,,

72:$5'6 $ )81&7,21$/ 6<17$;

II asked about the history oI structural linguistics, most people concerned would probably
say that it all began with the phoneme. When structuralists chose Iirst to concentrate on
what we have called the second articulation, they certainly Iound the correct approach
towards greater rigour in the treatment oI linguistic problems: the discrete nature oI
linguistic units is ultimately based upon the discreteness oI the phoneme, and phonology
was the Ioundation we needed Ior any Iurther progress. Still, dealing so long with
phonemes beIore attacking the more intricate Iield oI the Iirst articulation had an
unIortunate consequence: having achieved outstanding success in 'phonemics', linguists
were legitimately induced, when tackling the actually Iar more complex study oI the
signiIicant aspect oI language, to use a similar pattern. This accounts Ior Irequent
terminological pairings such as phoneme-morpheme, phone-morph, allophone-allomorph,
and so Iorth, and, on a diIIerent level oI abstraction, the isomorphism oI the
glossematicians with its strict parallelism oI the two planes oI expression and contents
*
.
The most Iundamental objection to this practice oI identiIying the patterns on two diIIerent
planes derives Irom the obvious Iact that, in language, something which is not maniIest,
variously called meaning or experience, is maniIested by means oI something else. This, no
doubt, implies a one-to-one equivalence, that oI the signifiant and the signifie, but not
necessarily an identical behaviour oI the minimal, signiIicant unit, which 1 call the
moneme, and oI the minimal distinctive unit, the phoneme. The moneme is
1
a Saussurian
sign, a unit with a meaning and a phonic shape, i.e. one which combines sOII1ething that is
not maniIest with its outward maniIestation. It belongs to the two planes oI expression and
contents, and it is the smallest segment that does. The phoneme has a phonic shape, but no
meaning. It is pure maniIestation and belongs exclusively to the plane oI expression. The
moneme-and-phoneme approach to linguistic analysis-and, Ior that matter, the widespread
morpheme-and-phoneme one too-does not coincide in the least with the two-plane pattern
oI glossematics: considering an utterance like give me the book, glossematicians will put,
on one side, the individual 'cenemes' like /g/, /i/, /v/, the significants, either minimal (/giv
mi ) or complex (/giv mi buk/), and their graphic equivalents (g, give, give me the book)
as well; on the other side the signifies, either minimal ('give') or complex ('give me the
book'). The basic glossematic dichotomy can be represented as Iollows:

/g/ give`
/giv/ 'give me the book'
/giv mi buk/

Linguists, who; explicitly or implicitly, operate according to the double-articulation pattern,
will keep their phonemes alone, and put signifiants and signifies together according to the
Iollowing schema:

*
CI. Jerzy Kurylowicz, 'La Notion de l'isomorphisme', Recherches structurales (Travaux du
Cercle Linguistique de Gopenhague 5), 1949, pp. 48-60.
1
Pag. 39.
^ a `
/giv/ give`

/g/
/giv mi buk/give me the book`


This lopsided diagram conveniently illustrates how the moneme stands on a Iar higher level
oI complexity than the phoneme and why any eIIort towards pairing them is bound to result
in distortion.
The Iundamental diIIerence between distinctive and signiIicant units must ultimately
account Ior one very important
2
discrepancy in the syntactic comportment oI phonemes and
monemes: a phoneme IulIils a Iunction in a deIinite position.
Ig we want to identiIy a signifiant, e.g. that oI the word lake, it is not enough to say that it is
made up oI three phonemes /1/, /e
i
/, and /k/ because the same phonemes are those which
characterize the words clay and kale; one must speciIy: /1/, /e
i
/, and /k/ in that order. In
other words, when pronouncing lake, speakers have to choose, in initial position, /l/ and
oppose any inclination to say /k/ Ior cake, /t/ Ior take, &c. Postponing /l/1 till the end oI the
word and anticipating the choice oI /k/ would not do, because we would thus get kale which
is not what we mean. All this, which sounds trivial, is, in Iact, basic Ior the establishment oI
the phonematic pattern oI the language. .
The situation is diIIerent with monemes or signiIicant elements gene rally ; the relevancy oI
order is Iar Irom general: it is Iairly immaterial whether I say the one I like is Paul or Paul
is the one 1 like; the implications are diIIerent iI I say with Paul, I went to Rome and I went.
to Rome with Paul, but they do not aIIect the identiIication oI the moneme group with Paul,
and the same applies to yesteraay in I went . . . yesteraay and yesteraay, I went. . . .
Certainly, the respective position oI monemes is oIten determined by tradition .or by the
need to distinguish between utterances with diIIerent meanings: it just isn't done to say
Paul with went 1 Rome to, and it is Iar Irom immaterial whether I speak oI root hair or hair
root, not to speak, oI course, oI the diIIerence between the man kills the bear and the bear
kills the man But it is clear that whereas phoneme classes can be established by listing all
the phonemes that appear in a given context, this cannot be done indiscriminately with
monemes, and it might seem that the Iirst step with them should be to determine the
situations where their presence results Irom an exclusive choice, as is normal with
phonemes.
It is, however, preIerable by Iar to Iorget the phonemes Ior a while and try to discern what
comportment we may expect
3
on the part oI monemes as the products oI the Iirst
articulation oI language. Let us start again Irom the experience that has to be
communicated. II the communication is to be linguistic, that experience will have to be
analysed into a number oI elements Ior each oI which the language under consideration has
an equivalent, a moneme with its meaning and its phonic Iorm For simplicity's sake we
shall assume at this point that the phonic Iorm oI every moneme is always neatly
circumscribed. The communication will thus take the Iorm oI a succession oI monemes,
each corresponding to so me deIinite element oI experience. But, oI course, the choice and
nature oI the elements oI experience will vary Irom one language to another. Let us suppose

