The Setting

In the previous chapter, we reviewed some of the evidence for a specific kind of structural relationship as the basis for the form of English sentences. The fact that neither meaning nor simple linear order can account for certain well-attested syntactic patterns led us to consider a more abstract relationship – structural constituency – as a possible source for these patterns. Displacement and replacement emerged as ‘gold standard’ probes for what turns out to be a hierarchical structure, in which words combine into syntactic units and such units in turn combine into still larger units, corresponding ultimately to the whole sentence. Phrase structure trees were introduced as a convenient graphic representation of this hierarchical mode of organization, and a systematic and workmanlike procedure was demonstrated to guide the reader in constructing arguments for structural analyses of any given sentence.

Having argued for the existence of phrase structure, we now turn to the question of its origins. What determines the class of trees that correspond in English (or in any other human language) to exactly the set of sentences constituting that language? One often sees arguments to the effect that the familiar fact of speakers being able to distinguish sentences of their languages, regardless of length, from nonsentences using the same set of vocabulary, means that whatever it is that corresponds to that knowledge must be finite (since it models a cognitive ability which is itself necessarily finite) but with no specified upper limit on the number of separate structures, and therefore word strings, that it can define as well-formed. Putting it quite roughly, our knowledge of language points to some kind of formal system with a finite set of components but no inherent limit on the size of its output. But this line of reasoning asserts an identity between a theory of the sentences of a language and our knowledge of those sentences, which is quite problematic (after all, a theory of the integers is going to look quite different from a psychologically plausible theory of our ability to count and carry out arithmetic operations).