Oxford University Press, Amen House, London E.C.

4
GLASGOW. NEW YORK TORONTO MELBOURNE WELLINGTON
BOMBAY CALCUTTA MADRAS KARACHI LAHORE DACCA
CAPE TOWN SALISBURY NAIROBI IBADAN ACCRA
KUALA LUMPUR HONG KONG

BY

ANDRE

BEING

THE WA YNFLETE LECTURES

DELIVERED IN THE COLLEGE OF

ST. MARY MAGDALEN, OXFORD

19 6 1

OXFORD
THE CLARENDON PRESS
19 6 2

FONETISKA INsrrITUTION
 TO

AUSTIN GILL

© Oxford University Press I962

PRINTED IN GREAT BRITAIN

 P E CE

TLE is rarely quite as explicit as its author intends

A it to be, and a preface is no unnatural place for some

glossing.
The 'functional' label, which appeared in the title of my
1949 booklet Phonology as Functional Phonetics and which I am
using here again, is probably no less forbidding to large
sections of the philological public than the more usual 'struc-
tural' tag. I am quite aware of this, and if I persevere it is
because I am less inclined to flatter the preconceptions of
some of my readers than to stress what, in my opinion, is
conducive to sounder descriptive methods and a better under-
standing of language.
When reading, with full appreciation, the books and essays
of some contemporary descriptivists, I cannot help wondering
at times whether, in their striving for scientific detachment
and their laudable efforts to do away with vague idealistic
phraseology, they are not missing what Sapir, hampered
by the pervading psychologism of his age, gropingly
referred to as 'a basic plan, a certain cut to each lan-
guage'. Fighting 'mentalism' should not consist in denying
the existence of well-established facts, but in showing what
palpable realities stand behind loose prescientific phrasing:
it is fair play, for contemporary structuralists, to ridicule a
phrase like 'the spirit of a language' because it is more likely
to evoke some winged supernatural being than a set of
internal relations; but it is bad policy to ignore or neglect
the fact that any utterance or any segment of an utterance
becomes a linguistic object only inasmuch as it has been
identified as belonging to a given language. Physically, a
glottal stop isjust a noise, but in a segment identified as Arabic
it will be perceived as a fully distinctive unit; if the context is
known to be German it will be conceived as a delimiting
 PREFACE
trick; in a utterance, it will be cough and nothing
else.. What differentiates a language from another language
is not, as still widely believed, another way of combining
c E s
ready-made sounds (or letters!) forevoking the same things
or concepts, with specific but marginal deviations called I .. REALISM VERSUS FORMALISM page I
'accent' and 'idioms', it is a sui generis organization where
particular sound types are given a certain role in shaping II . TOWARDS A FUNCTIONAL SYNTAX 39
specific signs . Each of these corresponds to one of the elements
III . LINGUISTIC TYPOLOGY
into which experien.ce has to be analysed before it can be
communicated by means of language and is to be classified IV.. LINGUISTIC VARIETY I
according to the particular functions it assumes in linguistic
communication.. Describing a language consists in pointing V .. LINGUIS1'IC EVOLUTION 134
out what makes it different from all others.. Now, since speech
INDEXES 161
organs; perceptive and thinking abilities would seem to be
much the same throughout mankind, what makes it different
is less the substantial nature of the units it operates with than
the way these units function or, in other words, contribute to
communication..
For offering me the opportunity of presenting these views
of mine, first to an audience of distinguished scholars and
advanced students at Oxford, and ultimately to a wider
English-speaking public, my thanks are due to the President
and Scholars of the College of St . Mary Magdalen who in-
vited me to deliver a series of six Waynflete lectures in the
winter of 196 I and to the Clarendon Press for their willingness
to print an expanded version of these lectures.
I should also like to address my thanks to nlY old friends
Eugene and Sylvia Dorfman for reading and improving the
text of the lectures, and to Mr. John Ross for his careful read-
ing of both the manuscript and the proof..
To Austin Gill at whose instigation the invitation was
launched, who made lny stay at Oxford a congenial
experience, and who read my manuscript before it went to
press, this book is dedicated..
A.. M.
 ISM RSUS F ................ y ................... IS

T
H E most remarkable achievement of contemporary
linguistics is probably the final assertion of its legiti-
macy as a completely autonomous discipline with its
own object, aims, and methods. Whereas 'philology' had
never severed the ties that linked it to old texts and classical
education, 'linguistics', as a practically new term in English,
labels its contents as free from any dependence or servitude.
On the Continent, where classical scholars were wont to
distinguish between their philological and their linguistic
pursuits, 'linguistics' was, until Saussure and long after the
publication of his Cours, I largely identified with comparative
grammar, and, accordingly, the difference between the former
and the present status of 'linguistics' is the more striking.
Yet this newly won autonomy is frequently lost sight of,
since many linguists are prone to stress, less the unity and
recent self-sufficiency of their discipline, than its multifarious
connexions with other branches of research, old and new,
humanistic or scientific, such as psychology, logic, anthro-
pology, cybernetics, and electronics. This, of course, mainly
results from the fact that, in the scientific world of today, few
linguists are just linguists and nothing else. Just as, in the
past, a linguist was necessarily a philologist and, more often
than not, the student of some literature, he is now likely to
be an anthropological linguist, a mathematical linguist, a
statistical linguist or, of course, still, a philological linguist.
Plain linguists are few and far between. Being one of them,
I may perhaps be excused for concentrating on language as
my sole subject, limiting my excursions outside of that domain
I Now available in English: Course in General Linguistics, trans!' by Wade

Baskin (New York, 1959).

811928 B
2 REALISM VERSUS FORMALISM REALISM VERSUS FOR!v.IALISM 3
to cases where may be to a better understand.. SO(:lOeLO~:lC(tl factors
ing of my own field. at all times, prepare, out of the language of today, the lan-
It is not by chance that the establishment of general guage of tomorrow.
linguistics, i.e, the study of language for its own sake, No wonder 'structure' has become the rallying sign of
coincided with the proclamation of the view that the study contemporary descriptivists. It points, of course, .to a co-
and analysis of etats de langue should necessarily precede any herence of the components, but, at the same time, to the
other linguistic endeavour. Language had, until that time, aloofness of the whole from all the rest. 'Function', which is
been mainly considered in its evolution, and, accordingly quite fashionable in some quarters, has no great appeal in
no one had tried to conceive of it outside the context of linguistics. It is obviously redolent of the uses to which
ever-changing human needs that, at all times, jeopardize the languages are put; it suggests contacts with the world at
balance of its economy and lead to the restoration of that large, those very contacts we have had to disregard in order
balance in a new form. I At one time identified with logic or to achieve self-sufficiency. Yet, since we are all agreed that
reason, language had, by many, been viewed as a product language works as an instrument towards certain goals, we
of collective thinking and was consequently placed at the can hardly deny that the functioning of that instrument
intersection of psychology and sociology. That double depen- should be one of our major concerns. Actually, all 'struc-
dence accounted for its perpetual instability, nay, made of turalists' reckon with the function of linguistic units: setting
this instability one of the basic features of language and apart a feature as 'distinctive' implies that its function
thereby hopelessly blurred the boundaries between language, suffices to make it an object of interest and assign, it to a
as we conceive of it today, and what acts upon it from the definite class. But becoming conscious of the paramount
outside. Only the strictly synchronic approach recommended importance of function in linguistics will normally lead to a
by Saussure could lead to the foundation of an autonomous greater respect for reality. And what is meant here by 'reality',
science of language through a drastic severance of all its ties is not any physical or semantic trait which happens to be
with physical and psychic reality. singled out, but linguistic reality, that which is recognized as
Decades of effort toward perfectly static descriptions, such because it belongs to a given language where it exerts
culminating with Hjelmslev's purely formal analysis," have a definite function. Function supplies the linguist with a
failed to convince us that they represent an ideal starting- scale of values that will stubbornly resist any attempt on the
point for linguistic research. But without Saussure's figment part of the theorist to make facts submit to the requirements
of the transversal cut," the perfect autonomy of linguistics of a method. Function may help in bridging the gulf between
might still be a long way off. To that figment we owe the the so-called progressive and traditional groups. It may
conception of a linguistic structure perfectly distinct from the bring about contacts that will dispel, on the one side, the
IOn language economy, see George K. Zipf, Human Behavior and the Principle belief that only sluggishness and vested interests are pre-
of Least Effort (Cambridge, Mass., 1949), pp. 19-22, and Andre Martinet, venting one's opponents from discovering the truth, on the
Economic deschangements phonetiques (henceforth Economies (Berne, 1955), pp. 94-
97, and Elements de linguistique generale (henceforth Elements) (Paris, 1960), pp.
other side, the conviction that the new trends are incorn-
182-7- patible with careful observation and respect for facts.
2 See Prolegomena to a Theory of Language, trans!' by Francis J. Whitfield
The most deeply rooted objection to 'structuralism' is,
(Suppl. to I]AL, vol. 19, no. 1) (Baltimore, 1953).
3 Course, pp. 87-88. indeed, that concentration on 'structure', whatever that may
4 REALISM VERSUS FORMALISM REALISM VERSUS FORMALISM 5

be, will necessarily lead scholars away from a close scrutiny tures, but one who makes them. I This is, ofcourse, an extreme
of observable facts and make them disregard whatever stands attitude, but it clarifies the more average position according
in the way of their attempts to set up theoretical construc- to the actual of structure is, at least, not
tions. In fact, no one among those acquainted with the postulated.
linguistic practice of the last decades would deny that there This highly formalistic approach underlies the practice of
is some truth here. We all remember so-called descriptions in a probable majority of contemporary descriptivists, although
which, at every point, facts were culled from the most various it is professed only by a handful of theorists for whom con-
sources, without any regard for consistency, so as to make sistency is a fundamental requirement.
them fit the author's preconceptions. Should this be struc- The realistic conception of linguistic structure as a feature
turalism, no serious scholar would want to be a structuralist. oflinguistic reality is apt to be mistaken for the naive assump-
But, even if we leave aside that sort of irresponsible juggling, tion that all that is physically present in speech is part of that
it cannot be denied that any effort towards establishing a reality, irrespective of whether it has a function and what
single method for the treatment of all linguistic facts or the that function is. Actually, structure can be found in language
description of any language whatever, will, almost inescap- only, as it were, as an aspect of its functioning. The varieties
ably, result in giving the same status to things which differ, of function establish among data a hierarchy which involves
not only physically-which would be quite in order-, but distinguishing between cores and margins, for whose de-
also in their role in the economy of the language. scription different methods have to be used. Function is the
Language reality is far more varied and far less homo- criterion of linguistic reality. Our duty is to describe that
geneous than many descriptivists would be willing to con- reality, and it should be no cause of alarm if one of our
cede. At many points it gradually merges into other aspects operational devices is found to fail us at a certain point.
of reality, which explains why it has taken such a long time These devices, such as phonemes, for instance, do correspond
to secure the autonomy of linguistics. For a linguist who is, to definite aspects of linguistic reality as shown by the
above all, intent upon notjeopardizing this painfully achieved speakers' comportment, and we value them in so far as they
autonomy, the normal reaction to a situation where the do, but no further .
limits between language and non-language seem to be We need not be ashamed of presenting marginalities as
blurred is to {j:proceed arbitrarily and draw clear-cut distinc- such in our descriptions, because they are the proof of the
tions even across uncertain ground. Once his domain is thus latter's truthfulness. Natural scientists, who approach their
delimited, he may proceed to submit all its parts to one and own problems far more soberly than do our theorists, arewell
the same treatment. Yet he may, all the time, be painfully aware of marginalities in those aspects of the universe it is
aware of the fact that this type of procedure will more or less their duty to investigate and to describe. Zoologists, for
distort the picture he will be drawing. From this he may con- instance, have gone on registering all phenomena as they
clude that there is no such thing as 'structure' in language entered the field of their observation and have never tried to
itself, that what is so called is nothing but a frame invented do away with the platypus. Why cannot those who study
by the linguist in order to help him to classify the data. In man's behaviour become reconciled to its entire range?
other words, a structuralist is not one who discovers struc- Our first illustrations will be borrowed from the field of
I Cf. W. S. Allen, On theLinguisticStudy ofLanguages (Cambridge, 1957), p. 14. phonology.
6 ALISlVI VERSUS FORl\,fALISM REALISM VERSUS FORM.A.LISM 7
According to the phoneme theory, every utterance, in requirement of a subsequent statistical study. But he will
any language, is totally analysable into a succession of dis- explain the reason for his choice, and stress for the reader
tinctive units; these units are and their the factual impossibility of equating, say, the case of jaij
in the language, is strictly determined. Methods have been with that of a clear-cut phonematic duality, as in the case of
devised for the analysis of utterances into phonematic seg- jklj in clear, and a clear-cut phonematic unity, as in the case
ments. It is clear, for instance, that the Spanish word mucho of the first vowel of father.
is made up of four successive phonemes: the segment corre- If the analysis of utterances into a number of successive
sponding to the ch group of the spelling is one and the same phonematic segments raises some questions whose solution is
phoneme, although it begins as a stop [t] and ends as a not quite obvious, the next phonological step, namely the
fricative [8], because that fricative never occurs in Castilian identification as one phoneme of segments appearing in
Spanish without the preceding stop, which implies that this different contexts, is so full of epistemological pitfalls that
stop, as such, is automatic and not distinctive.. The reasons most structuralists prefer to sidetrack the problem altogether
for conceiving of the consonantal complex at the beginning by declaring that phonetic likeness should be the only guide.
of English chip as a single phoneme are undoubtedly very This, of course, sounds beautifully sensible and is, in practice,
cogent, although not quite as obvious as in the case of highly satisfactory until one comes across cases where a
Spanish eli. But there are many cases on record where no sound appearing in a given context seems to stand just half-
agreement has ever been reached as to whether one should way between two of the sounds that are attested elsewhere.
reckon with one phoneme or two successive phonemes: Here again, some arbitrary decision is inescapable, unless
some would analyse the word ice into two, others, into three one is ready to restate the problem in functional terms: is
phonemes. If ice presents two phonematic segments, an jaij the distinction that generally obtains preserved or not in
phoneme has to be listed among the phonological units of the position at stake? At this point, the rank and file of
the language; this will probably entail a similar decision Bloomfieldians will shudder and exorcize the spectre of the
regarding jauj as the former of the two segments of out. It: archiphoneme by pronouncing the magic formula: 'once a
on the contrary, ice is made up of three segments, out follows phoneme, always a phoneme l'
suit and English loses two phonemes.. 1 -If we now come back to reality, we soon discern that while
Faced with such a problem, the formalist will reach an no one can question the fact that there is between big and
arbitrary decision. But since he wants to be consistent, he beg a difference, lying somewhere in the middle of the two
will formulate his verdict so that it can be used as a prece- utterances, which suffices to distinguish them, it is far more
dent if some similar case arises. This, however, should not risky to state that the two vowels of kitchen are the same. For
be construed as if he were looking in the data themselves for a satisfactory functioning of the language, at any rate, it is
a justification of his comportment. essential that the vowel of big and that of beg should be kept
The realist may also have to decide on one interpretation distinct, but it is immaterial whether the two vowels of
or the other, because of some practical necessity, such as the kitchen should be the same, or not the same at alL In theory,
I Contrast Trubetzkoy's position in Grundziige der Phonologie (Prague, 1939), there is no reason for the system of word initial consonants
pp. 108-9, with the average Bloomfieldian analytic approach as, for example, to present any point of similarity with that of intervocalic or
in George L. Trager and Henry L. Smith, Jr., An Outline of English Structure
(Norman, 1951), pp. 19-2 2 • word-final consonants.. We could easily imagine a language
8 REALISM VERSUS FORMALISM REALISM VERSUS FORMALISM 9

where all word initials would be articulated in the front part infancy onward, to detect in the speech of others what corre.. .
of the mouth, and all the final consonants at the back and in sponds to distinct choices on their part. Learning to speak is
the throat, so that it would be meaningless to try to pair learning to make the choices current in one's community.'
every unit of one system with one of the other. What prevents Phonemes are such choices. They do play their distinctive
this is, of course, economy: people who make distinctive use role precisely because they are specific choices and recognized
of labial stop articulation in one position would be stubborn as SUCI1.
monsters if they resisted, throughout millennia, its extension It is at the same time normal that there should be different
to other positions.. systems for vowels and for consonants and that the various
Where economy requires that systems of distinctive units consonantal or vocalic systems of a given language should
should differ from one position to another, systems are differ- tend to present similar distinctions, but no one should try to
ent, as is shown by the coexistence ofvocalic and consonantal spirit away any discrepancy between the distinctive pattern
systems. in one position and that in another. Yet formalists have
It is quite essential to try to understand why we all insist rationalized some of their arbitrary decisions into a principle
on calling the [p] of pat and that of tap one and the same of simplicity, according to which the best language descrip-
phoneme. We do so for no cogent theoretical and abstract tion is the one which presents the fewest possible units. When
reason, but only because both appear to be the product of the they come across a situation where one type of distinction is
same muscular habit with just the necessary adaptations found in position A and another type in position B, they never
imposed by the respective contexts. Both are prevented from stop before they have convinced themselves, if not ofhers, that
drifting by the same inhibitions determined by the necessity the two types are nothing but avatars of the same distinction.
of keeping them distinct from other articulatory habits, the The case of French [e] affords a nice illustration offormal-
same ones for the [p] of pat and that of tap. If the competing istic and realistic practice in a case of that sort: Parisian
articulatory habits were not the same for initial [p] and speakers of my generation are apt to distinguish between un
final [p], it: for instance, the language under consideration metre, 'a metre', and un maitre, 'a master' ;2 the distinction lies
were not English, where we have tab alongside tap, but one, in the vowels: both are normally said to be [£], with or
like German or Russian, where no such pair could be found, without minor tamber variations, the difference being one
some of the drift-preventing inhibitions existing initially of duration or whatever gives that impression. In word-
would be found missing finally. Strange as it may sound, final checked syllables, such as metre, maitre, the only other
true realists, in such a case, would object to the complete front retracted vowel units are li/ as in mitre, 'mitre', and lal
identification of initial [p] and final [p] even if these could as in quatre, 'four', which gives a partial system that could be
be shown to be physically identical. What is decisive, in represented as follows:
language, is achieving communication, and this is secured if, 1£/ (m)etre
at every point in the utterance, the unit chosen is kept distinct Iii (m)itre lei (m)aitre [e] (qu)atre
from the ones that could have been used, in the very same I Cf. Elements, pp. 3 1-3 2 •
context, in order to make a different message. But how close 2 Cf. Andre Martinet, La Prononciation du francais contemporain (henceforth
this unit is to one actually used elsewhere is of secondary Prononciation) (Paris, 1945), pp. 126-9; the distinction is being abandoned, as
shown in Ruth Reichstein's 'Etude des variations sociales et geographiques
importance. It is quite obvious that we are trained, from des faits linguistiques', Word, xvi (1960), p. 61.
10 REALISM VERSUS FORMALISM REALISl\1 R.SDS FORMALISM II

III absolute final position, the same French usage distin- -aw of English law. 1 In another of pronunciation, which
guishes four vocalic qualities, Iii as in riz; lei as in re, lEI has greater prestige, the tamber difference is much smaller,
as in rais, lal as in rat. the partial system looks as and it is eked out a difference in length, so that the vowel
follows: of quatre can be said to be short and front, that of pdtre long
Iii riz lei re 181 rais lal rat and slightly back. If we 110W consider our vowels in the
frame of the articulatory quadrilateral, it should be clear
We have thus four distinct units in both positions, and the
that the two short vowels, that of metre and that of quatre, are
principle of simplicity will convince formalists that the final
about as distant as the two long vowels, that of maitre and
lei of re must, by hook or by crook, be identified with either that of patre, the former pair lying a little higher and more to
one of the two IE/,s of metre and maitre. Pure and conscious
the front than the latter, as shown on this diagram:
formalism will favour an arbitrary solution. Mitigated for-
8
malism will try to buttress its decision by reference to minor
tamber variations. After long hesitations, Roman J akobson \8"
a a
decided in favour of equating the vowel of re with the short
lEI of metre, whereas the long vowel of maitre was identified This is the type of fact on which linguistic realists are likely
with the lEI of rais whose duration, in the dialect being to capitalize. In the present case, even in a purely static
discussed, is the normal short one of word-final vowels. 1 presentation of the vocalic structure of French, their aware-
Should instrumental observation reveal, in many speakers, ness of the importance of the relationships among the units
a slightly closer vowel in metre than in maitre, this would which really form a system, i.e, among which speakers have
certainly not point to any tendency to make it conform with to choose at each point, if communication is to be secured,
the close vowel of re, because if such a tendency existed it is will prevent them from ascribing too much importance to
hard to see why it should not 'be more successful. the relationships among units from one system to another
Instead of looking from one local system or inventory to and indulging in largely arbitrary speculations as to how
another, we shall, more realistically, consider what happens maximal descriptive simplicity is to be achieved.
in a definite position, namely before word-final -tre, where The case of what is called the French 'mute e'2 gives a
vocalic phonemes have, of course, to be kept distinct from more dramatic illustration of the contrast between what I
one another if linguistic communication is to be secured. am tempted to call formalistic totalitarianism and realistic
Next to our two IE/,s, in the system, we find two [a] phonemes, discrimination. English spelling is full of silent e's, as in case,
those of quaire and pdtre; the former is often said to be front mute, and give. Since they never correspond to any audible
and the latter back, and here lie the essentials of the dis- articulation, they do not raise any phonological problems.
tinction in strictly local Parisian pronunciation where the In French, some e's are found in the spelling where, accord-
vowel of pdtre might, in the 1920'S, have sounded like the ing to context or style, either nothing is heard, or a more or
less centralized vowel is sounded, which we will henceforth
I The present writer was used as a native informant, and his testimony was designate as [d]. It is important to notice that such an alterna-
solicited first in favour of the identification of the vowel of re with that of
maitre, then in support of the grouping which was eventually retained in
tion of zero and [d] is never found initially in an utterance or
R. Jakobson and John Lotz, 'Notes on the French Phonemic Pattern', TtTord, I Cf. A. Martinet, 'C'est jeuli Ie Mareuc', Romance Philology, xi (1958),
v (1949), p. 154· p. 35 2 . 2 Cf. Prononciation, pp. 37-7 0 .
 12 REALISM VERSUS FORMALISM REALISM VERSUS FORMALISM

a word, but always after a consonant.. In normal everyday different forms, Blin and Belin which are phonemically quite
unsophisticated spoken Parisian French, the alternation of distinct: some people are called Blin and others Belin, and
zero and [d] is in ninety-nine cases out of a hundred auto- confusions are not likely. With ordinary nouns, which are
matically regulated by the context, and this irrespective of norrnally preceded by an article, the situation is different,
whether some e appears in the spelling or not.. Phoneticians and I have heard children wonder why a small fish (l'ablette)
usually say that the appearance of the [g] vowel is regulated had the same name as the weasel (la belette) .
by the 'law of three consonants', I which prescribes that [g] At the age of six, children go to school and begin to learn
should be pronounced where it will prevent clusters of more fables and other pieces of verse. Now, French metrics are
than two consonants, thus, for instance, in arquebuse and based upon the language as it was pronounced at the
contre-maitre, where it finds its expression in the spelling, and beginning of the sixteenth century when all 'mute e's' were
in ours blanc [ursobla], and arc-boutant [arkobuta] where it sounded or, at least, made a difference, unless they were
does not.. This amounts to saying that 'mute e' is no phoneme elided. A line like
since its appearance is automatically determined by the con-
Et Je sais que de moi tu midis l' an passe
text, and that for every consonant phoneme of French we
should reckon with two allophones: a purely consonantal one, which would amount to ten syllables in normal speech
the one we find for jdj in dans, and one, partially vocalized, ([E3-sEk-dg-mwa-ty-me-di-la-pa-se]) is recited with the regu-
which we encounter in devant ([dgva], phonemically jdvaj). lar twelve syllables of the alexandrine. This contact with
All this is, of course, a greatly simplified picture of reality: classical poetry is, with the average child, too incidental for
there actually are cases in which the pronouncing or not him to learn where to put a 'mute e' and where not to put it.
of 'mute e' may make a difference in the interpretation of What he remembers is that one has a right to insert a 'mute e'
what is said. after any consonant if metric necessities require extra syllables..
First of all, French children in the course of their slow and And this is indeed what happens when children sing their
protracted learning ofa difficult language, eventually manage own words to a tune; as, for instance, On t' emmene ce soire ..
to distinguish between two different treatments of words instead of the traditional Va chez la ooisine. . . . In popular
beginning with a vowel, the one which prevails in the poetry, the 'mute e's' of the spelling are sounded or not,
enormous majority of cases and which children tend to make according to the occasional needs of the author. A protracted
universal, and the other which is reserved to words which contact with classical verse will, however, finally condition
used to be pronounced with initial [h]. Hence l'eau [10], but people in that respect. Apart from the initial consonant, I
le haut [loo] , l' etre [lstr], un etre [&n£tr], but le httre [ld£tr], un pronounce soir and poire alike; but I would no longer accept
hare [&£tr J. As is shown by some of these examples, the them as rhymes.
phonic difference often lies in the presence or absence of the The above is a realistic sketch of a most intricate problem.
so-called 'mute e', This implies that [g] cannot be equated It is realistic because it indicates certain aspects of the lin-
with zero before a word-initial vowel. guistic comportment of French speakers without trying to
The name of the ram, in the medieval epic of Reynard the account for them all within one and the same formula.
Fox, was Belin, and this survives as a surname, but in two A totalitarian approach would consist in stating that since
I Maurice Grammont, La Prononciationfrancaise, 7th ed. (Paris, 1930), p. 105. there are cases when [g] instead of zero makes a difference,
 REALISM VERSUS FORMALISM REALISM VERSUS FORMALISM

l'dl is a and be marked as It IS this gradual elimination of [d] wherever the result was not
actually or potentially there, 'Potentially', of course, is meant too heavy a cluster, biconsonantal groups of all kinds became
to take care of traditional poetic diction. This type of descrip- pronounceable: gibeciere, 'satchel', formerly a four-syllable
tion, if it deserves to be called a description, has definite word, was reduced to two with the group /bs/ in between.
practical advantages: it gives the language a monolithic From that time on, speakers could yield to the suggestion
appearance which is thought, in many quarters, to make the of the spelling and sound the -b- in absent. Therefore, the
approach to French as a second language easier; it meets /-bs-/ cluster in absent is not, in ordinary speech, different
with the approval ofthe mass ofed ucated Frenchmen who are from the same group in gibeciere.
inclined to identify culture with the written form of their From an educational standpoint, it may suffice to state
classical heritage; but its main advantage in the eyes of the that no one can hope to understand spoken French before,
formalist consists in the vistas it opens toward a further consciously or unconsciously, he has learned to disregard
simplification of the pattern: it is clear, for instance, that if 'mute e' as a distinctive segment except in the very specific
we can analyse cane, 'duck', which is pronounced [kan] , as situations which have been noted. It is not our purpose, here,
/kang/, nothing but phonetic reality (but who cares about to examine how this could be achieved.
that?) can prevent us from analysing quand, which is pro- A last illustration will also be borrowed from French, but
nounced [ka] , as /kan/, and in general analysing nasal this time I shall consider the case of a grammatical category,
vowels into oral vowel plus nasal consonant. We thus get rid namely gender. Here is a difficult problem which we can
of the four distinctive units we would have otherwise to posit only hope to clarify if we refuse to let ourselves be .rushed
in vent, oin, long, and un. It is true that a word like canton will into accepting from the start a totalitarian approach.
thereby require six successive phonemes (/kanton/) instead My choice of French is largely determined by the fact that,
offour (/kat5/). I But it would seem that formalistic simplicity since there are only two genders in that language, it will be
applies exclusively to the number of units in the system, not easier to conceive of the gender system in its totality, if in-
to that in the utterance. deed we are entitled to speak of it as a system. It goes without
The main disadvantage of this type ofinterpretation is that saying that the conclusion at which we shall arrive regarding
it gives a completely distorted picture of present-day natives' French does not necessarily apply to other languages, even
linguistic comportment. Toward the end of the fifteenth those as closely akin as Italian or Spanish.
century, French was a language in which, with very few French distinguishes between a 'masculine gender' and a
exceptions, all syllables ended in a vowel, very much like the 'feminine gender'. The formal behaviour of nouns belonging
Old Macedonian of Cyril and Method. z Consonantal groups to the masculine gender is simpler. If all nouns behaved
were unpronounceable unless they existed word-initially: in like them, there would, of course, be no gender problem,
a word like absent, the b must have been mute in normal quite exactly no gender. We may therefore assume that the
speech. On the other hand, [g] was the only vowel found in feminine gender is the key to gender. We might have done the
totally unstressed position, and consequently it could be reverse and considered normal the behaviour of feminine
dispensed with without any resulting confusions. Through gender nouns: then, the masculine gender would have been
the key to the gender problem. But OUf actual choice is
I cr. Knud Togeby, Structure immanentede la languefrancaise (Travaux du Cercle

Linguistique de Copenhague 6), 1951, p. 58. 2 cr. Economic, pp. 326-8, 34-9-52. operationally preferable.
16 REALISM VERSUS FORMALISM 17
REALISM VERSUS FORMALISM

