acquisition). Moreover, if he himself is to produce acceptable versions of the speech sounds he perceives, he also has to learn the idiosyncrasies of his own vocal tract. He must learn to control certain stylistic factors - speaking rate, attitudinal intonation and so on, both in production and perception. Finally, he may need to learn certain global phonetic properties of his language, e.g. its 'articulation basis' (Heffner 1950: 98-9).

In brief, we distinguish four distinct components underlying speech perception and production:  1) inborn phonological capability,  2) acquired phonological competence in one's language,  3) inborn phonetic capacity,  4) acquired phonetic skill. 7  For the study of these various components, speech synthesis by rule has certain impressive advantages.

First, we can hope to gain real understanding of the component of interest to us only by attempting a highly formal account; yet any nontrivial formal account will doubtless be quite complex: this is already apparent for phonological and phonetic capacities and particularly so for phonological competence -- as a glance at the summary of rules in Chapter 5 of  The sound pattern of English (Chomsky and Halle 1968) will confirm -- and must certainly prove true for phonetic skill as well. Much can be done to reduce apparent complexity by suitable notation. But, as in many other fields which make use of highly formal systems, checking the consistency of the formalization is most easily done by computer simulation. Linguists are in fact turning increasingly to computer simulation to check the operation of syntactic and phonological rules (Fromkin and Rice 1970).

Second, various dependencies exist among the components. An account of the phonetic skill of a particular speaker must begin with some assumptions about his phonological competence in his language and his phonetic capacity. Only in terms of the former can idiolectal variations be defined; only in terms of the latter can speaking rate be discussed. Similarly, the rules by which we try to characterize phonological competence must be stated in a form determined by phonological capacity. Phonological capacity, finally, depends on the choice of a set of features, the interpretation of which is a matter of phonetic capacity. Given this kind of dependency, it seems extremely risky to try to form hypotheses about the nature of one component without being quite specific as to the assumptions being made about the others on which it depends. Yet this is an ever-present temptation. Speaker variation, for example, is investigated without specification of precise phonetic and phonological models. Structural linguists rightly incurred the censure of generative phonologists because they formulated their phonemic inventories without proper concern for phonological capacity; generative phonologists, in turn,
__________
7.  Tatham (1969a) has recently used the term 'phonetic competence' to mean approximately what we mean by 'phonetic capacity'; otherwise we might have used the former term rather than the assymetrical 'phonetic skill'. Tatham's paper (see also Tatham 1969b) contains some cogent arguments not only for the existence of phonetic capacity but also for its importance in the formulation of phonological rules in a natural way.