On the interaction of phonological rules of various types

Discussions of linguistic structure generally accept a distinction between rules of sentence structure and rules of sound structure, but there is very little agreement on further subcategorization beyond this general dichotomy. Indeed, recent discussion in the area of syntax and semantics has been dominated, at least superficially, by the question of whether it is possible and/or desirable to distinguish within the rules of sentence structure between syntactic rules and semantic rules, and perhaps to impose further subcategorizations on each of these; see, for example, the contributions by Chomsky and Postal to Peters (1972). Within syntax and semantics, however, there is probably much less disagreement than meets the eye: most would agree that a desirable goal for linguistics is to isolate and distinguish as many types of rule as possible, and to associate each type with a set of maximally restrictive formal and substantive constraints.1 Although much less attention has been concentrated on this issue in phonology in recent years, the tradition of structuralist or autonomous phonemics took fairly explicit positions on a number of issues concerning the possibility of subdividing the study of sound structure. Figure [i] is a composite representation of the internal organization of that part of a structuralist grammar which deals with phenomena below the level of word structure: Thus, on this view, the conversion of a morphemic representation into sound involves two distinct intermediate levels: a level of morphophonemic representation (for those who believed in the utility of morphophonemics), and a (taxonomic or autonomous) phonemic representation. Each is defined by the set of rules which relate it to those on either side. Since Halle's celebrated argument against the phoneme was first presented in the late 1950s, generative phonologists have generally felt that they had freed themselves from the strictures imposed by the positing of such intermediate levels. Halle demonstrated, that is, that a representation with the characteristics of the autonomous phonemic level could not be maintained as a condition of adequacy for grammars, since this would lead to a loss of generalization in the description of sound structure. The same rule would sometimes have to be