Gestural overlap of stop‐consonant sequences: Evidence from analysis and synthesis

Abstract

This study uses an analysis‐by‐synthesis approach to discover possible principles governing the coordination of oral and laryngeal articulators in the production of English stop‐consonant sequences. Individual recordings were made of two male and two female native American‐English speakers reading phrases which include voiced and voiceless stop consonants in word‐initial (V♯CV) and word‐final (VC♯V) positions, as well as in VC♯CV stop–stop consonant sequences. Articulatory timing estimates were made based on analyzing acoustic data including formant movements, closure durations, release bursts, and spectrum shape at low frequencies. Based on the gestural estimates, the same consonant sequences were generated using HLsyn, a quasiarticulatory synthesizer. The synthetic utterances were acoustically and perceptually compared to the actual utterances in order to verify and refine the articulatory timing estimates from which possible principles could be derived. Preliminary results agree with earlier findings of more overlapping of oral gestures in sequences with front‐to‐back order of place of articulation than those with back‐to‐front order [Chitoran, Goldstein, and Byrd, Lab. Phonology 7, 419–448 (2002)]. Furthermore, overlapping of laryngeal gestures is suggested by the smaller acoustical loss at the glottis in vowels after voiced–voiceless sequences than voiceless–voiceless sequences.