¹ Besides Noam Chomsky’s classic book Syntactic Structures (The Hague, 1957), see also the bibliography given by Emmon Bach in his An Introduction to Transformational Grammars (New York, 1964).

² In Syntactic Structures, complex underlying structures were viewed as being derived by generalized, or double base, transformational rules which operated simultaneously on two sentences and embedded one into the other or conjoined one to the other. In a series of public lectures given at the Linguistic Institute at Indiana University in the summer of 1964, Chomsky gave a number of convincing arguments to show that abstract underlying structures, whether complex or simple, should be developed in the expansion rule component of the grammar and that actual, phonetically interpretable sentences should be derived exclusively by singularly transformational rules operating on the abstract structures.

³ The transformational component in a generative grammar is, of course, motivated by formal considerations and not by the observation that native speakers react in one way or another towards similarities and differences among sentences. I will cite a simple but convincing example. Native speakers of English surely know that there is a relationship between The man is going there, and Is the man going there? The first is a statement and the second is the corresponding question, and the two sentences contain roughly the same morphemes. These sentences should be related formally and explicitly in a generative grammar of English, not because native speakers say they are related but rather because of certain formal facts about such sentences in English. Notice, for example, that the number agreement between subject and auxiliary is the same for both sentences; the question order, with is first, is grammatical in the same way that the statement order, with is following the subject, is grammatical; and the question * Are the man going there? is ungrammatical in precisely the same way that *The man are going there, is. Questions and statements differ in the order of subject and auxiliary, but the number agreement is exactly the same for both. If we allow our descriptive grammar to contain a rule, a transformation, which derives one ordering of constituents from the other, then we need state the agreement rule only once. That is, we apply the agreement rule before the permutation rule, and the correct agreement is inherited by the output of the latter. This achieves a tremendous saving not only in the formulation of agreement rules of this type but also in the formulation of “selection” restrictions which ensure correct co-occurrence of so-called major morphemes, nouns and verbs, in the output of the grammar, i.e., rules which insure that such sentences as I amazed John’s son. be included in the output of the grammatical rules, but that a sentence like *I amazed John’s beer-drinking, be excluded or, at least, identified as deviant. The transformational form of grammar avoids unnecessary repetition of statements of agreement and selection, and this formal motivation is quite compatible with a native speaker’s intuition about relationships among sentences.