The method of semantic description

COMPONENTIAL AND DEFINITIONAL APPROACHES

In the componential approach each word is semantically described in terms of a number of basic components or features; the features involved in the semantic description of the words of a language may be arranged in a number of systems.¹ For instance man might be describable through features ‘human’-from the system {human, non-human} - ‘ male ’ - from system {male, female} - and ‘ old ’ - from system {young, old} - see Hjelmslev and Uldall (1957.45-6). This method of semantic description has been applied most successfully to kinship terms (for instance Goodenough 1956, Lounsbury 1956, 1964a, 1964b). In the definitional approach the meaning of a word is defined in terms of the meanings of other words, related together through any of the grammatical constructions of the language (this method does not deal in terms of elements of meaning smaller than the meanings of words). This approach has been employed - though not always very consistently or well-in monolingual dictionaries; for a discussion and critique see Weinreich(1962).

The componential approach has the advantage of breaking down word meanings into more basic units. Words of similar meaning are now directly related through their semantic descriptions having one or more features in common. This approach is unsatisfactory in that it does not refer much to the grammatical resources of the language, and the grammatical patterning is after all the rich heart of a language, that we would expect a semantic description to refer to and to exploit in a variety of ways. A further drawback of the componential approach is that very many semantic features must be posited - a number of the same order as the number of words to be semantically described.² Many features occur just in the semantic description of a single word. A componential description thus appears to gain little in economy over a simple list of the words of the language, each considered as a semantic primitive in its own right.

The definitional approach has the advantage that it does exploit grammatical relationships. It does not however in all cases clearly demonstrate the relationship between synonyms, near-synonyms and antonyms, as does the componential approach. The most severe drawback to the definitional approach lies in vicious circles - in following up the definitions of the words occurring in the definition of a given word, and so on, one will sooner or later be confronted with a word that has been encountered earlier in the chain; an advantage of the componential approach is that vicious circles are avoided.

¹ The features of a system are referentially complementary; also, each semantic description can choose one and only one feature from a given system. Systems typically enter into dependency trees, involving conditions of the type: if a semantic description chooses feature xx from system {X₁, X₂} then it must go on to make a choice of one feature from system {Y1, Y2, Y3}; if X2 is chosen then a choice must be made from system {z1} z2j. For an example of such a dependency tree - but involving syntactic features - see Chomsky 1965.83. The idea of linguistic system, in the sense used here, has been stressed by Firth (1957) and Halliday (1961); it has been made much of in recent work by Halliday (1967-8) - where the term ‘ system network ’ is applied to dependency trees. Little or no use is made of the idea of system in the work of Lamb (1964), Katz and Fodor (1963), for instance, and these writers appear to set up rather ad hoc features in dealing with the examples they consider. Compare Weinreich’s comment (1962.30): ‘the consistency of lexicography could be improved if dictionary makers were held to the assumption that the terms of a language are, on the whole, complementary. This assumption implies that the most important case to deal with in semantic description is one in which, where the signification of one term ends, that of another begins. On the whole, a semantic description should not aim at “absolute” definitions, but at definitions which delimit the meaning of a term from that of terms with similar meanings (synonyms).’ It should be noted that what is here called a ‘ feature ’ corresponds to a ‘ value of a feature ’ in Jakobson- Halle-Chomsky terminology, our ‘ system ’ corresponding to their ‘ feature ’. Thus where we would refer to a binary system with features ‘to’ and ‘from’, they would instead talk of a feature ‘ to ’, with possible values + to ’ (corresponding to our feature ‘ to ’) and ‘ — to ’ (corresponding to our ‘ from ’). This difference is entirely terminological. A non-trivial difference is that we do not restrict ourselves to binary systems. Empirically, most systems do have only two features but there are some with three or more features: thus {sit, stand, lie}, which cannot be reduced in any non-arbitrary way to a pair of binary systems.

² Cf. Lamb 1966.566: ‘it seems obvious that the number of such minimal [semantic] components for any real language will number in the thousands’.