ABSTRACT
Within current Transformational Generative Grammar a grammar is regarded as a set of rules relating sound to meaning. These rules must generate all and only the grammatical sentences of a language, and they must assign the correct meanings and phonetic properties to the sentences generated. The earliest model of Chomsky (1957) attempted to predict all and only the grammaticality or wellformedness facts of English and allowed for the generation of phonetic representations, but was not concerned with the assignment of meanings, in the form of semantic representations, to the syntactic strings generated. Not only was there no semantic component with the purpose of assigning such semantic representations — as there was in Chomsky (1965) — but it was further claimed that considerations of meaning were quite separate from the kinds of syntactic facts on the basis of which all and only the grammatical sentences of a language were to be predicted. Syntax was, therefore, autonomous and independent of meaning.
