Propositions, [Nice1951] 046-0990



(l) Semantics i Term and Domain

The term semantics is used ty many linguists to denote exclusively the "substance" of linguistic content as opposed to its form, or is even confi- ned to facts of la parole a Its place in linguistics would then be that of an auxiliary science such as phonetics, This would remain a mere question of terminology if there were any well-recognised term omawering to phonemics in the way that semantics is made to answer to phonetics, to denote the science of linguistic content as established on principles of relevance analogous to those applied in phono- logy. But there is no such term. Such attempts as those to distinguish between semasiology and semantics have generally been made by scholars not yet familiar with newer structural principles and have in any case had lit- tie influence on current usage. Now when a "case vide" is left waiting for a term, it is only too li- kely to be filled by some already existing notion which does not belong here at all. We find explicity in the works of some scholars, implicitly in those of many more, the equation : phonemat ics .‘phonetics = morphology: semant ics , This unhappy comparison is a fruitful source of error, not least with linguists who would be the last to recognise the equation as representing their view. Instances could be cited from works proceeding from any of the principle linguistic schools. An independent linguistic branch dealing with the semantics of la langue remains therefore to be founded. The questions with which it will be con- corned are at present dealt with undor the heading of morphology or dismis- sed as mere affairs of "substance", so far as they have even received atten- tion* The missing discipline will bear the same relation to morphology on the one side as phonematics does on the other. Morphology, so far from being irrelevant, will have a decisive role in the identification of units (the same r$le that it plays, though often without explicit recognition, in the identification of phonematic units). But just as the phonematic system, once established in conformity with the principles of relevance, can be treated independently of morphology, so also can the "semematic" system, without prejudice tc the solidarity of the different levels0 It is this system that semantics is used for below. (2) Semantics and Morphology, Morphological categories are not, as such, a f^Lt object, fqryS^mant^iCy

for these categories, apart from the general phonemic system of tho language concerned. The logical, absurdity is in both cases the same. This is not to say that me aphology is irrelevant to semantics, Morphe«« mic identity Is a valuable clue to the discovery of sublogical identities beneath superficially different meanings. But it is never a proof that there is, or ever was, such an identity. The assumption of such identities may hinder that of others. For ins- tance it is conceivable that ono moaning of a preposition might be identical with one meaning of a verbal aspect, when all the differences ar^tomatically


- 2 -

consequent on the other units in the combinations concerned were accounted for. The same semantic units may occur, like the same phonemes, in a great variety of morphological categories. If we begin be attributing to each ca- tegory its own single and unique r51e, the way to the discovery of such iden- tities is effectively closed. This is the error of paradigmatic atomism. It is usually accompanied by the error of syntagmatic atomism, the assumption that the meaning of a syn- tagm (after all contextual accidents have been deducted) is necessarily ac- counted for completely by the meanings of the individual morphemes and of their sequence. It is forgotten that the method of commutation, which serves to demonstrate a relevant distinction of meaning, cannot serve to localise this distinction at one point of the chain. That the chain cannot be split up into a series of discrete units follows from a form of cumul seldom or never mentioned under that heading, Cumul is understood to consist of indivisible signifiants answering to divisible signif iés. This presents no difficulty, But the position is usually more complex : we have rather a morpheme A in which the semantic units al, a2 etc, may be localised, a morpheme B with units bl, b2 etc., in a syntagm AB yiol- ding also the units c, d, e which cannot bo localised in either A or B inde- pendently, nor yet in the pattern of their combination, but is spread in inde- terminable proportions over each. We may succeed in reducing these units by putting the burden on the linguistic context, but the fact of a morpheme- division does nothing to assure that this reduction is possible. In brief, neither syntagmatic nor paradigmatic divisions in morphology are a guarantee of semantic division. All arguments to the contrary move in a circle, (3) Semantics and Syntax. In the older linguistics syntax and semantics tended to be confused. In the newer linguistics they tend to be contrasted. Neither tendency is justi- fied. We deal only with the latter. In a much simplified form, the view may be stated as follows : "Two do- finitions of the adjective are conceivable,, The one is syntactic; the adjec- tive is defined by its normally presupposing a substantive in the same syntagm, by its subordination to the substantive etc. Substantive and subordination are also defined syntactically (e.g. subordination is proved by the inability of a -unit to enter the same combinations as the syntagm of which it is part). The other definition would be semantic : the adjective normally expresses an epithet. But this criterion is either secondary, or quite irrelevant". Such a statement is acceptable. The adjective cannot be defined semanti- cally, for the simple reason that it is not a semantic unit. It can be defi- ned syntactically since all units (including the phonemes) can be defined syn- tactically, And since it belongs to no single liguistic plane, but to the Janus-faced class of signs, it can not be defined on any of these planes, but only in terms of syntax, as applicable to all planes. But the notion of a contrast between syntactic and semantic definitions is quite false; since there are also semantic units, and these are capable of a double definition, one syntactic and the other semantic; just as the pho- nemes are capable of a double definition, without anybody having spoken of a contrast between phonemics and syntax, (For instance in many systems the vowel may be defined, as against the consonant, by its non-presupposition of another


