CNL Lunch

Maria Chait

Binding Mechanisms in Speech Processing

Thursday April 3rd, 12:30pm, 3416 Marie Mount Hall

The basic units into which the acoustic speech signal is perceptually segmented is an issue of central importance for speech research. A growing body of research points to the perceptual reality of both the phonetic segment and the syllable during the course of speech processing. These findings have led to two kinds of popular models of speech segmentation: In one, the speech stream is initially segmented into phonemic segments which are later combined to create supra-segmental units. The second model assumes that the syllable is the basic unit of speech perception with the phone serving as a secondary unit of analysis.
Each model accounts for a fraction of the, seemingly contradictory, experimental results in the field . At the same time it is clear that neither model can accommodate the full spectrum of experimental data.
Here we propose a third form of model - one that attempts to reconcile these findings and suggest a new method of systematically examining the extraction and subsequent combination of the informational constituents of the speech signal.
The current study demonstrates that intelligibility crucially depends on both the slowly varying as well as the rapidly varying components of speech and suggests a binding process, in which a conjunction of these creates an emergent representation that forms the basis for successful speech processing.