Spatial uncertainty reduces the benefit of spatial separation in selective and divided listening

Abstract

Spatial separation between two talkers can improve speech intelligibility (i.e., cause spatial unmasking) when attending to either one talker (selective listening) or both talkers (divided listening). Such spatial unmasking occurs by (1) allowing listeners to hear otherwise inaudible portions of the signal(s) (energetic masking) and (2) reducing central competition between similar talkers (informational masking, IM). Here, we examine how a priori knowledge affects spatial unmasking in selective and divided listening in IM-dominated stimulus conditions. Performance was measured in blocks where the locations of the talkers were either constant (fixed) or varied from trial to trial (spatially randomized). Spatial separation improves performance in both selective and divided tasks when locations are fixed. In the fixed conditions, the acoustic signal-to-noise ratio at the better ear for the less-intense talker partially accounts for the spatial unmasking; however, there is also a large residual spatial unmasking, similar to that seen in previous studies. In the spatially randomized conditions, spatial unmasking is much reduced. In addition, in the divided listening, randomized conditions there is little evidence for any better-acoustic-ear effect. Results suggest that the way in which subjects utilize spatial cues depends on their knowledge of the source locations. [Work supported by ONR Grant No. N00014-04-1-0131.]