a descriptive science as we have discussed it above, is natu-
rally concerned with the patterns which a logical science may
recognise in language substance. Logic will recognise patterns
extending from language to non-language parts of observation and
so the relevant component of linguistic science is the interlevel of
context. This interlevel exists to correlate the formal features of
language with the behaviour of language in relation to its func-
tions in our everyday life. One of these functions is clearly to be
the vehicle for ‘reasoning’. Thus a consideration of those pieces
of logical raw material which involve an appreciable amount of
language is, in any case, a concern of our interlevel of context.
Where the language element is significant a linguistic approach
is clearly the appropriate one — we give central importance to the
most important component, the language part, and then consider
its interrelations with the rest of the logical isolate.
And so we have reached the remarkable conclusion that a for-
mal examination of a particular type of logical raw material is
an implicit function of the context interlevel of linguistics. Insofar
as this particular type of raw material is concerned (and it is the
type considered most important by logicians in the past) logic is
seen to be a necessary part of a unified theory of linguistics.
What now of logic as a science in its own right? Some logical
patterns can be completely accounted for by a descriptive lin-
guistics. Other types of logical raw material will be described by
other sciences, actual or potential, which deal with other aspects
of human behaviour. According as the central component of an
isolate is of one sort or another that isolate will be describable in
terms of the science whose function it is to describe the central
component. Each such description will be made with full reference
to the other features of the logical isolate — all such features being
part of the situation of the science in question. Finally our ‘ideal
logic’ will have to gather together its distinct parts, each of them
100 LINGUISTICS AND LOGIC

obtained from a relevant science dealing with some aspect of

human behaviour: they will not be ‘discrete parts’ but rather over-
lapping treatments of different sections of the complete logical
corpus. Just as grammar and lexis serve to complement each other
but do not have mutually exclusive sets of patterns to describe,
so a logical isolate which contained a reasonable, but not large,
language component, would be fully describable only in terms of
linguistic theory and in terms of the sciences which deal with its
other substantial components. The distinct treatments would not
be mutually inconsistent if they belonged to parts of the same gen-
eral scheme and dealt with different aspects of the same general
human behaviour. They would rather emphasise different aspects
of logical patterning and a complete theory of logic would need
to combine their separate statements. A theory of logic which
limited itself to isolates which have a preponderance of a particu-
lar scientific component could be completely contained in that
particular science.
Logic thus appears to be not so much a complete science in
itself but rather a kind of scientific super-category which cuts
across each of the sciences which deal with relevant aspects of
human behaviour. Its patterns are always contained within
some more basic scientific patterns; but if it is to be a true model
for all logical, things it will transcend the boundaries of any one
particular science. We mentioned that the other sciences were
‘actual’ or ‘potential’: sufficient reason for our failure to indicate
their number or even to describe some of their particular natures.
Here our only concern is with a linguistic consideration of logic;
it may well be that linguistics is the only one of these sciences
which is at a ripe enough stage of development to devote itself
to a consideration of its contained logic. And it is very necessary
that this should be done: we have already mentioned that this is
an important function of the interlevel of context. Contextual units

102 There is here a case for considering a particular group of sciences,

concerned with human behaviour, as being especially close to each other
merely by virtue of their each containing some ‘logical’ patterns. Lin-
guistics is one of the group.
 TOWARDS A ‘DESCRIPTIVE LOGIC’ 101
will have to be set up, hypotheses tested and a theory evolved.
Needless to say a contextual unit will not generally be coextensive
with any formal unit. But there will be considerable interrelations
between the patterns of this logical theory at the interlevel of
context and the patterns of the theories of grammar and lexis at
the level of form‘ — this will lead to a ‘feedback’ to grammar

