Perception of prosodic phrasing by hard-of-hearing listeners

Abstract

Severely to profoundly hard-of-hearing (HOH) listeners (n=12) were tested for their perception of prosodic phrase continuity in clear and conversational speech. In two experiments, fluently spoken sentences were compared with utterances artificially assembled from isolated words (experiment 1) and prosodic phrases (experiment 2) that replicated the sentences’ lexical composition. The difference in the overall duration of the assembled and fluent utterance was eliminated via signal processing, but the amplitude envelope and the intonational contour differences were left intact. Randomized repetitions of each utterance pair (72 pairs, 12 sentences × 2 speaking styles × 3 speakers, in experiment 1, and 40 pairs, 10 sentences × 2 speaking styles × 2 speakers, in experiment 2) were presented to each listener over several experimental sessions. In a 2IFC procedure, the listener identified the utterance that sounded more like fluent speech. Four normal-hearing (NH) controls were tested in white noise (S/N ratio=−20 dB). NH listeners identified the fluent sentences with accuracy greater than 90%. Across HOH listeners, identification accuracy varied from completely random to less than 75% correct, and was dependent on a listener’s sensation level and preferred communication mode (oral versus manual). [Work supported by NIDCD.]