WORD

ISSN: 0043-7956 (Print) 2373-5112 (Online) Journal homepage: www.tandfonline.com/journals/rwrd20

Elements of a Functional Syntax

André Martinet

To cite this article: André Martinet (1960) Elements of a Functional Syntax, WORD, 16:1, 1-10,
DOI: 10.1080/00437956.1960.11659716
To link to this article: https://doi.org/10.1080/00437956.1960.11659716

Published online: 04 Dec 2015.

Submit your article to this journal

Article views: 2420

View related articles

Full Terms & Conditions of access and use can be found at

https://www.tandfonline.com/action/journalInformation?journalCode=rwrd20
ANDRE MARTINET-----~-----

Elements of a Functional Syntax

Whoever is concerned with the welfare of linguistic research in general

and tries to transcend current parochialism must have wondered whether
the label "structural linguistics" implies anything beyond the rejection of
the traditional approaches to the study of language. If we refuse to be
fooled by the general use of certain terms, we may be apt to conclude that
the various "structuralistic" schools have little in common, except the
attempted reduction of the phonics of the language to discrete units, the
phonemes: a "linguistic structure" cannot be the same object for a dis-
tributionalist, for whom only the respective position of the elements in
the utterance is decisive, and for a paradigmatist, who concentrates on the
type of relationship prevailing among commutable units. Structuralism, in
linguistics, cannot be said to be a misnomer since all the structuralists
operate with some sort of structure. But the use of this term by people with
widely divergent outlooks certainly hampers, rather than promotes, mutual
understanding and wide-spread cooperation.
Yet, discussions with specialists of other branches of learning and also
with linguists of "non-structuralistic" persuasion will suggest that one
basic assumption is shared by those we might call the progressive linguists,
namely that nothing may be called "linguistic" that is not manifest-or
manifested-one way or another in that segment of the communication
circuit that lies between the mouth of the speaker and the ears of the
listener. This circumstance makes contacts among members of the different
structural schools worthwhile and fruitful as soon as good will is secured,
whereas it is apt to prevent lasting cooperation between progressive and
tradition-bound linguists.
Contrary to what seems to be felt in some quarters, this assumption does
not entail that linguists should restrict their field of research to the audible
part of the communication process and disregard what is not observable
there: speech can only be interpreted as such, and not as so much noise,
because it stands for something else that is not speech. Few linguists would
maintain that an acceptable analysis of utterances could be achieved with-
out positing that linguistically relevant variations in speech correspond to
variations in meaning. Yet, the ideal of most structuralists has been to
w.-1
2 ANDRE MARTINET