First, it is essential to distinguish feminine gender from Gender is quite frequently lumped together with number,
reference to female sex: when English speakers say she, as two frequent nominal categories. But this should not
instead of he or it, it is because they intend to refer to a induce us to put them on the same level of analysis.. Number,
female being or, exceptionally, to some machine they choose to the plural number, for instance, is, in French, a morpheme:
consider in the same relation as a female. The language offers if instead of la montagne Ilamotan/, I say les montagnes Ilemo..
three possibilities. The speaker is determined to choose one tan I, I choose to say I. . . e " . .1instead of I. . . a .... / because
or another by non-linguistic factors. On the contrary, when I want to stress that several mountains are involved, not one.
French speakers say elle instead of il, they do so because they Nothing, in the context, compelled me to say les instead of
cannot do otherwise. Apart from highly exceptional cases lao The plural was a new choice.
which we shall see later, there is, in such a case, no trace of It might be argued that French adjectives are frequently
a choice on the part of one who pronounces elle. The speaker used by themselves, both the masculine and the feminine
is forced to use this or that pronoun because he is referring forms, with the power and function of a noun, and that, in
to a linguistic unit that requires him to say elle instead of il. such a case, the marks of the feminine stand by themselves
I~ as I believe we should, we refuse to identify as a linguistic
without the support of the feminine noun. But, of course, the
unit a segment that does not correspond to a new choice ofthe adjective, here, is in the feminine because it refers to a sub-
speaker, we must declare that gender in French is no mor.. stantive which is understood, and could be furnished on
pheme, The situation is perfectly clear when, leaving aside request. In most cases of that kind, we would find the sub . .
the personal pronoun, we concentrate on the marks of stantive to be actually present if we cared to include the
gender in the nominal group. When I say la grande montagne whole conversation in the context, so that here again the
blanche, I choose to use the definite article and not the feminine would not correspond to a new choice.
indefinite une; I choose to qualify the mountain as great; I All this might induce us to state that the class of feminine
choose to speak of a mountain and not, for example, of a substantives in French is nothing but a group of nouns which
curtain; I choose to qualify it as white. But I never choose share certain formal characteristics. Theoretically, the case
feminine instead of masculine. This does not mean that I would not be different from, say, that of the English sub-
refuse to characterize the mountain as a female: I certainly stantives ending in -ow, such as shadow, widow, morrow,
do. It means that I am not given a chance to choose a gender, which have nothing in common except that final . . ow. Of
because as soon as I say, or foresee I am going to say, mon.. course, the signifiants of French feminine are discontinuous
tagne, I cannot avoid giving the definite article and the and their elements are dependent on the choice of accom-
accompanying adjectives the fuller, so-called 'feminine', panying adjectives, whereas the -ow of shadow can never be
form . My choice of montagne implies a number of formal separated from shad.. and is quite independent of the sur...
accidents which tradition labels as feminine forms. In Saus- roundings. But once these purely external features are set
surian terms, we may say that the signifiant of the feminine aside, the contents of the feminine gender boil down to zero,
noun montagne is not limited to that form, but that it is dis.. French feminines having hardly more in common than
continuous and emerges at other points of the utterance. In English . . ow substantives have.
our particular instance, it shows as the / .... e] of la, the Other circumstances have to be taken into consideration,
however, before we may venture to draw our conclusions:
I. · · dl of grande, the I · · · JI of blanche.
811928 C
18 REALISM VERSUS FORJVfALISM REALISM VERSUS FORMALISM :::9
it must be kept in mind that when a formally ferninine noun nouns, the gralnmatical concord is : la sentinelle
is applied to some animate being, that being is normally a elle. . . . The use of il in such a case would imply that the
female. There are, of course, a good many exceptions: specific reference to the soldier as a sentinel had been for-
baleine 'whale' is a feminine, which does not imply that the gotten. It would seem that the lack of gender agreement in
French believe whales are exclusively females. Still, the the case of le docteur ... elle ... is a recent development and
feminine gender of baleine cannot be equated with that of can still be considered a strictly marginal affair But it points
0

chaise or table. When referring in English to a chair or a table, to the fact that the use of elle instead of il may be governed
I, a native speaker of French, am not tempted to use she. But, by non-linguistic and this is a first step on the road
in reference to a whale, she would naturally be on the tip of followed by she in the course of the history of the English
my tongue. For the French, the unicorn is a female for the language.
purely formal reason that unicorne was interpreted as une A totalitarian description of the French language would
icorne before becoming une licorne. Consequently, it is difficult capitalize on this still sporadic phenomenon, stress the formal
for me to get reconciled to hearing, as I recently did in a identities in the expression of sex and gender, and finally
film , a unicorn referred to as he. There is at least some hint of emerge with all the features of the traditional, highly
sex in the feminine gender when used in reference to ,what idealistic, presentation of grammatical gender where it was
has a sex. assumed that the exhaustive distribution of all concepts into
Furthermore, we should not disregard the fact that the two classes of masculines and feminines entailed the anima-
feminine form of many adjectives differs from the corre- tion and sexualization of all conceivable objects.
sponding masculine in exactly the same way as the designa- A discriminating and functional approach will, of course,
tion of the femalediffers from that of the male: formally, stress the actual lack of function in a large majority of cases,
belle is to beau what chamelle 'she-camel' is to chameau. J ocu- namely when the substantive, whose gender entails some
larly, one could and does derive from bourreau, 'hangman', formal modification of articles and adjectives, is present
a word bourelle for a 'she-hangman'-or should I say 'a among them. It will take into consideration the somewhat
hangwoman'? We really cannot argue that we have, here, a 'pronominal' value gender marks will assume when the
pure and simple case of homonymy, because both uses of the substantive is understood or remote in the utterance, and it
-eaul-elle alternations actually fade into each other as shown, will take care of occasional lacks of gender agreement. But no
for instance, by the substantival uses of beau and belle. formulation shall be striven for, which, in an effort to cover at
Last but not least, elle is sometimes used in reference to once all known facts, would infallibly blur and even conceal
females who, in the same context, have been designated by the most specific, enduring, and characteristic features of
means of a masculine noun. A female physician is normally gender.
referred to as un docteur, but this never entails the use of the
masculinepersonal pronoun il. In a sentence like Le docteur It might be tempting, at this point, to indulge in a simile
est arrive; elle est dans le salon, there is no doubt that the pro- and say that a formalistic description is to the language what
noun refers to the sex of the person and not to the gender of a geographic projection is to the actual shape of the con-
its nominal designation. However, it is worth noting that, tinents, in the sense that the centre of the nlap is fairly
in the case of male beings designated by means of feminine accurate, but the margins are distorted, sometimes beyond
20 REALISM VERSUS FORMALISM REALISM VERSUS FORMALISM 21

recognition. But this would give too flattering a picture of Languages serve many purposes. They certainly help us to
the achievements of formalistic linguistics. What really hap- think. They give us an outlet for our pent-up feelings. We
pens, in too many cases, is that the inclusion of margins warps use them as artistic mediums. But they are first and foremost
the whole description in their favour at the expense of what used for communication, i.e, transmission of experience from
is, linguistically, really vital, as when the basic non-distinc- one person to another. Communication is, of course, in-
tiveness of 'mute e' in French is never even mentioned volved in the artistic uses we make of language, and what
because all centres around the exceptional cases where it is not communication there belongs to expression, a phrase
assumes a distinctive function. which, in technical parlance, we should reserve to self-
The danger inherent in the attempts to squeeze all facts, centred linguistic activity which does not aim at transferring
central or marginal, into the same pattern exists on all information from speaker to hearer, but to give the former
planes of linguistic description. It lurks, from the start, in relief from internal pressures and tensions of all sorts. Soli-
attempts to define the object of our science. In theory at loquy, which is pure expression, is normally frowned upon as
least, the first question a linguist should ask and try to answer unsocial behaviour, and those who want to 'express' them-
is: What is language? We may, in true Saussurian fashion, selves, will have to cheat and procure a victim with whom
come to the provisional conclusion that language is a system they can sham communication. This explains why linguistic
ofsigns. But, ofcourse, the question arises whether all systems evolution is entirely determined by the communicative needs
of signs are languages. Offhand, honest and sedate linguists, of man: soliloquy, ifit were not aping communication, would
who are trained to operate on such languages as Latin, soon result in the annihilation of language.
Russian, Chinese, but hardly on traffic signals, are inclined It may not be out of place to mention here that when I
to give a negative answer; but faced with the complexities speak of experience as being transmitted by means of lan-
of human languages, with central cores and marginalities guage, I give these terms their broadest possible meaning:
that they dare not identify as such, they finally yield to the experiencing a wish or a craving is part of experience, and
pressure of 'annexationists', those who want to put the lan- its transmission may assume various forms: plain statement,
guage label on as big a chunk of semiology as they can grasp. request, or command.
This leaves us without a definition of what it is our duty and Experience as such, prior to all attempts to transmit it to
our aim to investigate, namely human language properly so others, is not couched in words, except, of course, if it is
called in its different forms, the languages as actually spoken gained through linguistic communication. A very immediate
by men. type of experience, such as pain, is a good starting-point for
Ifwe want to know what language is, we should not try to understanding how and at what point language enters the
list all the features we may have come across when studying picture. Vocal reaction to pain may be purely reflex: a
the most diverse languages, and draft a definition which will groan. The groan may also be willed and meant as com-
somehow vindicate them all. We should rather try to deter- munication, but it is no linguistic communication: cats do
mine what all the languages we know, all the communica- communicate with their meows, yet we do not want to
tive instruments we want to call 'languages', really have in include meows in language. Language begins when the
common, so that we would not be willing to call 'language' homogeneous unanalysed feeling is interpreted into a suc-
some semiotic system which did not present that minimum. cession of definite vocal stretches, each of which can be used
22 REALISM VERSUS FORMALISM REALIS1v1 VERSUS FORMALISM
in the transmission of totally different of experience, weiss es nicht. differences in articulation show not only in
but which, when grouped and ordered as we hear them, the way monemes are combined into utterances, but also in
convey fairly specific information about what the speaker the range of people have at every point : where an
feels. Should I say, for instance, I have a headache, I would use English speaker may choose among blue, green, and grey for
five vocal stretches, namely I, have, a, head, and ache, each of conveying his experience, a Welshman will have to be con-
which can be found in totally different contexts for the tent with a single colour designation glas. All this points to
conveying of entirely different things; they are thus most a fundamental feature of human language: its variation from
unspecific, but, when grouped, they reach a fairly high one community to another and its variation through time.
degree of specificity. Throughout the world, cats say meow because this results
It is not too clear what is meant when laymen describe from voice accompanied by a lazy opening and closure of
human languages as 'articulate'. It is likely that people who the jaw. Language varies because it suits the varying needs
use this phrase just repeat what they have heard without of man. It follows that any feature of speech that is auto-
having ever thought of the implication of that term. Yet it matically found in all communities must be considered non-
describes perfectly what characterizes human language, less linguistic or, at best, marginally so. It is not a matter for the
in contradistinction to various forms of animal cornmunica- linguist to deal with, but for the psychologist, the physio.. .
tion than ill contrast with human experience before it has logist, or both, since these scholars study man in general,
been analysed with a view to linguistic communication. assumed to be the same throughout mankind. Our aim, once
What characterizes linguistic communication and opposes it we have agreed on the features we want to find in an object
to prelinguistic groans is precisely this analysis into a number before we list it among languages, is to describe languages,
of units which, because of their vocal nature, are to be pre- i.e. to indicate what makes a given language different from
sented successively in a linear fashion. These are the units all others, whether these are actually described, known to
which many contemporary linguists call 'morphemes'. But in exist, or just conceivable. The necessity of taking into con-
view of the fact that a number of others use 'morpheme' for sideration languages unknown to us compels us to reckon
different purposes, I prefer calling them 'monernes'. Monemes with all the possibilities that are not explicitly or implicitly
are the smallest segments of speech that have some meaning ruled out by our definition of language.
attached to them. According to Saussurian terminology, they The articulation of experience into successive units is only
are minimal 'signs', with two faces: signijiant and signifie. one of the two features which we want to include in our
The way experience is analysed differs from one lan- definition of language. We have to reckon in language with a
guage to another. The set of habits we call a language second articulation, that ofthe phonic aspect ofevery moneme
suggests the breaking up of experience a number of into a succession of distinctive units, the phonemes. Every
elements for which the language in question has equivalents: one of the five units of our former example is formally made
a language may use, for headache, a specific moneme, some- up of one, two, or three sounds or phonic complexes, to
thing like migraine, instead of two. Where the English say which, as such, no meaning is attached, but whose choice and
lift, others would speak of a hoisting machine. The way people order fully characterize the moneme whose manifestation
proclaim their ignorance is very differently articulated in they are: head, for instance, is made up of the three phonemes
English I don't know, in French Je ne sais pas, in German ich [u], lei, and Idl in this order.
 REALISM VERSUS FORMALISM REALISM VERSUS FORMALISM

Linguists of the old school, Saussurians be sense of the word may be. The phonic environments may
tempted to forget or disregard the second articulation. They result in distortions or warpings away from the performance
would thereby miss some fundamental conditionings of of the phoneme in isolation. But the semantic context is
human communication. The obvious advantage of the second normally powerless. All this is, of course, nothing but, a
articulation is economy. The first articulation was economi- synchronic version of the regularity of phonetic changes:
cal in the sense that with a few thousands of fairly unspecific only in very particular cases, which should be especially
monemes, it was possible to shape an infinity of different accounted for, may meaning affect phonetic evolution.
communications. In the same way, the second articulation is This phonematic solidarity can only be explained if we con-
economical, since the judicious combination of a few dozen ceive of the phoneme as an articulatory habit, something we
phonemes enables man to keep distinct all the monemes he should always keep in mind, even if we choose to disregard
needs. In view of the great variety and richness of human it temporarily when concentrating on descriptive methodo.. .
communication, the double articulation was bound to be a logy.
feature of human language: let us try to imagine how we A problem which always arises when language is to be
would fare if we, both as speakers and hearers, had to dis- defined is whether the vocal nature of speech should or
tinguish the thousands of homogeneous grunts which we should not be included in the definition. On the one hand,
would need for everyone of our monemes if the second it is a fact that the languages linguists deal with are or were
articulation were unknown. It is clear that the lexical ex- primarily spoken, even when we can only approach them
pansion made necessary by the progress of mankind would indirectly through written texts, and that, for a deeper
have been unthinkable without the tremendous economy understanding of those written texts, we should always try
entailed by the breaking up of signifiants into phonemes. But to imagine the spoken medium upon which the written
there is more to it than economy" Ifthe form ofevery moneme literary form was based" On the other hand, written lan-
was an unanalysable grunt, there would be a complete guage has a structure of its own, the study of which comes
solidarity between sense and vocal form. Meaning would within the realm of the linguist's preoccupations, and in-
exert a direct influence on form, and form on meaning, the cluding the term 'vocal' in the definition of language might
result being that at every instant, every speaker would be be interpreted as unduly narrowing the field by excluding
tempted to adapt his pronunciation to the particular shades areas where linguists feel perfectly competent. Besides, some
of meaning he would want to convey to his audience. As a scholars always want to grab the largest possible share of the
final result, both form and meaning would be in a permanent epistemological cake. Yet, there is one important argument
state of wavering, and this would prevent the establishment in favour of including vocal nature in our definition, namely
of discrete meaningful units, what the monemes of our that vocal quality is directly responsible for the linearity of
languages actually are thanks to their well-defined and speech and the consequent linearity of script. It is clear that
stable forms" The articulation of signifiants into a succession if signs were visual and presented on a surface, there would
of phonemes practically excludes the meaning of a given be no need for language to be manifested by a succession of
word from exerting any influence on its specific form. There items: the painter who is presenting his message on canvas
exists a solidarity among all the performances of the same must necessarily delineate his figures one after another, but
phoneme which tends to preserve its identity whatever the the recipient of the message may collect it in any order he
 REALISM VERSUS FORMALISM REALISM VERSUS FORMALISM

chooses, casting a general glance or concentrating on parti- It is clear that this in spite of its unwieldy
cular features as he sees fit. Should the message be the man is length, does 11.0t list all the types of features that may enter
killing a bear, onlooker might easily grasp it at first sight, the fabric We may even be sure that it does not
whereas the hearer will perceive it successively. The history include elements which probably play some role in all known
of writing begins with painting, i.e. a type of message which languages. There is no mention anywhere of speech melody,
is completely independent of language and speech, and leads which, in the minds of some educators, should embody the
through stages to alphabetic writing which implies complete most salient features and decisive aspects of the language they
conformity to the double articulation of language. But writ.. teach. We might dispose of an objection coming from those
ing becomes distinct from painting as soon as objects and quarters by arguing that the vocal nature of speech, covered
figures are shifted ever so little from the respective positions in our definition by the word 'phonic', implies the use of
which they would be expected to occupy in space, in order the so-called organs of speech, among which are the vocal
to suggest a succession reminiscent of the one our language chords; the vocal chords necessarily vibrate at a certain
units have in speech. It is true that we could easily imagine musical height, and the melodic continuum that results from
some other code, a gesture code or, why not?, an olfactory a continuous vibration of the glottis is precisely what we
code whichwould also impose successiveness. But it is hard referred to as speech melody, a physical reality some people
to imagine all the implications the particular nature of hastily identify with intonation. We shall revert to this later.
various bearers of meaning would have had forman's com- We should first answer another objection which affords
munication, The vocal nature of human language is cer- a very welcome illustration of the divergences between the
tainly no peripheral aspect of it, but a basic feature, without formalistic approach to linguistics and the one which t would
which linguistic organization might be fundamentally dif- recommend. I must say, at once, that I do not remember
ferent from what we know it to be. having heard it formulated, except perhaps by myself. In
All of this points to a definition of 'language' which might any case, there is enough of a formalist left in me to anticipate
run as follows: A language is a medium of communication it. It is not true that any utterance can be analysed into a
according to which human experience is analysed, differently neat succession ofmonemes, each with its own nicely wrapped-
in each community, into units (monemes) with a semantic up meaning and clear-cut segment. When I say he cut, where
content and a phonic shape. This phonic shape, in its turn, is my segment corresponding to the meaning 'past'? In
is articulated in distinctive and successive units (phonemes). French elle va au marche, with au a single phoneme /0/, what
whose number in a given language is fixed and whose nature slice should I ascribe to my preposition and what slice to my
and mutual relations also vary from language to language. definite article? As regards our second articulation, we have
This means that we should reserve the term 'language' seen before how difficult it is to analyse into phonematic
for a medium of communication which is doubly articulated, segments words like ice and out, and I have suggested that our
and whose outward manifestation is vocal. Apart from this failure may reflect less the imperfection of our methods than
common core, nothing can be said to be linguistic which a factual indeterminacy.
cannot differ from one language to another. This is how we Our answer will be that we have not said, or implied, that
should. understand Saussure's dictum that linguistic features the whole of language, as represented by corresponding
are arbitrary or conventional. speech, could be exhaustively reduced to successions of
 REALISM VERSUS FORMALISM REALISM VERSUS FORMALISM

monemes and phonemes. We have said simply that what we sounds very much like common sense, and although I, for
want to call a language makes use of monemes and pho- one, am convinced of its fallacy, I will not reject intonation
nemes; whether it adds to them other tricks which may at from linguistics. indeed, be made to play a role
times blur or distort some features of double articulation is similar to that of monemes, in such close co-operation with
another matter. the most central and abstract of linguistic tools that any
If we consider double articulation as the core of language, linguistic description would present wide gaps if all reference
and would see in the rest just margins, it is because this to intonation were to be avoided.
removes language farthest from indiscriminate, unanalysed, Actually, no clarity can be achieved in our discipline
interjectional, prehuman, or, should we say, proto-human, without establishing some sort of functional hierarchy. We
forms of communication. This alone can secure for language shall, no doubt, have to state that some feature or some aspect
the stability and rigour that result from the use of discrete of speech is not linguistic: this will be said, for instance, of the
units. We all, at one time or another, may get impatient initial rise of a speech melodic curve; this rise is due to the
with double articulation because some intonation or some fact that the glottis, starting from a state of rest, will have
interjection will enable us to achieve what we want far more to reach some degree of tension, and that the speaker is not
quickly and at a much lower cost. But this means nothing likely to wait until that degree is reached. Yet we shall have
more than the observation that it often consumes more time to distinguish different levels of linguistic relevancy: some
and energy to get someone to do something through asking French speakers pronounce their r's as a tongue-tip trill,
him than through giving him a push. Using language is a others as a dorsal fricative. This is found in Germany,
very complex and abstract procedure which is well adapted Holland, and in some other linguistic communities. But it is
to a certain degree of sophistication, but which, in a number far from universal. It is a culturally conditioned trait, not an
of trivial circumstances, may economically be replaced by inescapable result of the working ofman's psychic and physio-
some more direct means of communication such as gesturing, logical make-up. It has no function, if we reserve this term
either with the hands or the shoulders, or with the glottis. to cases where some choice of the speaker is involved. It may,
In which case, if the use of language is concomitant, we however, inform the hearer about the origin, urban or rural,
speak of intonation. of the speaker and, consequently, colour his interpretation of
The advantage of calling intonation 'gesturing' is that it the message. Some mention of all this must be included in a
removes it at once to the far periphery of the field oflanguage. description of French or German, and how could we totally
But of course no one will accept this, except perhaps as deny the epithet 'linguistic' to something which it is the duty
a metaphor, like the Lautgebiirden of a former generation of the linguist to describe?
of psycho-linguists. A gesture properly so called may be The double-articulation theory and any definition of lan-
accompanied by noise, a snapping of fingers, for instance, but guage based upon it leaves a wide margin, for which the
it cannot be sheer noise. Only a convinced functionalist name 'prosody' is today a widespread designation. Any-
could accept the view that a noise is not first and foremost a thing may be said to be prosodic that does not fit in the
noise. Since intonation is produced by the vocal chords, and monematic and phonematic segmentation; so that the Ameri-
since the vocal chords are par excellence the organs of speech, can 'suprasegmental' is not a bad substitute. I just think
intonation cannot be anything but speech. This dictum nothing is gained by speaking of 'suprasegmental phonemes'.
 REALISM VERSUS FORl\1ALISM REALISM VERSUS FORMALISM 3I

It must, in any case, be kept in mind that some prosodic and komma 'come'. In a language where syllabic nuclei
features like intonational contours are 110t distincti;e in the always coincide with vocalic phonemes, considering tones as
sense in which a is directly mean- distinctive vocalic features, together degree of aperture,
ingful: a rising interrogative contour onyou like it? has about lip-retraction, or nasality, might not be objectionable. But
the same function as the do of the more traditional do you like where this is not the case, we have to separate tones and
it? other words, it might equated with a moneme, not 'phonemes, although they function alike, and put the former
with a phoneme. In Saussurian terms, it is a sign, a minimal with accent and whatever remains ofthe melodic curves when
sign, with a signifiant, the and a signifi«, 'inter- tones have been extracted.
rogation'. We should, of course, by no means define the domain of
The distinction should thus theoretically be made between prosody by reference to the physical nature of the features we
double articulation and prosody and not, as is traditional, want to include: should, for instance, nasality prove supraseg-
between phonematics and prosody. Yet this latter practice is mental, we should not exclude it from prosody on the ground
in many respects amply justified because it is very excep- that it normally appears as a phonematic characteristic; -in
tional to find as clear-cut a signifie as the one corresponding many languages a glottal stop or catch functions as a tone
to the rising interrogative contour, intonation generally lack- and is regularly regarded as such although the glottal stop
ing the discreteness of n1any gestures. If we consequently is, in many languages, a phoneme, just like [u], another
agree to lump together phonematics and prosody under glottal product. It is, however, interesting to notice that the
phonology, we must point out that the distinction between features which are normally made use of in prosody are
the two branches of phonology is based on differences in those that are necessarily present in all utterances: stress, for
segmentation, and that this dichotomy may at times conflict instance, conceived of as the degree of energy with which a
with the classification based on function. I For instance , tones , spoken stretch is articulated, is always there; as soon as voice
or as some people call them tonemes, have exactly the same is heard, the vibrations of the glottis must have a given
function as phonemes: they are distinctive, which means that frequency, which results in melodic height; duration is, of
the speaker, at a certain point in the message, will have to . course, unavoidable in speech, since speech exists through
choose between a number of them in order to say just what he time. It is therefore understandable why speakers rarely get
wants to say . It is, ofcourse, perfectly immaterial whether the a chance to oppose these features to their absence at a certain
choice is conscious or not. If tones are not considered dis- point in the utterance, but only to choose between their
tinctive features of vocalic phonemes, it is because they are modalities, which may vary from one part of the utterance
usually found to affect, not a vowel phoneme as such, but a to another. Consequently, they are not so well adapted to
syllabic nucleus, often made up of two or more phonemes characterizing discrete units as others, like, for instance,
or even more than one syllable. This is shown by Swedish or nasality or dorsal occlusion, which mayor may not appear in
Norwegian, where the distinction between the two tones a given utterance: if I say What shall we do today? I make no
requires at least two syllables to be preserved: kbmma 'comma' distinctive use of either of these features, while I cannot help
giving my utterance a duration, using a varying amount of
I Regarding the methodological conflict between segmentation and func- energy from the first to the last phoneme, and whenever
tion, cf. A. Mart~net, 'Accent et tons', Miscellanea Phonetica, ii (London, 1954),
voice is there-and voice must be there-giving a certain
pp. 13- 1 4, and Elements, pp. 54-57.
32 REALISM VERSUS FORMALISM REALISM VERSUS FORMALISM 33

frequency to my glottal vibrations. The needs of communica- a drop is interpreted as .the end.
tion, acting on pre-existing, traditional linguistic patterns, the lower the drop, the more final the statement will sound.
may yield strange results which seem to challenge articula- cannot be said that a rapid and deep fall 'means' con-
tory and acoustic economy, as when, for instance, a lan- tempt, but it is normally indicative of some such feeling in
guage with a series of glottalized consonants extends it to the speaker, or, at least, since everybody knows how to play
the labial position, in spite of the fact that this involves using the game, it is what the speaker wants his audience to believe.
the cheeks, which are the worst possible organs to use when The less abrupt the fall, the friendlier it will sound. The
what is at stake is securing in the mouth as high a pressure frequency of statements with level final contour in polite
as possible. In general, however, language economy will tend British usage bears witness to an unmistakable effort not
to eliminate such quirks and to restrict the play of the dif- to give offence. The least indication of a final rise will suggest
ferent organs to those to which they are best adapted; this non-finality. The interrogative final rise is just one type of
accounts for the preference evinced for stress, pitch, and those non-final contours. Within the unit of utterance itself,
duration in the prosodic domain, and their limited employ- the sentence, a slight rise will normally precede any pause,
ment outside of it. Within prosody, however, pitch features since otherwise a pause might be interpreted as the end of
seem to be well adapted to a number of different functions, the sentence. In Ghiliak, there is a speech segment with the
and there is perhaps no better illustration of how the same sound [4>uru] with no specific function except the support of
physical fact can be used linguistically for totally different a melodic rise which might be interpreted as follows: 'the
purposes than a precise functional analysis of speech melody. preceding pitch fall indicates the end of a sentence, but
We should remember, first of all, that speech melody the present rise indicates that it is not the end of the story
cannot be left out, because voice is a normal ingredient of yet; please, don't clap your hands'. I French political oratory
speech, and voice cannot exist without pitch. Before speech makes use of a similar trick: the gasping orator, who does not
begins, organs are normally at rest, and some time will want to be interrupted by applause, lets his ,voice sink to-
elapse before they reach the degree of tension found to be wards the end of the period, but immediately tags the first
normal for a certain style. This accounts for the normal pitch word of his next sentence with a rising pitch, which will
rise at the beginning of an utterance. Voiceless stretches, secure him the breathing spell he needs.
usually very short, within an utterance do not seem to affect The picture we have just drawn of the main features of
the general tension of the glottis. When the end of the speech melody is not, as it might seem, exclusively derived
utterance is getting near, the speaker will tend' to antici- from a narrow, personal experience, but is, at least, but-
pate it and allow the organs to relax before speaking stops tressed by recorded observation. Nevertheless, it cannot be
altogether. This accounts for the speech drop at the end of assumed that no language ever shows deviations from this
utterances whenever there is no reason to check it. But it must general pattern. In sketching it, we have not included the
be kept in mind that, whereas the initial rise is physiologic- possibility that some stretch of the melodic curve might be
ally determined and consequently fairly stubborn, the final needed for the marking of a distinction, as is frequently the
drop can easily be replaced by some other contour. This case in so-called tone languages; we have disregarded the
being the general conditioning, it is easy to see how variations
I This illustration is taken from the material collected in Japan by my
of that curve can be put to use in human communication: colleague Robert Austerlitz.
811928 D
34 REALISM VERSUS FORMALISM REALISM VERSUS FORMALISM 35

fact that word accents may also affect the form of that curve; It would seem that an essential ingredient of English accent,
we have left out of the picture the existence in nlany, and traditionally but perhaps mistakenly called 'stress', is a
perhaps all, communities of favoured contours which may sudden and rise. or fall of the pitch. I In any case,
entail, on the part of the users, a slightly different interpreta- accent is likely to modify the normal course ofspeech melody,
tion of other contours. We mean this pattern to be nothing and SUCll accidents as it may cause in it will have to be
but a point of reference, which should prevent describers abstracted, just as tonal accidents were. A common error
from forgetting how much in speech melody is conditioned consists in ascribing to a vast ill-defined domain of intona-
by human physiology and psychology and consequently is tion a number of features which functionally belong to that
not and should not be the concern of the linguist. of accent. The use of 'stress', which refers to a physical
The first task of the linguist in prosodic matters is obviously reality, instead of 'accent' is apt to confuse even competent
to spot in the melodic curve all the stretches that have scholars and make them speak of intonation as soon as they
functions, in other words, the tones or tonemes. Once they fancy they are hearing pitch instead of stress. for many of
have been identified and classified, they should be abstracted them, the difference between to increase and an increase would
from the curve. Whether what is left, after they have been be due to a different placing of stress, while that between
disregarded as so many accidents, can be identified with the a moving van (a van used for moving furniture) and a moving
above-mentioned pattern seems to depend on the importance van (a van in motion) results from the use of a different into-
of tone distinction for the language, and the complexity of the national contour. Once accent is defined, not in reference to
tonal pattern: in African languages with two tones, a high one an alleged physical nature, but as prominence given to one
and a low one, the tonal succession does not seem to blur the syllable per word, or accentual unit, with a view to marking
general melodic pattern, so that the latter may be relied upon the respective importance of the units within the utterance,
to indicate whether the utterance is a statement or a question: it becomes clear that the difference between mooing van and
'high' and 'low' mean above and below what the general moving van is accentual and nothing else.
direction of the curve would lead one to expect if it were not What we should call intonation is therefore what remains
for distinctive tones. In Vietnamese, on the contrary, where of the melodic curve of speech once all tonal and accentual
a six-tone system is the norm, with a high and a low register, features have been extracted, and, as rightly pointed out by
with one rising, one punctual, and one interrupted tone for many scholars, most of what is of interest there centres
each, it is doubtful whether enough freedom is left to speakers around final contours. It has become usual to analyse these
in melodic matters for them to make use of the latitude contours by reference to three or four levels. No one will
offered by the general pattern. object to such a practice if those levels are presented and
If the language under description is not a tone language, used as a frame of reference with about the same value and
the describer will have to determine whether the language function as the vowel quadrilateral of Daniel Jones, with its
makes use of an accent, and, if it does, how accentual pro- cardinal vowels in relation to which a given vocalic sound can
minence is indicated and whether and how it affects speech be characterized. It: however, those levels are meant to be
melody. Recent investigations seem to indicate that pitch phonemes, there is bound to be some disagreement. It is,
plays a much more decisive role in accentual matters than
I Cf. Dwight L. Bolinger, 'A 'Theory of Pitch Accent in English', rVord, xiv
had previously been assumed for such a language as English. (195 8), pp. 1° 9- 49.
 REALISM VERSUS FORMALISM REALISM VERSUS FORMALISM 37
of course, perfectly true that if one of the levels is replaced implications contributed by everyone of its components,
by some other, the meaning of the utterance may be changed, the levels. This amounts to saying that the relation of levels
just as the meaning of I make is changed if I replace m by t, to contour is comparable with that of words to sentence,
But, of course, m and t are phonemes because interchanging in which both words and sentence are signs with signifie and
them modifies the identity of the moneme where this takes signifiant, and basically different from that between phoneme
place, and, thus, the utterance is affected by this interchange and word, or moneme. Should we insist on identifying the
only indirectly through the replacement of one moneme items we operate with in intonational matters, with the units
by another. Our problem here is whether a shift of intona- that are the frame of double articulation, we would have to
tional level is to be equated with the replacement of one say that levels are monemes, i.e, minimal units with meaning,
phoneme by another or that of one moneme by another. whereas contours are a succession of monemes. But I do not
Let us assume, for instance, that in a contour describable in think anything is gained through confusing the different
terms of the succession of levels 2-3-2, level 3 is replaced by planes whose distinction has proved instrumental in clarify-
level I. A peak of the curve will then be replaced by a trough. ing the working of language.
Level 3, which, all by itself, implied a rise, will be replaced The central problem of intonation, a problem which,
by level I, which, all by itself too, implies a dip. This shows having hardly been formulated, has, to my knowledge,
that everyone of the relevant successive levels is indicative never been tackled, is that of the relation between the levels
of one of the specific directions assumed by the melodic as posited by the language describer, and their linguistic
curve. Now, each of these successive directions .is contribut- reality, or, in other words, the behaviour of speakers, It
ing something to the significance of the total contour. This seems clear that they are not discrete units because it is
contribution is additive, as is the one of a moneme to the never quite immaterial for the message whether the 'level'
meaning of an utterance. It is not destructive, as that of is performed a little higher or a little lower; a level implying
a phoneme whose presence signalizes that any inference, as bitterness or contempt will be the more bitter or contemp-
to the meaning of the utterance, which might be drawn tuous if it is pronounced a little lower than may be usual;
from the context considered without it is wrong: if, to the with phonemes, which are typically discrete, the precise
statement it is good, I add very, I am just adding some addi- way a phoneme is pronounced cannot change the message:
tional information without deleting what was previously dab remains 'dab' and nothing else, whatever the amount of
there, but, if to the statement it is a roe I add a jdj phoneme, voicing of its initial jdj, as long as there is enough of it to
the statement becomes it is a road; one element of information keep dab distinct from tab. But, if intonational levels are not
roe is deleted and replaced by another one. Since a level is discrete, do they or do they not, in a given language, corre-
just a way of indicating a direction, and since every succes- spond to some habitual comportments? If they do, it means
sive direction would seem to add something to the whole, it that a certain type of situation normally entails the choice of
is permissible to state that the value one may wish to ascribe a certain level, which, as we have seen, determines a certain
to a given contour does not actually go beyond the values direction of the melodic curve. If this is the case, any per-
that might be attributed to everyone of its successive levels. ceptible deviation from this direction must be considered
In other words, inasmuch as the contour has a bearing upon and classified as an individual deviation from a norm. It
the sense of the message, it does so as the sum total of the is a meaningful variation, but still a variant of a type. The
 REALISM VERSUS FORMALISM

answer to query a mass

ofinstrumental recordings submitted to a statistical examina-
tion. This, of course, would have to be done for a large
number of the most diverse languages before a well-founded
general linguistic treatment could be presented. o s F C I s
That such intonational habits exist need not be further

I
substantiated. They are some of the features that often F asked a.bout the history of structural linguistics, most
enable people to state that some person has this or that people concerned would probably say that it all began
'accent'. But it remains to be determined how far such with the phoneme. When structuralists chose first to
habits actually hamper or prevent individual uses of the concentrate on what we have called the second articulation,
natural implications of speech melody. they certainly found the correct approach towards greater
What, in any case, is perfectly obvious even before any rigour in the treatment of linguistic problems: the discrete
world-wide research programn1e has been carried through, nature of linguistic units is ultimately based upon the dis-
is the linguistically marginal nature of intonation, which creteness of the phoneme, and phonology was the foundation
should be obvious as soon as a dispassionate examination has vie needed for any further progress. Still, dealing so long
deprived this term of its glamour and limited its application with phonemes before attacking the more intricate field of
to some residual uses of speech melody. This, to be sure, is the first articulation had an unfortunate consequence. having
very much the same as to give your dog a bad name and achieved outstanding success in 'phonemics', linguists were
hang it. But need we care what terms we use, provided the legitimately induced, when tackling the actually far more
analysis is correct and fruitful? complex study of the significant aspect of language, to use a
similar pattern. This accounts for frequent terminological
pairings such as phoneme-morpheme, phone-morph, allophone-
allomorph, and so forth, and, on a different level of abstraction,
the isomorphism of the glossematicians with its strict paral-
lelism of the two planes of expression and contents. I
The most fundamental objection to this practice of identify-
ing the patterns on two different planes derives from the
obvious fact that, in language, something which is not
manifest, variously called meaning or experience, is mani-
fested by means of something else. This, no doubt, implies
a one-to-one equivalence, that of the signijiant and the
signifie, but not necessarily an identical behaviour of the
minimal, significant unit, which I call the moneme, and of
the minimal distinctive unit, the phoneme. The moneme is
I Cf. Jerzy Kurylowicz, 'La Notion de I'isomorphisme', Recherches structurales

(Travaux du Cercle Linguistique de Copenhague 5), 1949, pp. 48--60.

