Perceptual space and learning in speech perception.

Abstract

When listeners are given training on synthetic speech, intelligibility improves and listeners become more efficient in using cognitive resources for the perception of speech [L. Lee and H. C. Nusbaum, J. Acoust. Soc. Am. Suppl. 1 85, S125 (1989)]. The present study investigated one possible account of the connection between changes in intelligibility and changes in efficiency in using cognitive capacity: Learning may shift cognitive resources away from the analysis of acoustic properties of synthetic speech that are not informative about phonetic structure to more informative acoustic properties. If learning changes the distribution of attention to the acoustic signal, a change should occur in the structure of the perceptual space used in recognizing synthetic speech. Multidimensional scaling was performed on consonant confusions before and after training in order to determine if and how the structure of perceptual space is affected by learning. The results are consistent with the hypothesis that perceptual learning of synthetic speech changes the way listeners focus attention on the acoustic properties of the speech signal. [Work supported by NIDCD.]