Differentiating 451 languages in terms of their segment inventories

Abstract

A “segment niche” for a language is a subset of its segment inventory that can distinguish this language from all other languages. This article gives the computer-generated segment niches for the 451 languages in the UCLA Phonological Segment Inventory Database (UPSID-451). It is shown that languages usually possess a multitude of segment niches, but not all of these alternatives are equally representative of a language's idiosyncrasy. A criterion is suggested for choosing among these alternatives: Preferring the niche for a language which contains segments with the smallest frequencies of occurrence in other languages (this niche, in effect, can be considered as more “typical” for the language). Some patterns emerge regarding the size and structure of the computed “typical” niches. © 2001 Blackwell Publishers Ltd.