A Perceptual Study of American English Diphthongs

Abstract

This study utilized synthetic speech for describing the phonemic boundaries and the effects of formant course on the identification of / i, ai, au/. In the first experiment, synthesized formant transitions appropriate to classes of / i-ai/ and /au-o/ were presented to ten phonetically trained listeners for purposes of phoneme labelling. Results showed that preferred / i/ courses were from [ ] or [u] to [i, y, I], /ai/ courses from [a] to [I] and /au/ from [a] to [ ]. The purpose of the second experiment was to determine whether the phonetic identity of the targets or the absolute course of the second formant transition serves as the primary identifying cue. These features were separated along the time dimension by synthesizing diphthongs whose second formant frequency course remained fixed but whose durations varied. Results of the listening tests showed shifts in perception from simple vowel to diphthong occurring as a function of duration rather than frequency onset or offset positions. Further, in the case of / i-ai/, second formant rate of change serves as the primary distinguishing cue. The results suggest / i, ai, au/ are characterized by an invariant second formant rate of change, whose onset and offset values vary across changes in duration.