2
Pag. 40.
3
Pag. 41.
that the experience to be conveyea coula be renaerea in English by means of Yesterday,
there was a riot in the village. One oI the elements oI the English analysis is the moneme
riot; riot applies to a deIinite happening, and it is immaterial whether it has previously been
identiIied or not: a riot in English is always riot, whether it is a riot or the riot; many
languages do not bother about distinguishing between previously mentioned and not
previously mentioned; English does, and this speciIication is treated as one oI the elements
oI experience; now, we could imagine a language which would have a diIIerent moneme
Ior 'a riot' and 'the riot', as iI people said; Ior example, riot Ior 'a riot' and brawl Ior 'the riot'.
This would, oI course, be most uneconomical, and when people really care to distinguish
between deIinite and indeIinite, they manage to procure articles which become elements oI
experience in their own right. In so me languages, the simple mention oI a riot may suIIice,
in the absence oI any restrictive moneme, to indicate the reality oI the riot; in our
hypothetical language, riot might mean not only 'a riot' but also 'there is a riot'. English, just
like a good many other languages, needs an actualizing phrase (there was) which we might
be allowed to consider a single moneme, were it not accompanied by a moneme indicative
oI time, in this case past time. This
4
could, oI course, be easily dispensed with in an
utterance like the present one, where yesteraay reIers speciIically to a well-deIined segment
oI past time. Consequently, there was a riot, with its Iour monemes, is the reIlex oI a
speciIically English analysis, into Iour elements, oI one aspect oI our experience, where
some other language might get along with only one.
Another aspect oI our experience is rendered by means oI the phrase in the village. The
moneme village represents, as an clement oI experience, a place, but not necessarily a place
where something is happening; village, preceded, oI course, by some article, as he re, could
be made to designate the place, or perhaps its inhabitants, as the agent oI some action, as in
the village aeciaea. or as its object, as in they saw the village. The Iunction would, in both
cases, be indicated by the respective position oI the elements. In our present case, the
circumstance that the village is the place where something happened is treated as another
element oI experience whose reIlex is the moneme in. It would not be diIIicult to imagine a
language in which the moneme designating a village would, at the same time, indicate that
the village in question is the portion oI space where the experience is located. In such a
language, village woula mean 'in (the) village', and this, oI course, all by itselI without any
case ending, since any case ending' would have to be considered the linguistic equivalent oI
another element oI experience.
The third aspect oI our experience corresponds to yesteraay. Were it not Ior literary and
poetic Iorms such as yesteryear and the like, we might consider yesteraay a single moneme
since, in that case, -day, as the automatic accompaniment oI yester-, could not be: counted
as a separate choice. For our present purposes, we shall take the liberty oI disregarding
yesteryear and congeners and oI treating yesteraay as a minimal unit, just as we would do
with German gestern or French hier. We have thus, in yesteraay, an exact parallel to what
we have just been imagining when we thought oI a moneme village with the sense
5
oI
'place where'; yesteraay, here, is not the day preceding this day, but that day as the segment
oI time in which something was happening.
We have been dividing our total experience into three aspects, as we called them, each oI
which was Iurther analysed into monemes. From a semantic point oI view, Iew people

4
Pag. 42.
5
Pag. 43.
would object to such an analysis. But our semantic reactions are, to a large extent, the
reIlex oI Iormal distinctions. The Iormal justiIication oI our initial break up is the Iact that
our three segments there was a riot. in the village. yesteraay are syntactically autonomous
units: every one oI them can be used initially, medially, or Iinally, without any diIIerence in
their own meaning, although, oI course, the choice oI this or that order may imply some
semantic diIIerence Ior the utterance as a whole. On the contrary, in normal contemporary
English, word order is Iixed within those segments. The reason Ior the syntactic Ireedom
enjoyed by those phrases is not Iar to seek: in every one oI them we Iind an unambiguous
marker oI its Iunction, Le. oI its relation to the rest oI the utterance: there was marks the
riot as the predicate, i.e. as the element around which others gravitate and in relation to
which their Iunction will be marked; in marks the village as indicating the place where the
riot occurred; yesteraay, as such, is the indication oI when the riot took place. The
relationships between the three main elements oI experience are thus precisely indicated,
and there is no need here to rely on word order to tell the hearers what these relationships
are. A language like Latin extended the practice oI explicit Iunction marking to situations
where English, and western European languages generally, make use oI word order, namely
in the indication oI subject Iunction and object Iunction. In Latin, a nominal subject was no
part oI the predicative autonomous phrase, neither was any nominal object: pater viaet
puerum, is made up oI three autonomous phrases; its English equivalent, the Iather sees the
child, is just one, since the respective position oI the segments is indicative oI their
6

Iunction, Le. in terms oI experience, their relations to one another, which prevents their
being shiIted at will without changing or impairing the message.
The criterion oI syntactic autonomy points to a threeIold distinction among monemes: we
have Iirst monemes that carry within themselves the indication oI their own Iunction and
which we shall designate as autonomous monemes: French vite. hier. aemain. aimanche in
il vienara aimanche, are autonomous monemes; in English there seem to be Iew clearly
autonomous monemes oI that type, but autonomous compounds such as last night. next
week are at least as Irequent as in French, where we have hier soir, la semaine prochaine.
Notice, in German, autonomous compound numeral s such as
neun:ehnhunaertneununaareissig with the meaning oI 'in 1939'.
Next, we have monemes that do not imply any deIinite relation to the rest oI the utterance
and will thereIore be available Ior several diIIerent Iunctions. OI course, every one oI these
Iunctions will have to be indicated somehow, either by position or by means 6I some
additional elements. These monemes could be called dependants; village is a dependant.
Last, we have monemes which secure autonomy Ior other monemes to which they are
attached, by indicating their Iunction, i1e. their relation to the rest oI the utterance. The
combination oI such a moneme with its dependants is an autonomous phrase. These we
shall call Iunctional monemes, Iunctional indicators, or just Iunctionals. Functionals
correspond to prepositions and conjunctions oI traditional grammar, but also to case
endings. The reasons Ior which people are so reluctant to lump prepositions and case
endings together are numerous: Iirst, the Iormer come beIore and the latter aIter the Iorm
they govern; second, prepositions can be separated Irom their substantives by various
additions such as an article and one or several adjectives, while case endings are
permanently glued to the word they characterize; third, there is normally one preposition