concentrate on the analysis of utterances with as few references as possible

to what they stood for and, consequently, to identify and classify linguistic
units on the basis of their distribution in the recorded speech segments.
This concentration on analysis represents, in my opinion, a wholesome
reaction against the introspective methods favored by tradition-bound
scholars. Yet, it has been detrimental when not eked out by a consideration
of the communicative process as a whole. The chief aim of language is to
convey information and, if-as I believe-the aim of general linguistics is
to understand what language is, how a language works, how it adapts
itself to new needs and, thereby, changes, no true linguistic science can
disregard this fact.
An illustration may be welcome at this point: if we overlook the com-
municative nature of language; in other words the fact that something to
be conveyed is made manifest by means of something else, we are likely to
miss some basic differences between the behavior of distinctive units and
that of meaningful units: if we consider two segments like tomorrow and
with the papers, we will know they are distinct because one can be replaced
by the other with an accompanying difference in reference, as in he' II come
tomorrow and he'll come with the papers; but they may also coexist in an
utterance such as: he'll come tomorrow with the papers or he'll come with
the papers tomorrow. They are mutually exclusive at a certain point in the
utterance, but they will be found in succession. All of this is attested in the
case of phonemes: pass and pat show /sf and /t/ to be distinct phonemes,
mutually exclusive at certain points in the utterance; past and pats show
that they also may coexist in succession.
There is however a basic difference between the two types of unit: using
pats instead of past, i.e. the succession /t+s/ instead of the succession
fs + tf results in either changing the nature of the message or spoiling it.
On the contrary, using with the papers after tomorrow or before tomorrow
has no effect on the nature of the message. This amounts to saying that,
in the case of phonemes, both their phonemic make-up and their respective
position in the utterance are distinctive, whereas in the case of meaningful
units or segments, the respective position of the elements in the utterance
is largely irrelevant. Of course, it is relevant in many cases: the man kills
the bear is clearly another message than the bear kills the man. The question
the syntactician will have to answer is How is it that what applies to
tomorrow and with the papers in their relation to each other or to the
rest of the utterance does not apply to the man and to the bear? In other
words, why is distribution relevant here and irrelevant there?
Please note that we are not concerned here with what has been called
referential meaning and carefully shunned by a majority of structuralists.
 ELEMENTS OF A FUNCTIONAL SYNTAX 3
Yet, in order to formulate the problem, a central problem of syntax, we
have had to keep in mind the basic communicative function of language.
We have had to remember that the vocal nature of our languages imposes
a linear articulation of non-linear experience: such an experience-an un-
pleasant experience-as a headache has nothing linear about it; but in-
forming the doctor about it implies the linear succession /-have-a-head-ache.
Speaking implies itemizing; it results from an analysis of experience into a
number of elements, each corresponding to a linguistic sign. The signs cor-
responding to a message are, with some exceptions, ordered in a succession.
The phonetic form of each sign is, in its turn, articulated into a succession
of distinctive units. All the languages that have so far been studied by
linguists present that double articulation, and, in my opinion, we should
agree to reserve the term "language" to such tools of communication as
have it.
Reverting to the first linguistic articulation, that according to which
experience is broken down into a succession of meaningful units, it is im-
portant to remember that the success of communication requires not only
that it should include the requisite number oflinguistic signs corresponding
to the various elements of experience, but also that the relationship obtain-
ing between the latter should find there some sort of linguistic expression:
such an experience as witnessing one person introducing some one to some
one else may be summarily conveyed by means of a succession of meaning-
ful units corresponding to such elements of experience as three different
persons (e.g. John, James, and Peter), a happening (the introduction), a
time (e.g. last night), provided the relationships between these elements be
somehow indicated: information must be given, not only about the parti-
cipation of, say, John in the process, but also about the nature of that
participation. If the participation of John consisted in being told by James
who Peter is, the nature of that panicipation should be indicated in English
by using to before the formal element designating the person, say, John,
hence to John.
When this type of relationship is expressed by means of a special segment
of the utterance, such as to in the case of to John, we may be tempted to
interpret to as the expression of just one more element of experience, and
it is hard to see what could be objected to that interpretation. But, lin-
guistically, the status of to is, as we shall see, basically different from that
of John because it combines information with the power of conferring
syntactic autonomy upon the segment that follows.
What, in language, corresponds to the relationships between the various
elements of experience is what has traditionally been called "function" when
we say, for instance, that this or that word functions as a subject or an
4 ANDRE MARTINET

object. Function is, of course, a purely linguistic concept. In other words,

function exists only in so far as it is expressed somehow in the utterance.
As we know, nothing should be called "linguistic" unless it is manifested
somehow in that stretch of the communicative process that lies between the
lips of the speaker and the ears of the listener. But we should not imagine
that the formal expression of function will always be clearly identifiable
as a given succession of phonemes. Function, as we know, is frequently
marked by the respective position of the elements of the utterance. But
even when its indication involves some phonemic difference, it may be
impossible to locate or delimit it exactly in the utterance: in a form like
Latin homini, we are at a loss to tell where the expression of the dative
function stops and where that of the singular begins; we know that homini
is a dative, but we also know it is a singular, because if it were still a dative,
yet no longer a singular, but a plural, we would have hominibus with a dif-
ferent choice of phonemes. In such a case, the indication of the function
corresponds to a meaningful unit with both meaning and form, although
we may find it difficult to formulate what that form is. The difficulties that
we, linguists, may experience in formulating something is no indication
that that something does not exist and that we cannot identify it and
scientifically operate with it.
Minimal meaningful units, both those indicative of function and others
corresponding to the various elements of experience, have sometimes been
called 'morphemes'. But since many linguists would resent applying this
term to a unit whose form cannot always be pinned down, I will rather use
another word, namely moneme, which has been used with that meaning by
some linguists of the Geneva school.

