AUDITORY FACTORS IN SPEECH PERCEPTION

这项研究的长期目标是发展我们的基础

The long-term objectives of this research are to develop our basic knowledge of how humans discriminate and recognize speech sounds, and to apply that knowledge toward the explanation of regularities in the sound patterns of languages. The proposed research will investigate a general claim about the phonetic realization of phonological feature distinctions. This claim, referred to as the auditory enhancement hypothesis, states that phonetic correlates of feature distinctions tend to have mutually enhancing effects which help to ensure the perceptual robustness of the distinctions. Phonological features typically have multiple individual phonetic correlates, but these correlates often can be grouped into coherent subsets based on relations of auditory commonality or enhancement. These subsets thus form "integrated perceptual properties" intermediate between the levels of the individual correlates and distinctive features. The auditory enhancement hypothesis is not only important for understanding speech perceptual performance, it also helps to account for various phonetic regularities that have resisted explanation in term of the physics and physiology of speech production. During the current grant period, much of the research has focused on the role of auditory enhancement in the perception of the distinction between [+voice] and [-voice] consonants and the distinction between [+high] and [- high] vowels. Most of the proposed research will evaluate applications of the enhancement hypothesis to a much broader range of feature distinctions. The distinctions to be investigated will include [+/-high], [+/-back], and [+/-coronal] among vowels and [+/-voice], [+/-continuant], [+/-sonorant], and sibilant vs. nonsibilant (where sibilant corresponds to [+strident] and [+coronal]) among consonants. For each feature distinction, specific claims about corresponding integrated perceptual properties, and their auditory basis, will be tested. In the case of vowels, a series of studies will evaluate the possibility that the "spectral center of gravity" effect provides a common auditory account of several different distinctions. In the case of consonants, the auditory hypotheses to be evaluated are generally specific to each feature distinction. Multiple converging operations will be used to test many of the particular hypotheses. These converging operations will include studies of speech production, listener identification and discrimination of speech stimuli and acoustically analogous nonspeech stimuli, and auditory modeling. Clear progress in understanding how speech is represented in normal auditory systems will contribute importantly to the development of models of speech perception in hearing impaired persons.