Yoruba Vowel Patterns: Asymmetries through Phonological Competition*

1 Introduction The analysis of Yoruba vowels has been the focus of several papers and has been a topic in numerous others. But while this body of work has led to a clear description of a wide variety of phenomena, it has also raised a number of problems of both analytic and theoretical interest. The core issue to be addressed here arises out of such work. Various patterns have been described that identify a range of special properties for high vowels in general and [i] in particular (e.g. etc.). In an attempt to account in a principled fashion for the special status accorded the high vowels, the analysis of these patterns was proposed in Pulley-blank 1988a to depend on a fundamental structural property, underspecification of the high front vowel. While successful for a certain range of data, subsequent work demonstrated the inadequacy of the underspecification hypothesis (Clements & S¢ onaiya 1990, Akinlabi 1993). There has been no account, however , succeeding in integrating accounts of the diverse range of phenomena into a unified analysis. A close examination of the argument against underspecification makes it clear that the overall issue of how to treat Yoruba vowels remains open. The problem is this. Asymmetric behaviour of high vowels necessitates some sort of special treatment. Underspecification was one possibility. When problems with an underspecified analysis of Yoruba were raised, the problems were not with the data patterns – it was agreed that high vowels were special – rather, the problems were with the underspecified analysis. Indeed, papers such as Clements & S¢ onaiya (1990) and Akinlabi (1993) showed that high vowels were special in still more ways, but that their special properties could not be accounted for by the device of underspecification. The problem of how to replace an underspeci-fied account by a more adequate account was left open. This paper proposes to resolve that problem. The development of an adequate theory of natural language depends crucially on integrating comprehensive analyses of individual languages with an understanding of cross-linguistic patterns. This paper constitutes a case-study of a class of vocalic patterns in Yoruba, attempting such integration. The central problem is how to account for a range of processes where vowels behave in an asymmetric fashion. Specifically, high vowels in general and the vowel [i] in