103 Work on a Descriptive Logic of British English is at a very early

stage, and we can only give indications of the sorts of results that may be
expected, rather than definitive examples. The following English conversa-
tional fragment (transcribed from tape) was part of a discourse between
two University lecturers (J and M). The relevant immediate situational
features are that M has been unable to find any tea with which to make
himself a cup of tea and has had to make a cup of coffee instead, which
he is at present drinking. M has just said: “As a cup of tea this is a fine
cup of coffee”.
1(M): //4 and there was//1 no/milk//1 either//
2): //13 looks quite/white from/here//
3(M): //1 well... I/rinsed it/out in the/bottom of a/milk bottle//
4): //1so there/was milk//
5(M): //1 A there/was milk//
6(M): //1 A there/was milk/in a//4 sense//
7(M): //1 (A there was a/whiff of milk//_
(The method of showing intonation used here is as described in Halliday’s
“The Tones of English” (to be published in ArchL, 15, i); underlining
denotes the tonic syllable i.e. the first syllable of the first foot of the tonic).
We can say that utterance J consists of a logical connective and...
either and a statement there was no milk. 2 consists of an observational
statement (of a feature of the immediate situation: the coffee in the cup).
3 is a connective well (with some sort of explanatory function, perhaps
predictable from J’s intonation of 2) followed by a statement IJ rinsed it
out in the bottom of a milk bottle. 4 consists of a different type of con-
nective (inferential), so, and the statement there was milk. 5 repeats this
statement, 6 repeats it with the addition of an adjunct in a sense, and 7
consists of the statement there was a whiff of milk. Within the wider
situation (British Culture) the lexical items milk and white will have situa-
tional correlation (one of the several factors demonstrating this will be a
fairly high probability of their mutual collocation). Thus the string of
seven utterances is delimited in one way by the occurrence of these items.
In this example there is a large amount of correspondence between gram-
matical and contextual patterns. In terms of the Grammar of Spoken Eng-
lish of Halliday and Sinclair, there was no milk is a clause of structure
‘SPC’ where the exponent of ‘S’ is there, of ‘P’ is was, and of ‘C’ is a
nominal group of structure ‘mh’; the exponent of ‘m’ (the modifier) is no
102 LINGUISTICS AND LOGIC

and lexis and may well result in amplification and clearer state-
ment of some formal categories. Until each part of a general lin-
guistic science has been properly formulated no other part can be
looked upon as really definitive — the parts are mutually depend-
ent and can only have existence relative to an explicit whole.
We can call an ‘ideal logic’ such as is indicated above: ‘a de-
scriptive logic’. This name will serve to separate it from earlier
unscientific, and often notional or prescriptive, studies. Relative to
linguistics we will study just one aspect of the projected ‘descrip-
tive logic’.1%4 But since this is the first aspect which appears to be
and of ‘h’ (the head) is milk. there was milk (4, 5) has the same structure
except that there is now no modifier (previously with exponent no) in the
nominal group acting at ‘C’ in clause structure. In addition (and less sig-
nificant, here) the was in I is an exponent of the term non-contrastive,
whereas was in 4 is an exponent of the term contrastive in the verbal
system contrastive/non-contrastive; the tonic in J is on milk and that in 4
is on was. 6 has clause structure ‘SPCA’ where in a sense is the exponent
of ‘A’. 7 has structure ‘SPC’ and here milk, which was exponent of the
head of the nominal group acting at ‘C’ in J, 4, 5 and 6, is exponent of
the head of a nominal group which is rank-shifted to act at ‘q’ in the
structure, ‘mhq’, of the nominal group which acts at ‘C’ (the exponent
of ‘h’ is whiff and of ‘m’ is a). The original statement, J, is indirectly
commented upon by J in 2 and clarified by M in 3. J then deduces 4,
a revised version of 1, which, at that degree of delicacy of focus, is agreed
upon by M in 5. After qualifying 5 by means of 6 (talking about 5 rather
than restating it) M restates 5 with rather more delicate focus as 7. (For
a description of the scale of ‘delicacy of focus’ see J. O. Ellis, “On Con-
textual Meaning”, to be published in the volume of essays in commemora-
tion of J. R. Firth.) Only extended textual examination will show whether
all the implications which have been made above can be given a scientific,
distributional basis.
104 Just as in linguistics we have “language turned back on itself”
(Firth, Papers, p. 190) so here we will have “logic turned back on itself”.
If we are examining logic in use of the English language we will be most
likely to be writing out theories, hypotheses and observations in the Eng-
lish language. This is our normal everyday language. The concept of a
‘meta-language’ distinct from an ‘object language’ is of abstract interest
only —it is a rather hypothetical means of overcoming a ‘paradox’ arising
from too definitive ideas regarding ‘truth’ and ‘universality’: the dichotomy
is inapplicable to practical scientific work. This whole question is a deep
one and of considerable interest to logicians, linguists and general scien-
tists. It is regretted that further discussion would be outside our present
scope. Our descriptive logic will contain statements of a ‘more/less’ type,
in preference to the ‘either/or’ statements insisted upon by traditional logic.
 TOWARDS A ‘DESCRIPTIVE LOGIC’ 103

