a certain inborn linguistic capacity: 'the innate organization that determines what counts as linguistic experience and what knowledge of language arises on the basis of this experience' (Chomsky 1968: 24). This capacity is what makes it possible for him to learn the rules and to use them in making grammatical judgments; it is reflected in some universal and quite severe constraints on the form and content of grammatical rules. In the case of phonology, not only are the forms of the input and output highly determined, but the set of phonetic features by which the output of the phonology may be represented is the same for all languages; moreover, the language-specific set of 'classificatory' features which are used to represent the lexical items at the input to the phonology are related in a significant though complex way to a subset of universal phonetic features with the same names. Phonological rules are constructed out of phonetic raw material.

Conventionally, the linguist's concern ends with the phonetic-feature representation which is the output of the phonology. But our account of the speaker-hearer's inborn capacity is incomplete, for we have said nothing about his knowledge (quite unconscious, but no less psychologically real) of the relationship between the phonetic-feature representation and the acoustic signal. The speaker-hearer can produce an acoustic representation of an utterance, given the feature representation (speech production), and he can apparently recover the feature representation for an utterance produced by someone else (speech perception). Neither process is trivial; to be able to produce and recognize speech, he must possess a definition of each feature in sufficient detail to enable him to assign a possible value to it, given the acoustic signal. He must therefore have sufficient information about the anatomy, physiology and acoustics of the vocal tract to permit such a definition; and he must also understand how the apparently discrete representation of an utterance, at the output of the phonology, as a series of phonetic segments, is translated into a continuous representation in the acoustic signal. Perhaps it would not be inappropriate to picture the speaker-hearer's knowledge as consisting of a dynamic neural simulation of the vocal tract, the state of which is determined by the values of the features, and which guides his production and perception of speech (Mattingly and Liberman 1969). If we assume that both the features and the general structure of the human vocal tract are universal, it seems highly likely a priori that this knowledge of the speaker-hearer's is inborn; moreover, some experimental evidence for such a view has recently appeared (Moffitt 1969; Eimas et al. 1971).

Just as we characterize phonological and syntactic capacity by assigning to them formal and substantive properties, in the form of conventions and features, so, to the extent that we can describe this simulated vocal tract, we characterize what may be called 'phonetic capacity'. It is to phonetic capacity which Chomsky and Halle (1968: 294-295) allude when they observe:

The total set of features is identical with the set of phonetic properties that can in principle be controlled in speech: they represent the phonetic capabilities of man and, we should assume, are therefore the same for all languages.