Michelle R. Molis
Dept. of Psychol., Univ. of Texas, Austin, TX 78712
Stockholm Univ., Stockholm, Sweden
Univ. of Texas, Austin, TX 78712
ENST, Paris, France
Ohman [J. Acoust. Soc. Am. 39, 151--168 (1966)] reported superposition of a consonant closure gesture on a vowel-to-vowel transition was sufficient to describe VCV coarticulation; however, other researchers [R. McAllister and O. Engstrand, Fonetik, 115--119 (1992)] have found a language-dependent articulatory trough in the movement of the tongue during some VCV sequences. Such a trough would limit the possible extent of coarticulation. In this study, one male speaker each of American English, French, and Swedish produced VCV sequences. Vowels included /i/, /a/, and /u/. For each of three stop consonants (/b/, /p/, and /d/), an index of coarticulation was obtained through calculation of a locus equation. In addition, coarticulation indices were obtained from the output of an acoustic tube model that uses superposition to generate VCV sequences. Preliminary results indicated that superposition alone predicted coarticulation of unaspirated stops in all languages, but was not sufficient to explain the reduced coarticulation observed for aspirated stops. Reduction in degree of coarticulation corresponded qualitatively with the language dependent extent of the articulatory trough. With modifications intended to simulate an articulatory trough, the model successfully simulated the qualitative change in cross-linguistic coarticulation.