Time-varying information for vowel identification in visual speech perception

Abstract

The relative salience of time-independent and time-varying dimensions during vowel identification accuracy was examined for visual speech perception. Six different conditions were created using nine vowels /o,I,i,e,ae,a,u,U,uh/ in the consonantal context /bVb/: control syllables, variable center syllables, silent center syllables, initial syllables, and final syllables. Control syllables were full syllables. Variable center syllables were produced by deleting the beginning and end transitions, leaving 50%–60% of the vowel. In silent center syllables, 50%–60% of the vowel was removed, leaving the beginning and end transitions. Initial syllables contained the beginning transition, and final syllables contained the end transition. Results revealed that silent center vowels were identified as accurately as the control syllables. All other conditions showed a significant decrease in visual vowel identification accuracy. A follow-up experiment revealed that the silent center syllable accuracy was greater than the variable center syllables even when their durations were similar. These results are analogous to findings in auditory speech perception [J. J. Jenkins, W. Strange, and T. R. Edman, Percept. Psychophys. 33, 441–450 (1983)] and suggest that the time-varying aspects of the syllable may be the primary source of information for visual vowel identification.