6
Pag. 44.
per phrase, irrespective
7
oI how many articles or adjectives are added, whereas case-
endings are likely to be Iound aIter every one oI the additional elements; Iourth, in the case
oI prepositions, the Iunctional usually Iorms a clear-cut segment oI the utterance, in
contradistinction to what we Iind, Ior instance, in Latin caseendings, where the indication
oI case, i.e. Iunction, is Iormally conIused with that oI a totally diIIerent type oI moneme,
namely number, and where it is not always clear what belongs to the substantive moneme
and what to the ending: is the nominative ending oI puppis, 'poop', -is or just -s as in urbs?
All Iour reasons should not be dismissed as sheer prejudice. But we should never allow
them to blur the Iunctional identity oI prepositions and cases. All oI them are Iinally
reducible to the same phenomenon: monemes, which, Ior some reason or other, are
Irequently or constantly in contact and will tend to merge. The merging will be the more
likely and the more intimate iI the element whose Iunction is indicated comes Iirst, and the
Iunctional indicator last. This is due to the Iact that, in any language, the number oI distinct
Iunctions 1s very much smaller than that oI elements capable oI perIorming them; these
elements, the so-called lexical items, are more inIormative and, accordingly, generally
given a preIerential treatment: they may be provided with an accent which gives them
prominence, and their initial phonemes are, as a rule, articulated with particular care so as
to Iacilitate their early identiIication in the Ilow oI speech.
A Iormal merging oI two or more monemes, 1 call an amalgam: he cut, as contrasted with
he aamittea, can be said to be an amalgam and to result Irom amalgamation. Fr. au, in au
marche, as contrasted with a l in a hspital, is also an amalgam. We may also, iI we
choose, call sang, the preterite oI the verb to sing, an amalgam, although it is not likely to
have resulted Irom a process oI amalgamation. This process is not necessarily carried
through so that it becomes .impossible to distinguish one signifiant Irom another: in Slavic
8

languages, amalgamating processes have, at diIIerent periods, begun to blur the boundaries
between the radical moneme ruk- 'hand' and Iollowing derivational monemes or case-
endings; this has resulted in yielding diIIerent Iorms Ior the radical: in Czech, Ior instance,
ruk- in the nominative singular ruka, ruc- in the locative ruce, and ruc- in the adjective
rucni.
Semantic amalgamation is common too, as when black mail becomes blackmail. In
winaow, or its Danish equivalent vinaue, originally wina eye, amalgamation, both Iormal
and semantic, has resulted in reducing two successive monemes to one. When, as in a Iorm
like blackmail, the semantic amalgamation is complete, since, synchronically, there is no
hope oI ever identiIying the meaning oI blackmail as the sum oI the meanings oI black and
mail, it is certainly advisable to consider blackmail a single moneme, because, Irom a
purely synchronic standpoint, the homonymy oI blackmail and black mail is purely
accidental. But when a Iormal amalgam is still identiIiable as limited to a deIinite Iormal
context, as is the case with Fr. au Iound only beIore consonants and not beIore vowels, or
when the amalgamated Iorms combine monemes with perIectly distinct Iunctional values-
as with Latin case-endings where, Ior instance, nominative is a Iunctional indicator, but
plural is nothing but a modality oI the noun-it is imperative to preserve the distinction. We
shall speak then oI two diIIerent monemes even iI an analysis oI the Iorm into two
successive segments proves arbitrary and, as in the case oI Fr. au, downright impossible.
The Latin Iorm homin conveys three elements oI experience: 'man', 'dative', and 'singular'.

7
Pag. 45.
8
Pag. 46.
Should we insist on chopping it up into successive slices, we might rephonemicize it as
/hominii/ and say that 'man' corresponds to homin-, 'dative' to -i-, and 'singular' to -i. This
analysis might be supported by a comparison with the dative plural hominibus, where both
homin- 'man' and -i- 'dative' would reappear, -bus being then an 'allomorph oI the plural
morpheme'. But, oI course, homin- means
9
'man' only in combination with a given type oI
endings, and -i is the mark oI the dative only in combination with certain nouns in the
singular. ThereIore it is more accurate and, oI course, less arbitrary to say that homin
means at the same time 'man', 'dative', and 'singular' than to try to segment it. Linguistic
articulation may be blurred, and it is our duty to describe it as we Iind it: we should never
try to disentangle Iormal units which happen to overlap or be conIused; yet we should
never deny the existence oI units, as singular and plural, in Latin" whose existence is
always secured and attested by some Iormal diIIerence, but in so intricate a Iashion as to
deIy analysis.
Our distinction oI three types, autonomous monemes, dependants, and Iunctionals, is based
upon syntactic autonomy. But this leaves out one type oI moneme or segment which is
independent rather than autonomous. Reverting to our Iormer example, we would say that
yesteraay is an autonomous moneme, and in the village an autonomous phrase, but there
was a riot is not only syntactically autonomous, since we can place it initially, medially, or
Iinally at will, but also independent, since we can use it all by itselI and obtain a complete
utterance, which is not the case with yesteraay, nor with in the village. In our hypothetical
language, riot, which should mean as much as '.there is a riot', could be used all by itselI as
a selI-suIIicient utterance. We would then say it is used with predicative Iunction. The
situation, in our English example, is somewhat more complex. There was a riot is deIinitely
a predicative phrase characterized as such by what we have called its independence. We
took the liberty, in what precedes, oI considering there was an actualizing phrase and a riot
the predicate, but others might preIer another kind oI analysis and see in a riot the subject
oIt he predicate there was. This view is supported by the observation that an analysis in
terms oI subject and predicate seems to be universally applicable in English. Every
complete utterance, in that language, centres around a core oI two monemes, one oI
10

