On the specification of input units in speech perception

Abstract

The first part of this paper considers the experimental evidence concerning a primary recognition unit in speech decoding. Considerations of general human information processing abilities lead to the suggestion that this primary unit must be a fairly long, but clearly identifiable, stretch of speech. Further evidence for the need of a primary recognition unit arises from a consideration of human abilities to identify the order of sounds in a repeated sequence of nonspeech sounds. In spite of the obvious ease with which the order of elements is perceived in speech, listeners have a great deal of difficulty determining the order of sounds in a repeated sequence of nonspeech sounds. Yet there is quite compelling evidence that speech and the perception of order are functions of the same cerebral hemisphere, and, further, that aphasic deficits are accompanied by deficits in the perception of temporal order. The data in the literature suggest that syllables, and phrases defined by suprasegmentals, might function as primary recognition units. In the second part of the paper, the results of an experiment are reported, showing that if a sequence of nonspeech sounds is provided with organization analogous to the organization provided by suprasegmentals in speech then normal subjects' performance on the task of determining the temporal order of the sequence is improved. Aphasic patients, however, appear to be unable to take advantage of such organizing parameters since their performance is not significantly affected by providing organization of the stimulus.