Introduction

This chapter presents the results of an analysis of the vowel systems of the 317 languages in the UCLA Phonological Segment Inventory Database (UPSID). It shows that deviations from the patterns predicted by a theory which proposes that vowels are dispersed in the available phonetic space are relatively infrequent and, for the most part, confined to matters of small scale, falling into a few definable classes. It will be argued that in most of these deviations from the predicted patterns there is nonetheless evidence that vowels tend toward a balanced and wide dispersion in the available phonetic space.

Preliminaries

A few basic vowel inventories and a few basic configurations show up time and again in natural languages, while other, no more complex patterns are rare or totally absent. The most prevalent patterns seem to be the so-called “triangular” systems, particularly those of average size, and notably the 5-vowel systems. For example, over a quarter of the 209 languages in the Stanford Phonology Archive have a triangular 5-vowel system consisting of /i, ε, a, o, u/, while less than 5% have any of the other 5-vowel configurations; the “square” 4-vowel and 6-vowel systems combined total less than 10% (Crothers 1978).

Several attempts to explain these patterns invoke a principle of vowel dispersion, proposed in slightly differing versions by Liljencrants and Lindblom (1972), Lindblom (1975), Terbeek (1977), and Maddieson (1977).