Spectral weights for localization and speech-in-speech recognition with talkers separated on the horizontal plane

Spatial hearing conveys information about source location and supports speech recognition in the context of spatially separated background talkers. However, some data indicate a dissociation between localization and spatial release from masking, suggesting that these abilities could be mediated by different auditory mechanisms. The present experiments evaluated spectral weights for these two tasks by filtering stimuli into 1-octave-wide bands and dispersing those bands on the horizontal plane. Target stimuli were 100-ms bursts of speech-shaped noise or single-syllable words, and maskers (when present) were sequences of words. Young normal-hearing participants were asked to judge target location or report back the target word; for speech-in-speech recognition, participants had to rely on spatial cues to perceptually segregate the target from the masker. Localization data were similar for noise bursts in quiet, words in quiet, and words embedded in a sequence; in all cases, weights peaked around 500–1000 Hz. Weights for correct word recognition also relied predominantly on spatial cues in the 500- and 1000-Hz bands, but trial-by-trial data suggest that correct recognition was not closely related to differences in perceived location of target and masker speech for all participants.