A Prosodic Theory of Vocalic Contrasts

Phonetic transcription allows us to put in square brackets many things that languages do not actually make use of, such as palatalized velar glides [ɰj] or velarized palatal glides [jɣ]. It also allows us to posit unattested contrasts like prevs. post-palatalized nasals [jn∼nj] and to entertain what seem to be purely orthographic contrasts like [pja∼pja]. We argue here that natural language does not use such refined distinctions and offer a more restrictive theory of vocalic features that treats them as properties of syllable margins (onsets and codas) rather than properties of individual consonants. Following Ladefoged & Maddieson (1996, 2), our study focuses on the elements "that are known to distinguish lexical items within a language", i.e., on minimal-pair contrasts involving labialization, palatalization, and velarization within single morphemes.1 The facts we present here suggest that natural languages allow at most a single unordered set of vocalic features per syllable margin, whatever the number of segments in that domain. For this reason, we propose that