Abstract:
Previous studies of speech perception by cochlear-implant users have employed stimuli in which both waveform envelope and spectral (frequency-specific) information is present. It is unclear as to how much spectral (i.e., frequency-sensitive) information cochlear implant users may utilize in their perception of speech stimuli. In this study, the identification and discrimination of CV tokens are contrasted by normally hearing and cochlear-implant subjects in two conditions: CV tokens (digitized and normalized) and their speech-correlated-noise counterparts (SCN). In the SCN stimuli, frequency-sensitive information has been eliminated while waveform envelope information has been preserved. Preliminary analysis of the identification data indicates that normal-hearing subjects were able to identify unprocessed (UP) stimuli at a level of nearly 100% correct (96.1%), while the identification of cochlear-implant subjects was lower, but above chance (30.6%). However, absolute identification of consonants was better for cochlear-implant subjects than normal-hearing subjects in the speech-correlated-noise condition. Overall, the difference between the two groups in making manner, voicing, and place-of-articulation decisions indicated that these cochlear-implant users may have primarily utilized nonspectral information in their identifications and discriminations, although they are able to utilize some spectral information. [Work supported by the Deafness Research Foundation.]