Paul M. Postal
We all have a certain familiarity with the problems of grammar from our schooling. Unfortunately, the approaches of school grammar, for either the native speaker of English or the individual learning English as a second language, are far from the concerns of modern linguistics. In fact, they are sufficiently remote from the research of contemporary linguistic scholarship that this familiarity may be as much a hindrance as a help in understanding grammar as studied within modern linguistics. Perhaps then you will forgive me if I ask you to suspend your familiarity with grammar for what follows.
To a modern linguist, the chief problem of grammar for a particular language, say English, is this. Each speaker has the ability to produce and understand a literally endless variety of sentences, the overwhelming majority of which he has never encountered as wholes before. Our amazing ability to deal with previously unencountered linguistic entities tends to obscure the fact that normal linguistic behavior does deal with unfamiliar sentences. You probably have not noticed at all that you have never before encountered these sentences. The problem of linguistics is thus to explain this amazing "creative" property of linguistic activity, to explain the fact that knowledge of language permits us to create novel sentences and to understand the novel sentences created by others, provided only that these are drawn from a language we know.
It is clear that with only the finite storage capacity of the human brain one cannot have learned an infinite set of sentences directly, the way most words are learned. It follows that