4° TOWARDS A FUNCTIONAL SYNTAX TOWARDS A FUNCTIONAL SYNTAX

a Saussurian sign, a unit with a meaning and a phonic shape, discrepancy in the syntactic comportment of phonemes and
i.e, one which combines something that is not manifest with monemes: a phoneme fulfils a function in a definite position.
its outward manifestation. It belongs to the two planes of If we want to identify a signifiant, e.g, that the word lake,
expression and contents, and it is the smallest segment that it is not enough to say that it is made up of three phonemes
does. The phoneme has a phonic shape, but no meaning. Ill, leil, and Ikl because the same phonemes are those which
It is pure manifestation and belongs exclusively to the plane characterize the words clay and kale; one must specify: /1/,
of expression. The moneme-and-phoneme approach to leil, and Ikl in that order. In other words, when pronouncing
linguistic analysis-and, for that matter, the widespread lake, speakers have to choose, in initial position, III and
morpheme-and-phoneme one too-does not coincide in the oppose any inclination to say Ikl for cake, ItI for take, &c.
least with the two-plane pattern of glossematics: considering Postponing III till the end of the word and anticipating the
an utterance like give me the book, glossematicians will put, choice of Ikl would not do, because we would thus get kale
on one side, the individual 'cenemes' like Ig/, III, [v], the which is not what we mean. All this, which sounds trivial, is,
signifiants, either minimal (/gIv/, Imi/) or complex (/gIv mi in fact, basic for the establishment of the phonematic pattern
Od btrk/) , and their graphic equivalents (g, give, give me the of the language..
book) as well; on the other side the signifies, either minimal The situation is different with monemes or significant
('give') or complex ('give me the book'). The basic glosse- elements generally; the relevancy oforder is far from general:
matic dichotomy can be represented as follows: it is fairly immaterial whether I say the one I like is Paul or
Paul is the one I like; the implications are different if I say
IIgl
gIV I
) 1'.1
('giVe'
,. ,
with Paul, I went to Rome and I went to Rome with Paul,
but they do not affect the identification of the moneme
jgrv mi O;:J buk/ gIVe me the book
group with Paul, and the same applies to yesterday in I went . . .
Linguists, who, explicitly or implicitly, operate according to yesterday and yesterday, I went. . . . Certainly, the respective
the double-articulation pattern, will keep their phonemes position of monemes is often determined by tradition or by
alone, and put signifiants and signifies together according to the need to distinguish between utterances with different
the following schema: meanings: it just isn't done to say Paul with went I Rome to,
and it is far from immaterial whether I speak of root hair or
/gIv/-'give'
hair root, not to speak, of course, of the difference between
jgj the mankills the bear and the bear kills the man. But it is clear that
} r-.;
{ [gtv mi O;:J bukj-'give whereas phoneme classes can be established by listing all the
me the book'
phonemes that appear in a given context, this cannot be done
This lopsided diagram conveniently illustrates how the indiscriminately with monemes, and it might seem that the
moneme stands on a far higher level of complexity than the first step with them should be to determine the situations
phoneme and why any effort towards pairing them is bound where their presence results from an exclusive choice, as is
to result in distortion. normal with phonemes.
The fundamental difference between distinctive and signi- It is, however, preferable by far to forget the phonemes for
ficant units must ultimately account for one very important a while and try to discern what comportment we may expect
 TOWARDS A FUNCTIONAL SYNTAX TOW.f-\RDS A FUNCTIONAL SYNT.A.X 43
on the part of monemes as the products of the first articula- could, of course, be easily dispensed with in an utterance
tion of language. Let us start again from the experience that like the present one, where yesterday refers specifically to a
has to be communicated.. If the communication is to be well-defined of past there was
linguistic, that experience will have to be analysed into a a riot, with its four monemes, is the reflex of a specifically
number of elements for each of which the language under English analysis, into four elements, of one aspect of our
consideration has an equivalent, a moneme with its meaning experience, where some other language might get along with
and its phonic form. For simplicity's sake we shall assume only one.
at this point that the phonic form of every moneme is always Another aspect of our experience is rendered by means of
neatly circumscribed. The communication will thus take the the phrase in the village. The moneme village represents, as an
form of a succession of monemes, each corresponding to some element of experience, a place, but not necessarily a place
definite element of experience. But, of course, the choice and where something is happening; village, preceded, of course,
nature of the elements of experience will vary from one by sorne article, as here, could be made to designate the
language to another. Let us suppose that the experience to be place, or perhaps its inhabitants, as the agent of some action,
conveyed could be rendered in English by means of Yesterday, as in the village decided . . ., or as its 0 bj ect, as in they saw the
there was a riotin the village. One of the elements of the English village. The function would, in both cases, be indicated by the
analysis is the moneme riot; riot applies to a definite happen- respective position of the elements. In our present case, the
ing, and it is immaterial whether it has previously been circumstance that the village is the place where something
identified or not: a riot in English is always riot, whether happened is treated as another element of experience whose
it is a riot or the riot; many languages do not bother about reflex is the moneme in. It would not be difficult to imagine
distinguishing between previously mentioned and not a language in which the moneme designating a village would,
previously mentioned; English does, and this specification at the same time, indicate that the village in question is the
is treated as one of the elements of experience; now, we could portion of space where the experience is located. In such
imagine a language which would have a different moneme a language, village would mean 'in (the) village', and this, of
for 'a riot' and 'the riot', as if people said, for example, riot course, all by itself without any case ending, since any case
for 'a riot' and brawl for 'the riot' . This would, of course, be ending would have to be considered the linguistic equivalent
most uneconomical, and when people really care to dis- of another element of experience.
tinguish between definite and indefinite, they manage to The third aspect of our experience corresponds to yesterday.
procure articles which become elements of experience in their Were it not for literary and poetic forms such as yesteryear and
own right. In some languages, the simple mention of a riot the like, we might consider yesterday a single moneme since, in
may suffice, in the absence of any restrictive moneme, to that case, . . day, as the automatic accompaniment of yester-,
indicate the reality of the riot; in our hypothetical language, could not be counted as a separate choice. For our present
riot might mean not only 'a riot' but also 'there is a riot'. purposes, we shall take the liberty of disregarding yesteryear
English, just like a good many other languages, needs an and congeners and of treating yesterday as a minimal unit,just
actualizing phrase (there was) which we might be allowed as we would do with German gestern or French hier. We have
to consider a single moneme, were it not accompanied by thus, inyesterday, an exact parallel to what we have just been
a moneme indicative of time, in this case past time, This imagining when we thought of a moneme village with the sense
 44 TOWARDS A FUNCTIONAL SYNTAX TOWARDS A FUNCTIONAL SYNTAX 45
of 'place where' ; yesterday, here, is not the day preceding this function, i.e, in terms of experience, their relations to one
day, but that day as the segment of time in which something another, which prevents their being shifted at will without
was happening. changing or impairing the message.
We have been dividing our total experience into three The criterion of syntactic autonomy points to a threefold
aspects, as we called them, each of which was further analysed distinction among monemes: we have first monemes that
into monemes. From a semantic point of view, few people carry within themselves the indication of their own function
would object to such an analysis. But our semantic reactions and which we shall designate as autonomous monemes:
are, to a large extent, the reflex of formal distinctions. The French vite, hier, demain, dimanche in il viendra dimanche, are
formal justification of our initial break up is the fact that our autonomous monemes; in English there seem to be few
three segments there was a riot, in the village, yesterday are syn- clearly autonomous monemes of that type, but autonomous
tactically autonomous units: everyone of them can be used compounds such as last night, next week are at least as frequent
initially, medially, or finally, without any difference in their as in French, where we have hier soir, la semaine prochaine.
own meaning, although, of course, the choice of this or that Notice, in German, autonomous compound numerals such
order may imply som.e semantic difference for the utterance as neunzehnhundertneununddreissig with the meaning of 'in
as a whole. On the contrary, in normal contemporary Eng-
1939'·
lish, word order is fixed within those segments. The reason Next, we have monemes that do not imply any definite
for the syntactic freedom enjoyed by those phrases is not relation to the rest of the utterance and will therefore be
far to seek: in everyone of them we find an unambiguous available for several different functions. Of course, everyone
marker of its function, i.e. of its relation to the rest of the of these functions will have to be indicated somehow, either
utterance: there was marks the riot as the predicate, i.e, as the by position or by means of some additional element. These
element around which others gravitate and in relation to monemes could be called dependants; village is a dependant.
which their function will be marked; in marks the village as Last, we have monemes which secure autonomy for other
indicating the place where the riot occurred; yesterday, as monemes to which they are attached, by indicating their
such, is the indication ofwhen the riot took place. The relation- function, i.e, their relation to the rest of the utterance. The
ships between the three main elements of experience are thus combination of such a moneme with its dependants is an
precisely indicated, and there is no need here to rely on autonomous phrase. These we shall call functional monemes,
word order to tell the hearers what these relationships are. functional indicators, or just functionals. Functionals corre-
A language like Latin extended the practice of explicit func- spond to prepositions and conjunctions of traditional gram-
tion marking to situations where English, and western Euro- mar, but also to case endings. The reasons for which people
pean languages generally, make use of word order, namely are so reluctant to lump prepositions and case endings to.. .
in the indication of subject function and object function. gether are numerous: first, the former come before and the
In Latin, a nominal subject was no part of the predicative latter after the form they govern; second, prepositions can be
autonomous phrase, neither was any nominal object: pater separated from their substantives by various additions such as
oidet puerum is made up of three autonomous phrases; its an article and one or several adjectives, while case endings
English equivalent, thefather sees the child, is just one, since are permanently glued to the word they characterize; third,
the respective position of the segments is indicative of their there is normally one preposition per phrase, irrespective
 TOWARDS A FUNCTIONAL SYNTAX TOWARDS A FUNCTIONAL SYNTAX 47

of how many articles or adjectives are added, whereas languages, amalgamating processes have, at different periods,
case-endings are likely to be found after everyone of the addi- begun to blur the boundaries between the radical moneme
tional elements; fourth, in the case of prepositions, the func- ruk- 'hand' and following derivational monemes or case-
tional usually forms a clear-cut segment of the utterance, in endings; this has resulted in yielding different forms for the
contradistinction to what we find, for instance, in Latin case- radical: in Czech, for instance, ruk- in the nominative singu-
endings, where the indication of case, i.e. function, is form- lar ruka, rue- in the locative ruce, and rue- in the adjective
ally confused with that of a totally different type of moneme, ruini.
namely number, and where it is not always clear what Semantic amalgamation is common too, as when black-s-
belongs to the substantive moneme and what to the ending: mail becomes blackmail. In window, or its Danish equivalent
is the nominative ending of puppis, 'poop', -is or just -s as in vindue, originally wind eye, amalgarnation, both formal and
urbs? All four reasons should not be dismissed as sheer pre- semantic, has resulted in reducing two successive monemes
judice. But we should never allow them to blur the functional to one. When, as in a form like blackmail, the semantic amal-
identity of prepositions and cases. All of them are finally gamation is complete, since, synchronically, there is no hope
reducible to the same phenomenon: monemes, which, for of ever identifying the meaning of blackmail as the sum of the
some reason or other, are frequently or constantly in contact meanings of black and mail, it is certainly advisable to con-
and will tend to merge. The merging will be the more likely sider blackmail a single moneme, because, from a purely
and the more intimate if the element whose function is in- synchronic standpoint, the homonymy of blackmail and black-s-
dicated comes first, and the functional indicator last. This is mail is purely accidental. But when a formal amalgam is still
due to the fact that, in any language, the number of distinct identifiable as limited to a definite formal context, as is the
functions is very much smaller than that of elements capable case with Fr. au found only before consonants and not before
of performing them ; these elements, the so-called lexical vowels, or when the amalgamated forms combine monemes
items, are more informative and, accordingly, generally with perfectly distinct functional values-as with Latin case-
given a preferential treatment: they may be provided with endings where, for instance, nominative is a functional in-
an accent which gives them prominence, and their initial dicator, but plural is nothing but a rnodality ofthe noun-it is
phonemes are, as a rule, articulated with particular care imperative to preserve the distinction. We shall speak then
so as to facilitate their early identification in the flow of of two different monemes even if an analysis of the form into
speech. two successive segments proves arbitrary and, as in the case
A formal merging of two or more monemes, I call an of Fr. au, downright impossible. The Latin form homini con-
amalgam: he cut, as contrasted with he admitted, can be said veys three elements of experience: 'man', 'dative', and
to be an amalgam and to result from amalgamation. Fr. au, 'singular'. Should we insist on chopping it up into successive
in au marche, as contrasted with al' in a Thopital, is also an slices, we might rephonemicize it as /hominii/ and say that
amalgam. We may also, if we choose, call sang, the preterite 'man' corresponds to homin-, 'dative' to -i-, and 'singular' to
of the verb to sing, an amalgam, although it is not likely to -i. This analysis might be supported by a comparison with
have resulted from a process of amalgamation. This process the dative plural hominibus, where both homin- 'man' and
is not necessarily carried through so that it becomes im- -i- 'dative' would reappear, -bus being then an 'allomorph
possible to distinguish one signifiant from another: in Slavic of the plural morpheme'. But, of course, hemin- means
 TOWARDS A FUNCTIONAL SYNTAX 49
TOWARDS A FUNCTIONAL SYNTAX

'man' only in combination with a given type of endings, which, as Sapir says, I is 'something to talk about', and
and -i is the mark of the dative only in combination with another which is what 'must be said about this subject of
certain nouns in the singular. Therefore it is more accurate discourse'. Since, in English, the subject is never an autono-
and, of course, less arbitrary to say that homini means at mous phrase as, for instance, in Latin when a noun, the
the same time 'man', 'dative', and 'singular' than to try predicative phrase is always made up of at least a subject
to segment it. Linguistic articulation may be blurred, and and a predicative moneme, whose only possible function is
it is our duty to describe it as we find it: we should never the predicative function, and which we call a verb. The sub-
try to disentangle formal units which happen to overlap or be ject, defined as what necessarily accom.panies the predicate,
confused; yet we should never deny the existence of units, is one function of certain classes of monemes which, in the
as singular and plural, in Latin" whose existence is always wake of tradition, we could call nominal and pronominal
secured and attested by some formal difference, but in so monemes. The other functions assumed by these monemes
intricate a fashion as to defy analysis. are those of the so-called complements. In a language where
Our distinction of three types, autonomous monemes, the predicative moneme need not be actualized by means
dependants, and functionals, is based upon syntactic auto- of a nominal moneme endowed with a specific function, we
nomy. But this leaves out one type of moneme or segment should not speak of a subject. What we would be inclined to
which is independent rather than autonomous. Reverting to label 'subject' because it is rendered by a subject in a trans-
our former example, we would say that yesterday is an autono- lation, is nothing but one of the complements.
mous moneme, and in the village an autonomous phrase, but We may want to define 'function' as the linguistic counter-
there was a riot is not only syntactically autonomous, since part of the relationship between one element of experience
we can place it initially, medially, or finally at will, but and the whole of experience, so that we could speak of
also independent, since we can use it all by itself and obtain function in the case of anyone of the marginal elements, but
a complete utterance, which is not the case withyesterday, nor hardly in the case of the predicative core; the predicative
with in the village. In our hypothetical language, riot, which function would then be no real function. This, after all, is
should mean as much as 'there is a riot', could be used all by a matter of convention. But even if we decide to speak of
itself as a self-sufficient utterance. We would then say it is function only in reference to a definite type of relationship,
used with predicative function. The situation, in our English we should be ready to ascribe different functions to the pre-
example, is somewhat more complex. There was a riot is defi- dicate at least in those languages-English is one of them-
nitely a predicative phrase characterized as such by what we where speakers have at their disposal two or more forms of
have called its independence. We took the liberty, in what the predicate indicating different types of relationships with
precedes, of considering there was an actualizing phrase and the participants of the action; if opening is the action, and the
a riot the predicate, but others might prefer another kind of participants a gardener and a gate, I may choose to say
analysis and see in a riot the subject of the predicate there was. either thegardener opens the gate or thegate is opened by thegardener.
This view is supported by the observation that an analysis We will have to distinguish between an active function, and
in terms of subject and predicate seems to be universally a passive function, these functions being, of course, nothing
applicable in English. Every complete utterance, in that but our traditional 'voices'. It is not easy to tell what the
language, centres around a core of two monemes, one of I Language (New York, 1921), p. 126.
811928 E
5° TOWARDS A FUNCTIONAL SYNTAX TOWARDS A FUNCTIONAL SYNTAX

functional indicators are in opens and is opened; probably . .s in given positions, with a comparatively restricted number
for one function, is .... ed for the other. But, once again, we of other monemes. The average frequency of grammatical
should not let our analysis be too narrowly determined by monemes like of,for, with or Lat. 'genitive', 'dative', 'ablative'
the difficulties we may experience in analysing forms. is considerably higher than that of lexical monemes such
If we leave out, as a special type, the functions of the as man, rich, or eat. Functionals are grammatical monemes,
predicate, we should distinguish between primary functions Among primary dependents some ll1ay be lexical (nouns)
and non-primary functions. Primary functions are those of and some grammatical (pronouns). Some determinants are
elements which are directly connected with the predicate. lexical (an adjective like great), and others grammatical (an
In a sentence like yesterday, the head of the department dictated a adjective like my, the article the, or the 'plural' moneme),
four-page letter to the secretary he had Just engaged, the four Grammatical determinants can be designated as modifiers.
elements yesterday, the head of the department, a four-page letter, Among modifiers should also be listed such grammatical
and to a secretary he hadJust engaged have some primary func- primary dependents as are part of the predicative phrase.
tion; the function of thedepartment.four-page, he hadJust engaged These include modes, tenses, aspects, and persons in so far
is not primary, since they are not directly connected with the as their signifiant is not syntactically autonomous: in I'll do it,
predicate dictated. Within a so-called subordinate clause such I, 'll, and it are modifiers.
as he hadJust engaged, the same functional hierarchy obtains With the setting up of modifiers as a specific type of
as in the so-called main clause, but we should not speak of moneme neatly distinguished from another type of gram..
a predicate there, but of a 'predicatoid', and the functions of matical element we have called functionals, we are definitely
elements directly connected with it should be considered at breaking with tradition. Stressing, as we have done, the
best primary-like. syntactic autonomy of certain: units or phrases, is, of course,
Among dependent monemes-those that are neither inde- nothing but pointing out a feature that sets apart adverbs
pendent, autonomous, nor function indicators-one should and adverbial phrases in so far as they are used as verbal
distinguish the ones that assume some primary functions adjuncts. Labelling predicative monemes as 'independent'
from the ones whose function is not primary. The former does not go far beyond what generations of grammarians have
could be designated as primary dependents, and the latter as stated about the nature of the predicate. But dividing the
marginal dependents, or determinants; in the above sentence, mass of grammatical elements into two basically distinct types
head is a primary dependent, the (in the head) and department -the functionals and the modifiers, placing them at two far
are marginal dependents, the (in the department) is, of course, ends of our chart, and enforcing thereby a strict segregation
marginal too, but marginal to the already marginal department. -may well be felt as verging on scandaL We are so used to
Another possible distinction is that between grammatical listing as belonging together such categories as tense, aspect,
and lexical monemes. In order to distinguish between them, mood, and voice, that it may be shocking to hear someone
one should set up the inventories of the monemes which are maintain that whereas tense, aspect, and mood are closely
found in specific contexts, within autonomous phrases, where akin and belong together, voice is as different from them as
the respective position of elements is functionally relevant. different types of monemes can be. We are used to consider.. .
Lexical monemes are those which belong to non-limited ing gender, number, and case as the three pillars on which
inventories. Grammatical monemes are those that alternate, any decent nominal system is built, and we may be tempted
52 TOWARDS A FUNCTIONAL SYNTAX TOWARDS A FlTNCTION AL SYNT .A.X 53

to any that has no Cla."""~"",,,""Ta PVjlC't-t::>nr'~ reference to what the experience really is that he wants to
and that number is a local accident of limited importance, communicate. If: in some cases, the range of the choice hap-
whereas case belongs to the constitutive frame ofany utterance. pens to be limited, it will be so on account of the nature of the
Yet the distinction between functionals and modifiers is unit that is modified, as when a certain noun is never used
fundamental, and it has lately been noticed and pointed out in the plural, but not because of some limitation imposed
in various quarters. I If in a phrase such as with a smile, the by the structure of the sentence as a whole.
primary dependent smileis considered the centre of the phrase, In spite of such fundamental differences, functionals and
the grammatical determinant a is centripetal, the functional modifiers have so far generally been confused, and it is easy
with centrifugal: a is connected with the rest of the sentence to ~n?erstand why linguists who, in their overwhelming
only through smile, which it helps to specify; with connects majority, had been taught Latin grammar first, were
smile with the rest of the sentence, and since the connexion tempted, at best, to consider them two aspects of the same
is thereby established, the speaker is free to place the phrase linguistic reality. In Latin, as in all Indo-European languages
with a smile anywhere he pleases. Syntactic autonomy is thus that have preserved the old declensions, the expression of
the criterion which, in all cases, and particularly in formally case and that of number are hopelessly mixed under the form
complex ones, will prove the presence or the absence of a of case-endings that resist any formal analysis into successive
functional. In a context like the hunter was killing a bear with segments. In other words, amalgams of a functional and one
his spear, neither the nor a are function.als, since they do not modifier, or more than one, are practically the rule there,
grant hunter or bear any syntactic autonomy: exchanging the and must have been felt by generations of linguists to be a
place of the hunter and a bear will result in conveying a totally normal feature of any self-respecting language structure..
different experience; with, on the contrary, makes it possible As a matter of fact, it cannot be considered a strange quirk
to place the phrase with his spear practically anywhere with- of Indo-European, because amalgams of functionals and
out changing its relation to the rest of the sentence. Whether modifiers are by no means absent elsewhere. It is quite clear
this or that functional mayor may not be used is partially that the signifiants of both are likely to appear in close
determined by the communicative needs at that point of the succession in the utterance and are therefore constantly
utterance: after he is distributing tickets, I mayor may not exposed to amalgamation.
specify to whom. But it also depends on which predicative But the existence of such amalgams would not suffice to
moneme I choose, whether I mayor may not use a to-com- explain the traditional disregard of the difference between
plement. It is what is referred to when we say that a given centrifugal and centripetal grammatical elements: after all,
verb governs this or that case. This amounts to saying that, the existence of complex case-endings never prevented gram-
to a large extent, the choice of a functional is predetermined marians from distinguishing between case and number.. TIle
by that of the verb of the clauseo On the contrary, the choice main reason why the distinction was not made lay in the
of the modifier is free, i.e, the speaker is determined to use a or existence of concord. It was generally believed that the justi-
the, the plural or the singular, at a certain point by direct fication for the somewhat clumsy repetitions which charac-
I e.g. by Richard S. Pittman, starting with A Grammar of Tetelcingo (lvlorelos) terize this phenomenon was to be found in the way it
Nahuatl (Language Diss., No. 50) (Baltimore, 1954), pp. 6-8, and Georges allowed the speaker to connect the successive elements of the
Gougenheim, 'Morphologie et fonctions grammaticales', Journal de psychologic
(1959), pp. 4 17-26 . utterance. This was professed even by a scholar like Otto
 TOWARDS A FUNCTIONAL SYNTAX TOWARDS A FUNCTIONAL SYNTAX 55
54
had patience with roundabout ways of would readily bear witness . Everything would be simpler
expressing oneself and denounced concord as a most awkward if the nominative case were always unambiguously distin-
approach to sentence building. Now, concord which amounts guished from the other cases. There then never be
to using several discontinuous segments for one and the any need to resort to the mark of the plural agreement to
same signifie is attested for functionals and modifiers alike. indicate which noun is the subject. The intricacies of Latin
This circumstance was, of course, one more reason for identi- are no proof that the distinction between functionals and
fying them. But what is really basic here is the fact that, modifiers is not fundamental,
thereby, modifiers would seem to be made to function as Concord is redundancy, and contrary to what could be
connective elements, a role which we have been denying expected, redundancy results, as a rule, from least effort:
them and entrusting exclusively to functionals, The answer people do not mind repeating if mental effort is thereby
to this is that, if what is normally just a modifier, redundantly reduced; if adjectives are quite freely and frequently used as
and discontinuously expressed in different positions through nouns, as was the case in older Indo-European languages,
the utterance, happens to act as the sole indicator of a it will be indispensable for them to carry the mark of their
function, it has to be interpreted, when this is the case, as function if nouns do; therefore a word likefortis, 'courageous'
an amalgam combining modification of the accompanying or 'courageous fellow' is inflected just like civis, 'citizen'. If I
monemes and indication of their relationship to the rest of the mean something like 'the citizen, the courageous one', there
sentence: in a Latin context like viri vident, viri contains is some justification in presenting the mark of their function
a modifier, the 'plural' moneme, and a case, the nominative twice, since the two words are more or less equals. But when
which marks its subject function; in vident the final -nt is fortis is really nothing but an attribute of civis, it would be
exclusively a part of the signifiant of the 'plural' moneme we more logical either to mark it as such, or to use the bare
have just identified in viri. It is worth remarking that the stem, letting its vicinity to civis indicate its relation to it. But
amalgam of the two monemes 'plural' and 'nominative' is it is certainly much easier to let the adjectives keep in all
only partial, since it could be said that the former's signifiant cases inflected forms that have had to be memorized any..
is -i ... -nt, while the latter's is just -i. But if our subject way. In popular French the equivalent of lnyfather says is mon
belongs to a type where the nominative ending is identical pere it dit (pronounced [idi]) instead of the traditional mon
with the accusative ending, if, for example, our context is pere dit; since I have to use [idi] in reference to my father's
homines animal vident, the -nt of vident becomes instrumental in saying something when I need not specify that the speaker
identifying homines as a nominative, since it indicates that the is my father, why shouldn't I save myself the trouble of
subject is in the plural and, in this context, points to homines choosing between [di] and [idi] depending on whether I
as the only form which can be interpreted as a plural subject. expressly mention my father or not? Since I can't help using
Thus, owing to the ambiguity of the nominal ending, the -nt [idi] at times, it is handier to stick to it, whatever the context,
of the accompanying verb may be made to play the role of the than to reckon every time with the context; pronouncing an
elements of a discontinuous moneme indicative of the sub- extra phoneme is nothing in comparison with the output of
ject function of the neighbouring plural noun. All this may mental energy required by the choice between [di] and [idi].
sound rather tricky, but the facts themselves are so, not our There are situations where it is not usual to speak of con-
interpretation, and this is something to which any schoolboy cord or agreement, but in which speakers also have recourse
 TOWARDS A FUNCTIONAL SYNTAX TOWARDS A FUNCTIONAL SYNTAX 57
to redundancy for obvious reasons of economy: in our and structurally similar languages. This might lead readers
former Yesterday, there was a riot, the notion of 'past' is to believe that our analysis is based upon the observation of
expressed once precisely, inyesterday, and then rather a certain type of language and, that it cannot
vaguely, in the preterite was. Now, it may be useful to have make any claim to universal validity. But this assumption
at one's disposal, together with a set of references to some would not be correct: starting from our definition of language
definite periods of the past, such as yesterday or last week, as doubly articulated, we have actually been proceeding in
some means of labelling the experience as a past one without a purely deductive way, and the existing forms we have been
any further specification. Of course, if we start with a precise quoting and even operating with were never meant as sup-
reference such as yesterday, there is no need to add a vague port for the theory, but simply as illustrations ofthe categories
reference to the past as the one included in was. The lazy we posited as necessary for the functioning of linguistic com-
solution, however, consists in letting the context, yesterday, munication. What has led us to stress the importance of
determine the choice of the tense. In similar cases, one might syntactic autonomy is the realization that this autonomy is
easily be tempted to characterize the tense moneme of was the test that a segment of an utterance corresponding to a
as relational, since it sluggishly- establishes some connexion given element of the experience contains all that is needed for
between different parts of the utterance. In a similar way, it marking its relation to the rest. It is a guarantee that the
could be argued that such a modifier as the definite article hearers will be in a position to reverse the process through
implies some relation with what precedes, since it normally which the speaker breaks down his previously unanalysed
labels its noun as something previously mentioned. Such experience into a number of elements for which the language
facts are what Sapir probably had ill mind when he labelled he uses offers equivalents. It would seem that there exist only
all grammatical items as relational. I three ways for the speaker to indicate the relation of a seg-
It has now become clear that redundancy, grammatical- ment to the whole:
ized or just lexical, is a basic feature of human communica- First, the linguistic equivalents of the elements of experi..
tion, and it may in particular instances be resorted to in ence may be connected by means of units, in all respects
order to indicate the function of some segment. But, once the similar to them, which we might consider new elements of
uses of the respective positions of monemes have been dis- experience in their own right: in John's hat, 's is but a handy
counted, only such elements as secure syntactic autonomy way of expressing 'belonging to' which is part of the experi-
are to be considered functional indicators. Whether this ence just like 'John' or 'hat' ; in the same way as some elements
syntactic autonomy is frequently or exceptionally made use of experience may be represented by prosodic features, the
of is immaterial. Even if it could only be shown by means of marking of relations may be secured by means of an intona-
some unidiomatic syntactic shift, it still would have to be tional or accentual feature"
deemed valid, if it were proved thereby that the identifica- Second, the relation between two elements may be ex-
tion of functions is not affected. pressed by means of the respective positions of their linguistic
equivalents in the speech continuum: in many languages,
Most of the preceding examples were borrowed from a subject is marked as such because of its position before the
English, with a sprinkling from other genetically connected predicate; in John's hat the respective position of John and
lOp. cit., pp. 86-92. 's marks John as the owner"
 TOWARDS A FUNCTIONAL SYNTAX TOWARDS A FUNCTIONAL SYNTAX 59
Third, the relation of an element to the experience as a fire!, not to speak of monomonematic or polymonematic in-
whole may be included in the 'meaning' of its linguistic junctions such as go!do! getaway! give himashilling! Traditional
equivalent: yesterday corresponds to an element of experience grammarians do not any difficulty in such matters
whose relationship to the whole is never in doubt. But, beside because it is obvious for them that, once injunctions are set
such perfectly autonomous segments as yesterday, we find aside, only subject-predicate cores can be dubbed 'normal'.
cases where the 'meaning' of the moneme may, in certain But this is precisely what we do not accept, because we want
contexts, imply a given function, as when Russian stol 'table', to consider the possibility of languages using utterances
as an inanimate masculine, will tend to be considered a made of one (predicative) moneme in exactly the same situa-
grammatical object, in the absence of any specific indication tions where other languages, such as English, cannot dispense
of that function. with the complex subject-predicate construction.
Proceeding in a strictly deductive way, it is our duty to It is by .rcference to our distinction between what is
determine as we have just tried to do, what possibilities are properly linguistic and what is marginally SOl that we may
afforded by the linear form of speech for the linguistic ex- hope to give a sound foundation to the concept of 'normal
pression of the various functions corresponding to the relations utterance'. What we consider properly linguistic is what
assumed to exist among the elements of experience. But we is achieved, in matters of communication, by means of the
have no right to posit the existence of relational universals: double articulation pattern: double articulation is what pro-
offhand, we may be inclined to believe that the with-type of tects the linguistic frame against interference from outside,
relation, or with-function, exists in all languages ; but observa- what makes it really independent and self-contained. But
tion reveals that, in many languages, the equivalents of do it .communication by means of double articulation is an expen-
with a hammer and he came with a friend make use of different sive procedure which man will tend to avoid when his needs
functionals; even in a language like French, wherefais-le avec can be satisfied through the use of simpler, more direct
un marteau and il est venu avec un ami show the same equivalent means, such as gestures, by themselves or supplemented by
of with, there are so many specific contexts where avec does speech. Another, very effective, way of reducing the output
not correspond to with that it would not be commendable to of energy involved in communication is to rely on the situa-
equate their respective functions. Even on a more funda- tion in which the interlocutors are placed: very nice! pooh-pooh!
mental level, we should be wary of following Sapir when he no! make excellent sense all by themselves among people
presents the subject-predicate pair as the necessary basis of who look at the same thing or witness the same event. This
linguistic communication. I Here again, we have to envisage reliance on situation is so general that all languages have
various possibilities and try to imagine what could constitute developed several classes of monemes whose interpretation
a minimal utterance, and how such an utterance could be is always dependent on situation. Such are demonstratives
further expanded. like this, that, except when used in reference to context, time
The first problem that arises in connexion with the minimal references like now,yesterday, today, last night or the 'preterite'
utterance is whether we should distinguish between normal moneme, and personal pronouns like I andyou. The situation
utterances, the ones which, in English, involve a subject and generally makes it so obvious who the second person subject
a predicate, and curtailed ones such as over there! the scoundrel! of imperatives is, that its expression is the exception rather
lOp. cit., p. 126. I See above, pp. 28-29.
60 TOvV ARDS A FUNCTIONAL SYN1'AX TOWARDS A FUNCTION AL SYNTAX 61
than the rule. All these economical tricks are very welcome but the 'here and now', corresponding to no linguistic units,
in linguistic practice, but they undoubtedly detract from are no parts of the communication. Still it could be argued
the ideal of human communication, is self-sufficiency. that and now' is in the of the
This ideal finds its expression in out-of-situation uses of moneme for 'rain' unless specifically excluded by the addi-
language as realized, e.g. in gossip, where the reference is to tion of some complement like 'yesterday' or 'beyond the
absent people, in narratives generally, and in literature, hill' . Yet, there is no doubt that languages in which the
where use is seldom made of the actual situation, namely normal way of saying it is raining or there is afox is by means
the author at his typewriter and the reader with his printed of a single moneme 'fox', deserve not to
pages and, in between, the long-drawn processes of editing be classed indiscriminately .with languages in which this is
and composition. It is true that authors do create situations not possible. The more so if it can be shown that this lTIOnO-
where their characters are found to use, with perfect rele- monematic utterance may be normally expanded by means
vancy, such terms as I, you, today, or this week. But these of various complements, whereby it is revealed as the
situations are actually contexts, and therefore a purely potential core of an unlimited syntactic complex '[here
linguistic accompaniment. comes a] fox' > '[there came a] fox, last year, that was killed'.
It . is in reference to out-of-situation uses of language that Even if the minimal utterance in a language cannot be
normal syntax can be defined : normal syntax is that which unambiguously shown to coincide with a single monerne, one
is attested in such uses, i.e. when communication is achieved should not jump to the conclusion that it must necessarily
by purely linguistic means. This should, of course, not be be one of the subject-predicate type. It is easy to understand
construed as excluding from normal syntax such segments as why so many languages have made it a rule never to use one
I tellyou: it is clear that I tellyou, as a syntactic pattern, is moneme by itself: even when centring his attention on the
identical with they tell them, which Inay contain no reference existence of a single being, thing, or process, a speaker will
to situation; in other words, I tell you is normal syntax, in normally not be satisfied with the mere mention of that item,
spite of the fact that this segment contains references to but will be inclined to locate it in time or space, or to connect
situation, since it conforms to patterns attested for segments it with himself or his interlocutors. These additions are
that do not contain SUCll references. often conceived as actualizers, which seems to imply that the
This situation criterion is certainly valid for determining moneme by itselfis an abstraction whose anchoring in reality
what could be called syntactic normalcy in a language, like can only be achieved by means of some element endowed
English, where it agrees with the consensus of generations of with just that function. This view is supported by the nature
grammarians that syntactically normal utterances contain of the subject, which is a moneme, likely to be found else-
a subject-predicate phrase. application to languages that where in a variety of complemental functions, used here as
we suspect do not demand a bimonematic core may not be the necessary accompaniment of a predicate, with a function
quite as decisive: there are many languages where a moneme perfectly characterized either by some privileged position
meaning 'rain' (with no possibility of deciding whether verb as in English, or some functional mark as in Latin. But the
or noun) is used by itself when the English say it is raining. fact that some languages make the actualization of the predi-
In such an utterance, there is no linguistic reference to any cate a definite function, does not imply that actualization is,
situation; it usually indicates that it is raining here and now, in principle, more than one aspect of the basic linguistic
62 TOW ARDS A FUNCTION AL SYNT AX TOWARDS A FUNCTION AL SYNTAX 63

