Broad binaural fusion impairs segregation of speech based on voice pitch differences in a ‘cocktail party’ environment

Abstract

ABSTRACTIn the normal auditory system, central auditory neurons are sharply tuned to the same frequency ranges for each ear. This precise tuning is mirrored behaviorally as the binaural fusion of tones evoking similar pitches across ears. In contrast, hearing-impaired listeners exhibit abnormally broad tuning of binaural pitch fusion, fusing sounds with pitches differing by up to 3-4 octaves across ears into a single object. Here we present evidence that such broad fusion may similarly impair the segregation and recognition of speech based on voice pitch differences in a ‘cocktail party’ environment. Speech recognition performance in a multi-talker environment was measured in four groups of adult subjects: normal-hearing (NH) listeners and hearing-impaired listeners with bilateral hearing aids (HAs), bimodal cochlear implant (CI) worn with a contralateral HA, or bilateral CIs. Performance was measured as the threshold target-to-masker ratio needed to understand a target talker in the presence of masker talkers either co-located or symmetrically spatially separated from the target. Binaural pitch fusion was also measured. Voice pitch differences between target and masker talkers improved speech recognition performance for the NH, bilateral HA, and bimodal CI groups, but not the bilateral CI group. Spatial separation only improved performance for the NH group, indicating an inability of the hearing-impaired groups to benefit from spatial release from masking. A moderate to strong negative correlation was observed between the benefit from voice pitch differences and the breadth of binaural pitch fusion in all groups except the bilateral CI group in the co-located spatial condition. Hence, tuning of binaural pitch fusion predicts the ability to segregate voices based on pitch when acoustic cues are available. The findings suggest that obligatory binaural fusion, with a concomitant loss of information from individual streams, may occur at a level of processing before auditory object formation and segregation.