receiving serious consideration and, moreover, is quite likely to

be the most vital component of a future complete logic, ‘descrip-
tive logic’ will for the time being be used as a name for the purely
linguistic component. This should not be allowed to obscure the
fact that parts of logic need not be at all linguistic and that our
study will only be a fraction of the needed whole. In the remainder
of this section we will mention some of the errors of previous
logical investigators in order that we can guard against making
similar ones ourselves, in the future.
Eventually we shall want to set up a general theory of descrip-
tive logic at the interlevel of context, within the general frame-
work of which particular theories for specific languages can be
constructed. Before we can construct a general theory we shall
need to examine a number of different languages and perhaps set
up a particular descriptive logic for each one. Then a general
theory can be obtained from a representative sample of particular
ones. In particular, we must guard against generalising on the
basis of just one language, There may well turn out to be a larger
number of common features in particular descriptive logics than
there are in particular grammars — so that our general theory of
logic will be all the stronger. But only a scientific investigation
can show whether this is actually the case or not. Any particular
language part of a logical isolate is necessarily governed by the
grammar and lexis of the language in question. To say that there
are ‘universal logical categories’ in our ‘minds’ which can obtain
expression in any language whatsoever is on a parallel to the
mediaeval view that “there is one grammatical system fixed and
valid for all languages, which the philosopher alone is able to dis-
cover and justify”: so that “the observed differences between lan-
guages must be treated, apparently, as a matter of vocabulary not
of structure”.1% A piece of logic only has existence in language
(or some other observable phenomenon) and can only be studied

And we should not limit our consideration to examples of reasoning which

are of synthetic, procedural kinds.
105 R. H. Robins, Ancient and Mediaeval Grammatical Theory in
Europe (London, 1951), p. 79.
104 LINGUISTICS AND LOGIC

relative to the language. What degree of universality there is will

be shown by scientific investigation; complete universality can
most certainly not be accepted as a basic assumption.'°6 Aristote-
lian syllogisms were obtained from the Greek language and could
only be said to have theoretic relevance to that language and its
culture. It is, to a certain extent, an accident, and also a reflection
on the family relationship of the languages concerned, that these
syllogisms could be readily translated into such languages as Latin,
German, English and French.1°7 We can not generalise on this and
say that they could be easily expressed in every language.1°8
Aristotle considered his logical propositions and syllogisms to
be autonomous and independent of context and he did not allow
for any instances of reasoning in which language played an im-
portant but not total part. In addition to this, traditional logic
made intuitive use of what we can call ‘logical meaning’ — this has
been extended and applied to a wide variety of different tasks; it
has even, unfortunately, been used in doing linguistics! As an
example, it might be said that “I'll buy it because it’s cheap” has
the same (logical) meaning as “Since it’s cheap, I’ll buy it”. This
idea is also employed in the use of ‘propositions’ for representing
‘reasoning’. The standard types of propositions by no means cover

1066 The small amount of actual universality of logical categories is

commented upon by Whorf (See Language, Thought and Reality, Boston,
New York, and London, 1956).
107 Here the distinction between God’s truth and hocus pocus logicians
can be nicely brought out. A particular type of God’s truth man will begin
by saying that there are universal laws of thought and that therefore they
can be expressed in any language. He will so express them, if necessary
by some complicated means or other, and will disregard his scientific
colleagues’ remark that they have quite different effects within the differ-
ing contexts i.e. that he has brought about ‘distortion’ by his sledgehammer
techniques. Thus the God’s truth position appears to imply that we can
‘think’ without using language. This whole question is very speculative and
quite irrelevant to any scientific enquiry.
108 J. R. B. Love remarked in “An Outline of Worora Grammar”
(Oceania Monograph, No. 3, p. 119): “Alternative propositions do not
occur. The Worora man does not say, Shall I do this or that? He says,
Shall I do this? No. I shall do that. If asked in English, Will you have
this or that? he will invariably reply, Yes, desiring both.” Worora is a
language in the Northern Part of the Kimberley Division, West Australia.
 TOWARDS A ‘DESCRIPTIVE LOGIC’ 105

