Binaural release from informational masking: Results from a speech identification task

Abstract

This study examined how binaural cues can reduce informational masking (IM) in a speech identification task. Target and masker sentences were processed into non-overlapping frequency bands, thus limiting ‘‘energetic masking,’’ and were presented over headphones. Listeners identified key words in the target sentence. In a baseline condition, target and masker were presented monotically (SmMm), producing large amounts of IM. Binaural release from IM (i.e., improved performance re. SmMm) was observed when the target was presented monotically and the masker diotically (SmM0), suggesting that the shift of masker image away from target led to the reduction in IM. Creating large interaural differences in level (ILDs) showed that for a monaural target, binaural release, though related to ILD, occurred even when the masker image should have been lateralized at the target ear. For differences in time (ITDs) up to 600 μsec, the amount of binaural release was completely independent of ITD. For a diotic target and binaural masker, however, release only occurred for large ITDs or ILDs. These results suggest that for monaural targets and binaural maskers, IM in a speech task can be reduced through a binaural cue that is present across the entire range of biologically plausible ITDs and ILDs.