ABSTRACT
Abstract
Speech perception has evolved on a perceptual substrate that allows us to recognize complex sounds in a noisy and frequently reverberant environment. This substrate involves mechanisms that allow different sound sources to be separated and that allow sounds that are masked or spectrally changed to be heard for what they originally were. It is unlikely that speech sounds do not invoke them. Such mechanisms we can refer to as "auditory" in that they apply indiscriminately to speech and other sounds. They do not necessarily lead to a conscious percept but produce a representation of sound that can usefully make contact with knowledge about particular sound types and categories, be they speech or not. The speech module can apply its speech-specific knowledge to such a representation, as can processes or modules that lead to a conscious percept of sounds other than speech.
I will describe experiments on the perception of simple speech sounds in the presence of other sounds that demonstrate some of the auditory (in the previous sense) processes that might be used to construct such an abstract representation of sound. Properties such as local frequency continuity and onset-time constrain the alternatives that speech-specific knowledge considers.
