- •Chapter 7
- •7.0 Introduction
- •202 The formalization of sentence-meaning
- •204 The formalization of sentence-meaning
- •206 The formalization of sentence-meaning
- •208 The formalization of sentence-meaning
- •210 The formalization of sentence-meaning
- •212 The formalization of sentence-meaning
- •214 The formalization of sentence-meaning
- •216 The formalization of sentence-meaning
- •218 The formalization of sentence-meaning
- •220 The formalization of sentence-meaning
- •7.5 Montague grammar
- •224 The formalization of sentence-meaning
- •226 The formalization of sentence-meaning
- •7.6 Possible worlds
- •228 The formalization of sentence-meaning
- •230 The formalization of sentence-meaning
- •8.0 Introduction
- •8.1 Utterances
- •8.2 Locutionaryacts
- •9.0 Introduction
- •9.1 Text-sentences
- •9.6 What is context?
218 The formalization of sentence-meaning
(8) 'My friend frightened that it was raining'.
Arguably, although I have represented them as sentences, each of them is both grammatically and semantically ill-formed. Their ungrammaticality can be readily accounted for by saying that 'exist' is an intransitive verb (and therefore cannot take an object) and that 'frighten', unlike 'think', 'say', etc., cannot occur with a that-clause as its object. (Such examples are handled by Chomsky in Aspects in terms of what he calls strict subcategor-ization.) The fact that they do not make sense - that they have no prepositional content - can be explained by saying that it is inherent in the meaning of 'exist' that it cannot take an object, and that it is inherent in the meaning of 'frighten' that it cannot take as its object an expression referring to such abstract entities as facts or propositions. But which, if either, of these two explanations - syntactic or semantic — is correct?
The question is wrongly formulated. It makes unjustified assumptions about the separability of syntax and semantics and ignores the fact that, although natural languages vary considerably as to what they grammaticalize (or lexicalize), there is, in all natural languages, some degree of congruence between semantic (or ontological) categories and certain grammatical categories, such as the major parts of speech, gender, number, or tense. Whether one accounts for categorial incongruity by means of the syntactic rules of the base component or alternatively by means of the blocking, or filtering, mechanism of the projection-rules is, of itself, a technical issue of no empirical import. What is important is that, whatever treatment is adopted, the details of the formalization should distinguish cases of categorial incongruity from (a) cases of contradiction and from (b) what are more generally handled in terms of selection-restrictions.
Contradictory propositions are meaningful, but necessarily false. Expressions whose putative semantic ill-formedness results from the violation of selection-restrictions can often be given a perfectly satisfactory interpretation if one is prepared to make not very radical adjustments to one's assumptions about the nature of the world. Categorially incongruous expressions are
7.4 Projection-rules and selection-restrictions 219
meaningless and they cannot be interpreted by making such minor ontological adjustments. (A classic Chomskyan example of a sentence containing a mixture of contradictory and catego-rially incongruous expressions is 'Colourless green ideas sleep furiously': see 5.3). These boundaries may be difficult to draw in respect of particular examples. But the differences are clear enough in a sufficient number of cases for the distinctions themselves to be established.
Let us now return to the Katz-Fodor projection-rules. We have seen how they distinguish meaningful sentences from at least one class of meaningless, or allegedly meaningless, sentences. They also have to assign to every semantically well-formed sentence a formal specification of its meaning or meanings. Such specifications of sentence-meaning are described as semantic representations.
It follows from what has been said so far that a sentence will have exactly as many semantic representations as it has meanings (the limiting case being that of meaningless sentences, to which the projection-rules will assign no semantic representation at all). It also follows that sentences with the same deep structure will have the same semantic representation. The converse, however, does not follow; in the standard theory of transformational-generative grammar, sentences that differ in deep structure may nevertheless have the same meaning. This is a consequence of the existence of synonymous, but lexically distinct, expressions (see 2.3) and of the way in which lexicalization is handled in the standard theory. We may simply note that this is so, without going into the details.
But what precisely are semantic representations? And how are they constructed by the projection-rules? These two questions are, of course, interdependent (by virtue of the principle of compositionality). A semantic representation is a collection, or amalgamation, of sense-components. But it is not merely an unstructured set of such components. As we saw in section 4.3, it is not generally possible to formalize the meaning of individual lexemes compositionally in set-theoretic terms. It is even more obviously the case that sentence-meaning cannot be formalized I in this way. If a semantic representation were nothing more