A theory of acoustic invariance claims that the universal set of phonetic features used in natural language can be characterized by invariant properties derived directly from the acoustic signal, that the perceptual mechanism is specifically tuned to extract these properties, and that they form the primary perceptual attributes used in ongoing speech perception. It is the goal of the proposed research to explore these issues. To this end, we plan to continue to investigate place of articulation in stop consonants by exploring the role of spectral shape as a perceptual cue to the labial and alveolar place of articulation and by studying the acoustic properties of another place of articulation, dentals in stop consonants. Our methodology will consist of acoustic analysis of natural speech from English and othe languages as well as perceptual studies utilizing both natural and synthetic speech stimuli. In addition, we plan to explore the feature contrast, stop-glide, in relation to a theory of acoustic invariance. Our methodology will consist of attempting to isolate the invariant property after detailed acoustic anslyses of natural speech, providing an operational definition of that property, testing its effectiveness on a new set of natural speech data, and then exploring the perceptual consequences of this property using both synthetic and natural speech stimuli for the identification of the particular property. Our ultimate aim is to characterize the inventory of acoustic properties corresponding to the finite set of speech sounds used in natural language. Such study should help characterize the nature of the speech processing system and its underlying mechanisms, as well as provide an essential base-line for delineating the developmental processes for speech in the young child and the effects of brain-damge on such processing abilities.
