ASA 124th Meeting New Orleans 1992 October

4aSP7. Effect of token variability on discrimination between vowel sounds II.

D. E. Ronan

R. M. Uchanski

L. D. Braida

Res. Lab. of Electron., MIT, Cambridge, MA 02139

Uchanski et al. [J. Acoust. Soc. Am. 90, 2254 (A) (1991)] reported significant sensitivity differences between discrimination experiments in which each vowel was represented by only four tokens (utterances) and experiments in which the same four tokens were embedded in a larger set of 16 tokens. In present experiments, the set size was fixed at 8 but the specific tokens representing one vowel were changed while those representing the other vowel were fixed. Listeners attempted to discriminate between the vowels /ae/ and /ah/ occurring in distorted [M. R. Schroeder, J. Acoust. Soc. Am. 44, 1735--1736 (1968)] /h/--V--/d/ syllables. Sensitivity was found to be more highly correlated ((rho)(approximately equal to)0.8) across conditions for a given subject than across subjects for a given condition ((rho)(approximately equal to)0.5). This suggests that each subject weighted the available discrimination cues in a fairly stable fashion across conditions, but that the weights varied from subject to subject. Multiple regression analyses indicate that sensitivity can be accounted for by a small set of physical parameters of the undistorted syllables: F2, F0, and syllable duration.