Harvey M. Sussman
Depts. of Linguist. & Speech Commun., Univ. of Texas, Austin, TX 78712
Locus equations, regression fits to coordinates obtained by plotting F2 of a variety of vowels at their midvowel nuclei in relation to F2 at the acoustic onset of the CV syllable have been shown to phonetically index place of articulation within oral stop categories across speakers of diverse languages---English, Swedish, Cairene Arabic, Thai, and Urdu. The purpose of this study was to determine if locus equations best encode featural or segmental entities. Several consonantal manner classes were assessed. The inventory consisted of CVt tokens having initial approximants---/w/, /j/, /r/, /l/; fricatives---/v/, /(edh)/, /z/, /(large hook)/; nasals---/m/, /n/. In addition voicing contrasts were examined using voiceless stops (s/C[sub 1]/VC[sub 2]) with C[sub 1]=/p/,/t/,/k/. Ten medial vowel contexts were used. Locus equations were found to acoustically distinguish place contrasts within each manner class, with nasality and voicing not affecting the place feature. However, similar place features across stop--fricative--approximant classes, did not correspond in terms of slope and y-intercept values. It was concluded that locus equations best characterize segment-based rather than feature-based, entities.