The study of function, as previously defined, is, in my opinion, the cen-

tral problem of syntax, and the first task of the general syntactician consists
in uncovering and listing all possible ways of expressing the function of a
linguistic segment. There seem to be three different types, all attested in the
sentence Yesterday the President spoke in the auditorium. The first type is to
be found in yesterday; for simplicity's sake we shall assume that yesterday
is a single moneme like its French, Spanish, and German equivalents; this
moneme corresponds to a given element of the experience which is being
communicated, the one that might be defined as 'the day before this day',
but it also implies that that segment of time is the one in which the related
event is to be placed. In other words, yesterday is quite parallel to and inter-
changeable with a phrase like in 1950, where the function of 1950 is ex-
pressed by means of the specific moneme in; in yesterday we have a single
moneme whose linguistically unanalyzable meaning implies a given func-
 ELEMENTS OF A FUNCTIONAL SYNTAX 5
tion; since yesterday can be placed in other positions than the initial one,
after spoke for instance, it cannot be assumed that its function is implied
by its position in the utterance. Units of this type I suggest to designate as
autonomous monemes.
The second type of function marking is to be found in the following
stretch, namely the President spoke. Here the function of the President is
the well-known subject function; the same phrase could be used with other
functions, as in I saw the President or I spoke with the President, and conse-
quently we cannot assume that President with or without the carries both
its meaning and the indication of its function, which was what we found in
the case of yesterday; the is no indicator of function, since replacing it by
a as in a President spoke will not change the function of President. The only
circumstance that enables us to identify it as a subject is its position before
spoke; the President is thus to be labeled negatively as a non-autonomous
phrase.
The third type of function marking is the most obvious one, the one we
find in in the auditorium where in can be defined as a functional moneme; in
the auditorium is an autonomous phrase with the same type of distribu-
tional latitudes as yesterday. Functional monemes may be "words" such
as in, i.e. forms which may be found separated from those whose function
they mark by other forms such as the in in the auditorium, or they may be
involved in inflexions, as is the case with the functional moneme "dative"
in Latin.
All this leaves us with spoke, which is the nucleus of a predicative phrase
which includes further the non-autonomous subject the President. The
predicative phrase should be defined as what cannot be eliminated without
destroying the utterance as such. There are languages where the nucleus of
a predicative phrase, the predicative moneme, always stands in the same
relation to the other monemes of the same utterance. If this predicative
moneme indicates an action, the participants of the action (such as the
agent, the patient, the beneficiary) will be expressed as such according to a
pattern that leaves no choice to the speaker. These are the languages that
do not distinguish between an active voice and a passive voice. In such
cases, we may speak of a predicative function characterized by no positive
mark: the predicate is the moneme by reference to which other monemes
mark their functions. In a language like English, the predicative moneme
may entertain different relations with the other elements of the utterance,
as shown by the two utterances the man speaks the language and the langu-
age is spoken by the man; in which case function marking is achieved
according to type three (in the auditorium) by means of some functional
moneme. The two functional monemes are formally very complex in our
6 ANDRE MARTINET

present English examples. They are much simpler in their Danish equiva-
lents: manden taler sproget and sproget tales af manden, glottal stop and -r
in one case, -s in the other case.