process according to which communication can be made languages, English among them, subject function can be
more specific by means of additional elements. A distinction ascribed to such monemes (nominals) as are found elsewhere
may be made between optional specification and compul- to assume the functions of complements ; in some, such as
sory addition, as of a subject, for which the term 'actualiza- Malagasy, monemes with subject function must be marked as
tion' could by convention be reserved. But it should be kept previously known or mentioned, which shows them to be
in mind thata compulsory subject does not really actualize informationally marginaL The possibility that some languages
a predicate more than an optional complement would. do 110t clearly distinguish between two successive statements
There may be languages in which one-moneme utterances +
and the succession subject predicate cannot be ruled out.
cannot be considered normal out-of-situation syntax, but in In languages where actualization is needed but can be
which the predicate can be actualized by means of any com- achieved by means of any complement, the predicate is
plement, the actualizing function being added to any other obviously the one moneme which is not marked as perform-
function the complement or complements may assume. ing any complement function.
Traditionally, the term 'predicate' is defined in reference All this, and what could be added about the different
to the subject-predicate complex and would seem to designate levels of complementation, amounts to establishing a hier-
everything in the clause that is not the subject, or some archy of syntactic functions. This hierarchy is, no doubt, set
dependent of the subject. Besides, 'predicate' implies some up with a view to differences between languages, but it
assertion, so that a question or an order would not contain should be clear that, even if it were carried through, it would
any predicate. If the term is to be retained by contemporary never account for all the varieties of linguistic structure.
linguistics, we shall want to use it in reference to mono- It is quite essential to know all the different functions that
monematic segments which, by themselves, may constitute characterize a language, but it is equally important to deter-
a complete, out-of-situation utterance, and also to the same mine for each language, what monemes are qualified to per-
segments when accompanied by various expansions (com- form this or that function. No language is known to allow
plements), but independently of them. Within the subject- everyone ofits significant units to perform all of the functions
predicate complex, 'predicate' should, in a similar fashion, it provides. It is even difficult to imagine how such a language
apply to any segment that is, jointly with the subject, con- would work. II: in such a language, functions should be
stitutive of the minimal utterance, thus excluding from it its marked by means of functional indicators, these, being signi-
various complements. ficant units in their own right, should be able to assume the
A subject is different from a complement only because it is role of functioning elements, and, conversely, all functioning
constitutive of the minimal utterance. Therefore, whenever elements would also have the role of functional indicators.
we speak of a subject we are referring to a linguistic situation But how could speakers make clear that a given moneme
in which subject and predicate are both equally indispens- is used as a functioning element here, and a functional
able, since our criterion is indispensability, and the problem indicator there, ifonly the respective position of the monemes
arises of how we can tell what is the one and what is the in the speech continuum could be relied upon? Many
other. If our terminology makes any sense, the subject should languages are recorded in which the same moneme is used
be the one which should somehow stand closer to the either as a functioning element with the meaning of 'to give'
marginal and optional elements of the utterance: in many or as a functional indicator with a dative value. But in a
64 TOW A R D S A FUN C T ION A L S Y N T A X TOWARDS A FUNCTIONAL SYNTAX 65

language like Vietnamese, where this is the case, not all and function: small, for instance, implies not only 'smallness',
'verbs' function as 'prepositions', nor vice versa, and this but also the function of determination, as in a small car, and
determines significant contexts that enable the hearer to also in he is small if we agree to locate the predication is.
identify 'give' as a predicate or as a dative function marker. It is even more frequent to find that a class of monemes
There are languages, and Vietnamese is again a case in point, (verbs) is restricted to predicative function, although this
in which position plays a great role in function marking; the does not necessarily imply that only verbs can be predicates.
main function here can be labelled 'determination', and this Contrary to what was generally assumed, this is a domain
is shown by postposition of the determinant; what looks like where languages are found to vary and which, in consequence,
a subject-predicate relationship might possibly be inter- should playa great role in typological matters.
preted as a case of determination. But if the core A receives
two determinants, Band C, how can the hearer know that,
in the succession C is not a determinant of the im-
mediately preceding B, but one of concomitantly with B?
Some marker, segmental or prosodic, will be needed, and
there goes our monematic omnivalence!
A language in which all monemes would be of the yesterday
type, i.e, syntactically autonomous, because the indication of
the function is part of the meaning of the term, would be so
uneconomical that we may as well rule it out as a practical
impossibility. Function being, in such a hypothetical case,
diluted in lexical meaning, would, of course, cease to exist as
such.
We are thus induced to accept, at least as a pragmatic
assumption, the view that there exists in all languages some
distinction between monemes as regards the extent to which
they may assume the various existing functions. In no
language are all monemes used indiscriminately as function-
endowed and function-marking. In other words, there is no
language without grammar. But once unambiguous function-
marking is secured, there is no universally valid reason why
any moneme, except one that is specifically a function-
marker, should be excluded from any function, whether
predicative or non-predicative, Still, specialization is very
widespread. Many languages have, for example, a class of
adjectival monemes more or less restricted to certain specific
functions and which, accordingly, tend to identify meaning
81192B F
 LINGUISTIC TYPOLOGY

and
""rf~"'TY'.... 't-,,,,.....

III still must be considered the necessary basis of reference for

all typological attempts, we may well understand why so
many linguists remained sceptical and why it has never been
LINGUIS r o TYP L used.
Our fundamental objection to previous attempts at classi-

W
H A T contemporary linguists somewhat pompously
fication is, of course, that they were bound to be random
call 'typology' is not basically different from what
and intuitive, since so little was really known about the
a long line of thinkers have attempted to do when
languages which we so glibly labelled isolating, agglutinative,
they classified languages, not according to their antecedents
or inflexional. If many linguists are convinced that it is high
and genealogy, but with respect to their directly observ~ble
time to tackle typological problems, it is because they think
characteristics. It is not difficult to understand why so Iittle
that we are in a position to give a generally valid description
has been achieved along these lines to date: for more than
of any language, one which is not biased by the background
a century, practically all responsible and reputable linguists
and previous experience of the describer. I Furthermore,
were engaged in looking for traces of linguistic kinship, i.e.
they deem that we have at our disposal reliable descriptions
for evidence pointing to the fact that different languages were
of a good many languages, and, last but not least, that we
divergent forms of one and the same older form of speech.
know exactly what we can expect from a language, because
We need not go too far back in time to find and, as far as
we ourselves have decided what we want to call a language.
I am concerned, to remember a period when attempts to
This does not entail, by any means, that reliable descrip-
classify languages on any other basis than genetic relation-
tions will give us the clue to the problem of establishing a
ship were frowned upon as a sheer waste of time and energy.
typology which will be found acceptable by anybody but the
Let me only mention the fact that the second edition of Les
one who sets it up. If so little has been done about typology
Langues du monde.' published as recently as 1952, has retained
during the last decades, it is because no one knows how to
genealogy as the well-nigh exclusive principle of classifica-
establish a hierarchy among the various items isolated by
tion. This, no doubt an extreme case, obviously results from
linguistic analysis. Sapir, it is true, taught us that we should
lack of information, on the part of the surviving editor,
not try to reduce the complexities of a language to a single
concerning the mid-century status of descriptive linguistics.
label.> But if we try to do justice to all the features that seem
But the well-known reluctance of Meillet for any non-genetic
to deserve mention, we may end up, not with what we ex-
comparison was perfectly justified at a time when there was
pected to obtain, namely a number of classes among which
no proof yet that linguistic description could be performed
with as rigorous or more rigorous methods than the ones pp. 86-126. J. Greenberg in 'A Quantitative Approach to the Morphological
available for genetic comparison. Although Sapir's typology," Typology of Language', I]AL, xxvi (1960), pp. 178-94, translates Sapir's
scheme into currently fashionable jargon.
I In this respect, Charles Bazell, Linguistic Typology, An Inaugural Lecture
I The first edition, published in 1924 in Paris, was entitled Les Langues du

monde, par un groupe de linguistes, sousla dire~tion de A. M~illet et ~arcel Cohen. Delivered on 26 February I958 (London, 1958), seems unduly pessimistic, parti-
2 As Chapter VI ('Types of Linguistic Structures) of hIS ,?ook Langua?e
cularly in connexion with the phonological level (p. 5); actual descriptions,
(New York, 1921), pp. 127-56~ but it.is based upon the analy'sls presented ~n as distinct from programmatic illustrations of a descriptive method at the
Chapter V of the same book ( Form In Language; Grammatical Concepts), expense of a given language, manage to present traits such as everyone con-
cerned would agree to consider relevant. 2 Op. cit., p. 128.
68 LINGUISTIC TYPOLOGY LINGUISTIC TYPOLOGY 69
could be distributed, but with a long
1..,.",nr"1"1..,.nrt:l.('l and its grammatical structure. what is meant thereby is
list of criteria which, when applied to a language, would that a change on one of these two planes is likely to have
probably set it apart from all others. In other words, what we repercussions on the there is ample proof that such
could achieve would be a method for trimming our descrip- a solidarity exists: the connexion between umlaut and Ger-
tions to their essentials rather than a handy tool for rapid manic morphology is obvious, and umlaut as a phonological
orientation. shift has resulted in umlaut as a morphological device as
Linguistic analysis yields, for each language, a most com- used, for instance, for the forming of new plurals. Yet, if it
plex picture with a list of phonemes, a prosodical pattern, is true that any phonological change may involve some
a set of morphological alternations, a grammar which some morphological change, it is also true that the nature of the
would still divide into a morphology and a syntax, and a phonological change will not determine the nature of the
lexicon, a kind of rambling and unstable structure with which corresponding morphological change: the confusion of e and
linguists do not know too well how to operate. The ideal a may, in a given language, entail the confusion of the indica-
solution would be to reduce all these aspects to one spectrum. tive and the subjunctive; but this is no indication that there
In order to achieve this, we might be willing to sacrifice should be a natural and permanent connexion between the
certain features felt to be of secondary importance. But, preservation of e and a as distinct phonemes and the reten-
even so, it is doubtful whether we could ever succeed: tion of a distinction between the indicative and the sub-
linguists have lately been repeating that a language is a junctive. On a strictly synchronic level, it is likely that, while
structure, or, maybe, a structure of structures, and if this is the morphology and the lexicon of a language require the
true, we should expect to find inner connexions extending existence of a number of phonological distinctions, what
from one end of the complex to the other end. This would be those distinctions actually are is of little or no importance.
true if a language were one of those tools or machines that This amounts to saying that it is hard to imagine how a
work with perfect accuracy and without any appreciable: linguist who knows the phon.ological pattern of a language
delay in transmissions. But this is not the case. As we shall could guess what the main features of its grammar are or,
see later, I every language retains features which result from its vice versa, how, having been told what the grammatical
functioning several millennia ago: French il est, ils sont, German categories of a language are, one could tell what its phonemes
er ist, sie sind, preserve to this day an alternation determined, must look like . It is clear that if the latter were the case we
in Proto-Indo-European, by the accentual pattern that must need not worry about the role phonology should play in our
have characterized the verbal inflexions at a remote stage of typology, since the essential of it should be deducible from
that language. Redundancy, a normal and necessary feature a consideration of the morpho-syntactic categories. In fact,
of all language, constantly acts as a buffer, retarding reac- however, phonology will have to be represented in a linguistic
tions to such an extent that it may take thousands of years for typology, unless we decide that what it has to offer has no
all the implications of one push to spread to all parts of the bearing upon what our typology is meant to elucidate.
structure. It is a fact that those who have tried to classify languages
The question has often been raised, whether there was any so far have not attempted to make use of phonological
solidarity between the phonological pattern of a language criteria, unless they were professional phonologists who, in
I See below, p. 138. their turn, were not trying to go beyond phonematic or
 LINGUISTIC TYPOLOGY LINGUISTIC TYPOLOGY 71
prosodic patterns.. Traditionally, classifiers concentrate on research with no other aim than the expansion of knowledge..
morphological traits: they probably believe a linguistic Still, we cannot work here without criteria for determining
classification should first and foremost at grouping the is what is essential. These criteria
languages of people who share the same outlook.. Now, they cannot be chosen arbitrarily if we want to reach agreement
deem the phonology of a language to a matter of chance, among linguists; they are likely to be generally accepted only
and the lexicon to reflect the world as it But its morpho- if the final result opens new vistas and reveals new fields for
logy, which ultimately governs its syntax, is supposed to research. This means that we should try to determine the
mirror the psychic activity of its users, When scholars possible reasons why two languages which are not genetically
distinguished among isolating, agglutinative, and inflective related, i.e, derived from one and the same language and
languages, they certainly believed they were transcending therefore likely to preserve substantial likenesses, should
mere linguistic reality and reaching deeper into the psyches belong to the same type.
of speakers.. Some of them-e-but should we call them 'SCllO- When we come across two languages which present striking
lars' ?-even went so far as to connect isolation, agglutina- resemblances and which, for some good reason, we do not
tion, and inflexion with some fundamental, permanent, and believe to be genetically related, we are inclined to assume
inherited features of different strains of men, as when one of a process of convergence determined either by protracted
them declared that it was a crime for an inflecting woman to contacts between two communities or by some common
marry an agglutinating man.' substratum. When we have to deal with a number of lan-
All this trend of reasoning is, of course, hopelessly out- guages which we suspect to have converged, it may prove
dated, less because of a weakening of racial prejudice useful to operate with the sort of concentrated characteriza-
among scientists, than on account of the widespread con- tion which we call a typology. It: for instance, we consider
viction that language is more than the reflection of the world the languages spoken in the northern half of Asia, which
in the mind of man.. There is something we have a right to certainly belong to different stocks, we may imagine how
call linguistic reality which combines signifie and signifiant handy a reliable way of characterizing languages would
and which is distinct from both the phonic and the mental prove for those who try to unravel the linguistic history of
realities. Language is an institution. A language is a set of that part of the world through an evaluation of the duration
habits that the child-or the adult in the case of a second of convergence. or divergence. If this is the kind of use we
language-acquires by imitation of those who surround have in mind when considering a linguistic typology, it is
him. If we want a linguistic typology today, it is certainly clear that phonological characteristics art likely to be just
not with a view to establishing a hierarchy of human races as informative and essential as any of those selected from
or discovering distinct varieties of human minds. grammar or lexicon: in a non-genetically orientated com-
This leads to the question: for what purpose do we want a parison of the languages of South Africa, the clicks of Zulu
linguistic typology? It would probably be more truly scienti- should playa major role.
fic not to consider ulterior motives. At the present time, It is, however, frequently assumed that convergence is
as applied linguistics gains more and more importance, it likely to breed purely outward resemblances, the ones that
is imperative to insist on the necessity of promoting free result from direct imitation of some sounds, or from borrow-
I Quoted from Sapir, op. cit., p. 131, n. 2. ing of lexioal elements or loose grammatical items, which,
 LINGUISTIC TYPOLOGY LINGUISTIC TYPOLOGY 73
since they are not welded to radicals, are easiest to isolate tive bulwark against random distortion.. t But communica-
and to transfer from one language to another. There is some tion begins with a first analysis of experience in the frame of
truth this, although I am of the opinion that there is no the morpho-syntactic of language, and the way
limitation to the extent to which two languages can con- this first analysis is performed seems far more fundamental
verge. Convergence will show in trimmings before it mani- than the materials which are made use of for its transmission
fests itself in fundamentals. Therefore, if similarity is found to other members of the community. Ifcarried through to its
exclusively in the structural cores of the two languages, we ultimate consequences, this statement implies that not only
may be induced, when genetic relationship is ruled out, to phonemes, but also the vocal form of meaningful units
reject as an explanation the sort of convergence that results should, in a linguistic typology, be subordinated to the
from contacts. We may prefer to work with a theory which articulations of experience considered apart from the signi-
has division of labour and increasing social complexity as the fiants which are needed for their transmission: what is essen-
.JI1Clin factors of linguistic evolution! so that, if by any chance tial about the plural of English is not the choice of the
their rhythm happens to be the same in two distinct com- [z] phoneme for its most usual rendering; it is not the fact
munities, linguistic evolution may follow very much the that [z] alternates with lsI and /IZ/ in certain phonic situa-
same tracks in both. We might, for instance, assume that tions; it is not either its being eked out by /dn/ in oxen, a
on a certain cultural and economic level we stand a good vocalic alternation in men, zero in sheep or deer, all this being
chance of coming across what has been called the ergative buttressed by verbal agreement with the subject of the clause.
construction." On another level, which we might consider These circumstances can all be lumped together as formal
more advanced, the ergative construction will probably have accidents. No doubt, they may ultimately be instrumental
given way to an active-passive verbal opposition. in shaping the fate of the plural category. But, from a
Now, ifsuch a theory should be found to tally with at least synchronic standpoint, they are just side issues in comparison
part of the observed facts, it would entail a typology based with the basic fact that English, unlike some other languages,
upon such linguistic features as are likely to be most directly distinguishes, for certain units, between a singular and a
affected by social and economic changes: these obviously plural. This is an essential point to which we shall revert
include lexical items, but also the elements that I would be below, because even the best among our predecessors have
tempted to call the joints, and which are the essential parts generally missed it and concentrated on formal accidents
of grammar.. This would certainly strengthen the traditional rather than the articulation of experience. Yet even if I
view according to which a non-genetic classification of disagree with the narrowly formal approach which has been
languages should be based upon morpho-syntactic traits. the rule in non-genetic linguistic classifications, I do not
Even if we grant that the incentives to linguistic changes are reject the accidents of vocal forms from linguistic typology..
to be found among human communicative needs, still changes Nor, for that matter, should phonological typology be dis-
will not take place irrespective of the existing vocal forms, sociated from one based upon other aspects of linguistic
and the distinctive units of the language will prove an effec- reality. Some day, linguists may raise again the question of
whether a single spectrum can be found for characterizing
l[ See below, pp. 136- 8. a language as a whole, But we shall have to be content with
2 Cf. A. Martinet, 'La Construction ergative et les structures elementaires
de I'enonce', Journal depsychologic normale et pathologique, 1958, pp. 377-9 2 • I Cf. above, pp. 24-2 5.
74 LINGUISTIC TYPOLOGY LINGUISTIC rrYPOLOGY 75
separate treatments, As Charles Bazell has aptly stated, opposed as fronted to non-fronted and in which, in the vocalic
'structural classification must start from small systems and pattern, front is not opposed to back, but rounded to non-
not from languages as wholes, since the postulate of soli- lui is ([u]), but some-
darity remains to be proved'. I times front ( [y] ), and the Iii phoneme is either front ( [i] ), or
articulated farther back ( [1] ).
Phonological typology was one of the central preoccupa- This type of connexion, because of its exceptional nature,
tions of the so-called Prague School in its early days. Among only underlines the normal autonomy of the two phonematic
the important contributions of Trubetzkoy to the Travaux du patterns. if it were shown that the efforts of some
Cerele linguistique de Prague, we find two classificatory surveys, linguists 1 to identify, on the plane ofimpressiOllistic acoustics,
one for vocalic, the other for consonantal systems." It is, the distinctive features of vowels and consonants, do not
indeed, symptomatic of the difficulties inherent in the estab- ultimately arise from an a priori binaristic conception of
lishment of a linguistic typology that, even in the neatly language systems, and are supported by sufficient evidence,
circumscribed field of phonematics, we should have to set it would not indicate, by any means, that what is good for the
up two distinct systems for vowels and consonants. There may vowels is either good or bad for the consonants and vice
seem to exist some inverse ratio between the number of versa: should it be true, in one sense or another, that [a] is to
vowels and that of consonants in a given language: in a [k] what [u] is to [p] and [i] is to [t], this would not mean
language with one or two vowel phonemes such as the ones that a language that presents la/ as a phoneme should either
we find in the north-western Caucasus," we expect the number definitely exclude Ikl from its consonantal inventory or
of consonant phonemes to exceed 50, whereas the type of necessarily have it.
French I use has barely 18 consonants to match its 16 If, on the paradigmatic plane, little is to be gained from
vowels. Yet American Spanish, with 17 consonants, gets trying to connect vocalic and consonantal systems, an attempt
along with only 5 vocalic phonemes, and Polynesian has very to characterize languages by reference to the way they
short lists for both types." Consequently, languages cannot balance vowels and consonants in the speech continuum
be labelled either 'consonantal' or 'vocalic'. Besides, where- may prove more successful: as is well known, some languages,
ever we seem to discover some connexion between two Polynesian, for instance, never use a consonant without
patterns of consonants and vowels, it is hardly anything tagging a vowel after it (type evcv), whereas others are not
but numerical. It is only when consonants are found to make averse to consonant clusters both before and after vowels
use of distinctive features usually restricted to vowels that we (cf. eeevee in English strict, eecveee in German sprichst). Some
may observe some definite influence ofone system on the other, linguistic mediums (Old Church Slavic is a good example)
since we usually notice that the vowel pattern ofthe language present consonant clusters before vowels, but end all their
then resorts to other features: this is true, for instance, in syllables in a vowel, and they could be characterized as
a language like Russian in which many consonants are open-syllable languages. Modern French has a way of
Proceedings of the 6th International Congress of Linguists (Paris, 1949), p. 116.
I

2 TCLP, i (1929), pp. 39-6 7; iv (1931), pp. 96-116.

restricting most of its clusters to ee by inserting a generally
3 Cf. W. S. Allen, 'Structure and System of the Abaza Verbal Complex', non-distinctive [~] wherever it is needed to split a larger
TPS, 1956, p. 129.
4 Cf. A. Haudricourt, 'Richesse en phonemes et richesse en locuteurs', I Cf. R. Jakobson, C. Fant, and :LVI:. Halle, Preliminaries to Speech Anafysis

L'Homme, i (1961), PP' 5-10. (Cambridge, Mass., 1952 ) , pp. 33-34.

 LINGUISTIC TYPOLOGY LINGUISTIC TYPOLOGY 77
cluster of consonants ;' a phrase like Je me le demande (phone- order to be considered normal, the phonological pat-
matically /3mldmad/, i.e, eeeeeve) is, phonetically, [3~mlgd­ tern of a language should include two series of stops, affri-
mad] or [3ffigldgmad], i.e, [cvccvccvc] or [ccvccvcvc]. cates, and fricatives opposed as voiced to voiceless, lenes to
For typological purposes it should not prove too difficult to fortes, or, in the case of stops and affricates, as non-aspirated
devise either a set of formulas of the cvev type or a short to aspirated. Partaking in these two series, we should expect
repertory of labels: Georgian, with its frequent clusters of up to find from five to ten articulatory types (orders) such as
to five successive consonants.> could be dubbed a heavy- a bilabial type, a labiodental type, and so forth. It should
cluster language, Modern French and Classical Hebrew be fairly immaterial whether stops and fricatives mayor may
would be 'shwa' languages, and so forth. Suchlike procedures not be conceived of as forming two parallel systems, whether,
would probably be handier and more informative than for instance, a pair like /8/ and /0/ should be classed as
indications on a percentage basis (e.g. e 55 per cent.- 'apicals' with /t/ and /d/, or made a separate order. Unless
v 45 per cent.), But, of course, languages are by no means there were some gaps in the pattern, this should yield,
always homogeneous in this matter, and many of them which, as normal, ten to twenty stops (including affricates) and
in a large majority of cases, give preference to largely vocalic fricatives. English with eight pairs related as shown on the
syntactic patterns, may, in a few words, indulge in very following diagram
heavy clusters: how many words like extra [/ekstra/, i.e . p t tJ k
veeeev) would we have to count in French before we label it b d d3 g
a heavy-cluster language? f 8 s J
This seemingly casual but probably fairly efficient classi- v 0 z 3
ficatory scheme suggests the use of an equally casual-looking
or French with twelve stops and fricatives forming six orders
procedure for dealing with the specific features of phone-
matic systems. It would consist in establishing, on the basis p f t s f k
of our present-day phonological experience, a fairly loose b v d z 3 g
norm in reference to which the phonological system of a conform, in this respect, to the norm as established here.
given language could be characterized by indicating in what Two to four nasal consonants, two to four 'liquids' in..
respect it deviates from the norm, either because some of the eluding vibrants, laterals, and also weakly articulated spirants
features of that norm are missing, or because it presents like English /r / and the weak dorsal continuant called
phonological types that do not appear in it. Such a norm 'Parisian r', should be considered normal.
should result from a survey of a large number of the most Vocalic systems with five to ten vowel phonemes would be
varied languages of the world and should be set up in such considered average: five corresponds to the Spanish system
a way as to achieve the tersest possible characterizations. with three degrees of aperture and two series, ten to the
The tentative one that follows is not meant to furnish much Danish system with four degrees of aperture and three
more than an illustration of how such a system might series (if the length distinction is disregarded), in both cases
function. with one phoneme only (fa/) for the widest degree of aper-
I Cf. above, pp. I I-IS.
2 Cf. Hans Vogt, 'Esquisse d'une grammaire du georgien moderne', NTS,
ture. The existence or status of 'semi-vowels' need probably
ix (1938), pp. 15-16. not enter into consideration here.
 LINGUISTIC TYPOLOGY LINGUISTIC TYPOLOGY 79

In reference to such a norm, the phonematics of a number Trubetzkoy's main distinction among vocalic systems was
of known languages could be characterized as follows: between triangular and quadrangular patterns: Spanish
with its single open vowel was said to have a triangular
Classical Arabic: 'emphatic' consonants, three vowels.
pattern; Finnish, with its distinction between front lal and
Finnish: a single series of consonants, gemination.
back [e], a quadrangular one . The French system was said
Russian: consonantal opposition between 'hard' and 'soft' .
to be quadrangular, but it is currently losing the distinction
French: nasal vowels.
between front lal and back [a] and is therefore becoming
English: two sets of diphthongs.
triangular. Since the physiology of speech organs implies
German: affricates as a distinct series.
that there should be fewer distinctive possibilities when the
Italian: gemination.
jaws are wide open, triangular patterns should be considered
Castilian Spanish: stops and non-liquid continuants
normal and are actually more frequent. Well-balanced
grouped in three series.
quadrangular patterns are rare and should be dealt with as
Such a norm, if widely accepted, would certainly render exceptional cases rather than put on the same footing as
some service. But we may wonder whether we have a right triangular ones. The main co-ordinates of vocalic patterns
to include it in a typological scheme, since it is a characteriz- are, of course, as was pointed out by Trubetzkoy, the degree
ing device pointing to deviations rather than a way of of aperture and the combined play of tongue and lips deter-
distinguishing between a limited number of types. It is, mining the length and shape of the front mouth cavity. The
no doubt, likely that if our sampling were large enough, we progress of acoustic phonetic research does not imply any
would find a number of languages which deviate from the revision of these basic facts. Tongue-and-lip action is excep-
norm in just the same way: it would not be difficult to find tionally of a single type, as in Adyghe, more commonly of
languages that could share with Russian the consonantal two types (generally front-retracted versus back-rounded,
opposition between 'hard' and 'soft'. But, still, the first step hence, for maximal oral closure Ii u/), as in Spanish or
consisting in disregarding some differences because they are Italian; three types when lips and tongue act independently
found to be very widespread, is in conflict with what is of each other with, as the result, Ii y ix] as in French or
expected from a typological approach. German, or Ii ill n] as in Rumanian; exceptionally of four
A real typology would probably concentrate precisely on types, viz. Ii y ill ul as in Turkish. English, with two types of
what we disregarded when establishing our norm, namely tongue-and-lip action, but with some central vowels, pre-
the organization of the cores' of our vocalic and consonantal sents another kind of pattern. The absence of any distinc-
pattern. In this field, the pioneering efforts of'N, S. Trubetzkoy tion whatsoever among degrees of vocalic aperture should be
must still be considered the necessary point of departure.> quite exceptional; two degrees are well attested (in classical
I The notion of 'core system' is borrowed from a paper presented by Eugene
Arabic, for instance) though not frequent; three degrees are,
Dorfman at the Chicago Meeting of the Modern Language Association on probably, to be considered a norm; four degrees are not rare,
28 Dec. 1959. It designates the occlusive-fricative section of the phonematic but hardly very stable; since there is less space back than
pattern, with whatever other consonants fit in the series and orders of that
section.
2 In YCLP, i (1929), pp. 39-67. Among recent attempts, cf. that of C. Voe- pp. 444-53. Our information regarding the phonological systems of languages
gelin in For Roman ]akobson, pp. 598-9, and C. Voegelin and J. Yegerlehner, quoted in what follows is generally taken from N. S. Trubetzkoy's Grundziige
'The Scope of Whole System and Subsystem Typologies', Word, xii (195 6), der Phonologic, Prague, 1939 (= YCLP, vii).
80 LINGUISTIC TYPOLOGY LINGUISTIC TYPOLOGY 81

front, fewer distinctions could be expected there, which is tions of the world. If their pattern is parallel to that of oral
the case in well-attested patterns such as that of seventeenth- vowels, as seems to be the case in numerous African languages,
century French ;' more than four degrees are not likely to we could think of such a formula as '23N' for a pattern
endure as such, neighbouring units being apt, in such made up of Ii e a 0 u I e a 0 fij; archaic Parisian French,
patterns, to become distinct by means of some other dif- with IE & a 5j, would need a separate nasal formula:
ference than sheer aperture.. 'N3 2 ' .
Were it not for accidents such as two distinct phonemes In many American varieties of English, one would have to
for the greatest aperture or a different number of relevant take into consideration retroflex vowels, as in park, court, hurt,
apertures front and back, it would be easy to devise formulae which might also require a separate formula with R as the
such as '23' for Spanish or Russian, '24' for Italian, '34' for marker. Glottalization is also to be considered a vocalic
Danish, where the first digit would indicate the number of characteristic in some languages.
distinct tongue-and-lip actions, and the second digit the What is just as important as the actual number of vocalic
number of relevant apertures. Rumanian, with [in] as head phonemes is the extent to which the distinctions among them
of its third tongue-and-lip type, would need some mark function: in Italian, for instance, the '24' formula given above
distinguishing its first digit, 3, from the 3 of the Danish is only true of accented vowels; elsewhere the formula is '23'
formula, e.g, '323', whereas Danish would be '313', or as for Spanish. In some varieties of English, it could be said
'3 ill3' vs, '3Y3' that all vocalic oppositions are blurred in totally unaccented
Many languages distinguish between long and short syllables. But, as indicated by the last formulation, it may be
vowels or between tense and lax ones, the passage from one preferable to deal with this type of restriction in prosodic
type of distinction to the other being often gradual, so that typology.
both perceived duration and degree of tension contribute, As regards consonants, it will be useful to distinguish
for some time, to the distinction. Should we use L for dura- between the core system and a margin. The core system is
tion, and for tension, LjT for a combination of both, made up of a number of proportions whose usual diagram-
we could present a formula such as '33LjT' for northern matic presentation offers a number of horizontal series, each
German. a succession of phonemes produced at different points of
In some languages, like English, there is no sharp boun- the articulatory channel, but with a concomitant feature in
dary between long vowels and diphthongs; the vocalic common. What I call an order is precisely the class of
nucleus of beat, bait, court, and cart being treated as long or phonemes articulated at the same point and with the same
diphthongized depending on individual or regional usage. organs at that point; orders appear as vertical columns on the
Here again, we could devise a formula with and D 2 for charts. For core consonants, we could devise formulas of
different types of diphthongs. But I doubt whether this the very same type as the ones we have been suggesting for
would be worth while, since the English vocalic system is the vowels: a first digit would indicate the number of series;
quite exceptional, a second digit the number of orders. If we leave out the nasals
Nasal vocalic phonemes are frequent only in certain sec.. and, for Greek, the fricatives, the consonantal core of classical
Greek would be designated as '33' (jp t kj, b d gl jph t h khj), that
I Cf. A. Martinet, 'Note sur la phonologie du francais vers 170 0 ' , BSL,

xliii (I 947), p. I 7. of French as '26' (jp ft s Jkj, jb v d z 3 gj).

