Influences of formant bandwidth and auditory frequency selectivity on identification of place of articulation in stop consonants

Abstract Can supra-normal spectral contrast in an acoustical speech signal compensate for the degraded frequency selectivity accompanying hearing losses of cochlear origin? To answer this question, two sets of stimuli were synthesized: whispered exemplars of “bet”, “debt”, “get” (the initial set) and “bib”, “bid”, “big” (the final set). Spectral contrast was increased by narrowing the bandwidths of the first five formants; contrast was reduced by broadening their bandwidths. Subjects were adults either with audiometrically-normal hearing or with moderate hearing impairments of cochlear origin. For both groups of listeners, accuracy of identification decreased as formant bandwidth increased. In the mean data of both groups for syllable-final identification, there was a tendency for accuracy to increase when format bandwidths were set to half their nominally normal values. Psychoacoustical measures of frequency selectivity correlated robustly with accuracy of identification, but again only for final consonants. These results suggest that factors in addition to reduced frequency resolution and reduced absolute sensitivity set limits on the accuracy of speech identification in cases of cochlear hearing loss, particularly for syllable-initial consonants. Possible candidates, not explored here, are reduced temporal resolution and an increased susceptibility to backward masking. The very limited benefits for speech identification which resulted from reducing formant-bandwidths may reflect the inability of that transformation to compensate for impairments to temporal auditory processing, as well as for extreme impairments to frequency resolution.