which, as Sapir says,
*
is 'something to talk about', and another which is what 'must be said
about this subject oI discourse'. Since, in English, the subject is never an autonomous
phrase as, Ior instance, in Latin when a noun, the predicative phrase is always made up oI at
least a subject and a predicative moneme, whose only possible Iunction is the predicative
Iunction, and which we call a verb. The subject, deIined as what necessarily accompanies
the predicate, is one Iunction oI certain classes oI monemes which, in the wake oI tradition,
we could call nominal and pronominal monemes. The other Iunctions assumed by these
monemes are those oI the so-called complements. In a language where the predicative
moneme need not be actualized by means oI a nominal moneme endowed with a speciIic
Iunction, we should not speak oI a subject. What we would be inclined to label 'subject'
because it is rendered by a subject in a translation, is nothing but one oI the complements.
We may want to deIine 'Iunction' as the linguistic counterpart oI the relationship between
one element oI experience and the whole oI experience, so that we could speak oI Iunction

9
Pag. 47.
10
Pag. 48.
*
Language (New York, 1921), p. 126.

in the case oI any one oI the marginal elements, but hardly in the case oI the predicative
core; the predicative Iunction would then be no real Iunction. This, aIter all, is a matter oI
convention. But even iI we decide to speak oI Iunction only in reIerence to a deIinite type
oI relationship, we should be ready to ascribe diIIerent Iunctions to the predicate at least in
those languages-English is one oI them where speakers have at their disposal two or more
Iorms oI the predicate indicating diIIerent types oI relationships with the participants oI the
action; iI opening is the action, ,and the participants a gardener and a gate, I may choose to
say either the garaener opens the gate or the gate is openea by the garaener. We will have
to distinguish between an active Iunction, and a passive Iunction, these Iunctions' being, oI
course, nothing but our traditional 'voices'. It is not easy to tell what the
11
Iunctional
indicators are in opens and is openea; probably -s Ior one Iunction, is . . . ea Ior the other.
But, once again, we should not let our analysis be too narrowly determined by the
diIIiculties we may experience in analysing Iorms.
II we leave out, as a special type, the Iunctions oI the predicate, we should distinguish
between primary Iunctions and non-primary Iunctions. Primary Iunctions are those oI
elements which are directly connected with the predicate. In a sentence like yesteraay. the
heaa of the aepartment aictatea a four-page letter to the secretary he haa fust engagea. the
four elements yesteraay. the heaa of the aepartment. a four-page letter, and to a secretary
he had just engaged have some primary Iunction; the Iunction oI the aepartment, four-page,
he haa fust engagea is not primary, since they are not directly connected with the predicate
aictatea. Within a so-called subordinate clause such as he haa fust engagea, the same
Iunctional hierarchy obtains as in the so-called main clause, but we should not speak oI a
predicate there, but oI a 'predicatoid', and the Iunctions oI elements directly connected with
it should be considered at best primary-like.
Among dependent monemes-those that are neither independent, autonomous, nor Iunction
indicators-one should distinguish the ones that assume some primary Iunctions Irom the
ones whose Iunction is not primary. The Iormer could be designated as primary dependents,
and the latter as marginal dependents, or determinants; in the above sentence, heaa is a
primary dependent, the (in the heaa) and aepartment are marginal dependents, the (in the
aepartment) is, oI course, marginal too, but marginal to the already marginal aepartment.
Another possible - distinction is that between grammatical and lexical monemes. In order to
distinguish between them, one should set up the inventories oI the monemes which are
Iound in speciIic contexts, within autonomous phrases, where the respective position oI
elements is Iunctionally relevant. Lexical monemes are those. which belong to non-limited
inventories. Grammatical monemes are those that alternate
12
, in given positions, with a
comparatively restricted number oI other monemes. The average Irequency oI grammatical
monemes like of. for. with or Lat. 'genitive', 'dative', 'ablative' is considerably higher than
that oI lexical monemes such as man, rich, or eat. Functionals are grammatical monemes.
Among primary dependents so me may be lexical (nouns) and so me grammatical
(pronouns). Some determinants are lexical (an adjective like great), and others grammatical
(an adjective like my, the article the, or the 'plural' moneme). Grammatical determinants can
be designated as modiIiers. Among modiIiers should also be listed such grammatical
primary dependents as are part oI the predicative phrase. These include modes, tenses,

11
Pag. 49.
12
Pag. 50.
aspects, and persons in so Iar as their signifiant is not syntactically autonomous: in Ill ao
it, I. ll, and it are modiIiers.
With the setting up oI modiIiers as a speciIic type oI moneme neatly distinguished Irom
another type oI grammatical element we have called Iunctionals, we are deIinitely breaking
with tradition. Stressing, as we have done, the syntactic autonomy oI certain units or
phrases, is, oI course, nothing but pointing out a Ieature that sets apart adverbs and
adverbial phrases in so Iar as they are used as verbal adjuncts. Labelling predicative
monemes as 'independent' does not go Iar beyond what generations oI grammarians have
stated about the nature oI the predicate. But dividing the mass oI grammatical elements into
two basically distinct types -the Iunctionals and the modiIiers, placing them at two Iar ends
oI our chart, and enIorcing thereby a strict segregation -may well be Ielt as verging on
scandal. We are so used to listing as belonging together such categories as tense, aspect,
mood, and voice, that it may be shocking to hear someone maintain that whereas tense,
aspect, and mood are closely akin and belong together, voice is as diIIerent Irom them as
diIIerent types oI monemes can be. We are used to considering gender, number, and case as
the three 'pillars on which any decent nominal system is built, and we may be tempted
13

