Much of grammatical description has to do with classifications of linguistic forms. Conceivably, if classification were thorough and complete enough, it would contain enough information about the structure of the language to tell what does and what does not occur and would give what has recently been called a generative grammar. In practice, however, the usual classificatory categories found in grammars neither give all the essential facts of structure nor enable one to produce all those and only those forms which are grammatical in the language analyzed. As Wang Lih has said, “Whether we regard English ‘than’ as a conjunction or a preposition, or, with Jespersen, as a particle, these are all unimportant questions. What is important is to tell people under what conditions it is to be followed by the nominative case and under what conditions by the objective case, or whether either is possible. Our responsibility ends when we can make a foreigner to the language use the word ‘than’ without error. As for the argument over what to call it, it is as trivial as whether the chicken gets a chance at the worm.” (Liiluenn, I, 4).
Grammatical structure can be studied by analyzing it into hierarchical structure of immediate constituents, or ICs; by predicting new forms from given forms, such as attributive from predicative, passive from active, interrogative from declarative, negative from affirmative, and other new forms by the now well-known process of transformation; and by additional processes which together with transformations will serve to generate all those and none but those forms which will be grammatical in the language. Since no transformational or generative grammar has as yet been fully worked out for any language, we shall, for the purposes of the present work, take the conservative position of using hierarchies of ICs as the main approach to structure (Chomsky’s “phrase structure grammar”)，but will make such use of transformations or rules of generation as will be useful. Statements to the effect that you can say this and you can’t say that will be fairly frequent, but no attempt will be made here to build a system of generative grammar of Chinese.
On these various approaches, see for example Wells, “Immediate Constituents”; Chomsky, Syntactic Structures; and Victor H. Yngve, “A Model and an Hypothesis for Language Structure”, Proc. of the Amer. Philos. Soc., 104:444-466 (1960).
Chomsky, chap. 4.
Chao Y R. A grammar of spoken Chinese[M]. Univ of California Press, 1965.