Effects of high-frequency cue reduction on the comprehension of distorted speech

Forty-five subjects with sharply sloping sensorineural high-frequency hearing losses were examined for comprehension with a tape containing sentences that had been time-compressed (250 words/min), interrupted, (50 msec on-50 msec off), and masked with speech-spectrum noise (+2 dB S/N) in that order. All subjects yielded normal speech reception thresholds, and generally normal scores on the Northwestern University Auditory Test No. 6. Distorted-speech testing was completed at 40 dB SL. Subjects with losses at 2 kHz and above were able to comprehend only 50, 65, and 68% of compressed, interrupted, and noise-masked sentences, respectively. In contrast, subjects with losses at 3 kHz and above performed poorer than normal controls by 11.3, 12.5, and 8% points respectively, while subjects within normal hearing sensitivity at 3 kHz performed as well as controls (maximum drop of 4.6 points with noise-masking). The multiplicative hypothesis was upheld in that the performance of subjects with severe high-frequency deficiencies was much poorer than one would predict on the basis of what is known about performance on these tests with either filtering alone or other distortions alone. These subjects, most especially those with severe frequency deficiencies, outperformed by up to 29 percentage points the mean scores of groups of normals given the same test items and distortion conditions, but listening through filtering that simulated the hearing loss of subjects in this study. It was suggested that hypacusics with high-frequency hearing losses can learn to use residual cues efficiently and that it is not altogether permissible to model sensorineural high-frequency losses with normal-hearing subjects using frequency filtering.