to resist any suggestion that gender has no separate existence and that number is a local
accident oI limited importance, whereas case belongs to the constitutive Irame oI any
utterance.
Yet the distinction between Iunctionals and modiIiers is Iundamental, and it has lately been
noticed and pointed out in various quarters
*
. II in a phrase such as with a smile, the primary
dependent smile is considered the centre oI the phrase, the grammatical determinant a is
centripetal, the Iunctional with centriIugal: a is connected with the rest oI the sentence only
through smile, which it helps to speciIy;. with connects smile with the rest oI the sentence,
and since the connexion is thereby established, the speaker is Iree to place the phrase with a
smile anywhere he pleases. Syntactic autonomy is thus the criterion which, in all cases, and
particularly in Iormally complex ones, will prove the presence or the absence oI a
Iunctional. In a context like the hunter was killing a bear with his spear, neither the nor a
are Iunctionals, since they do not grant hunter or bear any syntactic autonomy: exchanging
the place oI the hunter and a bear will result in conveying a totally diIIerent experience;
with, on the contrary, makes it possible to place the phrase with his spear practically
anywhere without changing its relation to the rest oI the sentence. Whether this or that
Iunctional may or may not be used is partially determined by the communicative needs at
that point oI the utterance': aIter he is aistributing tickets, I may or may not speciIy to
whom. But it also depends on which predicative moneme I choose, whether I may or may
not use a to-complement. It is what is reIerred to when we say that a given verb governs
this or that case. This amounts to saying that, to a large extent, the choice oI a Iunctional is
predetermined by that oI the verb oI the clause. On the contrary, the choice oI the modiIier
is Iree, i.e. the speaker is determined to use a or the, the plural or the singular, at a certain
point by direct
14
reIerence to what the experience really is that he wants to. communicate.
II, in some cases, the range oI the choice happens to be limited, it will be so on account oI

13
Pag. 51.
*
e.g. by Richard S. Pittrnan, starting with A Grammar of Tetelcingo (Morelos) Nahuatl (Language Diss., No.
50) (Baltirnore, 1954), pp. 6-8, and Georges Gougenheirn, 'Morphologie et Ionctions grarnrnaticales', Journal
ae psychologie (1959), pp. 417-26.

14
Pag. 52.
the nature oI the unit that is modiIied, as when a certain noun is never used in the plural,
but not because oI some limitation imposed by the structure oI the sentence as a whole.
In spite oI such Iundamental diIIerences, Iunctionals and modiIiers have so Iar gene rally
been conIused, and it is easy to understand why linguists who, in their overwhelming
majority, had been taught Latin grammar Iirst, were tempted, 'at best, to consider them two
aspects oI the same linguistic reality. In Latin, as in all Indo-European languages that have
preserved the old declensions, the expression oI case and that oI number are hopelessly
mixed under the Iorm oI case-endings that resist any Iormal analysis into successive
segments. In other words, amalgams oI a Iunctional and one modiIier, or more than one, are
practically the rule there, and must have been Ielt by generations oI linguists to be a normal
Ieature oI any selI-respecting language structure. As a matter oI Iact, it cannot be
considered a strange quirk oI Indo-European, because amalgams oI Iunctionals and
modiIiers are by no means absent elsewhere. It is quite clear that the signifiant oI both are
likely to appear in close succession in the utterance and are thereIore constancy exposed to
amalgamation.
But the existence oI such amalgams would not suIIice to explain the traditional disregard oI
the diIIerence between centriIugal and centripetal grammatical elements: aIter all, the
existence oI complex case-endings never prevented grammarians Irom distinguishing
between case and number. The main reason why the distinction was not made lay in the
existence oI concord. It was generally believed that the justiIication Ior the somewhat
clumsy repetitions which characterize this phenomenon was to be Iound in the way it
allowed the speaker to connect the successive elements oI the utterance. This was
.proIessed even by a scholar like Otto
15
Jespersen, who had little patience with roundabout
ways oI expressing oneselI and denounced concord as a most awkward approach to
sentence building. Now, concord which amounts to using several discontinuous segments
Ior one and the same signifie is attested Ior Iunctionals and modiIiers alike.
This circumstance was, oI course, one more reason Ior identiIying them. But what is really
basic here is the Iact that, thereby, modiIiers would seem to be made to Iunction as
connective elements, a role which we have been denying them and entrusting exclusively to
Iunctionals. The answer to this is that, iI what is normally just a modiIier, redundantly anq.
discontinuously expressed in diIIerent positions through the. utterance, happens to act as
the sole indicator oI a Iunction, it has to be interpreted, when this is the case, as an
amalgam combining modiIication oI the accompanying monemes and indication oI their
relationship to the rest oI the sentence: in a Latin context like viri viaent. viri contains a
modiIier, the 'plural' moneme, and a case, the nominative which marks its subject Iunction;
in viaent the Iinal -nt is exclusively a part oI the signiIiant oI the 'plural' moneme we have
just identiIied in viri. It is worth remarking that the amalgam oI the two monemes 'plural'
and 'nominative' 1S only partial, since it could be said that the Iormer's signifiant is -i . . . -
nt, while the latter's is just -i. But. iI our subject belongs to a type where the nominative
ending is identical with the accusative ending, iI, Ior example, our context is homines
animal viaent, the -nt oI viaent becomes instrumental in identiIying homines as a
nominative, since it indicates that the subject is in the plural and, in this context, points to
homines as the only Iorm which can be interpreted as a plural subject. Thus, owing to the
ambiguity oI the nominal ending, the -nt oI the accompanying verb may be made to play
the role oI the elements oI a discontinuous moneme indicative oI the subject Iunction oI the