811928 G
 LINGUISTIC TYPOLOGY LINGUISTIC TYPOLOGY
Unfortunately, conflicts will arise: in English, all voiceless German, because it is quite obvious that no one would
consonants (the ones that have a voiced partner) could be bother to learn I10w to decipher formulas which would have
put in the voiceless series; but ItII and III are practically to be either incomplete and unreliable, or complete and
produced at the same point and with the same organs, and unwieldy. Yet we need not give up all attempts. According
this is nearly true for ItI and 181. Therefore they can be said to whether one addresses technicians or a wider linguistic
to belong to the same order. This suggests that we posit two public, phonematic characterizations could be secured either
different series for voiceless phonemes, one for stops and one by the presentation of charts with series and orders, or by
for fricatives, as shown on our diagram on p. 77. If we now listing the main distinctive traits and indicating the number
put Itl, Idl, 18/, and 101 in the same vertical row, and dis- of phonemes affected. For classical Greek, this would give
regard the ·difference between bilabial and labiodental, we either:
obtain, as a formula for non-nasal core phonemes, '45' in-
stead of '28' with gaps in the stop series for the sibilants and b d g dz r y ii
in the fricative series for the dorsals. If, being more fastidious, P t k s e 0
ph th kh h
we decided that 181 is not to ItI what III is to Itfl and insisted E. 5
on making bilabials and labiodentals two distinct orders, m n it
our pattern would be still more lacunar. or, less technically (and far less accurately) , 'voice and
The main objection to such formulas is the fact that they aspiration (12 ph.), 2 nasals, 1, r; y-, and u-types,
leave the prospective user in the dark regarding the nature of 4 apertures, length (II ph.)",
the distinctions between series and orders. We might, no None of these solutions is really satisfactory because we
doubt, use further indicators: e.g. for classical Greek, we have found no way of combining brevity and precision.
might write '3 VA3' where V would stand for 'voice' and A There is, in fact, among the phonemes of a language much
for 'aspiration', and for English '4VF5', where F would stand of the same sort of solidarity as we have found among the
for 'friction'. As regards the second digit, which indicates various planes of linguistic structure: a change here will
the number of orders, it would be fairly easy to interpret it, involve a change there and determine a whole chain of
in the case of Greek, as pointing to the three most obvious changes. But the existence of a given type will rarely imply
orders of labials, apicals, and dorsals, but 2 would mean 'no the existence of another type: on being shown, for instance,
labials' in Iroquois, 'no apicals' in Hawaiian, 'no dorsals' in the oral vocalic pattern of a language, there is no way of
Tahitian. Besides, as was the case with vowels, gaps would be telling whether or not that language has nasal vowels as
left unmarked and we would never dare multiply the second distinctive units. Only detailed information regarding the
digit by the first in order to get the total number of the distribution and contextual variations of oral vowel pho-
phonemes concerned. We could think of giving that total nemes might give a clue as to whether they have nasalized
between parentheses, hence, for English, 4VF5(I6), with 16 partners. A language with a 101 phoneme stands, of course,
the real total instead of 20 suggested by 4 . . . . 5. But this a good chance of having a lei phoneme too, and a series of
would not tell us where the gaps are to be found. voiced stops seems to imply a series of voiceless ones. But this
We need not consider the case of the nasals and that of does not lead very far. Our failure to contrive a set of preg-
marginal consonants such as 11/ and Ir I in English, French, or nant formulas need not affect us too much, however, since
 LINGUISTIC TYPOLOGY LINGUISTIC TYPOLOGY
the substantial nature of the phonemes of a language has Swedish, Vietnamese belong to the former type, English,
no direct bearing upon its actual functioning as a medium Russian, Arabic to the latter.
of communication. Even the absolute or relative number 2. A tone language may present tones on all successive
of vowels and consonant phonemes is fairly immaterial in intonable segments (usually syllables) or on some favoured
so far as the transmission of information is concerned: in ones only (as a rule accented ones): Southern Chinese dia-
many cases, French, for instance, will, on the phonematic lects, Vietnamese, and Ibo show no speech segment without
level, be satisfactorily characterized as a shwa-language with a tone; Northern (Mandarin) Chinese, Serbo-Croatian, and
nasal vowels. Lithuanian distinguish two or more tones, but only in syllables
It is not a mere chance if we have, so far, found it easier with some accentual prominence.
to devise terse characterizations in reference to syntagmatic 3· All tones may be punctual, i.e. fully characterized by
reality than for describing the relations of phonemes in the their relative pitch (high, medium, or low), all of them per-
system: coexistence in a system means distinctiveness and ceptibly so, or with some easily reducible exceptions (e.g,
consequently as much mutual independence as is compatible occasional rising tones being analysed into a succession of
with economy; coexistence in the speech continuum implies, Jow-l-high}; or they may be melodic, i.e, opposed as rising to
it is true, preservation of each successive unit's respective falling, as in Lithuanian, as glottalized to non-glottalized,
identity, but also adaptation to context, i.e, a much greater as in Danish, as simple to complex, as in Swedish.
measure of solidarity. This explains why it is comparatively 4· If melodic, the tones may be all on the same register as
easy to devise concentrated characterizations of the syntag- in Mandarin Chinese or in the languages just quoted ; or they
matic phonological features of a language, and why the field may be on different registers as in Vietnamese, where it is not
of prosody is the one where typological research has yielded enough to say that a tone is rising or glottalized, but where
the most satisfactory results so far.! one has to state whether it is high rising or low rising, high
Among prosodical features, intonation is far too directly glottalized or low glottalized.
conditioned by generally human, physiological and psycho- 5· Finally, the segment characterized by each tone may be
logical, factors to be of any use when what it amounts to the syllable, as in Serbo-Croatian, a segment smaller than
is distinguishing among linguistic types. Only properly dis- the syllable, the mora, as in many languages of Central
tinctive units, the tones, and the contrastive traits that can be Africa, a segment larger than the syllable, such as the (poly-
grouped under the heading of accentual prominence can be syllabic) word in Swedish or Norwegian.
used for typological purposes, but they do playa central role I. A language may have an accent, a prominent segment
there. In such matters, a fairly satisfactory characterization per word or comparable unit, something which is often called
can usually be achieved by answering, for each language, a 'stress' in E.nglish, but accent is a functional reality, which
few questions bearing on tones and a few others concerning may involve, for its actualization, stress, pitch, length,
accent. or any combination of these; or it has no accent: most
I. A language has tones (distinctive tones, of course, some- European languages have an accent; Vietnamese and many
times called 'tonemes') or it has no tones: Serbo-Croatian, Central African languages have no accent.
I Cf. A. Martinet, Phonology as Functional Phonetics (London, 1949), pp. 1I -
2. The place of the accent is either predictable and non-
15, and 'Accent et tons', Miscellanea Phonetica, vol. ii (London, 1954), pp. 13-24- distinctive, or non-predictable and consequently distinctive:
86 LINGUISTIC TYPOLOGY LINGUISTIC TYPOLOGY

it is predictable in Czech where the accent falls on the first be to develop a system of handy labels for rapid orientation
syllable of the word, in Polish where it falls on the penult, in regarding the prosodical pattern of languages: German, for
classical Latin where its place is determined by syllable instance, might be said to have an initial bound-accent on the
length; it is unpredictable in Spanish where a significant lexeme and no tone, if we exclude foreign elements; Italian
unit characterized by the phoneme succession /termino/ accent would be described as a 'word' accent, free, but nor-
may mean three different things according to the place of mally restricted to the last three syllables. Something could
accent. be added in both cases regarding the limitations of vocalic
3- The distinctiveness of the place of accent is either distinctions in unaccented syllables, and some indications
unlimited, as in Russian, or variously limited, as in Spanish concerning the make-up of both accented and unaccented
where the accent falls on one of the last three syllables of syllables would not be out of place.
a lexical unit. There would no doubt remain cases where some specific
4. If in a language tones only exist under the accent, it is information would have to be added, if we do not want our
not unusual to say that that language has as many different labelling to be misleading: to say either that French has no
accents as there are tone distinctions: thus Swedish and accent or that it has a bound word-group accent would be
Lithuanian have two and Latvian three accents. equally unsatisfactory, because the very tenuous prominence
5. Most accentual languages have one accent per 'word': continuing the Latin accent is one thing, the optional pro-
in Russian, nos 'nose' loses its accent when it combines with minence on the initial syllable is something different, and the
rog 'horn', in nosorog [nssarok] 'rhinoceros', which means emphasis on and around the first consonant of the word, is
that when two lexemes are agglutinated to serve as one, the different again: the true French accent is neither impossible,
resulting complex receives only one prominence. Other accen.. . with a very slight prepausal eminence, nor impossible, with
tual languages preserve the accent of each lexeme, irrespec- didactic and demarcative insistence on the initial, a very
tive of whether it is free or agglutinated; but some sort of normal feature of professorial diction, nor imp-possible with
hierarchy is established among the successive accents of the a dramatic lengthening of [p] meant to suggest anger or
same 'word': 'rhinoceros' is Nashorn in German, and Nas- passion; French has no accent in the sense English, or Russian,
(for Nase 'nose') and -horn preserve their respective promi- or Italian has, and to use one and the same label for function-
nence, with, however, a subordination of the one of . . horn to ally so different things would be a source of lasting confusion.
that of Nas-, English shares this feature with German and
extends a similar pattern to learned borrowings: energetic Once phonematics and prosody have been taken care 04
with two accents as iffrom "enner-s-getic. It is true that English what remains is the vast field of meaningful units and their
has one accent in midland, just like Russian in the equivalent combinations, in other words, lexicon, morphology, and
sredizemnyj, and Russian has two prominent syllables in konuso- syntax. Few linguists are likely to insist on establishing a
obrdznyj like English in the equivalent cone-shaped. But, lexical typology, not exactly because they feel that the
although it is only statistically valid, the distinction between lexicon of a language is too closely dependent on non-lin-
'word'-accenting and lexeme-accenting languages may be guistic reality, but-and it may amount to the same thing-
retained. because it is what remains of the language once its neatly
The preceding sketchy survey indicates how easy it would structured parts have been abstracted and dealt with, in
88 LINGUISTIC TYPOLOGY LINGUISTIC TYPOLOGY 89
other words, the domain of loosely connected units for which realized that beaucoup de must once have been identical with
it would be difficult to devise a wholesale characterization. the syntagm beau coup de. It is by no means certain that
There is, however, one criterion whose application might autostrade, with its unidentifiable element -strade is more
result in establishing interesting contrasts between languages. difficult to remember and to handle than its competitor
It is what might be called the amount of motivation 1 in the autoroute, with its obvious ingredients. If it were shown that
vocabulary: some languages make use of a comparatively German children have less trouble with Fernsprecher than with
small stock of monemes because they frequently resort to Telefon, it would probably not be on account of the lack of
composition and derivation; their vocabulary may be said motivation of the latter, but rather because Fern- and -sprech-
to be largely motivated: such a thing being called this or that sound familiar, irrespective of what they mean. The French
because it is this or that. Other languages have a relatively children of today, who constantly hear the phoneme com-
large number of unanalysable designations; their vocabulary binations /tele/ and /fJn/ in television and grammophone as well
is thus more largely arbitrary in the Saussurian sense of the as in telephone, are probably just as well off as their German
term: a thing is called thus for no discoverable reason except, contemporaries even if they do not understand tele- and
perhaps, for the etymologist. The traditional illustration of -phone. This, however, does not mean that motivation is to
this contrast is that of German, a language with a highly be rejected as a criterion for lexical typology, but rather that
motivated vocabulary, and French, with a largely arbitrary here is one more domain where one should not jump to
one: in situations where French makes use of the two mono- conclusions.
monematic and highly abstract terms monte(r) /mot/ and In so far as the form of minimal lexical elements ·raises
descend(re) /desa/ German uses combinations of two or three specific problems, because, for instance, they are found to
monemes like aufsteigen, heraufgehen, heraufklettern, some of these vary in different surroundings, this is normally taken care
monemes being very specific and descriptive. This illustration of in the morphological chapter as well as in the lexical
is further supported by the remark that French speakers are section of the description. This, after all, is as it should be,
generally satisfied with their thousands of loans from classi- since 'morphology' is the study of form, with no specification
cal languages, most of which are unanalysable and arbitrary that only the form of grammatical elements is involved. As a
for the vast majority who know neither Latin nor Greek. matter of fact, it is more or less understood that morphology
German, on the contrary, is prone to replace foreign elements deals only with the latter, and if the formal vagaries of
by indigenous compounds, as when Perron becomes Bahnsteig, lexemes come in into the bargain, it is because they normally
Telefon is displaced by Femsprecher, and Automobil by Kraft- result from their combination with certain grammatical
uiagen. items: the shift of radicals we find in I go, I went, and the
is quite probable that this contrast between motivated formal variations we observe in I keep, I kept are taken care
and arbitrary is something of which foreign linguists and of in all grammars of English even if lexicographers kindly
local purists are more keenly aware than the average users: include them in their dictionaries for good measure.
the present writer, a native Frenchman, had to read the Since morphology and syntax are legitimate parts of
diary of Ernst Junger at the age of thirty-four before he grammar, we may now state that what remains for us to
I Cf. F. de Saussure, Course, pp. 131-4, and Ch. Bally, Linguistique generale
investigate is grammatical typology. Traditional grammar
et linguistique francaise, znd ed. (Berne, 1944), pp. 137-9. distinguishes neatly between two main chapters: on the one
go LINGUISTIC TYPOLOGY LINGUISTIC TYPOLOGY 91

hand is the study of those combinations of significant ele- linguistic typology. Ifwe want to formulate what it amounted
ments that may involve some formal variations or accidents to in terms ofmonemes, we may say that, in the last analysis,
(cow, cows, but ox, oxen, child, children) work, worked, but keep, counted was hardly anything but the degree of formal
kept, sing, sang), which normally take place within the word; variation of individual units, for which we can set up the
this is called accidence or morphology; on the other hand, following scale:
the examination of the way separate words can be combined I. A moneme may have the same signijiant throughout,

into larger units, which is called syntax. But since there is whatever the combinations it enters: the relation marked, in
no way of defining the term 'word' in such a way as to make English, by without is always marked by jWlloautj. This,
the definition tally with the naive uses of it, contemporary which logically seems the most obvious solution to the prob-
structuralists are prone to employ it most sparingly and to lem of the formal relations of the signijiant with its signifie,
refuse to set up any universally valid linguistic unit between has appeared, to generations of scholars blinded by their
the moneme (often called 'morpheme') and the sentence. admiration for classical Indo-European languages or their
Formal accidents will be ascribed to monemes, not to 'words', ethnocentric prejudices, as a quaint and outlandish feature
and dealt with, for instance, as the allomorphs of a given when applied to the expression of functions and to gram-
morpheme. They will be taken care of in the course of the matical modifiers.
analysis of utterances into minimal significant units. Once 2. The signifiant of a moneme may vary from one context
the identity of these units, our monemes, is established, the to another, but it will always be identifiable with a definite
linguist will have to observe 110W they combine. Since he has segment of the utterance: Italian con, 'with', normally ap-
already listed and described the vagaries of moneme signi- pears as jkonj, but also as jkoj, e.g. before the masculine
jiants, it is by now immaterial whether the various monemes article (col, coi) and as jkolj, e.g. before the feminine article
of a given context are amalgamated, agglutinated, or form- +
(colla, colle); the analysis of colla into col la is obvious,
ally independent; the material aspect of their combination since la is the normal form of the feminine article, and
does not concern him any more. Should one insist on positing that of col into co+l (for *con+il) cannot be said to be
a unit intermediate between the moneme and the sentence, I arbitrary.
would propose the syntactically autonomous or independent 3. A moneme may, in some contexts, appear as a clear-cut
phrase. Such phrases would include practically all the in- segment, but, in others, be merged with the signifiant of some
flected words of Latin or Greek, and also 'real' phrases such other moneme (or monemes), as when Fr. a, which is
as with the dog and down the road. Still, I would not be inclined a
generally jaj, even before the masculine article in l'hopital
to distinguish an intraphrasal syntax, redolent of the old (jal ... . j), is amalgamated with that same article in au
morphology, and an extraphrasal one which would be syntax moulin (jo .. .j).
properly so called. 4. A moneme may never appear as a separate segment
Traditional non-genetic classification of languages was because its signijiant is always amalgamated with that of some
really based on accidence; the title ofP. S. Kuznecov's short other moneme (or monemes) ; still, the independent existence
survey, Morfologiieskaja klassifikacija jaeykoo, I is, in fact, quite of the monemes involved is never endangered: in Latin, the
descriptive of its contents and of the way people conceived of dative moneme never appears as a segment which is not
I In German Die morphologische Klassifikation derSprachen (Halle a.d.S,; 1956). at the same time the signifiant of the moneme of singular or
 LINGUISTIC TYPOLOGY LINGUISTIC TYPOLOGY 93

that of plural; but Latin speakers were never at a loss to of linguistic reality that afford some justification for the use
distinguish dative singular and dative pluraL of such an ill-defined concept. Yet a clear-cut distinc-
5. A moneme may have a discontinuous signifiani, its use by tion between inseparability, total or partial, and variation
the speaker implying some formal modifications in two or of signifiants is fully justified only on a strictly synchronic
more different places in the utterance: the Latin maneme plane. It is indeed clear that inseparability is conducive to
usually identified as sine, 'without', was necessarily accom- formal accidents which ultimately result in amalgamation:
panied by a specific (,ablative') ending of the nouns whose phonetic evolution will tend to merge successive elements in
function was being indicated, so that, in sine dubio, its sig- the utterance, and this can be counteracted only inasmuch
nifiant included Isine/, plus the I. . . 6j of dubio, in which, as the elements in question may, at any time, cease to be in
however, the signijiant of the singular moneme was amalga- contact; any limitation of the latitude to combine monemes
mated. The plural moneme of French may show as a single freely is the first step toward eventual coalescence, semantic
segment, e.g. as I. . • e .. .1in dans les champs Idalefaj (cf. the as well as formal. WIlen what is at stake is a general charac-
singular dans le champ Idalfaj) ; but it may also appear in the terization of languages, from a synchronic angle, no doubt,
form of several successive accidents, e.g. as I. . e . " . z .. " but on a wide temporal scope, widespread inseparability will
o . . " mj in les petits animaux dorment Ileptizanimodormj as be found to have bred some degree of formal overlapping
opposed to le petit animal dort jlptitanimaldor/. and amalgamation, so that the two features normally go
There is, however, another feature which has to be con- together.
sidered if we want to account for the pronouncements of Sapir's outstanding merit was to discover .and point out
former typologists: that is the degree to which two monemes, that the traditional approach to non-genetic classification
appearing in immediate succession and in a definite rela- was founded upon a fairly casual rating of the frequency and
tionship, may be separated by the insertion of some new degree offormal adhesion, overlapping, and coalescence. He
element. As a matter of fact, inseparability is one of the most therefore recommended a typology that would not be based
useful criteria for distinguishing what is formally one word exclusively upon features which might not be so fundamental
from what is a succession of different words.' In any case, it as many scholars had believed them to be. As a matter of
is the one that generations ofscribes and writers have adopted, fact, observation had revealed that most languages presented.
as a rule, throughout the centuries of alphabetic writing a mixture of various formal types, and comparative evidence
practice, when they have endeavoured to divide the written had indicated that formal features were subject, through
continuum of each language into those segments which con- time, to considerable variations either towards syntagmatic
stitute our graphic 'words'; German spellings like aufgeben, coalescence or away from it. Without rejecting what he
auJgebe, each written as one block in spite of the possibility called degrees of 'fusion' and the traditional 'synthetic' I"'J
of inserting something between auf and the rest iaufrugeben, 'analytic' opposition, Sapir concentrated on 'the nature of
ich gebe es auf), represent shocking exceptions for whoever is the concepts expressed by the language' and made it the real
not awed by the majesty of spelling traditions. In view of the foundation of his classificatory attempt. He thereby un-
paramount role played by the notion of 'word' in former doubtedly went farther and deeper than any of his predeces-
typological speculations, it is fair to single out all the aspects sors. A proof of this is that, to this day, forty years after the
I Cf. Proceedings of the 6th Intern. Congress, Paris, 1949, pp. 293-4. publication of Sapir's book, anyone who deals with language
 LINGUISTIC TYPOLO G Y LINGCUISTIC TYPOLOGY 95
94
classification necessarily uses Sapir's analysis either as a use the form bus as an abbreviation of autobus; this form was
starting-point or as a frame of reference. Yet, for all its probably used first in imitation of its English equivalent, and
perspicuity, it stands as a nearly tragic illustration of the pit- it is not in general use yet; for those use autobus is
falls of psychologism. a compound word of the autoroute 'motor-road' type; for those
Sapir's original contribution centres around a conceptual who do not, it is a derivative of the electrobus type, with bus
analysis of language elements starting from the acceptable as a suffix. The formal criterion of both derivation and
view that we should find, in all languages, significant units composition is, of course, a combinatory comportment of
that do not carry in themselves a mark of their relation to derivatives and compounds which is, in all respects, identical
the rest of the utterance, say, chair, lamp, and others, such as with that of corresponding single monemes, It is, no doubt,
with or for, that are expressly meant to indicate what sort of quite essential to determine whether a language makes use of
relations the former keep with each other. This leaves out derivation, or composition, or both: languages vary a good
the theoretical possibility of a language entirely made up deal in that respect. But this has little to do with the distinc-
of words of the yesterday type, which stand, at the same time, tion between monemes that do not indicate their relations
for some element of experience: 'the day before this day', to the context (dependents) and those that are meant to
and the relation of that element to the rest of the experience: mark those relations (functionals); both composition and
'as the segment of time when ...'. But we know that economy derivation yield units which may function as dependents
rules out any such language, and we may agree with Sapir (farm-yard, yellowish) or as functionals (on-to, Lat. in-ter,
that all known languages utilize monemes with the chair type in-tr-a, in-tT-v) . We have here a distinct type of linguistic
of function, and monemes having the with type of function. activity, which we could dub lexical, or, more generally,
Our criterion for distinguishing between one type and the paradigmatic expansion. Any typology will have to take this
other is not semantic, however, but positional: if chair by into account in connexion with the lexicon, a domain in
itself is to have a function in an utterance, it must stand in a which, as we have seen, the frequency of still analysable com-
well-determined position in relation to the other elements of pounds and derivatives could be characterized in terms of
the utterance; but this changes as soon as a moneme of the motivation.
with type is added to it: the phrase with (the) chair will not Sapir further distinguishes between 'pure relational con-
depend on its relative position for the expression of its cepts' of the with type, and 'concrete relational concepts', but
function. The importance of this reliance on formal criteria he never makes perfectly clear where the difference between
will appear in what follows. the two types actually lies. It would seem that this distinc-
As a third type of element, Sapir poses what he calls tion more or less parallels that between what we call func-
'derivational concepts', which formally correspond to affixes. tionals and modifiers, the grammatical elements which secure
There can be little objection to this, except that, of course, autonomy, and those which do not. But his conceptual
his conceptual approach prevents him from connecting com- approach misled him into expressing the difference in terms
position and derivation, which are such closely related of degree of abstraction, and made him fall a victim to the
phenomena that the analyst is often at a loss to say, in con- common illusion that some grammatical words or items (his
crete cases, whether a given moneme is an affix or an element 'pure relational concepts') have no meaning. If meaning is
of a compound word: in contemporary French, some speakers to be mentioned in linguistics at all, it should be defined as
96 LINGUISTIC TYPOLOGY LINGUISTIC TYPOLOGY 97
resulting from the necessity, for the speaker, of choosing at a as 'singular' or 'plural'. It summarizes some previous informa-
point among several units for the expression of some element tion, but gives no indication as to what the function of its
of experience.. Choice, on that level, implies meaning, and environment is the linguistic rendering of experience
meaning is there only if there has been a choice. Syntactic which is being communicated. Amalgams may make it
functions, such as the ones expressed by prepositions or difficult, in many cases, to decide which phonic segment is
cases, are, in a way, predetermined: a dative relation is to be ascribed to a functional, and which to a modifier.. But,
expected after a verb meaning 'to give'. But the choice often functionally, the two types are fundamentally different.. The
exists between the presence of the dative complement and its distinction is not, as Sapir has it, between material content
absence: advice can be given absolutely or specifically to and relation, but between functional monemes and non-
someone. A grammatical subject, as such, is an item whose functional ones, the latter including the bulk of what Sapir
presence does not result from a choice: the speaker does not puts under 'material content', namely lexicon, including
choose to use a subject or not, because, by definition, as it elements of composition and derivation, together with our
were, a subject is what must be added to a predicate to make modifiers.
a statement in those languages where we have a right to It is, as we have seen, by no means impossible to find
speak of a subject. Therefore, since the introduction of the functional justifications for everyone of Sapir's four con-
subject function is never a matter of choice, we may say it ceptual groups. But this does not mean that we can retain
is deprived of meaning. A formal consequence of this is Sapir's analysis as the foundation of a typology that tran-
a tendency to eliminate any mark of the subject function, so scends formal accidents. Once we have ruled out reference
that a zero signifiant should correspond to a zero signifi». But to things meant as a principle of linguistic classification, it
this is an extreme case. The difference between functionals becomes clear that the four types of facts are not comparable
and modifiers, i.e, the two types of grammatical monemes, and that it makes no sense to consider them along a scale
should not be expressed in semantic terms, although modi- leading from pure conceptual to pure relational, Should we
fiers are likely to be semantically richer than functionals. It try to classify monemes according to their function, we
should be based upon the part each type plays in the sen- should first put on a plane of its own the distinction between
tence. The notion of relation, which Sapir used in order to derivational monemes and the rest. The latter could not very
give a semblance of unity to his non-lexical units, is extremely well be dubbed non-derivational, since the fact of being
vague and misleading: there is no doubt that a definite derivational implies nothing but a limitation, in the distribu-
article normally involves some relation to what precedes, tion, to cases of paradigmatic expansion, in other words,
since one of its normal functions is to indicate that the a negative feature. We certainly would not want to have
accompanying being, object, or notion has been previously to specify 'non-derivational' every time we characterize
mentioned. But this role is in no way distinct from that of a moneme affected by no such limitation. Once the problem
pointing to the fact that the being, object, or notion is uni- of derivationals is set aside, our two criteria would be
versally known, as when we say the sun, the King, or the lion functional indication and grammaticality. By combining
for the species. The definite article is not indicative of any them, we would obtain four types again, as shown on the
relationship; it is the mark of some specific information diagram on P: 98, out of which only three (viz. I, 3, and 4)
added to the communication, just like any other modifier such could be found parallel to those of Sapir.
811928 H
98 LINGlJISTIC TYPOLOGY LING·UISTIC TYPOLOGY 99
2 language is likely to have monemes acting as
lexical grarnmatical Therefore we cannot follow Sapir when he classifies languages
without indication of function without indication of function
the according to whether they 'express concepts' of this or that
chair
group of his, which, in less-guarded parlance, would have
3 4 been expressed as 'having' this or that sort of units: beside
lexical grammatical
function indicating
his group I (of 'basic concepts'), which he sensibly assumes
function indicating
yesterday type with type to be universal, a given language should use, for the ex-
pression of relations, either his group (of 'pure relational
Two of them (I and 2) would not partake in functional concepts') or his group III (of 'concrete relational concepts') ;
indication or, in other words, would not confer, to themselves of his four, A, B, C, D, classes, two (A and B) combine I and
or others what we have called syntactic autonomy: the IV, and two (C and D) combine I and III; the absence of
ones (I) would be lexical dependents, the others (2) eit~er derivation distinguishes A from Band C from D. Groups IV
grammatical. dependents such as pronouns, or grammatIcal and III are thus presented as mutually exclusive, which rules
determinants, i.e, modifiers. The remaining two would be out any attempt to preserve some parallelism between our
autonomous monemes of the yesterday type (3) and autonomy types and Sapir's groups: when Sapir conceived of his group
conferring grammatical items, i.e, functionals (4). Predica- III, he must have had in mind concepts corresponding to our
tives, as such, would not all belong to the same one of the two modifiers; when he set up his group IV, he certainly wanted
lexical types, since they might be said to be dependent where to include in it elements corresponding to prepositions, which
they can be orientated (in languages with both active and are the most obvious of our functionals, and position, as a
passive voices), but autonomous, even independent elsewhere. mark of function, is expressely mentioned in this connexion;
Furthermore, there is a fundamental difference between but the lack of perspicuity inherent in the conceptual ap-
Sapir's groups and our types, which makes it impossible for proach led him to list English and French among languages
us to utilize the precedent. When we say that our types are that make no use of his group IV, thereby disregarding the
functional, we imply that a given item may belong to dif- fact that in both languages the relations of nominaIs with the
ferent types: in a large number of languages, the same rest of the utterance are, in the overwhelming majority of
moneme is used predicatively as the equivalent of 'to give' cases, indicated by means of position or prepositions. It
and as a functional with the value of 'dative'; this; of course, would be interesting to look for the features of these two
does not entail that all predicatives can be used as func- languages, such as widespread amalgamation, concord, and
tionals or all functionals as predicatives. In Basque, the discontinuous monemes of all sorts, that must have been
moneme -ko is used either as a functional ietxe-ko 'of the instrumental in obscuring for him such an obvious fact. But
house') or as a derivational use for lexical expansion (etxeko- this would lead us too far. It will suffice to point out here
a-k 'the domestics'; -a-k = 'the' -l-plural}; here again, this that, even in his conceptual classification, Sapir is so depen-
does not mean that all functionals of Basque can be used as dent on form and intent upon finding formal differences
derivationals, but it indicates that our types are not mutually supporting his conceptual grouping, that he indicates, as the
exclusive. We should not operate as if a given language had only normal expression of his 'concrete relational concepts'
or had not modifiers or functionals, but assume that any (group III), affixation and inner modification, the use of
100 LINGUISTIC TYPOLOGY LINGUISTIC 'TYPOLOGY 101

to that of 'pure relational

'Y'pC"p'Y''trpri predicative function the on the other hand, forms
concepts' (group IV). of speech in which the equivalents of 'tree' or 'leg' may per-
As soon as each utterance in any language is conceived haps be to act as predicates, whereas those of 'it
of as a succession of minimal signs, with or without over- stands' and 'it walks' can be predicate and nothing else.
lappings, it becomes clear that grammar can be presented Languages with a special 'verbal' class restricted to pre-
as the set of rules restricting the free combinations of all these dicative use can be further subdivided into languages in
signs. A grammatical typology is bound to start from the which non-verbal lexemes can normally be used with pre..
various types of restrictions that have been, inductively or dicative function and those in which this is formally im-
deductively, identified. We have seen, in a preceding chapter, I possible. Russian is a language in which lexemes endowed
that a language can hardly be conceived of without such with 'nominal' functions are also found as predicates. On the
restrictions, for instance, regarding the extent to which the contrary, English is a language in which 'verbal' monemes
various monemes may assume the various existing functions. have a practical monopoly of predicative function, and it
Some monemes must necessarily be identified as endowed illustrates a linguistic type where a given lexical item is
with functions and others as functional indicators. This always used either predicatively (verb) or non-predicatively
accounts for the fact that there is no language recorded in (noun or adjective); in regular, out-of-situation, syntax, no
which one should not somehow distinguish between gram- statement is possible without some verbal form.
matical and lexical items, whether these be opposed as empty Another fundamental distinction should be made between
words to full words, or as grammatical categories to basic languages in which the predicative moneme, once it has
concrete units. Restrictions on the free use of the different been selected, can only stand in one definite relation to every
lexemes in all the existing functions are probably to be found one of the other elements of the utterance, and languages in
everywhere, but in many cases these may result from the which the predicative moneme can be orientated in different
specific meaning of some lexemes: a thing may not be used ways in order to give prominence to its relationship with one
as an agent; some predicates cannot be passively orientated. of the other elements. This means, in traditional parlance,
But what may be peculiar to some languages and therefore that great importance should be attached to whether a
typologically relevant are wholesale restrictions resulting in language distinguishes or not, in what we may call its con-
splitting the lexical repertory into neatly contrasting classes. jugation, between an active voice and one or more so-called
In this respect the most fundamental dichotomy is the passive voices. This feature may be connected with the
one between languages in which all lexical monemes can be presence of a subject properly so called, i.e. the compulsory
found performing the same basic functions, and those, prob- use of an item accompanying the predicate and whose rela-
ably a clear majority in the world of today, in which some tion to it is accordingly established as preferential and
monemes, which we may designate as 'verbal', are specialized fundamental,
and restricted to predicative uses: on the one hand, we have All this amounts to saying that the foundations of a lin-
languages in which the equivalent of 'tree' and 'it stands guistic typology are to be looked for in the way in which each
upright', of 'leg' and 'it walks' are identical and distin- language community proceeds to analyse experience into a
guished only by the predicative use of the latter, some non- number of elements in such a way that the linear order of
I Cf. above, pp. 63-64. their succession will not prevent hearers from perceiving the
102 LINGUISTIC TYPOLOGY

nature of their mutual relations and thus reconstructing the

total experience. The basic syntactic patterning of each
language has to be presented in terms of minimal significant I
units, the monemes, each moneme being conceived as the
habit of identifying some definite facet of experience with LINGUIS to IE
some formal distinction. There are thus always two faces
to every moneme, It would be linguistically ideal to have

