A Data-Limit Account of Release From Masking During Speech-on-Speech Listening

Speech-on-speech listening involves selectively attending to a target talker while ignoring a simultaneous competing talker. Spatially separating the talkers improves performance, a phenomenon known as spatial release from masking (spatial RM). The same is true of spectral separation, that is, filtering the talkers into non-overlapping frequency bands (spectral RM). The relative benefit of spatial versus spectral RM is currently unknown. Furthermore, it is unclear how listeners’ ability to exploit spatial versus spectral cues is related to individual differences in cognition. The resource-limit account suggests that cognitive resources are required to support the processing of degraded speech, implying the strongest cognition/performance relationship when RM is limited or absent. However, an alternative claim, referred to as the data-limit account, suggests that cognitive resources cease to be useful when the target is severely degraded. In this study, participants (N = 240) completed a selective listening task in which they transcribed the speech of one of two simultaneously presented talkers. The speech was filtered into interleaved or overlapping frequency bands (spectral RM vs. no spectral RM) and presented dichotically or collocated (a proxy for spatial RM vs. no spatial RM). A battery of cognitive tasks was administered to assess working memory/attention. Spectral RM provided at least as much benefit as spatial RM, with the best performance when both RM types were present. Cognitive scores were significantly positively correlated with RM benefits. However, the weakest correlation between cognitive scores and performance was observed in the no-RM condition. The results therefore support an account of speech-on-speech listening that lies on a continuum from data-limited to resource-limited processing as a function of the quality of the target speech signal.