Identification of synthetic diphthongs with variable transition duration

Abstract

Traditionally, diphthongs have been described as consisting of an initial steady state portion, a transition, and a relatively short final steady state portion. Gay [“A Perceptual Study of American English Diphthongs,” Lang. and Speech 13, 65–88 (1970)] found that the presence of a transition without steady state portions is sufficient for the identification of synthetic diphthongs. This study investigated the contribution of transitions of variable duration, in the presence of steady state portions, to diphthong identification. The three American English diphthongs /ai/, /au/, and /oi/ were synthesized, employing fixed durations for the steady state portions and varying the duration of the transition from 0 (no transition) to approximately 150 msec. Two‐vowel sequences, identical to the steady state portions of the diphthongs and separated by pauses of varying durations, were also synthesized. Comparable sets of synthetic diphthongs and two‐vowel sequences, synthesized at different fundamental frequencies, were presented to different groups of subjects; the subjects were required to identify each stimulus item as one of the three diphthongs or one of the three possible two‐vowel sequences. When the transitions were short or absent, subjects were consistent in identifying the stimuli as diphthongs. For some stimulus items, even relatively long pauses separating two steady state vowel segments were ineffective in shifting the identification to a vowel sequence.