I
N this world of ours where, apart from pitiful exiles, every
a situation where the formal face of every moneme would man is supposed to owe allegiance to some ruler or State,
be always the same and correspond to a neatly analysable this allegiance is tacitly assumed to imply the use of one
segment. In fact, the signifiant of a given moneme may vary and the same language: an Englishman speaks English and
according to context, be discontinuous, or Inerge with neigh- a Frenchman speaks French. Belgians are widely believed to
bours, so that only functional identity salvages the unity of speak 'Belgian', a belief that is partially substantiated by the
the moneme, A fundamentally synchronic description of a existence of Flemish, It took two world wars to convince the
language should go beyond these formal accidents, and reach French that the British and the Americans were different
the essentials: the resources of a language as an instrument of nations, but the realization of their separateness carried along
communication. with it the by now well-entrenched belief in the existence
of a fully distinct American language. We all, in daily life,
speak, and sometimes act, as if there existed neatly circum-
scribed language communities wherein all members are
expected to behave linguistically in exactly the same way.
Those who do not, in all details, are said to speak with an
'accent' if their deviations from an assumed norm are mainly
phonic. They are supposed to speak a 'dialect' if their aber-
rance extends to grammar and lexicon, particularly if com-
munication is thereby somewhat impaired.
As long as linguists were 'philologists' who mainly concen-
trated on written, literary forms of human communication,
they were not inclined to take exception to this sketchy and
naive approach to socio-linguistics: 'accents' were hardly
ever noticeable as such in their texts, and 'dialects' were
best forgotten about except, perhaps, in the isolated case of
ancient Greek..
But even after they have become conscious of the basically
vocal form of language, after decades of efforts devoted by
many of them to the collection of dialectal material, they still
seem to take it for granted that there are well-identified
1°4 LINGUISTIC VARIETY LINGUISTIC VARIETY 1°5
objects called 'language communities' whose members speak believe alike in all respects, as between a new-
alike in all respects. At least most of them behave and write comer and the former residents a suburban district; but
as if they did. In this matter there is little difference between it takes among Russian and
'traditionalists' and 'structuralists': all would seem to believe fishermen who to settle in the same neighbourhood
in the perfect unity of each language. Among the latter, few along the and results in the development of
would hesitate to posit that French has so many (e.g, 34) a new form of speech.. 1
convergence will inevitably
phonemes, or that the /s/ of Spanish is apico-alveolar, even breed divergence: the new-comer who adapts his speech to
if they were ready to grant, when challenged, that many that prevailing in his new will thereby deviate
French speakers use less, and many others more, than thirty- from what had been his set of linguistic habits so far, and all
four phonemes, and that millions of Spanish speakers, in the quicker if the original linguistic differences between the
Spain and elsewhere, use dorsa-alveolar sibilants. two parties did not hamper immediate oral communication..
The homogeneity of linguistic communities is a useful As long as needs differ from one district to
pragmatic assumption, at least at a certain stage of linguistic another, complete linguistic homogeneity cannot exist.
research.. If languages are, first and foremost, instruments Ifit has so for such an obvious fact to be acknow-
of communication, it is fairly natural that we should assume, ledged, it is again because those who were interested in
at least as an ideal, that all people who use one of them share matters of language concentrated on literary forms of com-
the bundles of articulatory habits and vocal reactions to munication.. Those forms had early assumed a large measure
various stimuli whose sum total we call a language: com- of unification since they were used by somewhat restricted
munication would be best secured if all people concerned groups of literate people engaged in a sort of communication
spoke exactly in the same way. Some variety is no doubt which did not require the establishment of physical contact
welcome in human affairs. But when efficiency is at stake, and which, consequently, could easily spread to the most
relevancy is what counts exclusively. But unbiased observa- distant provinces..
tion shows that when people understand each other they do Once the familiar and comfortable idea of the homo-
so in spite of differences which affect not only irrelevant geneity of linguistic communities is abandoned, the world
trimmings, but sometimes fundamentals. It is true that appears as an ocean of conflicting attractions, convergence
when we look at things from a dynamic angle, the prospect here breeding divergence there, with new centres of attrac-
clears up a little: it is comforting to notice that when people tion developing at all times and threatening to disrupt
actually communicate they tend to identify their speech existing ensembles. this is true on all levels, national,
habits and linguistic reactions: communicating improves provincial, local, and familial. In the practice of descrip-
communication and does it apace. But this might eventually tive linguists the of this endless linguistic
result in the complete unification of linguistic communities variety has led to the position that if a linguistic description
only ifit worked within their confines exclusively, which is by has to be .consistent, it must be that of an idiolect, i.e. the
no means the case. Linguistic convergence is universal: just language as spoken by a single individual. But does this not
like charity, it begins at home and, just like charity, it extends contradict our assumption that language is above all an
to the whole of mankind; it takes place among those who
I See Olaf Brach, 'Russenorsk', Archivfur slavische Philologie, xli (1927), pp.
feel they belong to the same language and social group and 2°9-62.
 106 LING UISTIC VARIETY LING UISTIC VARIETY 1°7
instrument of communication? Besides, if we want to do least conscious of what they do, they can hardly be doubted
justice to all the implications of the ceaseless turmoil we have in the fields of syntax and lexicon. The practical conclusion
just sketched, we cannot be satisfied with limiting our obser- to be derived from this is that a linguistic description that
vation to one single individual, since that individual will not is not expressly that of a dated idiolect should be expected
handle his language tomorrow in exactly the same way as he to subsume divergent usages. If existing variations are not
does today or did the day before. The idiolect, as a frame sacrificed on the altar of descriptive simplicity and are duly
of linguistic description, needs to be precisely dated. presented, the result will be the establishment of some sort of
At this stage, and before we proceed, it may be useful to hierarchy among linguistic oppositions: some distinctions are
stress that the rather startling picture that has just been found to be universal among the members of the group under
presented is no fiction, but that it is derived from factual ob- consideration; others are found to be kept by some members
servation: when, in 1941, sixty-six French officers, born and only and to be disregarded by other members. If the language
reared in Paris, were asked some forty-odd simple questions, at stake is American English as a whole, it will be pointed out
such as 'Do you pronounce patte and pate alike?', which were that horse and hoarse, morning and mourning are kept distinct by
meant to reveal the main features of their respective vocalic some speakers, while others pronounce them alike, and this
system, no two of them agreed in all respects, although their type of distinction will not be put on the same level as that
phonological comportment as a whole contrasted with that between card and cord, lard and lord which seems to be univer-
of their 343 non-Parisian comrades who also answered the sal. In describing French in general, it cannot be said that
questionnaire. I pres differs from pres in the same way as pris differs from pres,
As regards the evolution an idiolect may undergo through although physically the three vowels are close to cardinal [i e £],
time, the present writer may be allowed to refer to a sketch because all French speakers distinguish pres from pris, but
of the system of French vowels he published in 1933.2 In that millions identify pres and pres.
first attempt to present a phonological analysis of that We can conclude from all this that the notion of linguistic
language, being still unaware of the existing latitude in such community is not only useful, but unavoidable in linguistics
matters, he identified his then system with that of French at as soon as a language is conceived as an instrument of com-
large, and posited two long phonemes jy:j and jre:j as dis- munication constantly adapting itself to the needs of the
tinct from jyj and [ex] on account of such minimal pairs as group who make use of it; 'communication' implies 'com-
sur-sure (/syr1"'-' Isy:rI), seul-seule (/srelj"-'Isre :1/). When recon- munity'. But in order not to let this term confuse linguistic
sidering his phonological practice, some years later, he issues, it is indispensable to stress a number ofwell-established
noticed that these distinctions of length had been wiped out, facts.
probably under the influence of his Parisian surroundings. I. No community is linguistically homogeneous. no two
If such variations from speaker to speaker and, with the persons use a language in exactly the same way; the same
same speaker, from one period to another, are attested in the situation will elicit different linguistic reactions from different
realm of phonology, a domain where people as a rule are onlookers; no two persons will use or understand the very
same vocabulary; even the highly structured aspects of
I Cf. Prononciation generally, and Phonology as Functional Phonetics, pp. 36-37. language, such as phonology and morphology, may differ
2 'Remarques sur Ie systeme phonoIogique du francais', BSL, xxxiv (1933),
pp. 19 1- 2 0 2 • in important matters from one speaker to another without
 108 LINGUISTIC VARIETY
LINGUISTIC VARIETY log
impairing mutual understanding and even without being
more phonemes than is strictly necessary to keep them
noticed by the interlocutors.
distinct from others in given contexts, and messages being
2. Many people belong to two, or more than two, com-
as a rule more profuse than logic a few
munities. This is, of course, the case in such well-known
divergences will not prevent nor even impair mutual under-
bilingual stretches as Brussels, Alsace, or South Africa. But
standing. People being normally intent upon understanding
this applies to many situations where both a vernacular and
what is said, and hardly upon observing the way it is said,
a standard language are alternately used by the same people
they will never notice any such divergence. Language works
with different interlocutors, From a linguistic standpoint
best when it is not noticed as such, and speakers will be better
we cannot make bilingualism depend on the amount of
off as long as they manage to forget about it. This is why they
prestige enjoyed by the two forms of speech in contact. The
are apt to get impatient and scornful when some 'accent' or
alternate use of two different phonological systems is probably
'dialect' feature reminds them that linguistic communica-
the least ambiguous test of a bilingual situation.
tion can be a problem. Unexpected deviations may prove
3· Many people use concurrently different styles of the misleading even if they do not affect the distinctive and
same language. The same Frenchman may use from one
significant pattern: if I have never heard my la/phoneme
minute to the next two totally different equivalents of 'shall
pronounced as anything but [a], an occasional [0] rendering
we go?'; the literate partons-nous? or the slangy on les met?
may be startling and may keep me wondering for a split
with the familiar plural first person pronoun on.
second whether that unexpected sound 'meant' something,
4· Many people who do not use more than one style or one even if the nearest vowel in the system is some distant /'J I
language understand different styles or different languages.
phoneme.. But if one is used to hearing something that
Passive knowledge of languages and unimpeded aural under-
deviates from one's own practice, that something will cer-
standing of various not actively used styles is of frequent
tainly pass unnoticed: if I pronounce my /a/phoneme as [a]
occurrence and should playa great role in the correct appre-
but am quite used to hearing [0] for the same unit, no [0]
ciation of socio-linguistic situations..
pronunciation can ever ruffle me except, perhaps, if it
Before we examine in detail cases where people cannot help
comes from someone from whom I do not expect anything
being aware of linguistic differences, it is important to revert
but [a].
to definitely unilingual situations where the different mem-
In this connexion I may perhaps be allowed to present an
bers of the community are not aware of any such deviation
illustration borrowed from my own experience: as a child I
in the speech of others or in their own as could be dubbed
knew that my mother's French was slightly tainted by her
'accent' or 'dialect', In these the communicative ideal of
early use of a Franco-Provencal dialect and that, on the other
language seems fully realized: it really does not matter in the
hand, when she spoke that dialect there was something that
least whether the speech of one man is physically identical
did not sound quite right. One day, about the age of twenty,
with that of another provided possible divergences do not
I was sitting in an orchard reading for an examination. My
interfere with the free and easy transmission of'experience. In
mother was conversing with her sister, about ten yards away.
this respect it is quite essential that all participants should be
At some point she happened to mention the name of the
intimately convinced that they speak 'the same language': since
famous political leader Marat and pronounced it [mana]
language is constantly redundant, words having generally
with a dorsal spirant. This startled me out of my reading and,
 LING UIS1'IC VARIETY I I I
110 LINGUISTIC VARIETY

halfjokingly, I told my mother she had no right to pronounce with in terms of dialects. Other terms such as patois, brogue,
that name otherwise than [mara], i.e, with a trill. It suddenly bable, Platt, refer to varieties which are ultimately presented
had dawned upon me that she used both 'rolled' and 'non- under the rubric of dialectology. attempt on the part of
rolled' r, which explained why her French sounded provincial specialists to use these terms without previous redefinition
and her dialect Frenchified, but that she usually had the trill is likely to be unsatisfactory because they are, in fact, quite
for single intervocalic r of the spelling and the dorsal spirant unspecific designations used, with a touch of contempt, by
for -rr-. Her use of the latter in Marat, probably as an imita- standard speakers in reference to any speech variety that is
tion of the teacher from whom she had first heard the name , not socially acceptable. The American College Dictionary defines
had suddenly made me realize that I expected from her patois as 'any peasant or provincial form of speech', which
a tongue tip [r] wherever the spelling had a single r. amounts to authorizing the use of that term for any language
Generally speaking everything one has grown up with variety except some metropolitan standard. Such a defini-
belongs to one's universe, and this applies to linguistic ex- tion, which reflects social prejudices rather than the realiza-
perience as well as to any other domain. Anything we have tion of real linguistic differences, makes the word useless for
heard in the course of the acquisition of our language is felt our purposes unless we decide to define it in our own terms.
to be part of that language. This does not, by any means, When we try to approach this problem from a linguistic
imply that we ourselves shall make use of everything we angle, it is disappointing to find that little can be done if we
readily accept from others: a number of words, certain pro- start from those achievements of which we are so proud, viz.
nunciations, we have heard from people we either do not our phonological and morphological patterns. Here is my
approve of or dare not imitate, and these we avoid. But this wife's vocalic pattern in French in absolute final position pre-
remains unconscious until, by chance, we are startled by sented side by side with mine:
hearing them in an unexpected situation or context. 1 y u 1 y u
Something similar applies to different generations in e o o e e o
each other's company: we hear in the speech of older people a e
many turns and traits which we would never use, but which a
still belong to our language. Yet our grandchildren will
never get used to them and these features will thus gradually These look very different. But it happens that contacts be-
disappear, first only from active use as long as we live, then tween people with such divergent vocalic systems have been
also as acceptable forms as soon as only those generations so frequent during, say, the last two centuries, that the
remain on earth that have never heard them. What we have French language is handled in such a way to render the con-
heard ever since we could perceive what was said around us fusion of lal and [a], lei and [e], 101 and Idl (as sur ce ==
is always part of our language as we conceive it, even if sur ceux) perfectly innocuous. Only someone intent upon
we may exceptionally become aware of the obsolescence of detecting phonological differences would think of my wife's
certain terms or elements. speech and my own as two distinct varieties. The only valid
criterion in such matters is less perhaps the ease of under-
Linguistic variety on a large scale, within one and the same standing, which we would not know how to measure, than
community, is, by both laymen and linguists, usually dealt the existence of bilingualism, in the widest sense of the word,
 112 LINGUISTIC VARIETY LINGUISTIC VARIETY 113

namely, the use same persons of two phono- ancient Greece, before the establishment of Athens' cultural
logical and morphological patterns depending on one's inter- hegemony, corresponds to our and reflects a
locutors.. need a say, ' to refer linguistic situation far more similar to the American than to
to linguistic forms used by unilinguals their oral com- the Italian one.
munications with any other member community, even It could be objected that if dialect, necessarily implies
those who use some other dialectj : the York City form bilingualism in contradistinction to dialect., it would be
of speech and the Chicago form of speech would thus be better and clearer to call dialect, a language, a term which is
labelled two dialects, of American since a speaker of implied in 'bilingual'. But a language is understood to enjoy
one would not hesitate to use his own form of speech when a status which can by no means be granted to many dialects,
addressing a speaker of the other, term, say 'dialect,', that only survive as the impoverished mediums of retarded
,":o~ld desi?,nate linguistic forms used as a vernacular by rural segments of a community. The widespread reluctance
bIlInguals In their communications with some .particular to speak of bilingualism in the case of situations involving
members of the community, whereas they use a dialect. with dialects, is due precisely to the impression that they are
the other members; the speakers of any dialect 2 are' in fact , granted thereby a status they do not deserve.
a smaller (provincial) community within the larger (national) The dialectal situation in France has long been the subject
community.. This type ofsituation will be amply illustrated in of careful and detailed studies, and although professional
what follows . dialectologists have widely disregarded some of the distinc-
It is clear that dialects., are likely to be more divergent than tions which we today find basic, these studies have largely
d~alectsl' since if they were not and i:f, accordingly, their use contributed to making France one of the best fields if we
did not hamper mutual understanding, speakers of dialects, want to illustrate linguistic variety within one and the same
would community.
. not take the trouble of learning a dialect1 which is not
t h eir vernacular.. this greater divergence is not included It cannot be said that France, as a political entity, coin-
in the definition because it is, as we have seen, too difficult to cides with a linguistic community, since millions ofunilingual
measure.. French speakers live in other countries such as Belgium,
Whether dialects. all enjoy the same prestige, or whether Switzerland, and Canada. But it cannot be doubted that any-
one of them, or a group of them, ranks higher and is, there- where in France people are supposed to be able to handle
fore, rIO longer considered a dialect, but the standard, is their public and private affairs in French: all children
again a different matter: prestige is difficult to measure and living in France are expected to attend schools where
~ad better not intervene in linguistic classifications.. Still, it is French is taught as a subject and at the same time is the
Important to be aware ofsuch as situation in the medium through which other subjects are taught; further-
United States, where no regional standard can be localized more, all able-bodied males spend an average of two years
and
. ' in an army in which French is the only official medium.
. that In Italy, where the Tuscan varieties, when de-
prived of their strictly local features, rank definitely higher As a result of this, non-French-speaking Frenchmen must be
than other dialects., which are; of course, and in contra- rare except among older people, chiefly women and mainly
dis~inction to dialects., local forms of the national language. in Alsace and the traditionally Frankish-speaking fringe of
It IS worth remembering that the original use of 'dialect' in Lorraine.
811928
114 LINGUISTIC VARIETY LINGUISTIC VARIETY 115

At the present day a definite majority of the 45 million one was still practised and understood over a large area so
of must unilingual French speakers, that there might remain, in some regiments, groups of
which, of course, does not mean that they all speak alike. All soldiers who found it easier to use it among themselves than
of them could be said to make use of dialects., i.e. different French. On returning home, those who for close on four
varieties of French, but this would by no means reflect the years had been using French almost exclusively insisted on
way the French react to such variations: deviations from speaking French to their children, and their wives eventually
what is felt to be the norm in matters of pronunciation are followed suit. This explains why, in large sections of present-
labelled 'accents'; other aberrances, if startling, might be day rural France, dialects, are frequently understood but
considered 'patois' by those who would be tempted to apply hardly spoken by people under forty. It is likely that the
this term to any departure from the norm. The few competent Second World War has sealed the fate of a good many of
people who have dealt with French dialects! refer to them as those which had fully survived the First. The very possibility
francaislocaux. Some of them have been summarily described. I of a dialect disappearing through a break in the transmission
They are certainly well attested in Gillieron's Atlas.' but is a clear indication that that dialect is a dialects. A dialect!
there is no way of knowing with certainty whether what we cannot disappear, since, in the case of unilinguals, it is the
find at a certain point is dialect! or dialects, In a radius only form of speech at their disposal. The effect of a war will
of sixty miles around Paris, Gillieron's forms are likely to be to make it closer, if not to the 'standard', at least to the
originate from unilinguals and therefore to represent local average form of speech.
French (dialectsj), In the southern half of the country all It is a great pity that Gillieron was not aware of the neces-
notations, with very few exceptions, should represent utter- sity of distinguishing between the two types of dialect and
ances from bilinguals who knew French at least well enough consequently did not try to devise a way of drawing the
to understand Edmont's questions and who translated the frontier between unilingual and bilingual France in his time.
French words into the local dialect.; But, in between, the It is true that the thing would not have been as simple as
forms attested in many places are of such a nature as to we have made it sound: there are socio-linguistic situations
authorize anyone of the two interpretations. where the shift from one form of speech to another does not
Unilingual France is expanding rapidly. The First World seem to be sudden, but graduaL In certain sections ofPi cardy,
War was fatal to dialects. in many sections of the country: before and during the Second World War, the linguistic
all fit men between the ages of eighteen and fifty went to war. situation could be characterized as follows: when first con-
At first they were in locally mustered regiments, but very fronted with strangers, civil servants, holiday-makers from
soon, heavy losses resulted in successive amalgamations, as a Paris, and the like, local peasants would use a sort of local
result of which soldiers from all parts of the country could be French close enough to the average, if not to the standard
found in the same unit. This meant that there was little use form, to be generally understood. This, they would even use
for dialects, at the Front and behind it, except when such a for a time among themselves in the presence of strangers,
I e.g. Le Francais parle a Toulouse, by Jean Seguy (Toulouse, 1950) (Biblio-
but very soon features from their dialect, would begin to
theque Meridionale publiee sous les auspices de la Faculte des lettres de Toulouse, crop up: etait, mangeait would gradually yield before letol,
rst series, XXVIII), 132 pp.
2 About the Atlas linguistique de la France by J. Gillieron and E. Edmont, see
Ima301 with the local 1-01 ending in the imperfect; French
Sever Pop, La Dialectologie, I. Dialectologie romane (Louvain, n.d.), pp. I 13-36. vocabulary items would persist a little longer, but eventually
116 LINGUISTIC VARIETY
LING UI8TIC VARIETY I 17
such a form as chaussette, pronounced locally as [Jose~t] (in
dialect.), would be replaced by the regular Picard Ik0 Jex], to area
The outcome would be a form generally impervious coast. It is still a matter of and when this par-
to anyone except the initiated. This type of stratification ticular segment of Romance-speaking area acquired the
suggests the possibility of a gradual word-by-word and form- measure of which must be postulated by anyone
by-form elimination of dialect. through stages where dif- who has discovered, behind the motley of contemporary
0

ferent styles would coexist just as they do, for instance, in patois, a bundle of fundamental likenesses. Dialect varies
unilingual French. Such a type may have been, or may still from village to village, at times from hamlet to hamlet and
o '

be, fairly general along the periphery of the Basin, but Its use secures unimpeded communication within a radius of
10 t~ 15 ~iles at best. Beyond that distance, peasants may
it is by no means the necessary intermediate step between
dialect. situations and unilingualism, since a sudden break is fi~~ ~t easier to use French (dialect.}. In a period ofincreasing
just as likely. d~vIslon ?f labour and geographic mobility, this means that
Dialectal bilingualism its neater form, i.c. one with dialect. IS doomed: people speak French better and better
clearly distinct dialect, and dialects, deserves to be dwelt upon and are. never .gi:ren a chance to eliminate pre-existing or
a little further, since professional dialectologists, who are developing variations among dialectsg; the range of forms
mostly antiquaries looking for old forms in any context (dia- and words that are understood and used if necessary, con-
Icct, or dialectj), have not been inclined to describe the stantly decreases: at Hauteville, I former generations accepted
socio-linguistic settings in which they operate. for 'bull' the form Ibar'ral beside the local Italre/; younger
The illustrations that follow are based upon observations people who ~peak French with those for whom Iborlral is
carried out in Franco-Provencal territory, more precisely the normal dialect, form never get acquainted with it. Had
in Savoy, a region where dialects, were doomed as the both parties tried to converse in dialect., both forms would
result of the contacts established in the course of the First have retained their wider currency. As a matter of fact,
World War. Although we shall use here the present tense, the any reference to a bull among people from the Ibarlra/ and
situation referred to is, generally speaking, that which pre- the Italrel zones will, from then on, be made by means of
vailed at the eve of that war. Dialect! is, with careful speakers, Fre~ch ta~r~au pronounced locally (dialect.) [tc'ro] (as
a fairly good approximation to average French, closer prob- agamst Parisian [touoj), It might be advantageous to reserve
ably than in most unilingual rural regions. This largely in technical parlance the word 'patois' for the designation of
results from the fact that people are aware that what is good minimal linguistic communities limited to a few thousands
for dialect (dialectj) is not good for French; French, of and even a few hundreds of rural speakers. But since then
• • 0 • 0 ' ,

course, is taught at school, and school attendance is universal. no patois situation IS Iikely to endure, it may not be advisable
Only very few, usually feeble-minded, old women speak only to keep a word such a transient phenomenon.
dialect, although they understand spoken French (dialectj), Dialect., alias local French, contains permanent and uni-
Dialect, is a variety of that type of Romance called versal devia:ions from average or standard French, most, but
Franco-Provencal whose domain, limited by Provencal in the not necessarily all, of them carried over from dialcctj: alleren
,I Cf. ~. Martin:t" 'Des.cription phonologique du parler franco-provencal
south and merging gradually into the Burgundian and
~ Hauteville .(SavoIe) , RL~R, xv, pp. I-85, revised and reprinted in La Descrip-
Frane-Comtois areas in the north, spreads from the upper tion phonologique avec application au parler franco-prooencal d'Hauteoille (Savoie)
(Geneva, 1956).
 LINGUISTIC VARIETY LING UISTIC VARIETY
118

champs les naches 'grazing the , for stanuaro process we

baufs, quand moi 'when I did', for standard 'en mime temps que has certainly not yet affected the whole of Romance-speaking
moi, poche 'ladle', (dialect, I'p08e!) for standard louche, &c. France, but it cannot be doubted that unilingualism is estab-
Most of these are normally used by both unilingual dialect! lishing itself in more and more rural districts at the expense
speakers (teachers and priests, for instance) and bilinguals. of Provencal, Languedoc, Gascon, Catalan, and Corsican
But, at any time, dialect., forms may, in the speech of the dialects. As regards .Catalan and Corsican, it is often felt
latter, be transferred to dialect. with just the necessary that their status differs from that of Gallo-Romance dialects
phonological and morphological adaptations. Children who because, diachronically speaking, they belong together with
pride themselves on their early handling of the standard Hispanic and Central Italian Romance respectively. What
language are particularly apt to effect this type of transfer: may count, socio-linguistically, is the fact that the use of
as a calque of dialect, [a'badda 'bje- le 'plo-ts pg ka-lbo- 19 Catalan and Corsican dialects enables some Frenchmen to
gJ1Aa], approximately 'take a big step to get over the puddle', establish linguistic contacts with foreigners with whom
a ten-year-old girl said once: Abade bien les plotes pour camber French is no recourse, i.e, people who do not have French
le goillat [abad bje le plot pur kabe 1 goja], the phonology of as a common, official language. Inasmuch as these contacts
which was that of dialect, (local French), the grammatical across the Spanish border or the Tyrrhenian Sea are lively
cement and the adverb bien were average and standard and profitable, they may retard or prevent the crumbling
French, but where the four lexemes abade, plote, cambe, and down of the dialects in question into a multitude of mutually
goillat were exclusively local. The closest standard French hardly intelligible patois, the prelude to their eventual
equivalent would have been Ecarte bien lesjambespour enjamber elimination. To what extent contacts across the border do
la flaque. It should be pointed out that this utterance, when consolidate the Catalan dialects spoken on French territory
heard, excited the merriment and the derision of the by- and make them more resistant to the disintegrating pressure
standers, with some comment to the effect that if the girl of the official language than their Languedoc or Gascon
could not speak French better, she had better use the patois neighbours, remains to be investigated.. In the case of Corsica,
which she would have handled to the general satisfaction. distance and the lack of territorial contiguity may largely
What deserves particular consideration here is, however, contribute to the conservation of the vernaculars. In those
less the extent to which a bilingual person may let one margins of France where non-Romance forms of speech are
language influence the other, than the fact that the sentence in common use, the possibility of using them for contacts,
with its local lexicon is readily and unmistakably identified across the border, with people who do not speak French
as French, ludicrous French indeed, but French, and not the must also be reckoned with as an asset for the vernacular.
local vernacular, because the phonology and the grammatical This undoubtedly plays a role in the case of Basque, with a
monemes are those of dialect., not dialectj, majority of Euzkarian speakers on the Spanish side of the
It would be interesting to determine whether some local Pyrenees and a lively intercourse from north to south, parti-
words such as abader, plote, camber, and goillat survive to date cularly in the region close to the Bay ofBiscay.. Breton, spoken
in the speech of ten-year-old children who, in the regions exclusively on its land's end by people who all learn French
concerned, are by now definitely French unilinguals. If at school, does not find any support outside and must be
some do, it probably will not be for long. receding rapidly, even on the rural front" in spite of the
120 LINGUISTIC VARIETY G UISTIC VARIETY 121

of a intel..
lectuals. Flemish should, in principle, stand a good chance
of resisting the pressure of French, since its French domain
is nothing but a fringe of a large area including north-western
Belgium and the Netherlands. Yet as long as French remains
a prestige language among the Belgian bourgeoisie of West and
Flanders, the Flemish dialects of France cannot withstand the of
pressure of the official language. The Alsatian situation, which of features." . . . . . . . ~, i. ... .. "- .................

also applies to the northern stretches of Lorraine, is, in fact, the a French Creole be con-
trilingual, with German dialects as vernaculars, and two com- sidered a form of speech, and in
peting national languages, the former official language, view of of the vocabulary, the
German, still extensively used by the Churches and in part of latter very a Yet if
the press, and French, which is by now the universal teaching we insist on as French the sentence abade bien les
medium in schools. Both German and French could be said plotes pour camber le lexemes are Franco-
to be used under the form of dialects., The local dialects, Provencal, on account the use of French phonology
are widely divergent, which would seem to bear witness to an and grammatical argue we could call
early use of a common standard, namely German, but their a Creole an What is decisive, however, on
position does not seem to be really threatened so far. a pragmatic is the fact that the abade bien _.. utterance
What seems decisive for the comportment of speakers of a was meant to and the that the speaker
local form of speech in conflict with a standard language is had such a of French lexicon prevented
less, as generally imagined, the degree of similarity or dis- it from being as SUCll. In the case of
similarity of their structure and vocabulary, than whether Creole, any utterance is meant to be Creole and nothing else,
the standard language is the same for all the speakers of a whatever the origin of the ingredients may be. It is
given vernacular or whether the vernacular is felt as a link never to Ewe or a lexicon.
with people who could not be reached through the language be from attempts to force the
one is taught at school and trained in in the army. Accord- into previously estab-
C111T'llI",·t"l1r'\"li'"ll

ingly, the socio-linguistic situation should be much the same From a standpoint, it
in Romance Gascony and Celtic Brittany, differences in differs from any 1J.L\\"'V!.il...JL'l-.l!.V . L.JL
'~~~1~,~~~"'M France in
sentimental reactions notwithstanding, the sense that no "I:TA1i"'''I''1lr:,,,,,, ... so much simi.. .
""r'\"I'Y'1ln,"I"li'"llIf.::'C"

The preceding rapid survey of linguistic variety in France larity on the

would not be complete unless something were said of the habits. There is >.J'U ....... 'V''U' ............ ~ were as thorough
language situation in the French West Indies, a few islands
I Cf. the thesis of Suzanne Sylvain, Le Creole haitien (Wetteren, 1936).
which are, politically, a part of Metropolitan France. Their 2 Best shown by Beryl Loftman in an unpublished ColumbiaUniversity
population, predominantly African in origin, are taught M.A. dissertation in 1953.
3 Cf. the debate in Word, with Douglas Taylor, Robert A. Hall, Jr., and
standard French at school, but normally speak among them-
Uriel Weinreich as xii (1956), pp. 407-14; xiv (1958), pp. 367-
selves what is dubbed a French Creole, namely a form of 79; xv (1959), pp.
122 LING UISTIC VARIETY LING UISTIC VARIETY 12 3

and as it is
"' ... .!I..!I.."-"'-il '''-<''V say
of Basses-Pyrenees, Creole users would achieve a complete never dare put on paper.. Consequently, written style is not
mastery of French far more easily than their Euzkarian spoken style. it is not always easy to distinguish, here,
fellow citizens. between style and language: did Gregory of Tours, when he
wrote his Historia francorum, some time in the second half of
The widespread beliefin the unity and homogeneity of the the sixth century, use a learned style of his own language, or
language of a given 'community' does not only conceal the another language? We shall be tempted to speak of one and
linguistic variations existing within the boundaries of each the same language as long as the different linguistic forms are
state, but tends to convince even the educated that the felt to be complementary because each situation requires one
language people speak and the language they write is neces- definite form and excludes all others, so that the user will
sarily the same. This is, in some cases, palpably untrue, and, never be faced with the necessity of making a decision as to
in others, open to question. When a linguistic community, what form to choose. No one will hesitate to speak of different
illiterate hitherto, gets acquainted with the art of writing, styles if there actually is at the user's disposal a whole gamut
this is used for the rendering of some other language. In such of different linguistic forms merging gradually into one
a case, dissociating the writing from the language requires another, as is practically the case with the contemporary
a power of abstraction that may be absent, so that the first national languages in which anything can be found in print,
persons who write do so in the foreign language. This may with a tendency for frontiers between genres to get blurred.
become institutionalized, so that literate people go on speak- It is often said that some difference between -spoken
ing their vernacular, sometimes exclusively, and do not language and written language is rendered inevitable be-
know how to write except in another language. In many cause so many decisive features of speech are not transferred
cases, the written language is that of a 'classical' literature or to writing. Whereby people mean the accentual hierarchy,
that of a liturgy, which was the case with Latin in medieval so imperfectly rendered by the occasional use of italics,
Europe, and is still so with Sanskrit in India and Koran intonation, whose salient features only may be suggested by
Arabic in Muslim countries. This, it is true, is not always our punctuation, and all those peculiarities that characterize
exclusive of efforts to apply writing to the vernaculars. But, individual elocution. This amounts to saying that writing,
as a rule, those efforts take centuries to bear appreciable with print as its ideal form, is a set of discrete visible symbols,
results.. each corresponding to some discrete audible unit of speech,
It could be objected that the fact that some people use a and anything that is not discrete will be sacrificed in the
foreign or ancient language as a written medium is a very transfer.. There is a large measure of truth in this: in the last
special case and that it does not mean that the written analysis, many of the features that differentiate written style
language of a nation is necessarily different from its spoken from spoken style can be traced back to a need, in writing, to
medium. Most people would concede, however, that there compensate for the loss of suprasegmental and individual
may be, between one and the other, a difference of style: the elements of speech. Through too much insistence on the
chief reason why people do not write as they speak is prob- blurring of contextual features, however, one is apt to forget,'
ably that since writing leaves permanent traces, whereas or at least to minimize, the importance of situation in spoken
speech, unless recorded, is lost for ever, writers are far more communication, and the necessity to compensate for its
12 4 LINGUISTIC VARIETY LINGUISTIC VARIETY 12 5

absence in : a speaker may at all times refer to what he which, in may appear as pagaye, pagaie, or pagaille;
has, hie et nunc, common with the hearer; the author has popular German I'fuzelj, 'spirit', cannot even be given a suit-
Fusel, is in rJ.I "."...,.r'\.",.."I.:>.l'l

nothing in common the reader but his text, and the fact
that they belong to the same linguistic community. It will would correspond to *I'ffizdlj and *Fussel to *I'fusdlj. The
not do to argue that the author constantly re-creates situa- reverse is far more common: hosts of words are so
tions so that he will use 'shifters';' words that are understood generally learnt through print no single pronuncia-
in situation only, like I, you, here, lastyear, just as easily when tion can permanently establish itself (gerrymander with [g] or
he writes as when he talks, very fact that he has to with [d3]?).
recreate his situations if he wants to make use of a large part In a comparison of the frequently distinct systems pre-
of the lexicon suggests that an appreciable part of his activity sented by the written form and the spoken form of one and
is devoted to descriptions and presentations from which the same language, one may, in order to make the contrast the
'shifters' are excluded except when, quite exceptionally, the more striking, combine in one pattern the archaisms of
author ('I') addresses his readers ('you') . As a matter of fact, the spelling and those of the literary style on the one hand, the
contemporary literature has done much to introduce spoken innovations of the spoken form and those of colloquial speech
style in written matter: the ideal of some writers is obviously on the other hand. As an illustration of how profoundly
to try to do away with presentations and descriptions and to different things can be subsumed under a single .language
let their characters 'speak for themselves' . But the difference label, the case of French may be dealt with at some length.
between written style and spoken style is not thereby wiped English spelling gives, in many respects, a less reliable
out: they speak in print, instead of speaking aloud, and the picture of the spoken form than French does: it is bad enough
ideal of a written style, self-sufficient, relying exclusively on for native speakers who have to discover that what they pro-
discrete, linguistically central, non-expressive elements, sur- nounce jredj is either red or read; it is a frightful nuisance for
vives, less perhaps in literature in the narrow sense of fiction, foreigners who have to pronounce read sometimes as jri·dj,
than philosophical and scientific writings. sometimes as jredj; but it does not give an inaccurate picture
The distinction between literary language, or style, and of the main lineaments of spoken grammar. French spelling
colloquial speech is, of course, not identical with that be- gives foreigners fairly reliable hints as to how to proIlounce
tween primary spoken form and secondary written form, as vowels; where it is under-differentiated (jaj or [a], jej or lei,
our use of 'written' and 'spoken' style may have suggested: and the like), Frenchmen as a whole do not really care about
contractions like don't, can't, ain't belong to various levels of differentiation in speech. The real trouble is for French-
'spoken' style; still, they have a written form, and the many men who waste the best years of their childhood trying to figure
words are used outside of written texts have a out they shall be content with the bare stem or when they
pronunciation of their own, if only because they may have shall write -s, or -nt, or -t.. Formerly, those children who went
to be read aloud. Still, there are a few words that are so to school had to learn grammar as a preparation for Latin
definitely spoken rather than written that there is no tradi- lessons to come.. In the world of today, where hundreds of
tion governing their spelling; such is French jpagajj, 'mess', millions of children go to school with no intention of ever
learning Latin, grammar, a highly abstract affair, perfectly
I So called by Otto Jespersen, Language, Its Nature, Development, and Origin