- 3 -

unit - a syntactic definition - or by its acoustic features, in the Trubetz- koyan style). These semantic units have not yet been isolated from the speech-continuum, not because this is impossible, nor even because it is diffi- cult, but rather because the climate of structural linguistics is opposed to the question being raised. When these units have been isolated it will be found that they have their own syntax, and that they have their definitions in terms of this syntax; that they are definable extensionally as well as in- tensionally, and have a structure worth investigating for its own sake, (4) The structural definition of semantic relations. Structural semantics is governed by the same general principles as pho- nemics, and in particular by the principle of relevance. Hence if two con- neoted semantic "units” stand in complementary distribution (thus never ans- wering to a distinction of expression) they must be regarded as variants of a single semantic unit. For instance the relations "possessor of (an object)" and "agent of (a process)" between which the difference is automaticaly regulated by the mea- nings of the semantemes to which the relation applies, are variants of the same semantic unit. (Both variants occur in nominal combinations, only the seoond in actor-action phrases, and so on. It is not of course asserted that these are the only relations holding between the terms in question). The structure of semantic relations, apart from their complexity, is similar to that of phonemic relations. Hence the same terms can be used, and illustrated first from the phonemic plane, where we have three principle re- 1 at ions : (i) Two asymmetric relations (ArB is incompatible with BrA) 1 (a) A relation with equipollent poles : sequence. (Equipollent, since an isolated unit is both before and after zero, (b) A relation with privatively contrasted polos : prominen00 (usually actualised by stress-differences as between syllables, or by diffo- rences of syllabicity as between phonemes). (Privative, since an isolated unit is "more prominent than" the surrounding zeros), (ii) One symmetric relation ; juncture (open, closed etc.) (ArB=BrA), Juncture seems always to be privative : an isolated unit has open juncture with surrounding zero. (Hence open juncture is the unmar- ked pole of the opposition. We return to the semantic relation cited above. This relation is obvi- ously asymmetric. It can also be described as privative, since the terms most normally in isolation (e.g, impersonal verbs) are processive and not agental, other terms being neither. The relation thus comes the heading of "prominence". (Though its definition is purely structural, the term answers well enough to our instinctive feeling that the predicate is more prominent than the subject, that the object possessed rather than the possessor is the "centre of attention" in genitival constructions, and so on. But it must be noted again that such terms as predicate and genitive do not belong to this level of analysis; they have semantic relevance but no semantic status). But it is obvious that a term such as prominence is insufficient to define the relation in question (which for want of a name may be called "participation") for the are many other semantic relations answering to the same definition :


- 4 - In fact superordinat ion in syntax (as usually understood) normally answers to prominence in semantics. Many other structural terms lie ready to hand for the narrowing of the definition : transitive and intransitive in the lo- gistic sense, commutational and permutational in the glossematic senses, and so on. (For instance the relations expressed by the cases are normally per- mutable, not commutable). Such terms have however hardly yet been exploited for the structural definition of relations, in the field of linguistics. It is not only in richness that the semantic relations exceed the pho- nomic. To take again the relation whoso variants have been united under the oommun label of »participation”; it will be easy to find this same relation expressed by a stem-morpheme, most commonly of the type have. But then in the group »A has X” two analyses will be necessary : on the one hand there is the relation of participation between A and the group has X (as in any other verbal group); on the other hand there is the same relation between A and X, the verb itself cumulating the semata of participation and other re- lations. These analyses (ArX and ArYX, in which have plays the r81es of r and Y respectively) are contradictory. These contradictory analyses must not be confused with merely indifferent analyses (for instance it is indifferent whether we regard an inflection as affecting a noun or a whole nominal group). Analyses are indifferent when the whole system can be described with equal economy and completeness one way or the other. But here neither analysis can be deduced from the other and both are necessary for a complete description of linguistic relations. The principle of non-contradictory analysis, which (though often some sacrifice of realism) may be maintained in phonemies, breaks down at the start on exa- mination of the semantic system.