Limiting spectral resolution in speech for listeners with sensorineural hearing loss

Consonant recognition was measured as a Function of the degree of spectral resolution of the speech stimulus in normally hearing listeners and listene rs with moderate sensorineural hearing loss. Previous work (Turner, Souza, and Forget, 1995) has shown that listeners with sensorineural hearing loss could recognize consonants as well as listeners with normal hearing when sp eech was processed to have only one channel of spectral resolution. The hyp othesis tested in the present experiment was that when speech was limited t o a small number of spectral channels, both normally hearing and hearing-im paired listeners would continue to perform similarly. As the stimuli were p resented with finer degrees of spectral resolution, and the poorer-than-nor mal spectral resolving abilities of the hearing-impaired listeners became a limiting factor, one would predict that the Performance of the hearing-imp aired listeners would then become poorer than the normally hearing listener s. Previous research on the Frequency-resolution abilities of listeners wit h mild-to-moderate hearing loss suggests that these listeners have critical bandwidths three to Four times larger than do listeners with normal hearin g. in the present experiment, speech stimuli were processed to have 1, 2, 4 , or 8 channels of spectral information. Results for the 1-channel speech c ondition were consistent with the previous study in that both groups of lis teners performed similarly. However, the hearing-impaired listeners perform ed more poorly than the normally hearing listeners for all other conditions , including the a-channel speech condition. These results would appear to c ontradict the original hypothesis, in that listeners with moderate sensorin eural hearing loss would be expected to have at least 2 channels of frequen cy resolution. One possibility is that the Frequency resolution of hearing- impaired listeners may be much poorer than previously estimated; however, a subsequent filtered speech experiment did not support this explanation. Th e present results do indicate that although listeners with hearing loss are able to use the temporal-envelope information of a single channel in a nor mal fashion, when given the opportunity to combine information across more than one channel, they show deficient performance.