(London, 1922), pp. 123-4.

unsuited for the budding intellect of the average ten-year-old
 LINGUISTIC VARIETY LINGUISTIC VARIETY 127

youngster, is doomed as a universal prerequisite. French modal combinations such as il doit aller, il peutfaire, and the
children, however, will have to go on carrying the old burden most striking development in this respect is the
as long as they have to decide when liJatl is to be spelled il spread of doubly compound forms such as (quand) il a eteparti
chante and when ils chantent. or il avait eu pris, whose ultimate source must have been the
What follows is a sketch of the verbal forms of French in necessity of distinguishing between the true perfect with a
what could be designated as 'unguarded' spoken usage. I present meaning, as in quandj'aijini, and the compound past,
This, in turn, may vary from one region to another, and what quandj'ai eujini, built from the former on the analogy ofj'ai
will be considered specifically is the type of speech observable faim, a present with, as a corresponding past,j'ai eufaim. Any
among people living in Paris. But it must be understood that study of the syntactic functioning of the verbal system will
this type has a much wider currency and that speakers of all have to place such a form asj'ai eujini among the constitutive
provinces tend to conform to it. Some of the forms below elements of that system. But, for us here, it is nothing but a
may, in their written garb, strike one as extremely vulgar and present form ai and two past participles eu andjini.
belonging to a popular rather than a familiar level. Yet all of Before we pass on to the analysis of our forms, the problem
them can be heard in the speech of all social classes. They of linkings (liaisons) had better be dealt with apart. French
are normally used by the present writer in his home. verbal endings are felt to be particularly tricky because,
Some forms that have been disregarded cannot be said according to whether the next word begins with a vowel or a
to be totally absent in informal conversations: the present consonant, a characteristic liaison consonant may be sounded
writer, even in his most slipshod linguistic behaviour, may or not. This is quite true for any sort ,of 'guarded' French
well be pushed into risking an imperfect subjunctive in a cas- which is to some extent affected by the spelling. But in the
cade of subordinate clauses. But he will hate himselffor being informal style we are considering here post-verbal linkings
tricked into it, since he isjust as likely to use a wrong I-sl form have been preserved only in some very specific situations, so
in the third person singular as the right, I-s/-Iess form. As that they should no longer be considered formal variations
a matter of fact, our examination will be restricted to those of the verbal forms, but parts of variant forms of certain pro-
forms which adult speakers of any social class know how to nouns. Most post-verbal linkings that are suggested by the
use and use without tremor. spelling are either fanciful or optional and actually excep-
By 'verbal forms', we mean the set or sets of verbal signi- tional in familiar delivery: pronouncing a I-z-j in tu donnes a
jiants, with their variants, that children have to learn before ..... would sound ridiculous although the liaison is acceptable
they can handle the language in a way that will satisfy their after sg. 2 in the case of the copula and a few very common
elders. Therefore it is not our intention to dwell here on verbs: contrast, for tu es une ... 'guarded' ItyEzyn/, familiar
compound forms as distinct from their components, the ItyEyn I, and slipshod ItEyn/; ils donnent une
4> .. .. 4> is /id~nyn
auxiliary on the one hand, the participle and the infinitive .. I, just like il donne une .
4> .. or lild~ntyn
4> ., I in careful
on the other. French, just like, and even more than, English, speech; the desire to distinguish between sg.. 3 and pl. 3 un-
has a host of compound forms, even if we disregard quasi- doubtedly contributes to making the I-tl linking here more
frequent than the I-zl linking of sg. 2: in the same style we
I For a fuller treatment of the same, see 'De l'economie des formes du verbe may hear j-tj linking in ils donnent un livre, but no linking in
en francais parle' in Studia philologica et litteraria in honorem L. Spitzer (Berne,
195 8) , PP' 309-26.
les parents donnent un livre, where the plurality of the subject
 LINGUISTIC V ARIE LINGUISTIC VARIETY 12 9

factors intervene which It is characterized by a zero ending in sg.. I, 2, 3 andpl, 3,

queer, /-0/ in pl. I and /-eJ in pl. 2 .. Informal French, however,
r-rvn r-ro tH/011T a e • • 1, normally employs I forms a zero /0/ as
. . . . . . . . . . .,. . . . :-. here is never obliga- the subject pronoun: /0 d'Jn/ on donne, instead of /nu d'Jno/,
tory as item is not a If it is a nous donnons, with /nu 0 d'Jn/, nous, on donne, as an insistent
pronoun in it is a third person pro- form. Here is the resulting paradigm:
noun (il, ils, elle, elles, on), and these are preceded by
a I-t-I, its is a regu- 3 don o don
lar part of the as in il ils donneni, ty don vu don-e
donnent-ils, or an ad hoc as in donne-t-on. Only im- i don i don
peratives may comple-
2. The imperfect, which we need not call 'imperfect in-
among en with a
dicative' since we do not reckon with any other imperfect. It
vowel; as a rnatter of fact by I-z-I
is characterized by an 1-8/ ending in sg. I, 2, 3, pl. 3, and
irrespective of whether the form is permanently
informal pl. I .T·he endings are /-i5/ for less informal pl. I and
spelled with -s (cours, cours:)) or not (va, vas:)). If we leave
/-ie/ for pl. 2. Hence the following paradigm:
the spelling out of the we may all this
as follows: the li(l), are pro- 3 don-e o don-e
nounced Iti (1), ti (1), tel, t5 I after the ty don-e vu don-ie
verb this is not common in current speech idon-e i don-e
where interrogation is indicated by means of lesk/,
est-ce que, rather by ; the (ai, en, and 3. The future which is marked by a combination of /-r-/
li/,y, are pronounced Izal and Izil immediately plus a vowel which is /-e/ or I-E/ I for sg. I, I-a/ for sg. 2,3, and
after an imperative. thus dealt informal pl. I, /-01 for traditional pl. I and pl. 3, /-el for
with in the frame of the of pronouns, they pl. 2. The ensuing paradigm
can be disregarded in an French.
we call of verbal 3 don-re o don-ra
ty don-ra vu don-re
form that follows as a radicaL An
ending may be of distinct i don -ra i don-ro
monemes: the -rais made up of is etymologically connected and, perhaps, synchronically
I-r-I, is found
is supported by that of the present of 'to have': /e 2 a a a av-e 0/.
in imperfects.
In of II-! .. "',cl>Ybron the only I Cf. Prononciation, pp. I 22-6: among Parisian speakers, 53 per cent. rhyme

are the following: serai with etd; most of the rest rhyme it with etait; the percentage for I-el sinks
moods and tenses we to 40 for the youngest third. Note that I-el is widely known to be the 'right'
I. The present, indicative'
v ...... ..,,·""C"1£l>'YllT
form.
since it is, in 1'Y1l-tA1"1r1l"l ':) tense positively
"Il'"'I.V'C.'C1.c::>"D"'l'f- 2 It would be worth while investigating to what extent the speakers who use

tenses. leI as the sg. I of the present of avoirare the same who pronounce future sg. I
marked as such f-h"'V'''''''"1l1oon
"Y'I ".....11'10- 'Y'1"AClA1Yll1"
as I-reI.
811928 K
 LINGUISTIC VARIETY LINGUISTIC VARIETY

It might be tempting to assume that its comparative com- fasse very early, so that these survive after young speakers
plexity has been instrumental in the promotion of compound have learnt to use analogy, and the subjunctive shows few
futures of the il va donner Ii va danel type. But it should be kept signs of obsolescence e

in mind that the present of the auxiliary aller is just as com- 6. The imperative has only three distinctive forms, those of
plex: IVE va va va al-e vb], sg. 2, pl. I and 2, with zero, 1-01 and /-e/ respectively, i.e, the
4. The conditional, whose status as a tense or as a mood we same endings as the present. Traditional pl. I imperatives in
need not discuss here. It is marked by I-r-I followed by the /-0/ seem seriously affected by the disaffection for such forms:
endings of the imperfect.. Hence: traditional chantons, 'let's sing', is likely to be replaced in
many situations by such phrases as allez! on chante or allons!
3 don-rs o don-re on chante, in which the status of allons is close to that of an
ty don-r e vu don-rie [donorjc] interjection.
i don-r e i don-r e The non-personal moods present three distinct forms .
those of the infinitive, the present participle, and the past
When it is used as the future in the past (il disait qu'il donne-
participle. The present participle is always in I-a/, but the
rait 'he said he would give') it is often replaced by a compound infinitive and the past participle have endings that vary
form parallel to il va donner (il disait qu'il allait donner), but the
from one type to another: most of the verbs that present the
I-r-f form is indispensable as the conditional proper: il don-
same stem throughout have /-el for both (/dan-e/); those that
neraitcher pour. . . . add their personal endings to different stems end their in-
5. The subjunctive, which corresponds to the present sub-
finitives in I-rf (/fini-r/, Ibat-r/), hut in the case of such verbs
junctive of traditional grammars. Its marks are zero for sg. I,
as deooirkivixes] or partir/partir/, we may either decide that
2, 3, informal pl. I and pl. 3, I-iol and I-iel for less informal
the ending is just I-rI and the infinitive stems are Idvua/ and
pl. I and for pl. 2 respectively. The informal paradigm /parti/ distinct from both /dua/, /par/ of Je dois Je pars, and
o dan /dv-/, /part-/ of vous deoez; vous parter, or posit that the infini-
vu don-ie tive endings are /-uar/ and /-irl respectively, the stems being
i dan in that case those of present pl. 2 /dv-/ and /part-/. Past parti-
ciples of verbs with /-r/ infinitives may either be said to end
is identical with that of the present with the exception of in zero (/fini/; cf /3 fini/}, in I-i/ (/part-i/; cf /3 par/) or in
pl. 2. Only such verbs as make use of different stems for pre- /-y I (/bat-y I; cf /3 ha/), or be all of them considered bare
sent and subjunctive manage to keep the two tenses distinct stems (/parti/, Ibaty/, just like lfini/).
throughout. The subjunctive is actually in most cases what But the infinitive and the participles are, of course, quasi-
has been called a 'grammatical servitude': its use is largely nominal forms that stand aside from properly conjugated
restricted to certain grammatical contexts from which other ones. The past participle and the infinitive are in French so
tenses are excluded: after il faut que, vous donne; and vous frequent that children often learn them as if they were in-
donnere; are excluded; the subjunctive form vous donnie; is dependent words with no formal allegiance to the personal
compulsory and consequently does not contribute any in- moods. This accounts for the fact that obsolescent French
formation. But children learn frequent forms like je sois, je verbs are as a rule verbs which people do not know how to
 LINGUISTIC VARIETY LINGUISTIC VARIETY 1 33

comucate. . . . . . ""Thl''''''''''''' ......." ..

+- ' "Y'11t- ... TT.c>. and are remem- start from the or more than two if the
bered and used if necessary. infinitive, the form under the verb is quoted, is taken
14 in consequence, we dissociate conjugated from non.. as a one is said to
conjugated forms and concentrate on the former, we may be productive, the one with endings in -e- (-e, -es, -e, -er, &c.) .
summarize our observations by saying that in spoken collo- It corresponds roughly to the one-stem verbs of spoken collo-
quial French all verbs are inflected in the same way, i.e . by quial French, coincidence between productivity and
means of the addition of the same endings to a stem, but that invariability of the stem is, of course, not fortuitous: a
whereas most verbs, such as Je donne (/3 d'Jn/), present the denominative canoter from canot or a loan like dribbler
same stem throughout, others distinguish between a short from English dribble cannot be expected to modify its stem,
stem (e.g. 13 ba/) and a long stem (/vu bat-e/), others still, unless the modification is somehow automatic and conse-
among those with the highest frequency, presenting short quently predictable; e"g./3 driblj with the stem pronounced
stems and different long stems with peculiar distributions as [daibl], future 13 pronounced [dibaloae]. But
(e.g, 13 VEl, Ity val, 13 i-t e], Ivuz al-e/, 13 aj/, for aller). Such this does not lessen the contrast between the two systems, one
isolated remnants as Inu s'Jml nous sommes or Ivu fet/ vous with distinct conjugations, the other with a single conjuga-
faites, where it made sense to identify an l-iu] ending and tion, but a distinction between those verbs whose forms can
a I-tl ending as long as the simple past, with its regular I-ml all be predicted if anyone of them is known, and those for
and I-tl endings in pI. I and 2 (/nuz ala.. ui] nous alldmes, Ivuz which this is not the case .
ala-t/ vous alldtes) was part of the system, must, in the frame Two so fundamentally different systems can be made to
of contemporary colloquial French, be considered unanalys- coexist as two styles of the same language only through a
able amalgams. painful and protracted training, which a nation will impose
The picture is completely different if we consider now on all its new members when it is felt that thereby a precious
literary French in its written form. There is probably little cultural heritage will be preserved.' This is not the place to dis-
need here to list forms that can all be found in any grammar cuss whether the burden could be lessened without affecting
of French. Two tenses, the simple past and the imperfect that heritage, for instance through some drastic spelling
subjunctive, that have been disregarded in the preceding reform. OUf only aim in presenting the two systems was to
survey, re-enter the picture, and thereby the connexion illustrate how different two linguistic forms that purport
between the quasi-nominal items and the conjugated form to be 'the same language' can be.
is largely restored: donne is predictable from Je donnai, parti
from Je partis, couru from Je courus, and vice versa in spite of
exceptions like Je cousis, cousu. But the main difference lies
in the fact that all verbs do not take the same endings: some
take -e, -es, -e in the singular of the present indicative, while
others take -s, . . s, -t. It is therefore necessary to posit different
conjugations, just as in Latin, Italian, or Spanish, and this is
actually what we find in traditional grammars; irrespective
of whether grammarians decide to list two of them, if they
 LINGUISTIC EVOLUTION 135

enter the realm of hypotheses, had had to make up their

minds regarding the nature of phonetic changes, and, in
order to reach a decision, they had been compelled to con-
LINGUIS r c EV LU ION sider the problem of how the obvious, if not total, regularity
of sound changes could be explained. This regularity, whose
reason is obvious as soon as speech is found to be analysable

I
T is customary, among contemporary descriptivists, to
consider that pre-structural linguistics dealt exclusively in terms of a definite number of discrete units, the phonemes,
with the history of languages. This is not quite accurate: was for several decades the subject of strenuous debates,
traditional linguistics, as practised during the last 150 years, which resulted in its being generally acknowledged, at least
has probably more often been engaged in comparing genetic- as a working hypothesis. But since speech sounds were at that
ally related languages and trying to account for some feature time usually considered not to belong to language proper,
of one by reference to some features of another, than in trying people were tempted to look for causes of change that lay
to determine how and why a given language had evolved outside of language, and preferably outside of man; hence
through the centuries. In other words, scholars have been the ever-recurrent references to climate, latitude, altitude,
more inclined to point out correspondences than to explain and the like.
them. It is true that no serious research can be pursued that We have now identified the outward manifestation of
is not based upon observation; history implies the study of language as a perfectly legitimate part ofit, and consequently
documents, and, even for the best known among languages, we are not inclined to account for its vagaries in any other
our documentation is full of gaps. Therefore, it may have way than the one we shall use for other linguistic changes. It
been scientifically safer in many instances to avoid historical would seem that if languages change, as we know they do, it
treatments altogether. is, basically, because the needs of their users change, and
Actually, most linguists, until the dawn of structural lin- this has been found to apply to phonology as well as to lexicon,
guistics, were not aware of any necessity of distinguishing morphology, or syntax. This, of course, involves a total
between diachrony and synchrony, and, quite frequently, they revision of traditional views regarding sound changes.
studied what Saussure called etats de langue without drawing Another basic contribution of contemporary research to
any clear boundary between comparative attempts and dia- the problem of linguistic change is the establishment of a
chronic references on the one hand, and synchronic obser- relation between the frequency of a linguistic unit and its
vation on the other hand. form, so that a change in frequency may involve a change in its
Today, after decades of conscious synchronic practice, it phonic aspect. This, a consequence of the theory of selective
is certainly easier to understand what historical linguistics information, applies to units of both articulations, to phon-
really implies. On the plane of general linguistics, it amounts emes as well as to monemes. It leads to the conclusion that
to determining how and why languages change through since the rise in the frequency of a unit is nothing but a rise
time, and this is what is meant here by linguistic evolution. in the number of times that unit is used, any use ofa linguistic
This problem was at one time a favourite with some unit contributes, even if only in an infinitesimal way, to an
language theorists, but mainly as far as sound changes were increase in its frequency. This automatically entails a lower-
in question. Sober scholars, in spite of their reluctance to ing of its contribution to information and a tendency to
 LINGUISTIC EVOLUTION LINGUISTIC E VOL UTION 137

reduce its cost, that is the output of energy it requires for its order to understand how the influence of technical,
memorizing and production. Such a reduction is likely to economic, or cultural changes may spread to the inner core of
physical aspect of the unit. This language, we have to remember what we do when we want
could be summarized by saying that, strange as it may sound, a linguistic communication to be more specific. If I want
a language changes because it is used. someone to pass me a certain book and if the request 'give me
People as a rule are not conscious of any change taking the book' does not make me get the desired object, I shall
place in their language. When they are told and shown how try to be more specific and say give methe blue book or givemethe
different it was four or five centuries ago, they are apt to blue book that lies on the far end of the shelf I might also replace
wonder how this came to pass and imagine some period of book by another in itself more specific term like octavo. But it is
rapid change from one period of stability to another, the far more usual to achieve specification through the addition
one they imagine they are enjoying. It is not difficult to of new elements to the sequence of meaningful units than
understand the reasons of the common illusion that one's by replacing one term by another. Besides, in the last analysis,
language is stable and homogeneous: people tend to identify a highly specific term like octavo is nothing but the residue of
language and its written form, and would naturally think that a phrase formerly used as a specification of book or some
nothing changes as long as spelling has not budged; as a rule equivalent. People do not, as a rule, invent new monemes,
they do not and cannot remember how they spoke ten or All this is in keeping with the analytic process I have called
twenty years before; everyone is used to and considers normal the first articulation, according to which experience is sub-
many forms and turns which he himselfnever uses, but which jected to an analysis manifested in linear form.
he seems always to have heard; he is likely to brand as Now ifmy request had been presented in a very simple and
'accent' any deviation to which he is not used or which is so somewhat primitive household, there would have been no
considered in his town or province; but as regards other need for any specification, because there would have been
features, his tolerance is boundless. This is all to the good; only one book there, namely the Bible or some almanac.
the chief aim of language is communication; it is greatly But life having become more complex, people need more
advantageous that on the one hand something that hampers books and more specific books. On the plane of language, the
communication should be resisted and denounced, and on result is that, while I might have got along in the past with a
the other hand that something that does not should be dis- simple utterance made up of the injunctive give accompanied
regarded or condoned. by two handy complements me and the book, the satisfaction
Yet no one will deny that under the pressure of technical of my needs will require now the addition of a relative clause
advance the vocabulary of a language is likely to expand; and that of an attribute blue, which is an original predicate
most of us remember a period .when radar and sputnik did used here with the same function as that of a relative clause.
not exist. We may also at times become conscious of the It is thus clear that an increasing complexity of social rela-
obsolescence of a word which was once of daily occurrence. tions will be accompanied by an increasing complexity
But it is more difficult to imagine how the changing needs of of syntax. Division of labour will involve the appearance
man or ·his communities may influence the more intimate of new forms of human and material relations which
fabric of his language, namely the morphology and, above will determine the appearance, in language, of new func-
all, the phonology. tions.
138 LIN G U 1ST ICE VOL u T ION LINGUISTIC EVOLUTION 139

A language like Latin, considered in itself and in its evolu- language and through a limited period, the various conflicts
tion towards its modern Romance representatives, affords and trends in the frame of the permanent needs of human
a fine illustration of how two types of function marking, communication.
corresponding to two successive periods in the evolution of the In order to understand how and why a language changes,
language, combined for a while, before the older type was the linguist has to keep in mind two ever-present and antino-
eventually eliminated by the expansion of the more recent mic factors: first the requirements of communication, the
one: on the one hand, an inherited, formally rigid and strictly need for the speaker to convey his message, and, second, the
limited system of case-endings no longer capable of taking principle of least effort, which makes him restrict his output
care of the expression of all the relations needed in Roman of energy, both mental and physical, to the minimum com-
society; on the other hand, a set of former adverbs gradually patible with achieving his ends.' Now the objection is often
promoted to the role of function marking prepositions con- raised that human activity in general, and linguistic activity
stituting a very handy and expendable instrument, which, in particular, may in many instances be an end in itself:
for a while, eked out the case system, but finally eliminated a play: talking, more often than not, contributes very little
it as an unnecessary burden. to information; many people talk because they like it, not
What this Latin example illustrates is not only the way because they have anything to communicate. But this does
increasing social complexity determines an expansion of the not imply that linguistic evolution is not determined ·by the
functional complexity of the corresponding linguistic medium, principle of least effort. Talking is often just a game, but a
but also the protracted resistance of the language against the game is only worth while if he who plays the game sticks to
reorganization required by new social and communicative the rules. Cheating makes sense only if the game ceases
needs. Inertia combined with redundancy delays the spread to be an end in itself, and there is, for chatterboxes, no reason
to all the parts of the language structure of the repercussions why they should cheat in the game of talking. The rules of
of some initial change. As a result of this, a language is the game of speech are laid down by its communicative uses.
necessarily the battlefield of conflicting trends, the linguistic At every stage, the structure of language is nothing but the
consequences of past social changes running against the im- unstable balance between the needs of communication,
plications of new innovations corresponding to new stages in which require more numerous and more specific units, each
the evolution of society and coming with them to terms of them of comparatively rare occurrence, and man's inertia,
which represent the structure of the language at every step. which favours less numerous, less specific, and more fre-
This implies that it is extremely difficult to trace linguistic quently occurring units. It is the interplay of these two main
causality back to its ultimate social antecedents. Linguists, factors that constitutes the essentials of linguistic economy.
once they have ascertained the decisive influence exerted by We shall therefore concentrate on language as a communica-
social factors on linguistic structure, should not try to do tive tool, since this use of language gives it a form likely to be
what they are not trained to do and what might lead them imitated in all its other uses. We may thus posit, as the basic
into the realm of unverifiable hypotheses, namely to examine principle of language economy, that the amount of energy
the details of that influence and venture into the field of
I The basic treatment of least effort, also in reference to language, is still
cultural history.
George K. Zipf's Human Behavior and the Principle of Least Effort (Cambridge,
Their real task is to observe and describe, within a given Mass., 1949).
14° LINGUISTIC EVOLUTION LINGUISTIC EVOLUTION

spent toward linguistic ends will tend to be proportionate to indication of the past contained inyesterday and dispense with
the amount of information to be conveyed. tense endings. The more so, ofcourse, ifwe have learned to do
might be as if no feature or unit in a so as . ., . ., .
language would endure unless it contributed to communica- Redundancy is indispensable for the transmission of
tion a share corresponding to the efforts required by its language to new generations of speakers: a good many words
memorization and production. But this is palpably untrue; are learned through identifying certain aspects of certain
every language carries along a heavy load ofmost uneconomic situations, say a given animal, the horse, with a certain vocal
forms which, for various reasons, have been retained through- product. But thousands of others, which do not correspond
out centuries. Some are learnt by children at an early age, to concrete objects or beings, are normally learned through
before they are able to use analogy, when the language they redundant contexts, hungry, for instance, iflearned by a child
use is so poor that its economy is bound to be very different when listening to such utterances as he's hungry; give him some-
from that of more-advanced users; others are forced upon us thingto eat, or I'm hungry; when do we have dinner? Learning new
by reasons of prestige; others still are preserved because no words through redundant contexts is a very common occur-
indication is to be found anywhere as to how they could be rence throughout life, whether this takes place in one's own or
adapted to current needs. But the main reason why the in some foreign language.
energy spent is not strictly proportionate to the amount of The importance of redundancy does not, of course, in-
information intended by the speaker derives from the fact validate the concept of language economy, but reminds us of
that communication hardly takes place in ideal situations; its complexity; what is easiest is not necessarily logically
some amount of noise, man-made or not, interferes with the simplest. It would be totally erroneous to identify informa-
identification ofthe successive linguistic units by the audience; tion meant and information actually conveyed.
besides, people often listen intermittently because they rarely All this means that linguists, in their efforts to understand
ascribe more importance to what is said to them than to their how language changes, should undoubtedly take advantage
own latent or conscious preoccupations. Depending on of information theory, but that they should thereby keep in
various conditions, messages will be more or less repetitious, mind that, in any speculation involving cost, i.e, in our case,
and this will be determinant for the average handling of energy spent toward storage and production, or, if we prefer,
language. Redundancy, in various forms, is a basic necessity memorization and actual use in speech, they will have to
of linguistic communication. As a matter of fact, human avoid attempts at mathematical formulation. There are two
nature is such that a total elimination of redundancy might, reasons for this: first and foremost, we have no way of quanti...
in many circumstances, entail a serious increase of the energy fying that energy except in a very approximate fashion; we
spent on speech; as popular practice indicates, it is probably may point out, for instance, that, everything else being equal,
easier to use double or triple negatives (1 haven't seen nobody ...) the addition of a new phoneme to a succession of such entails
than to order one's utterances in such a way that there the expenditure of extra energy: feed 'costs' more than fee.
should never be any repetition. Once the word yesterday Secondly, even if we could achieve some quantification,
figures in an utterance it is certainly less trouble to let the other factors than the ones involved in information theory,
aura of past action pervade all that follows and determine the factors which might, for instance, tie up with sociology,
choice of past tense than to take advantage of the precise would remain so potent that we would not know what to do
 LINGUISTIC EVOLUTION LINGUISTIC EVOLUTION 143

with our hard-won mathematical accuracy.. Therefore, in likely . This means that the information contributed by a
language dynamics, we shall have to be satisfied with deter- unit depends on its probability in the context where it
mining in what direction the variation of certain factors is occurs. more probable a less informative it is..
likely to cause that of others.. These variables are (I) the The probability ofa unit is presented in terms offrequency.
number of the units among which the speaker chooses at a The relations between number of units or frequency of
given point in the utterance, (2) the probability of units, units on the one hand and information on the other hand
which we shall in practice identify with their frequency, follow, as it were, from the very definition of what informa-
either in a definite position or generally, (3) the cost of each tion is. They are necessary and,should we say, automatic.
unit, and (4) its information.. If the frequency of an item decreases, its probability must
We call information whatever reduces uncertainty through necessarily decrease too, and nothing can prevent its informa-
the elimination of certain possibilities.. This means that in- tion from soaring. The relationship between information and
formation is not the same as meaning. If I say hehasp . . . . and cost is of a totally different nature; if the information of a
stop short, p has, ofcourse, no meaning, but it carries informa- unit is affected either because of some increase or decrease
tion because it excludes the possibility that the utterance may in the number of units with which it is in competition, or
have been meant as he has given or he has seen.. If I say he haspr because of some change in its frequency, we may expect that
....., r has no meaning, but, again, it contributes information speakers will be tempted to let the output of energy in their
because it eliminates hehaspushedor hehasplaced. This implies production of that unit be partially determined by its new
that what is going to be said about the dynamics of language informational role. In other words, people will be ready to
applies to all linguistic units, distinctive or significant, pho- pay more for more information to be conveyed and inclined
nemes and monemes alike. to pay less if information decreases. But it is by no means
As is well known, the more numerous the units among certain that they will, in all cases, be in a position to yield to
which the speaker may choose, the more informative each of that temptation: slurring, in case of a decrease in informa-
them will be: if someone is trying to locate a place on a map, tion, seems excluded as long as the unit in question, a word,
and it: in my directions to him, I am only allowed to use the for instance, is still in competition with a large number of
four cardinal points, telling him to look south will limit his similar units which have to be kept distinct. Only where the
research to one-fourth of the expanse; but, if I may also use choice is limited to two or three possibilities, as when only
compounds such as south-west and north-east, his research yes or no can be expected, will the barriers raised by phone-
will be limited to one-eighth of the map. His uncertainty will matic articulation crumble down, and grunts such as uh-huh
thus be reduced, and we may say that each one of the direc- take the place of a neat succession of distinctive units. If
tions in the eight-point system is more informative than each phonemes cannot be tampered with, the number of them
one in the cardinal, four-point, system. may, in specific instances, be reduced through abbreviation,
Now if, in a certain place, the wind blows from the west as when professor becomes prof and underground railway, under-
just as often as from any other direction, the information ground, or through replacement, as when tube replaces under-
contributed by west in reference to the wind will be that of ground. But there are cases and situations where, for various
each unit in a two-unit system; in other words, it will be reasons, nothing can be done: it seems that, in America, no
smaller than if north, south, east, and west had been equally one has been able to find a handy and generally acceptable
144 LINGUISTIC EVOLUTION LINGUISTIC E VOL UTION 145
substitute for the comparatively frequent elevator-operator, and It is quite that the of a language
there are many situations where a professor is always a pro- are of different degrees of complexity, and that this should be
fessor and never a prof. If, now, the information of a word in- one to understand
creases, i.e, if its frequency diminishes, the bulk of its signijiant dynamics of a phonological system. I But when dealing with
may well be preserved unchanged: the French for 'spinning- phonological complexity it is easier and safer to look for it along
wheel' is still the three phoneme rouet,just as it was in the days the line of successivities and operate with the assumption
when the word was in daily use, and this is likely to endure as that two successive phonemes require more energy than one. It
long as no homonymic conflict arises. is clear, on the other that two successive phonemes
Statistically, it seems true that, in all languages, the sum give far more information than one: we get much closer
of the phonematic segments needed for the 500 most frequent to I have praised if we perceive, after I have, two phonemes
words is smaller than that obtained for the next 500 lexical JprJ instead of just JpJ. come to the same conclusion if
items; which means that frequent words are, on the average, we start from JprJ as a whole whose frequency we compare
shorter than rare words. But if the relationship between fre- with that of, say, JpJ. We may assume that, in a corpus of
quency and cost were of the same type as that between fre- English utterances, there would be about twenty times as
quency and information, we should be able to state that, for many Jpl's without a following JrJ than Jprl's, which means
a given frequency, a: word should have n phonemes, no more that the information conveyed by JprJ would be considerably
and no less, and that a word of n phonemes should definitely higher. Therefore, it is worth while spending more energy on
belong to a given frequency range. This is, of course, not the JprJ than on JpJ. Now, let us assume that, for some reason or
case; conspicuous with its four syllables and eleven phonemes other, the frequency of JprJ increased so as to become equal
is so much more frequent than dinosaur with three syllables or comparable with that of JpJ. Speakers would certainly be
and seven phonemes. We have here a clear indication that a tempted to reduce the amount of energy devoted to the pro-
strictly mathematical treatment of the problems of language duction of JprJ and make it similar to the one used for JpJ.
dynamics is not practicable. But the unmistakable existence But dropping JrJ would not do, because JprJ and JpJ would
of an inverse relationship between frequency and linguistic then be confused and, on account of the high frequency of
complexity is a most precious discovery. both, misunderstandings would be so frequent, that speakers
This applies to all linguistic elements in so far as they con- who, in principle, want to be understood, would correct them-
vey information and to distinctive non-significant elements as selves at once, so that JprJ would never get a chance to be
well as to meaningful units, and it is' perhaps advisable to reduced. A reduction could only take place if either JpJ or
illustrate this relationship, first by a reference to some phono- JprJ could be induced to become something else without
logical phenomena: the advantage of phonological illustra- creating conflicts: if IpJ for instance, become JfJ with-
tions lies in the fact that everything is so much simpler there, out conflicting with any previous labial fricative, then JprJ
with a definite number of phonemes per language and no could become JpJ without creating any confusion.
necessity to reckon with meaning. This would seem to out- Now a sudden spate of Jprl's is something I have never
weigh the disadvantage resulting from the fact that many heard mentioned in any language at any stage ofits evolution.
philologists feel more at ease among meaningful than among
I Cf. G. K. Zipf, The Psychobiology of Language (Cambridge, Mass., 1935)
distinctive segments, among words than among phonemes. and Economic, pp. 132-8.
811928 L
 LINGUISTIC EVOLUTION LINGUISTIC EVOLUTION 147

cases proliferation of some consonant clusters are to notice that at the

on record in the case of the so-called 'geminates', i.e. the place in England, the as a nation had hardly any
same consonant at the end and at the beginning of two contacts the bulk of users of other Germanic languages..
successive syllables. This indicates that the adoption of that particular solution
In modern English, geminates do not exist except at the of the problem must have been by such general
juncture of two words, as in pen-knife /pen-naif/ or head-dress. features of the language pattern as the phonematic build of
But in many languages, such as Italian, Hungarian, or words and the of vowels which at the
Finnish, geminates are perfectly normal within the same beginning of the second millennium after Christ must still
moneme, Geminates do not seem to have existed in what is have been largely similar throughout the Germanic domain.
reconstructed as Proto-Indo-European. But they were a In Latin, the ofgeminates had begun earlier, I
normal feature of ancient Greek, Latin, and the older Celtic at a time when word final vowels had not been reduced, as it
and Germanic languages. It seems that, at some time or other, was in Germanic the process ofreadjustment started, and
the speakers of every one of these developed a habit of replac- this difference in the eCOl101l1Y of word forms con-
ing single intervocalic consonants by geminates in words felt tributes to the divergent paths followed in this con-
to be 'expressive":' the word knock, for instance, attested in nexion by Romance and Germanic. The assumption that
Old English as cnocian with a single c, appears in Middle Gaulish, the linguistic substratum of most of Western
English as knokke, with a geminate which accounts for the Romance, had a finger in the pie cannot be ruled out: the
quality of the modern vowel (from cnocian, one would have main initial features of the consonantal evolution of Western
expected *knoke rhyming with oak). In the course of centuries, Romance and Brythonic Celtic are so much alike that one
these geminates ceased to be optional, and, as shown by the is legitimately tempted to assume that Gaulish, another
ME. knokke, became perfectly normal features of the words Celtic form of speech, as it was being displaced by Latin, was
where they appeared. When added to the bulk of the ones undergoing what has been called 'lenition', and lenition,
that had developed through various phonologically regular a weakening of intervocalic consonants, would be nothing
processes, their frequency became comparable to that of but the outcome of a situation where geminates had begun
their simple counterparts. In other words, the information to be practically as frequent as their simple intervocalic
of intervocalic /-kk-/ became similar to that of intervocalic counterparts. TIle Romance consonantal shift
/-k-j. In English, as in the other Germanic languages, the would accordingly be nothing but a reflex of a Celtic shift.
burden of the distinction between /-kk-/ and /-k-/ was gener~ Yet the structural conditioning of the Western Romance
ally shifted over to the preceding vowel: OE.. socian became change must have been largely present in late Latin, and
ME. soke, with the vowel lengthened, hence NE. soak with Celtic may been just the last straw.
[o"], while ME. knokke preserved its short vowel, hence the The Western Romance outcome is a general simplifica-
non-diphthongized vowel of NE. knock. The distinction be- tion of the geminates, as a rule without any confusion with
tween j-kk-/ and /-k-/ could then disappear and the balance corresponding single consonants. In the case of stops and
between information and cost be restored. It is interesting
I Cf. A. Graul', Les Consonnes geminees en latin (Paris, 1935), where the expan-

