Formal Properties of Phonological Rules

Abstract

One of the salient features of generative phonology has been the emphasis put on formal questions. The study of abstract properties of grammars has been the distinctive concern of work carried out in the field and most of the research has been aimed at discovering formal universals. The fundamental methodological assumption has been that the investigation of formal properties of grammars would eventually lead to significant discoveries. At the same time, however, generative linguists have been aware that phonological theory, as a part of a general theory of language, should concern itself with substantive universals as well. In particular, it was apparent that the evaluation measure would have to incorporate an elaborate system of substantive constraints. For example, in The Sound Pattern of English (SPE),¹ the basic theoretical work of generative phonology, Chomsky and Halle remark that the formal evaluation procedures and the associated notational devices defined in the early chapters of their book give wrong results in many instances and that they must be supplemented by a set of conventions which takes into account the intrinsic content of phonological features and of phonological rules.² Of course, there is no contradiction in undertaking investigations of both types of universals, formal and substantive. The open, and empirical, question, however, is how greatly substantive conditions limit the class of grammars available to the language-learner: if very greatly, formal properties have little interest. In this paper I would like to provide a partial answer to this question, and to argue that the fundamental methodological assumption of generative phonology, as stated at the beginning of this introduction, is correct. I will suggest that phonological theory contains a rich core of formal. constraints. In particular, I will present some striking results from Halle et al. (1975).

I gratefully acknowledge the help of the following people: Jay Keyser, Morris Halle, Alan Prince, and Elisabeth Selkirk. This work was supported by C.N.R.S. grant E.R.A. No. 247.