all the distinct sorts of sentences which actually occur. The logi-
cian’s view has been that any propositional-type statement can be
rewritten in one of the set propositional forms without altering its
(logical) meaning. Again, the failure to recognise non-verbal
context leads to the belief that certain statements are ellipsis-forms
of longer ones. The actual statement is supposed to ‘mean’ the
same (through consideration of context, one supposes, although
this is seldom explicitly admitted) as the longer proposition which
is submitted for it. This logicalg use is an established sense of
‘meaning’. But it can have no place in a scientific treatment of
logic. As we saw in section 1.4 only ‘contextual’ and ‘formal’
meaning, as explained there, are relevant to a scientific investiga-
tion. After we have constructed a theory of descriptive logic with-
in the interlevel of context (which will account for logical ‘formal
patterns’) we will be able to include a ‘formal logical meaning’:
then two logical items will be said to have the same ‘formal logical
meaning’ if they receive identical descriptions from the theory of
logic. But we cannot talk about a ‘logical meaning’ until after we
have our theory. Certainly to begin with some quite notional ‘logi-
cal meaning’ and delimit the raw material in advance by applica-
tion of it is most unscientific. In our linguistic study of logic we
will begin by considering “Ill buy it because it’s cheap” as quite
distinct from “Since it’s cheap Il buy it”. Within our terms of
reference as an empirical science we cannot do otherwise.

109 Even with such devices, doubt has been thrown on the adequately
representational character of traditional logic. Sweet believed “that not only
is the order of subject and predicate to a great extent conventional, but
that the very idea of the distinction between subject and predicate is purely
linguistic, and has no foundation in the mind itself. In the first place there
is no necessity for a subject at all: in such a sentence as ‘it rains’ there is
no subject whatever, the it and the terminal s being merely formal signs
of predication. ‘It rains: I will therefore take my umbrella’, is a perfectly
legitimate train of reasoning, but it would puzzle the cleverest logician to
reduce it to any of his figures [i.e. through use of ‘logical meaning’]. Again,
the mental proposition is not formed by thinking first of the subject, then
of the copula, and then of the predicate: it is formed by thinking of the
two simultaneously...” (Collected Papers, p. 20). Whilst we would not
necessarily agree with Sweet’s method of presentation, his remarks are
penetrating.
106 LINGUISTICS AND LOGIC

Traditional logic had no theory of descriptive linguistics which

it could draw upon. Aristotle’s logic was purely notional and, of
course, he used his logical categories in a description of language.
His logic, through gross employment of ‘logical meaning’, and so
on, was only a very restricted case of ordinary language and so his
grammar only threw light upon the type of language that he used
in doing logic. His pragmatic logic and its derivative grammar led
to extremely normative ideas concerning language. His logicale
grammar naturally gave rise to the idea that ‘correct’ language was
that which was used in doing logic. All this may have been a mere
extension of earlier prescriptive ideas: but ‘doing grammar by
logic’ certainly made explicit prescription regarding language and
may well have been the major force behind its inception in the
West. Latham reversed the process when he used linguistics to
introduce and describe logic: but his task was essentially trivial
since his logic was largely Aristotelian and his grammar was of
the very traditional type which had been derived from that logic
in the first instance.11° Latham, himself, as we have mentioned
earlier, spoke against normative ideas in linguistics but he still
employed the essentially normative grammar and logic"! of his
time. Only now, when we have gone some way towards a descrip-
tive linguistics, can a study of logic in language be properly con-
templated.

