Neural dynamics of speech perception: Phonemic restoration in noise using subsequent context.
Sohrob Kazerounian, Stephen Grossberg · The Journal of the Acoustical Society of America · 2009
Phonemic restoration describes a class of auditory illusions during which broadband noise replacing a deleted phoneme can cause the perceptual restoration of the deleted phoneme. Phonemic restoration exemplifies the brain’s ability to complete and understand speech and language in noisy environments. It also clarifies how both past and future acoustical events can contextually guide completion of a percept that is occluded by noise, and highlights that conscious speech is due to bottom-up and top-down contextual interactions that can operate across hundreds of milliseconds. This work develops a neural model that quantitatively simulates restoration phenomena, including the forward development in time of a conscious speech percept even in cases where future events control how previously presented acoustic events are heard. The model clarifies how acoustic items are stored in short-term working memory, and how they interact reciprocally with unitized representations of item sequences, or list chunks, to generate a resonant wave of activation that embodies the consciously heard percept. Model simulations clarify why the presence of noise is necessary for restoration to occur, and why in the absence of noise a silence gap is perceived. These properties are traced to the brain's ability to rapidly and stably learn language.