A number of circumstances have contributed to delay the identification

of functional monemes as such: first, as we have seen above when trying
to analyze Latin homini, the formal expression of the functional moneme
is frequently amalgamated with that of some other moneme, so that who-
ever insists on identifying linguistic units by their forms will reckon with
only one unit per amalgam; the -i of homini will be considered one mor-
pheme, which will hamper the identification of its two components on the
plane of the contents. But even if this first hurdle is finally negotiated, the
grammatical nature of these two· monemes may induce the linguist not to
separate them. By 'grammatical' I mean, of course, belonging to closed
inventories, which is what we find for Latin cases and numbers.
This might suffice to confuse the issue: few linguists have become aware
of the fundamental difference between functional monemes that serve to
connect a segment with the rest of the utterance (to in to John, "dative" in
homini), whose character is, as it were, centrifugal, and other "grammatical"
elements, such as singular or plural, definite or indefinite, which help to
define the value of the segment to which they are attached: presidents, not
president, the plural in hominibus, not the singular as in homini; the man,
not a man or man. The latter are modifiers, with a centripetal character,
whose relationship to neighboring units is of the same nature as that of a
non-grammatical moneme like an adjective; a functional moneme is not
a modifier, but a link.
But what has prevented even a scholar of Sapir's acumen from acknow-
ledging the basic difference between functional monemes and grammatical
modifiers is the existence, in both cases, of concord.
Concord amounts to using discontinuous and redundant forms for the
expression of a certain meaning: in Latin fortes et acerbos hostes, the
plurality of the enemies is expressed three times, and in three different
places we are supposed to be informed of the nature of the participation of
those enemies in the action implied by an accompanying predicate; in dis-
cipulus uenit, the t of uenit, which points to a subject other than speaker
and interlocutor, adds nothing to what was previously implied by discip-
ulus;1 the same could be said of the -s of Engl. the man comes. Traditionally
it is assumed and often stated that concord has a connective function: it
should, as it were, help the listener put together what belongs together.
I or course, discipulus uenio is not impossible; cr. the famous qua/is artifex pereo. But
a substantive subject is normally expected to be a third person.
 ELEMENTS OF A FUNCTIONAL SYNTAX 7
This view is mainly fostered by a long familiarity with classical poetry,
where poets are found to have taken advantage of concord to let metrical
necessities determine the respective position of words in the text, so that
only concord will show, for instance, which nouns and adjectives belong
together: in incuruo terram dimouit aratro, concord, even more than com-
mon sense, tells us that incuruo belongs with aratro. Even if it is found that
incuruo aratro or aratro incuruo are far more normal constructions, it is
a fact that concord has precedence over proximity, because no amount
of syntactic ingenuity will ever connect two forms, such as incuruum and
aratro, that are shown not to agree.
In the case of discipulus uenit, the situation is somewhat different: here
the nature of the connection between discipulus and uenit would remain
perfectly clear without any personal ending. Yet there are quite a few con-
texts of this type where concord affords the only way to a correct under-
standing of the message: in an utterance like uenatores animal occidunt, we
are dependent on the -nt of occidunt for the understanding that it is the
hunters who kill the beast and not the beast who kills the hunters. It is
true that as soon as we speak of "beasts" instead of "a beast", concord is
no help: uenatores animalia occidunt gives us no clue as to who does the
killing, and we shall rely on the context or, if the context is ambiguous, fail
to register the message in its entirety. In an ambiguous context, uenatores
animal occidunt would still be understood, and we therefore must agree
that -nt may act as an indicator of function. This does not mean however
that we should consider it the formal face of one and only one "morpheme"
with a complex and best not analyzed "grammatical" function, but rather
as the amalgam corresponding to different monemes: permanent and fre-
quently redundant ones like "third person" and "plural," and a chance one,
namely the indicator of the subject function of some plural nominal sub-
jects such as uenatores whose ending fails unambiguously to mark the
function. In any case, -nt is the signifier df several different monemes, just
as are the nominal endings of Latin, and we have already seen that one and
the same of these endings normally stands for both a functional moneme
called case and a numeral modifier. The ablative ending of incuruo stands
redundantly for ablative case and singular number, both being concurrently
expressed by the same ending in aratro; but it is, by itself, an indicator of
adjectival function in connection with a substantive of non-feminine
gender "in the dative singular," i.e. accompanied by a singular modifier
and a "dative" functional indicator. To people used to the expression of
adjectival function by the simple syntactic trick of pre-position, the work-
ing of concord for identical purposes seems formidably intricate and hope-
lessly entangled with other grammatical processes. But if we concentrate
8 ANDRE MARTINET