I Cf. A. Martinet, La Gemination consonantique d'origine expressive dans les langues sion of gemination, supposed to be a Proto-lndo-European phenomenon, is
germaniques (Copenhagen, 1937), pp. 29-44, 104- 47. examined. On gemination in Romance and Celtic, see Economic, pp. 257-96.
 LINGUISTIC E VOL UTION 149
LINGUISTIC EVOLUTION
soon as phonological changes are considered within the
the jg"ellllIla tie
'I

tion: late Latin totta becomes toute, whereas the single frame of the language system, the whole of the consonantal
consonant is shifted: Latin tota yielding Spanish toda. Nasals of Western
' - ' ... \.AC"" ... '-J' ... .a. appears as ultimately
and liquids, for which, in the absence of any opposition of conditioned by the economical necessity of reducing the
voice, the conditioning was largely different, present diver- energy needed by the articulation of geminates to the amount
gent treatments which point to later evolution: Castilian has compatible with their informative value. It is not meant
hereby that other factors have not been at work. To say that
palatal reflexes of /-11-/ and I-nn-j and retains 1-1-1 and I-n-I
with their old values; Portuguese shifted the former to 1-1-1 the difference between the Castilian and the Portuguese
and I-n-I, while Latin single 1-1-1 and I-n-I were eventually treatment of /-11-/ is due to chance would amount to stating
eliminated as segments, although the nasal tamber of I-n-I that the factors that have determined its shift to /-1..-/ in one
and, possibly, the velar quality of 1-1-1, must have been car- language, /-1-/ in the other have not been identified yet. In
ried over to neighbouring vowels, as we see from siio-csanum. a number of cases where French deviates from its partners,
French shifted the burden of the distinction on to the pre- the more intimate contacts of that language with Germanic
ceding vowel, whose quality differed according to whether forms of speech must have at work; lengthening and
it was in a former open or checked syllable (OFr. pele 'shovel' diphthongization of accented vowels in open syllables seem
from Lat. pala, but balle from germanic balla, sain from sanum, to be good illustrations of these, But the diverging treatments
but pan from pannum). This is, of course, parallel to what we are nothing but locally determined reactions to the unbalance
have found for Germanic at a much later period. In the case deriving from the high frequency and low informational value
ofgeminates, '
of I-rr-I versus I-r-I, the process of simplification is still at
work in many quarters. In the case of French, the burden
of the distinction could not, in this case, be shifted to the The existence of an inverse relationship between frequency
preceding vowel because many /-rr-I were the reflexes of and linguistic complexity is abundantly illustrated on the
former I-tr-I and I-dr-I, as in pierre, from petra, and lierre plane of grammar, I Statistically first, when it is found that
«l'ierre), from hed(e)ra. Since I-tr-I and I-dr-I did not 'make grammatical items, each of which is on the average incom-
position' in Latin, the preceding vowel was in an open parably more frequent than is the case with lexical items,
syllable and was treated as such, hence the diphthongs of are as a rule much shorter than the latter. Minimal com-
pierre and lierre. But since the vocalism preceding the gemin- plexity, for a moneme, would seem to amount to formal
ated I-rr-/ in pierre was the same as that preceding the single inexistence, and this indeed is what we find in the case of
/-r-I ofjiere, fromftra, the burden of the distinction was left to such grammatical functions as are marked by position: in
the consonant, It remained there, in some form or another, in languages with a subject, i.e, a formally characterized com-
Parisian French, down to the seventeenth century, when pulsory actualizer of the predicate, no normal complete
/-rr-I must have passed first to a long uvular r which affected statement can exist without one. This determines a maximal
the tamber of some of the preceding vowels: today the inter- probability which reduces the information to zero. Of course
vocalic consonant is the same in marron and parer, but most the choice of a given subject among all the ones available and
older Parisian speakers have lal in the former, and I-a-I in I Cf. A. Martinet, 'Linguistique structurale et grammaire comparee'
Traoaux de l'Institut de linguistique, i (1956), pp. 7-21. '
the latter.
 LINGUISTIC EVOLUTION LINGUISTIC EVOLUTION

likely in the particular context may furnish a large amount or vowel, but dropped after a continuant (hence Greek
of information. But the presence of a subject as such affords animate nouns with final -Yjp, -wv, -Yj « -0: < *.. eH)), and
none whatsoever, and it is normal that it not be posi- a previous stage when . . s was the universal ending of the
tively marked by a specific speech segment or any formal nominative of animate nouns.
modification. Consequently, a zero ending for 'subject-case' Whether the preceding reconstruction is accepted or not,
is what should be expected. We find a similar situation when we are faced with the existence, in early Indo-European
we consider the 'object'. The object is not 'quite as compul- languages, of an -s nominative limited to the designation of
sory as the subject: eat is never without a subject, except ill beings or objects conceived as animate. This limitation sug-
the imperative, but it may appear without an object. Yet, gests that in Proto-Indo-European this -s marked a function
once a given 'transitive' verb is chosen by the speaker, the felt to be characteristic of animate creatures, namely that of
only information furnished by the object as such results acting, as opposed to the passivity of things. Such -s forms
from the fact that it might have been suppressed. But in could hardly be subjects, i.e, forms used with the function
contexts where an object is actually found in 99 per cent. of of compulsory actualizer of a predicate, because the zero
the cases, its information is close to nil if considered in its information of a subject as such would hardly be compatible
object-function, and not as a specific choice among a variety with a formally existing indicator. All this points to -s as the
of lexical or pronominal items. It is, therefore, normal that mark of an ergative case, the case of the agent, which must
the most frequent type of object should not be marked by any have coexisted with a true nominative, the form of the noun
specific segment, provided it is kept distinct from the subject used in order to introduce or 'nominate' a person or to pre-
by means of word order. sent a creature or an object. This true nominative, being used
This indeed corresponds to what we find in a majority outside of grammatically organized utterances, was in no
of languages with subjects. But there is an exception that definite relation to a predicate and had therefore no gram-
certainly looms large for classical scholars, namely the posi- matical function properly so called. It must have been
tively marked nominatives and accusatives of classical Indo- identical with the bare stem of the word and, as such, identical
European languages. If we leave out the plural cases which with the vocative which also corresponded to a use of the
raise special problems, we may state that, with the exception noun outside of grammatically organized utterances. In
ofsome neuters, the accusative, which is the case ofa majority languages that make use of an ergative case for the agent, I it
ofverbal objects, ends in -me The nominative, often identified is frequent, ifnot universal, to find the bare stem used for the
as the subject-case, ends in -s in the majority of masculine most obvious complement of the predicate; in an utterance
and feminine nouns. Ifwe oppose masculine and feminine to corresponding to 'the woman is washing the linen', the
neuter as animate opposed to inanimate, we may say that woman, as the agent, will be a complement in the ergative,
animate nouns ending in stops and prehistoric vowels present but the linen, as the 'most obvious complement', will be in
an -s nominative (Latin pleb-s, cioi-s, ficu-s); others, which the nominative. The nominative, here, is no longer the sub-
have what is often called a zero ending, have a stem-final ject case, but the bare stem as used in non-grammatical
+
consonant or an -a which must be the reflex ofvowel 'laryn- I Basque is an example; for an evaluation of the ergative construction in

Basque, see the author's paper 'La Construction ergative et les structures
geal' , the latter probably some fricative continuant. This elementaires de l' enonce', Journal de psychologie normale et pathologique (1958),
points to a prehistoric stage were <S was preserved after stop PP·377-9 2 •
 LINGUISTIC EVO N LINGUISTIC E VOL U'TION

contexts: in a word-for-word translation, the equivalent of tic form with a compulsory subject, which we find in attested
the former utterance would sound like washing [if] (the) languages, may have been determined by the develop...
linen by (the) woman or linen washing by woman. intran- ment of the a subject we cannot investigate here.
sitive predicates, 'the most obvious complement' is identical The use of the rS form instead of the bare stem may have
with what we call the subject: 'the man talks' is presented as expanded first to constructions with intransitive verbs and
talking [if] (the) man or man talking, where man is treated as the ultimately to the properly nominative uses. The gradual
normally expected complement, and consequently presented elimination of the bare stem from its vocative uses and its
without explicit mark of its function. replacement by the -s form is an historical process which is not
A similar situation must have existed at some stage of quite completed yet: in Czech, for instance, the vocative
Proto-Indo-European: the bare stem was used out of context clovece 'man!' is still distinct from the nominative c!ovek (whose
as a nominative properly so called and as a vocative and, with -s, by an irony of fate, has long since disappeared). Formally
a grammatical function, as the most obvious complement this expansion of the -s form resulted in a twofold anomaly:
corresponding to the direct object of our transitive and the the <S form was by now a subject, i.e. a form with no gram-
subject of our intransitive verbs; the -s form was used for matical information any more, but with a functional mark still,
marking the agent, but there could be an agent only if some and the bare stem of animate nouns, restricted now, in gram-
being or object (explicit or understood) was being acted upon: matical contexts, to the function of complement of transitive
in equivalent of man talks', the man was not treated verbs, was deprived of any grammatical sign although its
as an agent. The working of this is clarified if we think in occurrence was somewhat less automatic than that of the
terms of nominal predicates and parallel talking if man and formally marked subject. This latter aspect of the anomaly
washing oflinen. Nouns designating objects or beings that were was soon remedied through the replacement of the bare-
never conceived of as agents would not get a chance to stem by a to-case in -m, the attested accusative (cf. Spanish
receive the -s tag. We are so used to identifying agent and 'direct' animate objects in a: ueo a Pablo '1 see Paul'). This
subject, that it is difficult for us to remember that an instru- was analogically extended later to neuter o-stems.
ment as such can never be an agent. But the speakers of The other lack of balance between information and cost
Proto-Indo-European would not have been misled into using was more difficult to eliminate: -s forms, the bulk of the new
the designation of the sieve (Lat. cribrum, GE. hrider n.) in the nominatives, were extremely frequent, early learned, and
ergative -s case; a sieve in action would, of course, be used in consequently resistant: if word final s dropped out at all, it
an instrumental case. IE. neuters are properly the nouns could only be as the result of a regular phonological change,
which, on account of their meanings, were never used in the but not just in the particular cases where it was the nomina-
ergative. some lin.guists would put it, the absence of an -s tive function marker. Its regular prehistoric elimination after
form must have been unfait de parole before it became unfait sonants was welcome: it may well have been dropped only
de langue. A neuter noun like Latin mare is used as a bare stem in certain contexts (utterance finally, for instance, or before
both in mare videt 'he sees the sea' and mare patet 'the sea is to consonant ofa following word), but, ifso, the forms without
be seen', and this mirrors a former stage of Indo-European -s must have been eagerly extended by analogy to all posi-
syntax. tions. In thematic nouns, i.e, after the elo vowel, it is well
shift from a former ergative construction to the linguis- attested in many Indo-European branches. But, in later
154 LINGUISTIC EVOLUTION LINGUISTIC EVOLUTION

stages, whenever speakers are given a chance to choose, its frequency in speech became such that the cost of its full
among phonologically regular forms between an -s form and designation exceeded by far what corresponded to its, by then,
an -r-less form, they give preference to the latter. Still, in informational No curtailing was possible
many cases, centuries elapsed before phonological evolution since the combination of any two of its elements would have
offered any such chance: in French, one had to await the been misleading. The use of the three initials, probably first
elimination of declension before -s singular nominatives in writing, was certainly the most obvious and, in this com.. .
yielded to -s-less oblique case forms. plex world of ours, the most natural solution.
In common speech, where no spelling can suggest the use
In the lexical domain, abbreviation of segments whose fre- of initials, there may occur similar situations where none
quency is on the increase does not raise problems, as a rule, of the components can be left out, whatever the increase in
if those segments are phrases of the underground railway type frequency. Since the phonematic bulk cannot be reduced,
where the more general term can be left out and the specifier speakers resort to a reduction of the morphological com-
preserved as the equivalent of the whole. The reverse may plexity and coagulate some of the elements into a single
also happen, the more general term being preserved, if the maneme or, at least, handle these elements grammatically as
object that the phrase designates becomes so universal that if they were one. The French phrase avoir l'air 'to look (like)'
the general term may be used, without any danger of con- affords a nice illustration of coagulation in two successive
fusion, as its normal designation: in most countries, motor- stages. This phrase, in daily speech, must be about as frequent
cars were first referred to as 'automobile carriages'; this as its English equivalent and it has practically eliminated its
phrase was soon curtailed to automobile and, finally to auto. traditional and literary competitors sembler and paraitre. In
But nowadays, after the practical elimination of all horse- its most usual form, the third person singular of the present
drawn carriages, the usual designation of motor-cars is the il a l'air jilaler j, it cannot be said to be phonologically more
old word for 'carriage'; Fr. voiture, Germ. Wagen, and so complex than what its frequency would lead us to expect,
forth. If the designation is a learned word made up of but its semantic articulation is cumbersome, as is shown by
elements that were significant in Latin or Greek, but make the behaviour of speakers: in 'unguarded' colloquial French,
little sense for the average user of modern languages, the it is made to behave as a sort of copula; for 'she looks nice',
curtailing will be made at random, as far as etymology is con- one would expect elle a l' air gentil, with gentil agreeing with
cerned, but in such a way as to get a handy form whose size is the masculine air; what one actually hears is elle a l'air gen-
well suited to its frequency and which rules out any danger tille, with gentille agreeing in gender with the subject elle.
of homonymic confusion.. With many children, the process of coagulation goes further:
Another frequent trick is the use of initials, as in LCC or air is no longer identified as a noun, but as the second syllable
USSR. These are particularly favoured among specialists, of the stem alair-jaler-j of a regular -er verb *alairer, hence, in
but are often resented, not only by purists, but by large sec- the imperfect, fa m'alairait bon, 'it looked good to me', instead
tors of the general public. Still they often represent the of standard fa m'avait l' air bon.
normal way out of some informational quandary: when the We have, so far, been considering what happens to a form
London County Council became a generally identified con- when its frequency increases. In cases where it decreases, we
cept that had to be frequently referred to by many people, have pointed out that this does not necessarily involve any
 LINGUISTIC E VOL UTION LINGUISTIC EVOLUTION 157
formal change provided no homonymic conflict arises . But a remuer, bouger, demenager, all single-stem -er verbs; another
lowering of the frequency which, in itself, would not endanger literary equivalent of move is emouooir which survives, in
the survival of a form, may seal its doom if it is somehow speech, as an two participles: emauuant and'
isolated in the structure, if, for instance, it is a verb that is emu; but even these are threatened by the forms of the
not inflected according to a widespread analogical pattern. denominative single-stem emotionner. I
When children learn their language, they tend to imitate We have so far, in the case of significant units, been
comparatively long stretches they have heard before they are operating with general frequency. But in order to under-
able to analyse them, i.e . to use in other contexts the elements stand how language works and changes, it is quite essential
they contain. As long as they do not use analogy, which to operate with frequency or, better, probability in given
amounts to contriving new moneme combinations, they have contexts: book and walk are frequent words, but a combina-
no problems with irregular forms . But when they start doing tion like the book walks is quite unexpected and either non-
it extensively, which, with the average child, will begin about sensical or highly informative because it is quite improbable.
the age of five, they will be apt to replace an irregular and In other words, the informative value of the book walks is
isolated brought by an analogical brung, Irregular forms that higher than what a calculation based upon the general fre-
are. very frequent are, by that time, well entrenched, and quency of book and that of walk might lead us to expect. On
occasional slips like brung will eventually disappear. But the other hand, blue and sky are so frequently used together
those whose frequency is somewhat lower will be widely that blue sky contributes less information than the average
mishandled. If the community is not of a conservative type, probability of blue and that of sky would suggest. According to
children will be allowed to use their own forms, and ana- whether the speaker or the writer wants to startle his public
logical forms will finally get established. I~ on the contrary, or, on the contrary, to flatter his laziness and conservatism,
adults insist on correcting children, the latter will try to he will try to find unexpected combinations, or use time-
avoid the forms they do not know how to inflect to the satis- hallowed phrases and turns. Generally speaking, cultured
faction of their elders; these trouble-makers will become audiences or publics are those who, at the same time, are able
obsolescent and eventually disappear. English move was bor- to digest more information per second, and need more un-
rowed from French and was from the start inflected accord- expected combinations in order to get the same amount of
ing to most general analogical patterns. It is today one of information, since they hear new turns more often and con-
the most frequently used lexical items of the English language. sequently find them much sooner hackneyed and stale. It is
The original French mouvoir must have had, from early times, thus clear that style is largely a matter of density of informa-
a number of unusual inflexional features. In any case, there tion, and this covers to a large extent the subject-matter ofthe
is today no other verb conjugated just like it. It is about as experience, real or fictive, which is communicated, since the
peculiar in its inflexional comportment as pouvoir or couloir, understanding and appreciation of literary communication
yet it must always have been somewhat less frequent than implies previously shared experience, and shared experience
these quasi-auxiliaries, and the amount of irregularity which implies familiarity with a certain vocabulary. Whether com-
speakers could cope with in handling these proved a little munication is of a literary nature or just an everyday spoken
too high for mouvoir. As a result, mouvoir has disappeared from
I For an illustration of the working of formal analogy, see 'De I'economie
colloquial French; the normal equivalents of 'move' are des formes du verbe en francais parle', Studia ... in hone L. Spitzer, pp. 3 I 5 ff.
 LINGUISTIC EVOLUTION LINGUISTIC EVOLUTION 159

affair, it involves a constant renewal of the details of lin- seen that a similar frequency of geminates and corresponding
guistic practice, because everyone of us, author or speaker, single consonants is only one of the factors that determined
will find it necessary to emphasize some segments of his the simplification of the geminates: simplification can
utterances through additional information resulting from only occur if no dangerous confusion will ensue, which
unexpected collocations. If the trick proves successful, the largely depends on the phonic nature of the other units of the
inventor himself or some of his hearers will be tempted to system. All this means that applying the statistical technique
use it again, and the new turn may get established in the of informational research to language should not make us
language. But should it become too popular, it would soon forget that semantic and phonic properties of linguistic
lose all its efficiency and be replaced by some fresh innova- units cannot be disregarded when problems of evolution are
tion. There are, at all times, large sections of the vocabulary at issue. The importance of the implications for our research
that are in a constant state of unrest, and since all the parts of recently evolved methods of investigation should not entail
of a language condition one another, we find here a per- a disregard of previous efforts towards the understanding of
manent cause of linguistic evolution. semantic evolution. I
As regards phonology, whole-hearted attempts at estab-
It is tempting to try to reduce all linguistic facts to quanti- lishing a structural method of diachronic investigation are
tative data, as we have been doing. But we should not forget comparatively recent, and they have, if not from the start, at
that what we may call the nature or quality of the various least very early, involved informational considerations." A
linguistic units plays an essential role in the conditioning of basic problem, with phonological changes, is to explain how
linguistic evolution: what can be measured is the quantity, the changing needs of man can affect the second articulation
not the quality of information. If a new word enters a of language, one of the main functions of which is precisely
language, its appearance will necessarily modify the in- to make the phonic form independent of the semantic value
formational economy of that language. If there were n words of the message and of its significant components. The hypo-
before, there will now be n+ I words. In theory, everyone thesis that frequency could modify cost, i.e. exert an influence
of the n words should have become less frequent in the pro- on the form of the units, phonemes as well as monemes,
cess. But, in fact, the frequency of all but very few of the suggested that this was at least one of the channels through
n words will not be affected. Only those whose meaning is in which communicative needs could act upon the economy of
some way connected to that of the new-comer are likely to phonematic patterns. Yet there are other channels, and we
score a lower probability. The impact of the lexical expan- must keep in mind the role played by prosodical features
sion will be limited to a certain semantic domain. Besides, we such as accent and intonation which, being direct responses
cannot be satisfied with the statement that the appearance of to communicative needs, are most likely links between these
a new word A has reduced by so much the probability of a and the phonemes." But what is probably more important
formerly existing word B, because this does not tell us what than determining how external factors of unbalance pene-
semantic field remains allotted to it and what section of its I Cf. Suzanne Ohman's survey paper 'Theories of the "Linguistic Field" "

former domain has been encroached upon by A. This we Word, ix (1953), pp. 123-34.
2 See Economic, pp. 139-47.
need to know if we want to understand the further repercus- 3 cr. Manual of Phonetics, ed. by Louise Kaiser (Amsterdam, 1957), pp. 255-
sions of the adoption of A. In the field of phonology, we have 62.
160 LINGUISTIC EVOLUTION

trate the system is the examination of the repercussions of

that unbalance within the system. When trying to account
for past linguistic changes, on anyone of the planes of I E s
language, there will come a point where we will be at a loss
to go farther for lack ofprecise historical information. It is not accent, 34-35. dialectj, 112- 18.
'accent', viii, 38. dialects, 112-18.
suggested here that we should not try to cull useful hints accentual typology, 85-87. discontinuous signifiant, 92.
wherever we can find them, but we shall achieve the best accidence, go. discrete units, 24, 33, 37, 39·
accusative (IE.), 153. distinctive features, 30.
results if we concentrate on a domain we know well and are active, cf. voice (of verbs) . double articulation, 24, 26, 28, 30, 59.
best equipped to investigate, namely language structure In actualization, actualizing, actualizer, duration, 3 I .
all its aspects. 42 , 48, 61-63.
adjective, 64-65. economy (linguistic), 2, 139, 141.
amalgam, amalgamation, 46, 47, gl. ergative, 151.
analogy, 156. expansion (paradigmatic), 95.
arbitrariness, 88. experience, 21.
autonomous moneme, 45. expression, 39-40.
autonomous phrase, 44. expressive function (of language), 2 I •
autonomy (syntactic), 44-45, 52.
first articulation, 22-23.
bilingualism, 108, 113. France (socio-linguistic situation in),
bilingualism (dialectal), 116-20. cf. socio-linguistic.
frequency, 51, 135·
case-ending, 45-46. frequency/information, 143.
centrifugal, centripetal, 52. function, vii-viii, 3.
change (causality of), 136-44. function of monemes, 49.
choice,8-g, 16, 17, 30. functional indication, 97-98.
clause (main, subordinate), 50. functional moneme or indicator, 45.
coagulation, 155. functionals v. modifiers, 95-97.
coalescence (syntagmatic), 93.
colloquial speech, 124. geminates, 146, 154-7.
communicative function (of lan- gender, 15-19.
guage), 21. gestures, 28, 30.
community (language), 103-8. glossematics, 39-40.
complement, 49. glottal stop as a tone, 3 I, 85.
composition, 94-95. glottalized consonants, 32.
concord, 53-55. glottis, cf. vocal chords.
conjunction, 45. grammar, 100.
content, 39-40. grammatical monemes, 50-51.
contours (intonational), 30-3 2, 35. grammatical typology, 89.
convergence, 72, 104· grammaticality, 97-98.
core system (of phonemes), 78.
correcting children, 156. hierarchy among linguistic opposi-
cost, 136. tions, 107.
hierarchy (functional), 29.
dependent (moneme), 45.
derivation, 94-95. idiolect, 105-6.
determinants, 50. 'idioms', viii.
'determination', 64. inertia, 138, 139.
dialect, I I I • information, 46, 142-3.
811928 M
162 INDEX OF SUBJECTS INDEX OF CTS
informationjcost, 143-5. phoneme, 6, 23-25, 27--28, 30 , 35-
information theory, 135, 36, 39-4 1 •
inseparability, 92-93. phonetic change (regularity of), 25. vocal nature
intonation, 35-38. phonology, 30. universals (relational), 58. voice, 31.
intonational habits, 37-38. pitch, cf. melody (speech). voice (of verbs), 49, 51, IOI.
intonational levels, 35-37. position, cf. word order. variadon (as a characteris tic of Ian-
inventory (moneme), 50-51. predicate, predicative, predicative word, go, g2.
isomorphism, 39. function, predicative moneme, pre- 0.1"-"'.""1"<,<-,,,,,,9 1 - 93. word order (relevancy of), 4 I, 64.
dicative phrase, 44, 48-49, 62-63. one community, writing, 25-26.
language, 20, 26, 70. 'predicatoid', 50. 106-7· written v. spoken language, 122.
least effort, 55, 139. preposition, 45-46.
lexical monemes, 50-51. primary dependents, 50.
lexical typology, 87-89. primary function, 50.
liaisons, cf. linking. probability in context, 157.
linearity of speech, 21, 22, 23, 24, 25. prosodic, prosody, 29-32.
linguistic communication, 21-22. prosodical typology, 84-87.
'linguistic field', 159.
linguistic reality, 70. quadrangular vowel patterns, 79.
linking, 127-8. quality v. quantity, 158-9.
literary communication, 157-8.
literary language, 105, 124. redundancy, 56,68,140.

main clause, cf. clause. second articulation, 23.

marginal dependents, 50. series (of phonemes), 77, 81-82.
margins, 19-20, 28. 'shifters', 124.
meaning, 95-96. signifiant, 22, 24, 30.
melody (speech), 27-35. signifi«, 22, 30.
minimal utterance, 58-61. signs (linguistic), 20, 22, 30.
modifiers, 51-52. simplicity (principle of), 9-10, 14.
Inoneme, 22, 24,27-28, 30, 36, 39-41. situation, 59, 123-24·
'morpheme', 22. socio-linguistic situation in France,
morphology, 89-90. 113·
motivation, 88. solidarity between phonology and
'mute e' (in French), I I, IS, 20. grammar, 68-69.
spelling (English and French), I2S-6.
standard language, 112.
neuter (IE.), 152.
stress, 3 I.
neutralization (of phonological oppo-
'stress', 35.
sitions),8.
structure, 3-5.
nominative, 151.
subject,44,61,62-63,g6,IoI,I49-S0.
non-primary function, 50 ..
subject-predicate phrase, 48-49, 58.
number, 17. subordinate clause, cf. clause.
successivity of linguistic elements, cf.
object, 44, 150. linearity of speech.
operational devices, 5. suprasegmental, cf. prosodic.
orders (of phonemes), 77,81-82. syllabic nucleus, 30-3 I.
out-of-situation, 60-62. syntactic autonomy, cf. autonomy
(syntactic) .
passive, cf. voice (of verbs). syntax Chap. II, 8g-90.
patois, I I I, 117.
phonematic(s), 29-30. tonal typology, 84-8S-
phonematic typology, 74-84. 'tonemes', cf. tones.
 s s
Abaza, 74. Ibo, 85. Allen, vv. S., 4, 74. ____,... ' ,10.
Adyghe, 79. Indo-European (Proto-), 146, 151, Charles, 88. Andre, 2, 9, I I, 14, 30, 72,
African (Central . . . languages), 85. 152. Charles, 67, 73. 80, 84, 106, 117, 126, 129, 146,
Arabic, 78, 79, 85, 122. Indo-European languages, 150. .IlJV.'JliJlF''-'i, Dwight L., 35. 149, 15 1, 157·
Basque, 98, 119, 122. Iroquois, 82. 1
°5- Antoine, 66.
Breton, I 19. Italian, 78, 79, 80, 81, 87, 91, 146. Cohen, Marcel, 66. Suzanne, 159.
Brythonic Celtic, 147. Latin, 44, 86, 87, 91, 92, 122, 138, Dorfman, Eugene, 78. Pittman, Richard S., 52.
Celtic, 146. 146,15 0 • Edmont, E., 114. Sever, 114.
Chinese, 85. Latvian, 86. Fant, C., 75. .lL'- ..... Ruth, 9.
. 1\..-.1.I.:llA.•.1.1.1,

Creoles, 120- I. Lithuanian, 85, 86. Gillieron, J., 114· Sapir, Edward, 49, 56, 58, 66-67, 70,
Czech, 47, 86. Malagasy, 63. Oougenheim, Georges, 52. 93- 1 0 0 •
Danish, 47, 77, 79, 80, 85· Middle English, 146. Grammont, Maurice, 12. Saussure, Ferdinand de, I, 2, 22, 88.
Dutch, 29. Norwegian, 30. Graul', .A., 147. J.,II4·
English, 6,7,8, I I , 16, 17, 18,22,33, Old Church Slavic, 75. Greenberg, Joseph, 67. Suzanne, 121.
4 2-43,44,45,63,73,75,77,7 8,79, Old English, 146. Han, Robert A., Jr., 121. Douglas M., 121.
80,81,82,85,86,87,89,91,99,101, Picard, 115-16. Halle, Morris, 75. Togeby, Knud, 14.
107, 125, 156. Polish, 86. Haudricourt, Andre G., 74. Trubetzkoy, Nikolai S., 74, 78-79-
Ewe, 121. Polynesian, 74, 82. Hjelmslev, Louis, 2. Voegelin, Carl, 78.
Finnish, 78, 79, 146. Portuguese, 146. Jakobson, Roman, 10, 75. Vogt, Hans, 76.
Flemish, 120. Romance, 146. Jespersen, Otto, 53-54, 124. Weinreich, Uriel, 121.
Franco-Provencal, 109-10, 11'6-18. Romanian, 79, 80. Jones, Daniel, 35. Whitfield, Francis, 2.
French, 9- 19, 22, 29, 33, 45, 58, 74, Russenorsk, 105. Kaiser, Louise, 159- Yegerlehnerv ]., 78.
75,7 6,77,78,79,80,81,82,84,87, Russian, 8, 74, 78, 80, 85, 86, 87, 100. Kurylowicz, ]erzy, 39. George Kingsley, 2, 1 39, 145.
88-89, 9 1, 9 2, 94-95, 99, 104, 106, Sanskrit, 122. Loftman, Beryl, 121.
10 7, 108-9, I I I, 125, 126-33, Serbo-Croatian, 84, 85.
148-9, 154, 155, 156-7. Spanish, 6, 74, 77, 78, 79, se, 81, 86,
Gaulish, 147. 1 ° 4, 148 .
Georgian, 75. Swedish, 30, 85, 86.
German, 8, 22, 23, 29, 45, 75, 78, 79, Tahitian, cf. Polynesian.
80, 83, 86, 87, 88-89, 92, 120, 125, Turkish, 79.
154· Vietnamese, 64, 85.
Germanic, 146-7, 148-9. Welsh, 23.
Ghiliak, 33. Western-Romance, 147-9.
Greek (classical), 81, 82, 83, 146, 151. 'Volof, 121.
Hawaiian, cf. Polynesian. Zulu, 71.
Hebrew, 76.