15
Pag. 53.
neighbouring plural noun. All this may sound rather tricky, but the Iacts themselves are so,
not our interpretation, and this is something to which any schoolboy
16
would readily bear
witness. Everything would be simpler iI the nominative case were always unambiguously
distinguished Irom the other cases. There would then never be any need to resort to the
mark oI the plural agreement to indicate which noun is the subject. The intricacies oI Latin
are no prooI that the distinction between Iunctionals and modiIiers is not Iundamental.
Concord is redundancy, and contrary to what could be expected, redundancy results, as a
rule, Irom least eIIort: people do not mind repeating iI mental eIIort is thereby reduced; iI
adjectives are quite Ireely and Irequently used as nouns, as was the case in older Indo-
European languages, it will be indispensable Ior them to carry the mark oI their Iunction iI
nouns do; thereIore a word like fortis, 'courageous' or 'courageous Iellow' is inIlected just
like civis, 'citizen'. II I mean something like 'the citizen, the courageous one', there is some
justiIication in presenting the mark oI their Iunction twice, since the two words are more or
less equals. But when fortis is really nothing but an attribute oI civis, it would be more
logical either to mark it as such, or to use the bare stem, letting its vicinity to civis indicate
its relation to it. But it is certainly much easier to let the adjectives keep in all cases
inIlected Iorms that have had to be memorized anyway. In popular French the equivalent oI
my father says is mon pre il ait (pronounced |idi]) instead oI the traditional mon pre ait;
since I have to use |idi] in reIerence to my Iather's saying something when 1 need not
speciIy that the speaker is my Iather, why shouldn't 1 save myselI the trouble oI choosing
between |di] and |idi] depending on whether 1 expressly mention my Iather or not? Since 1
can't help using |idi] at times, it is handier to stick to it, whatever the context, than to
reckon every' time with the context; pronouncing an extra phoneme is nothing in
comparison with the output oI mental energy required by the choice between |di] and |idi].
There are situations where it is not usual to speak oI concord or agreement, but in which
speakers also have recourse
17
to redundancy Ior' obvious reasons oI economy: in our Iormer
Yesteraay. there was a riot, the notion oI 'past' is expressed twice, once precisely,. In
yesterday, and then rather vaguely, in the preterite was. Now, it may be useIul to have at
one's disposal, together with a set oI reIerences to so me deIinite periods oI the past, such as
yesteraay or last week, some means oI labelling the experience as a past one without any
Iurther speciIication. OI course, iI we start with a precise reIerence such as yesteraay, there
is no need to add a vague reIerence to the past as the one included in was. The lazy
solution, however, consists in letting the context, yesteraay, determine the choice oI the
tense. In similar cases, one might easily be tempted to characterize the tense moneme oI
was as relational, since it sluggishly establishes some connexion between diIIerent parts oI
the utterance. In a similar way, it could be argued that such a modiIier as the deIinite article
implies so me relation with what precedes, since it normally labels its noun as something
previously mentioned. Such Iacts are what Sapir probably had in mind when he labelled all
grammatical items as relational
*
.
It has now become clear that redundancy, grammaticalized or just lexical, is a basic Ieature
oI human communication, and it may in particular instances be resorted to in order to
indicate the Iunction oI so me segment. But, once the uses oI the respective positions oI
monemes have been discounted, only such elements as secure syntactic autonomy are to be

16
Pag. 54.
17
Pag. 55.
*
Op. cit., pp. 86-92.
considered Iunctional indicators. Whether this syntactic autonomy is Irequently or
exceptionally made use oI is immaterial. Even iI it could only be shown by means oI some
unidiomatic syntactic shiIt, it still would have to be deemed valid, iI it were proved thereby
that the identiIication oI Iunctions is not aIIected.
Most oI the preceding examples were borrowed Irom English, with a sprinkling Irom other
genetically connected
18
and structurally similar languages. This might lead readers to
believe that our analysis is based upon the observation oI a certain type oI language and,
consequently, that it cannot make any c1aim to universal validity. But this assumption
would not be correct: starting Irom our deIinition oI language as doubly articulated, we
have actually been proceeding in a purely deductive way, and the existing Iorms we have
be en quoting and even operating with were never meant as support Ior the theory, but
simply as illustrations oI the categories we posited as necessary Ior the Iunctioning oI
linguistic communication. What has led us to stress the importance oI syntactic autonomy
is the realization that this autonomy is the test that a segment oI an utterance corresponding
to a given element oI the experience contains all that is needed Ior marking its relation to
the rest. It is a guarantee that the hearers will be in a position to reverse the process through
which the speaker breaks down his" previously unanalyzed experience into a number oI
elements Ior which the language he uses oIIers equivalents. It would seem that there exist
only three ways Ior the speaker to indicate the relation oI a segment to the whole:
First, the linguistic equivalents oI the elements oI experience" may be connected by means
oI units, in all respects similar to them, which we might .consider new elements oI
experience in their own right: in John s hat, 's is but a handy way oI expressing 'belonging
to' which is part oI the experience just like John' or 'hat'; in the same way as some elements
oI experience may be represented by prosodic Ieatures, the marking oI relations may be
secured by means oI an intonational or accentual Ieature.
Second, the relation between two elements may be expressed by means oI the respective
positions oI their linguistic equivalents in the speech continuum: in many languages, a
subject is marked as such because oI its position beIore the predicate; in John s hat the
respective position oI John and 's marks John as the owner
19
.
Third, the relation' oI an element to the experience as a whole may be included in the
'meaning' oI its linguistic equivalent: yesteraay corresponds to an element oI experience
whose relationship to the whole is never in doubt. But, beside such perIectly autonomous
segments as yesteraay, we Iind cases where the 'meaning' oI the moneme may, in certain
contexts, imply a given Iunction, as when Russian stol 'table', as an inanimate masculine;
will tend to be considered a grammatical object, in the absence oI any speciIic indication oI
that Iunction.
Proceeding in a strictly deductive way, it is our duty to determine as we have just tried to
do, what possibilities are aIIorded by the linear Iorm, oI speech Ior the linguistic expression
oI the various Iunctions corresponding to the relations assumed to exist among the elements
oI experience. But we have no right to posit the existence oI relational universals: oIIhand,
we may be inclined to believe that the with-type oI relation, or with-Iunction, exists in al
languages; but observation reveals that, in many languages, the equivalents oI ao it with a
hammer and he came with a friena make use oI diIIerent Iunctionals; even in a language
like French, where fais-le avec un marteau and il est venu avec un ami show the same