110 R. G. Latham, Logic in its Application to Language (1856). The

content of this work is in fact: Language applied to Logic, the exact op-
posite of the title. Latham devotes a good deal of his preface to explaining
this curious choice of title: “Now Grammar being taught early, and Logic
Cif taught at all) late, the effect of such a title as Language applied to
Logic, would be to engender the notion that the work wherein the applica-
tion was exhibited was for adults rather than boys and youths. Its real
object is the reverse of this. Its real object is to get certain so-called points
of Logic studied as early as the elements of Grammar” (p. iv).
111 At about the same time as Latham’s work, Lewis Carroll exploited
the vast differences between traditional logical) autonomy and actual usage
in Alice in Wonderland and Through the Looking Glass. Without current
logical, ideas he would have been unable to point fun in many of the ways
that he did. The continued popularity of Alice is an indication that these
ideas concerning ‘correct’ logicaly reasoning and usage are still with us
today.
 TOWARDS A ‘DESCRIPTIVE LOGIC’ 107

Logicians have, of course, seldom had in mind that their sub-

ject should be a science in our sense: they were essentially God’s
truth men. When we judge them relative to our standards we do
not dispute their philosophical worth-— scientifically, we can at-
tempt no judgement of usual, notional philosophy — but rather the
worth of their work for our own purposes. We must demonstrate
their unscientific nature in order to clear the way for a ‘descriptive
logic’. Traditionally, logicians have been interested in the ‘truth’
or ‘falsity’ of propositions: and they have talked in terms of the
universal or absolute ‘truth’ of certain propositions and syllogisms.
We would rather consider everything relative to its context and
only look upon ‘truth’ in an extremely subjective way, relative to
each particular speaker or writer. Scientifically, we cannot ever
give definitive ‘truth’ labels to things, and moreover, we recognise
the individual and personal ‘right-ness’ of each person’s point of
view.112 This will, in turn, lead us to a broader scientific evalua-
tion of logic within the scheme of linguistic levels. Our contextual
units will be chosen so that they can be used for other purposes
than an investigation of ‘reasoning’ — they will be quite general.
Aristotle dismissed prayers, for example, as being neither true nor
false and so not a subject for any philosophical consideration. Our
purpose and our point of view differ radically from his and our
work will lead to different results. This may appear obvious: it
must always be kept in mind. It is unfortunate that the everyday
use, logical;, has been applied so widely, inasmuch as it tends to
promote the general idea that logical, theory was ever meant to
be a proper reflection of our ‘reasoning’, rather than just a peg
upon which to hang a philosophy of ‘truth’.
We have indicated a few of the developments so far (largely in
Britain) which seem to us to be leading towards a truly descriptive
linguistics; some small attention has been given to the differing
natures of certain other approaches. Finally we can point the way
for future work to proceed. Only large scale observational studies
can yield particular lexes for specific languages. Here new tech-
112 By ‘personal rightness’ of certain views we mean rightness relative
to the person who holds the views.
108 LINGUISTICS AND LOGIC

niques may have to be either invented or else ‘borrowed’ — not

least probabilistic and statistical methods. Each such ‘borrowed
technique’ must be overhauled and used with discretion for the
job in hand. Finally, much work remains to be done on ‘context’.
A theory of ‘descriptive logic’ is one of the most urgent tasks
awaiting us and it could best be achieved through a collaboration
of many diverse types and nationalities of scientists.
 JANUA LINGUARUM

STUDIA MEMORIAE NICOLAI VAN WIJK DEDICATA

Edited by Cornelis H. van Schooneveld

SERIES MINOR

ROMAN JAKOBSON and MORRIS HALLE: Fundamentals of Language.

