Pre-glottal vowels in Shanghai Chinese

This study examines the acoustic realization of two complementary sets of vowels in Shanghai Chinese: One appearing in open syllables only (OSVs) and the other in closed syllables with a glottal coda (PGVs). Two factors consonant onset and prominence level were controlled to test the null hypothesis that PGVs are reduced realization of the OSVs due to their short duration. Results showed spectral reduction of the PGVs but suggested that PGVs cannot be the reduced realization of OSVs as both onsetless and prominent PGVs failed to be consistently realized with a more expanded acoustic vowel space, which was predicted by the reduction account. We propose that the spectral reduction of the PGVs is an inherent feature of the vowel’s phonetic implementation. Introduction Shanghai Chinese is a Wu dialect spoken in the city of Shanghai. One interesting feature of the language is that it has two types of vowels: One occurs in CV syllables (i.e. open-syllable vowels, hereafter OSV) and the other in closed syllables with a glottal coda (i.e. pre-glottal vowels, hereafter PGV). It is clear that PGVs are much shorter than OSVs (Zee & Maddieson 1979). The phonemic status of the PGVs, however, has been of much controversy (Chen & Gussenhoven, submitted, and references therein). This paper seeks acoustic evidence to test the null hypothesis that PGVs are the reduced and more centralized realization of some of the OSVs, given the effect of short duration on vowel production (Lindblom 1963). Two factors, consonant onset and prosodic prominence, which have been commonly recognized as important determinants of vowel reduction, were therefore carefully controlled. Studies on the effect of consonant environment on vowel formant patterns show that vowels in CVC are in general more centralized than that in the /hVd/ context, the classical “null” environment (Peterson & Barney 1952), which is comparable to vowels in isolation (Stevens & House 1963, Hillenbrand & Clark 2001, among others). As for the effect of prosodic prominence, it has been shown that vowels occurring in a prosodically prominent position are longer and implemented with more effort, and consequently the underlying vowel target is better approximated (Fourakis 1991, van Bergem 1995, Moon & Lindblom 1994, among others). If indeed PGVs are the reduced and more centralized realization of some of ITRW on Experimental Linguistics Athens, Greece August 28-30, 2006 ISCA Archive http://www.isca-speech.org/archive