When a consonant emerges from vowels : The conditions of the audiovisual integration

Christian *ABRY, Institut de la Communication Parlée Université Stendhal

Abstract
We tested auditory and visual perception of the vowel-to-vowel [yi] gesture via the production of a [˙] epenthetic glide in-between. This epenthetic glide can gain in the course of linguistic change the status of a true represented segment, like v in French pouvoir, hence English power (from Old French t deletion of Latin potere): what we dubbed the power-effect. In experiment 1, we showed that the retraction movement in the off-gliding phase of the [y] vowel is misleading for the anticipation of the following [i] vowel, since [i] identification always comes after the minimal [˙] constriction event. Moreover the audio identification boundary is systematically ahead of the visual identification. In experiment 2, we tested if the epenthetic glide could give birth to a consonant, using an audiovisual McGurk paradigm. We evidenced that this glide needs to be sufficiently lengthened (i.e. maintained in a static phase) in order to be integrated with the sound and represented as a true consonant. These results are discussed in reference to the time-varying vs. stationary representational sound status, from recent computational modelling and neural data, especially the snapshot neuronal computation which fits with recent brain imaging data of stilled and moving speaking mouths.

Not available

Back to Abstract