1956. 97 pp. Gld. 6.—
EMIL PETROVICI: Kann das Phonemsystem einer Sprache durch
fremden Einfluss umgestaltet werden? Zum slavischen Einfluss
auf das rumanische Lautsystem. 1957. 44 pp. Gld. 4.—
NOAM CHOMSkKy: Syntactic Structures. Third printing 1963.
118 pp. Gld. 8.—
N. VAN WIJK: Die baltischen und slavischen Akzent- und Intona-
tionssysteme. Ein Beitrag zur Erforschung der baltisch-slavi-
schen Verwandtschaftsverhdltnisse. 2nd ed. 1958. 160 pp.
Gld. 15.—
BERNHARD GEIGER, TIBOR HALASI-KUN, AERT H. KUIPERS and KARL
H. MENGES: Peoples and Languages of the Caucasus. A Synopsis.
1959. 77 pp., map. Gld. 8.—
ERNST PULGRAM: Introduction to the Spectrography of Speech.
1959. 174 pp., 31 figs., 2 tables. Gld. 12.—
AERT H. KUIPERS: Phoneme and Morpheme in Kabardian
(Eastern Adyghe). 1960. 124 pp. Gld. 16.—
10. URIEL and BEATRICE WEINREICH: Yiddish Language and Folk-
lore. A Selective Bibliography for Research. 1959.
66 pp.
Gld. 6.—
ie E. and K. DELAVENAY: Bibliography of Mechanical Translation.
— Bibliographie de la traduction automatique. 1960. 69 pp.
Gld. 10.—
12: CARL L. EBELING: Linguistic Units. Second printing 1962. 143
pp. Gld. 12.—
13. SHELOMO MoRAG: The Vocalization Systems of Arabic, Hebrew
and Aramaic. Their Phonetic and Phonemic Principles. 1962.
85 pp., two folding tables. Gld. 15.—
14. DWIGHT L. BOLINGER: Generality, Gradience, and the All-or-
None. 1961. 46 pp. Gld. 5.50
iD? ALPHONSE JUILLAND: Outline of a General Theory of Structural
Relations. 1958. 58 pp. Gld. 7.50
16. Sens et usages du terme Structure, dans les sciences humaines
et sociales. Edité par ROGER BASTIDE. 1962. 165 pp. Gld. 16.—
17. W. SIDNEY ALLEN: Sandhi. The Theoretical, Phonetic, and His-
toric Bases of Word-Junction in Sanskrit. 1962. 114 Pp. ;
Gld. 16.—
il) WALERIAN S$WIECZKOWSKI: Word Order Patterning in Middle
English. A quantitative study based on Piers Plowman and
Middle English Sermons. 1962. 114 pp. Gld. 16.—
20. FINNGEIR HIORTH: Zur formalen Charakterisierung des Satzes.
1962. 152 pp. Gld. 15.—
22. E. F. HADEN, M. S. HAN and y. W. HAN: A Resonance Theory for
Linguistics. 1962. 51 pp. Gld. 7.—
23. S. R. LEVIN: Linguistic Structures in Poetry. 1962.64pp. Gld. 8.—
24. ALPHONSE JUILLAND and JAMES MACRIS: The English Verb
System. 1962. 81 pp. Gld. 8.—
Dill. LASZLO ANTAL: Questions of Meaning. 1963. 95 pp. Gld. 10.—

SERIES MAIOR
DEAN S. WORTH: Kamchadal Texts collected by W. Jochelson.
1961. 284 pp. Cloth. Gld. 58.—
. PETER HARTMANN: Theorie der Grammatik.
2. Zur Konstitution einer allgemeinen Grammatik. 1961.
136 pp. Gld. 21.—
3. Allgemeinste Strukturgesetze in Sprache und Grammatik.
1961. 136 pp. Gld. 21.—
4. Grammatik und Grammatizitat. 1962. 144 pp. Gld. 24.—
GUSTAV HERDAN: Type-Token Mathematics. A Textbook of
Mathematical Linguistics. 1960. 448 pp., 17 figs. Cloth.
Gld. 54.—
TATIANA SLAMA-CAZACU: Langage et Contexte. Le probléme du
langage dans la conception de l’expression et de l’interprétation
par des organisations contextuelles. 1961. 251 pp., 5 figs. Cloth.
Gld. 48.—
ALF SOMMERFELT: Diachronic and Synchronic Aspects of Lan-
guage. Selected Articles. 1962. 421 pp., 23 figs. Cloth.
Gld. 54.—
THOMAS A. SEBEOK and VALDIS J. ZEPS: Concordance and The-
saurus of Cheremis Poetic Language. 1961. 259 pp. Cloth.
Gld. 58.—
GUSTAV HERDAN: The Calculus of Linguistic Observations. 1962.
271 pp., 6 figs. Cloth. Gld. 42.—
10. Proceedings of the Fourth International Congress of Phonetic
Sciences, held at the University of Helsinki, 4—9 September
1961. Edited by ANTTI SOVIJARVI and PENTTI AALTO. 1962. 855
pp., num. figs. and plates: Cloth. : Gld. 125.—