The semantic theory KF: a critical analysis

SCOPE AND COMPONENTS

According to KF, the goal of a semantic theory is to account for certain aspects of human competence with respect to a language. This competence involves the production and understanding of expressions abstracted from the non-verbal setting in which they occur. The domain thus staked out for semantics is relatively narrow; it does not include the human ability to name objects correctly, to distinguish synthetically true statements from synthetically false ones, or to perform other referential tasks. In this respect KF follows the tradition of linguistics and saves the investigation of meaning from the sterile ‘ reductions ’ urged upon semantics in recent decades by philosophers of other sciences (cf. Wells 1954; Weinreich, 1968).

But what aspect of linguistic competence is a semantic theory to account for? Programmatically, it aims at nothing less than the ability to interpret sentences.

A semantic theory describes and explains the interpretative ability of speakers [1] by accounting for their performance in determining the number and content of the readings of a sentence; [2] by detecting semantic anomalies; [3] by deciding upon paraphrase relations between sentences; and [4] by marking every other semantic property or relation that plays a role in this ability. [KF, p. 176; bracketed numbers supplied.]

On closer examination, the subject matter of KF turns out to be far less broad. For example, paraphrase relations [3] are touched upon only in passing,¹ and no ‘ other semantic property or relation’, whose explication is promised under [4], is actually considered in the article.² Moreover, the theory cannot deal adequately with the content of readings of a sentence. In actuality, KF is concerned with an extremely limited part of semantic competence: the detection of semantic anomalies and the determination of the number of readings of a sentence.

To carry out this goal, KF visualizes a semantic description of a language as consisting of two types of components: a dictionary and a set of ‘projection rules The dictionary contains statements of meanings of words (or some other suitable entries), each entry being in principle polysemous. The projection rules³ specify how the meanings of words are combined when the words are grammatically constructed, and, in particular, how the ambiguity of individual words is reduced in context. To express the matter schematically, let us imagine a sentence consisting of words A + B + C. The dictionary gives two meanings for A, three for B, and three for C. By multiplying 2x3x3, we calculate that the sentence should be 18 ways ambiguous. In fact it turns out, let us say, that the sentence is only three ways ambiguous. The major function of the projection rules is to account for the reduction of the ambiguity. The limiting case is one in which there is no interpretation of a sentence, even though its components in isolation do have at least one, and possibly more meanings, each.

In an idealized semiotics signs are regarded as combining expressions and meanings in one-to-one correspondence; the polysemy of words in natural languages is but an awkward deviation from the model. KF conforms to the trend of modern lexicology to eschew this prejudice and to seize on polysemy as a characteristic, and even the most researchable, aspect of natural languages (cf. Weinreich 1963 b). KF is also comfortably traditional with regard to the role of context: the idea of contextual resolution of ambiguities has after all been a commonplace with Neogrammarian as well as descriptivist semanticists.⁴ But in assigning this concept so central a place, KF is guilty of two errors. In the first place, it takes no cognizance of the obvious danger that the differentiation of submeanings in a dictionary might continue without limit. In the second place, one would think, a scientific approach which distinguishes between competence (knowledge of a language) and performance (use of a language) ought to regard the automatic disambiguation of potential ambiguities as a matter of hearer performance.⁵ The KF theory only accounts for the construal of unambiguous (or less ambiguous) wholes out of ambiguous parts; it does not undertake to explain, and could not explain, sentences that are meant by the speaker to be ambiguous. In particular, it cannot represent the ambiguity between a grammatical and a deviant sentence (e.g. She is well groomed ‘ combed and dressed), since the theory contains a component (the projection rules) which automatically selects the fully grammatical interpretation, provided there is one. Thus the theory is too weak to account for figurative usage (except the most hackneyed figures) and for many jokes. Whether there is any point to semantic theories which are accountable only for special cases of speech - namely, humorless, prosaic, banal prose - is highly doubtful (Weinreich 1963a: 118).

Semantics can take a page out of the book of grammar. The grammar of a language, too, produces ambiguous expression (e.g. Boiling champagne is interesting, He studied the whole year, Please make her dress fast). But each such sentence, ambiguous at its surface, corresponds to two distinct, unambiguous deep structures. Its ambiguity arises from the existence of transformational rules which produce identical surface results from different deep sources, and from the simultaneous existence of words which can function in dual syntactic capacities (e.g. boil as both a transitive and an intransitive verb). But grammatical theory is not required to explain how a hearer of such ambiguous expressions guesses which of two deep structures is represented by a given occurrence of a surface structure, nor is the goal of grammatical theory limited to the calculation of such ambiguities. The preoccupation of KF with disambiguation appears to be an entirely unjustified diversion of effort. Semantic theories can and should be so formulated as to guarantee that deep structures (including their lexical components) are specified as unambiguous in the first place and proceed from there to account for the interpretation of a complex expression from the known meanings of its components.

¹ Katz and Fodor (1963: 195). The notion ‘paraphrase’ perhaps remained undeveloped because Fodor (1961) and Katz (1961) disagreed about it. See also the discussion of (86) and (87) below [pp. 448-9].

² Katz alone resumed the study of additional relations in a subsequent paper (Katz 1964b). [pp. 446 ff] for further comments.

³ This awkward coinage is based on an allusion to the fact that the speaker, to know a language, must project ‘the finite set of sentences he has fortuitously encountered to the infinite set of sentences of the language’ (KF, p. 171). Since every rule of grammar is involved in the projection mechanism, the term fails to identify its specific content. We suggest no terminological replacement, however, since we aim at a more radical revision of the theory [see esp. §3-5i].

⁴ Although one would not know it from reading KF. See Weinreich (1963a: n. 48) for references.

⁵ This was first pointed out to me by Edward H. Bendix.