18
Pag. 56.
19
Pag. 57.
equivalent oI with, there are so many speciIic contexts where avec does not correspond to
with that it would not be commendable to equate their respective Iunctions. Even on a more
Iundamental level, we should be wary oI Iollowing Sapir when he presents the subject-
predicate pair as the necessary basis oI linguistic communication
*
. Here again, we have to
envisage various possibilities and try to imagine what could constitute a mini mal utterance,
and how such an utterance could be Iurther expanded.
The Iirst problem that arises in connexion with the mini mal utterance is whether we should
distinguish between normal utterances, the ones which, in English, involve a subject and a
predicate, and curtailed ones such as over there! the scounarel!
20
fire!, not to speak oI
monomonematic or polymonematic injunctions such as go! ao! get away! give him a
shilling! Traditional grammarians do not experience any diIIiculty in such matters because
it is obvious Ior them that, once injunctions are set aside, only subject-predicate cores can
be dubbed 'normal'. But this is precisely what we do not accept, because we want to
consider the possibility oI languages using utterances made oI one (predicative) moneme in
exactly the same situations where other languages, such as English, cannot dispense with
the complex subject-predicate construction.
It is by reIerence to our distinction between what is properly linguistic and what is
marginally so
**
that we may hope to give a sound Ioundation to the concept oI 'normal
utterance'. What we consider properly linguistic is what is achieved, in matters oI
communication, by means oI the double articulation pattern: double articulation is what
protects the linguistic Irame against interIerence Irom outside, what makes it really
independent and selI-contained. But communication by means oI double articulation is an
expensive procedure which man will tend to avoid when his needs can be satisIied through
the use oI simpler, more direct means, such as gestures, by themselves or supplemented by
speech. Another, very eIIective, way oI reducing the output oI energy involved in
communication is to rely on the situation in which the interlocutors are placed: very nice!
pooh-pooh! no! make excellent sense all by themselves among people who look at the same
thing or witness the same event. This reliance on situation is so general that all languages
have developed several classes oI monemes whose interpretation is always dependent on
situation. Such are demonstratives like this, that, except when used in reIerence to context,
time reIerences like now, yesteraay, toaay, last night or the 'preterite' moneme, and personal
pronouns like I and you. The situation generally makes it so obvious who the second person
subject oI imperatives is, that its expression is the exception rather
21
than the rule. All these
economical tricks are very welcome in linguistic practice, but they undoubtedly detract
Irom the ideal oI human communication, which is selI-suIIiciency. This ideal Iinds its
expression in out-oI-situation uses oI 1angIJage as realized... e.g. in gossip, where the
reIerence is to absent people, in narratives generally and in literature, where use is -seldom
made oI the actual situation, namely the author at his typewriter and the reader with his
printed pages and, in between, the long-drawn processes oI editing and composition. It is
true that authors do create situations where their characters are Iound to use, with perIect
relevancy, such terms as I, you, toaay, or this week. But these situations are actually
contexts, and thereIore a purely linguistic accompaniment.

*
Op. cit., p. 126.
20
Pag. 58.
**
See above, pp. 28-29.
21
Pag. 59.
It is in reIerence to out-oI-situation uses oI language that normal syntax can be deIined:
normal syntax is that which is attested in such uses, i.e. when communication is achieved
by purely linguistic means. This should, oI course, not be construed as excluding Irom
normal syntax such segments as I tell you: it is clear that I tell you, as a syntactic pattern, is
identical with they tell them, which may contain no reIerence to situation; in other words, I
tell you is normal syntax, in spite oI the Iact that this segment contains reIerences to
situation, since it conIorms to patterns attested Ior segments that do not contain such
reIerences.
This situation criterion is certainly valid Ior determining what could be called syntactic
normalcy in a language, like English, where it agrees with the consensus oI generations oI
grammarians that syntactically normal utterances contain a subject-predicate phrase. Its
application to languages that we suspect do not demand a bimonematic core may not be
quite as decisive: there are many languages where a moneme meaning 'rain' (with no
possibility oI deciding whether verb or noun) is used by itselI when the English say it is
raining. In such an utterance, there is no linguistic reIerence to any situation; it usually
indicates that it is raining here and now
22
, but the 'here and now', corresponding to no
linguistic units, are no parts oI the communication. Still it could be argued that 'here and
now' is included in the meaning oI the moneme Ior 'rain' unless speciIically excluded by the
addition oI so me complement like 'yesterday' or 'beyond the hill'. Yet, there is no doubt
that languages in which the normal way oI saying it is raining or there is a fox is by means
oI a single moneme meaning 'rain' and 'Iox', deserve not to be classed indiscriminately with
languages in which this is not possible. The more so iI it can be shown that this
monomonematic utterance may be normally expanded by means oI various complements,
whereby it. is rcvealed as the potential core oI an unlimited syntactic complex '|here comes
a] Iox' > '|there came a] Iox, last year, that was killed'.
Even iI the minimal utterance in a language cannot be unambiguously shown to coincide
with a single moneme, one should not jump to the conclusion that it must necessarily be
one oI the subject-predicate type. It is easy to understand why so many languages have
made it a rule never to use one moneme by itselI: even when centring his attention on the
existence oI a single being, thing, or process, a speaker will normally not be satisIied with
the mere mention oI that item, but will be inclined to locate it in time or space, or to
connect it with himselI or his interlocutor. These .additions are oIten conceived as
actualizers, which seems to imply that the moneme by itselI is an abstraction whose
anchoring in reality can only be achieved by means oI some element endowed with just that
Iunction. This view is supported by the nature oI the subject, which is a moneme, likely to
be Iound elsewhere in a variety oI complemental Iunctions, used here as the necessary
accompaniment pI a predicate, with a Iunction perIectly characterized either by some
privileged position as in English, or so me Iunctional mark as in Latin. But the Iact that
some languages make the actualization oI the predicate a deIinite Iunction, does not imply
that actualization is, in principle, more than one aspect oI the basic linguistic
23
process
according to which communication can be made more speciIic by means oI additional
elements. A distinction may be made between optional speciIication and compulsory
addition, as oI a subject, Ior which the term 'actualization' could by convention be reserved.
But it should be kept in mind that a compulsory subject does not really actualize a predicate