on function rather than form, the mutual autonomy of the various formally
amalgamated monemes will eventually stand out.
Once established, concord may be extensively used for functional pur·
poses, but we should not imagine that it becomes established in a language
in order to help listeners put together what belongs together. Whenever we
have a chance to witness the appearance of new concord, we find that it
does not arise out of a need to clarify the connections between the different
segments of an utterance, but through the working of what is usually called
least effort and what I would prefer to designate as a language economy.
Language economy is, of course, prodigiously complex, and, off hand, we
might believe that concord, which requires the addition of redundant
elements, contradicts economy. But observation shows that, when school
and prestige do not interfere, speakers faced with the choice between a
shorter utterance involving adaptation to a specific situation and a longer
utterance without such adaptation will normally prefer the latter: the
French, who say nowadays if ne croyait pas qu'il puisse . .. instead of former
il ne croyait pas qu'i/ put . .. make use, in the subordinate clause, of seven
successive phonemes (/kilpyis/) instead of five (/kilpy/) for the same amount
of information, but they save themselves the trouble of deciding whether
they should use one tense or another. Concord usually results from sticking
to the same full form, whatever the context, whether it is repetitious or not,
because it saves the speaker the trouble of adapting form to context: when
reading was done by several people, the Romans would say legunt irrespec-
tive of whether the subject was specified or not. Substandard French /es
gens ils /isent le journal(/ . .. iliz .. .f) for /es gens /isent . .. (/ ... liz ... /),
which amounts to the same, is clearly a product of least effort.
All this means that, through concord, a segment may be made to stand
as the expression of a functional moneme, and the very fact that we could
fairly easily distinguish, in the rather complex Latin contexts we have
analyzed, between relational value and modifying role is a clear indication
that we should keep apart what a one-sided preoccupation with forms had
prevented many of our predecessors, both traditionalists and structuralists,
from separating.
Some of the statements that precede might perhaps be construed as if
the distinction established between functionals and modifiers was founded
upon a subjective evaluation of their semantic contents. Yet this is not
the case: the functional moneme may be identified as that which confers
syntactic autonomy on a moneme or a phrase that is not, by itself, auto-
nomous: the phrase the auditorium is not syntactically autonomous in the
sense that, unless it is accompanied by a functional moneme such as in, to,
or above, its place in the utterance will be determined by the function we
 ELEMENTS OF A FUNCTIONAL SYNTAX 9
want it to assume: if it is the subject function, as in the auditorium is full,
it has to precede the predicate; if it is the object function, as in they entered
the auditorium, it has to follow the predicate. In both cases, the phrase
combines with the predicate into a predicative phrase with full syntactic
autonomy: it is immaterial whether they entered the auditorium precedes
or follows such a syntactically autonomous complement as last night. It is
clear that using the modifier the or the modifier a before auditorium does
not confer to it any syntactic autonomy. But if we add in before that
phrase, we make it independent of its position in the utterance: we can put
it in one place or another without any appreciable difference in the message:
in the auditorium, there are many seats or there are many seats in the
auditorium. In other words, syntactic autonomy, a distributional feature,
is the criterion which enables us to distinguish between functional
monemes, which are connectives, and those specifications, called here
"modifiers", which because of their grammatical status and formal com-
portment have, as a rule, been confused with them.

The adoption of the point of view presented so far entails important

consequences, notably in typological matters. Within the present frame, it
is fairly immaterial whether a functional moneme always presents the
same form, as is the case with English in, or whether it has different allo-
morphs, like Italian in, which also appears as n- in nella for instance, or
whether it is occasionally amalgamated with some other moneme, as is the
case with French a, which combines with the masculine and plural articles
into the "portmanteau" au(x), or whether it is constantly amalgamated
with some other moneme, as is found in Latin cases. What is said here of
functionals also applies to other monemes, whether they be predicatives,
modifiers, autonomous, or non-autonomous.
Now, a survey of the various morpho-syntactic typologies presented so
far shows that even that of Sapir, which 1 would describe as the most
elaborate and refined one, is, in the last analysis, based upon two, largely
interdependent, formal features: first the amount of all om orphic varia-
tion; second the degree of inseparability of certain groups of units. It is
not my contention that none of this should be taken into consideration in
typological work. But this is, at most, but half of the story. It is all right to
consider how languages manage to express various relations, but it is prob-
ably more basic to determine what those relations are, and it is highly
desirable to set up language types based upon the existence or non-existence
of certain relations: it is worthwhile pointing out that, in order to dis-
tinguish between active and passive, some languages make use of special
inflexions, others of distinct suffixes, others still of special auxiliaries. But
10 ANDRE MARTINET

it is certainly more important to stress that some languages do distinguish

between a passive voice and active voice while others don't.
In short, we should never assume that what differentiates one language
from another is essentially and basically a different choice of the formal
means of expression, but rather the type of analysis of experience it repre-
sents and the type of relationship prevailing among the corresponding
linguistic articuli. In other words, whatever has to be retained of the
Whorfian hypothesis applies to syntax as well as to the other aspects of
linguistic structure.

Sorbonne, Paris