Ml. Erickson, Simultaneous effects on vowel duration in American English: A covariance structure modeling approach, J ACOUST SO, 108(6), 2000, pp. 2980-2995
The powerful techniques of covariance structure modeling (CSM) long have be
en used to study complex behavioral phenomenon in the social and behavioral
sciences. This study employed these same techniques to examine simultaneou
s effects on vowel duration in American English. Additionally, this study i
nvestigated whether a single population model of vowel duration fits observ
ed data better than a dual population model where separate parameters are g
enerated for syllables that carry large information loads and for syllables
that specify linguistic relationships. For the single population model, in
trinsic duration, phrase final position, lexical stress, post-vocalic conso
nant voicing, and position in word all were significant predictors of vowel
duration. However, the dual population model, in which separate model para
meters were generated for (1) monosyllabic content words and lexically stre
ssed syllables and (2) monosyllabic function words and lexically unstressed
syllables, fit the data better than the single population model. Intrinsic
duration and phrase final position affected duration similarly for both th
e populations. On the other hand, the effects of post-vocalic consonant voi
cing and position in word, while significant predictors of vowel duration i
n content words and stressed syllables, were not significant predictors of
vowel duration in function words or unstressed syllables. These results are
not unexpected, based on previous research, and suggest that covariance st
ructure analysis can be used as a complementary technique in linguistic and
phonetic research. (C) 2000 Acoustical Society of America. [S0001-4966(00)
01812-9].