22
Pag. 60.
23
Pag. 61.
more than an optional complement would. There may be languages in which one-moneme
utterances cannot be considered normal out-oI-situation syntax, but in which the predicate
can be actualized by means oI any complement, the actualizing Iunction being added to any
other Iunction the complement or complements may assume.
Traditionally, the term 'predicate' is deIined in reIerence to the subject-predicate complex
and would seem to designate everything in the clause that is not the subject, or some
dependent oI the subject. Besides, 'predicate' implies so me assertion, so that a question or
an order would not contain any predicate. II the term is to be retained by contemporary
linguistics, we shall want to use it in reIerence to monomonematic segments which, by
themselves, may constitute a complete, out-oI-situation utterance, and also to the same
segments when accompanied by various expansions (complements), but independently oI
them. Within the subject-predicate complex, 'predicate' should, in a similar Iashion, apply
to any segment that is, jointly with the subject, constitutive oI the mini mal utterance, thus
excluding Irom it its various complements.
A subject is diIIerent Irom a complement only because it is constitutive oI the minimal
utterance. ThereIore, whenever we speak oI a subject we are reIerring to a linguistic
situation in which subject and predicate are both equally indispensable, since our criterion
is indispensability, and the problem arises oI how we can tell what is the one and what is
the other. II our terminology makes any sense, the subject should be the one which should
somehow stand closer to the marginal and optional elements oI the utterance: in many
24

languages, English among them, subject Iunction can be ascribed to such monemes
(nominals) as are Iound elsewhere to assume the Iunctions oI complements; in some, such
as Malagasy, monemes with subject Iunction must be marked as previously known or
mentioned, which shows them to be inIormationally marginal. The possibility that so me
languages do not clearly distinguish between two successive statements and the succession
subject predicate cannot be ruled out.
In languages where actualization is needed but can be achieved by means oI any
complement, the predicate is obviously the one moneme which is not marked as perIorming
any complement Iunction.
All this, and what could be added about the diIIerent levels oI complementation, amounts to
establishing a hierarchy oI syntactic Iunctions. This hierarchy is, no doubt, set up with a
view to diIIerences between languages, but it should be clear that, even iI it were carried
through, it would never account Ior all the varieties oI linguistic structure.
It is quite essential to know all the diIIerent Iunctions that characterize a language, but it is
equally important to determine Ior each language, what monemes are qualiIied to perIorm
this or that Iunction. No language is known to allow every one oI its signiIicant units to
perIorm all oI the Iunctions it provides. It is even diIIicult to imagine how such a language
would work. II, in such a language, Iunctions should be marked by means oI Iunctional
indicators, these, being signiIicant units in their own right, should be able to assume" the
role oI Iunctioning elements, and, conversely, all Iunctioning elements would also have the
role oI Iunctional indicators. But how could speakers make clear that a given moneme is
used as a Iunctioning clement here, and a Iunctional indicator there, iI only the respective
position oI the monemes in the speech continuum could be relied upon? Many languages
are recorded in which the same moneme is used either as a Iunctioning element with the

24
Pag. 22.
meaning oI 'to give' or as a Iunctional indicator with a dative value. But in a
25
language like
Vietnamese, where this is the case, not all 'verbs' Iunction as 'prepositions', nor vice versa,
and this determines signiIicant contexts that enable the hearer to identiIy 'give' as a
predicate or as a dative Iunction marker. There are languages, and Vietnamese is again a
case in point, in which position plays a great role in Iunction marking; the main Iunction
here can be labelled 'determination', and this is shown by postposition oI the determinant;
what looks like a subject-predicate relationship might possibly be interpreted as a case oI
determination. But iI the core A receives two determinants, B and C, how can the hearer
know that, in the succession A-B-C, C is not a determinant oI the immediately preceding B,
but one oI A, 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 oI the yesterday type, Le. syntactically
autonomous, because the indication oI the Iunction is part oI the meaning oI 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, oI 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 so me distinction between monemes as regards the extent to which they
may assume the various existing Iunctions. In no language are all monemes used
indiscriminately as Iunction-endowed and Iunction-marking. In other words, there is no
language without grammar. But once unambiguous Iunction-marking is secured, there is no
universally valid reason why any moneme, except one that is speciIically a Iunction-
marker, should be excluded Irom any Iunction, whether predicative or non-predicative.
Still, specialization is very widespread. Many languages have, Ior example, a class oI
adjectival monemes more or less restricted to certain speciIic Iunctions and which,
accordingly, tend to identiIy meaning
26
and Iunction: small, Ior instance, implies not only
'smallness', but also the Iunction oI determination, as in a small ear, and also in he is small
iI we agree to locate the predication in is.
It is even more Irequent to Iind that a class oI monemes (verbs) is restricted to predicative
Iunction, although this does not necessarily imply that only verbs can be predicates.
Contrary to what was generally assumed, this is a domain where languages are Iound to
vary and which, in consequence, should playa great role in typological matters
27
.

25
Pag. 63.
26
Pag. 64.
27
Pag. 65.