Reconsidering [consonantal

Abstract

One of the fundamental claims of modern phonology is that generalisations about phonemes are based on a small, universal set of phonetically defined features. The theory of distinctive features is a very powerful idea, insofar as it predicts that sets of phonemes will act together in specific ways within and across languages, allowing the voiceless stops [p t k] to serve as a unified target or trigger class, but ruling out the possibility that [p r y a] will function as a unitary class. Much of the appeal of distinctive features lies in the fact that the rich array of segments and processes encountered in the world's languages can be understood in terms of a very small number of features. Numerous hypotheses have been advanced regarding what that set of features is – see Jakobsonet al. (1952), Chomsky & Halle (1968), Sagey (1990) and Clements & Hume (1995),inter al. The ideal set of features must be rich enough to express phonemic contrasts which actually exist, and must otherwise be capable of capturing generalisations about sound structure. Equally important, a theory of features should be maximally impoverished – it should use the minimum resources necessary to describe languages.