Speech intelligibility in reverberation is reduced during self-rotation

Spatial unmasking helps speech intelligibility in a cocktail party but its effects have been studied mainly for stationary participants. Here, we investigate behavior and speech intelligibility during active self-rotation of standing participants and we assess the impact of motion and visually presented location cues. We employed a spatialized speech test in a controlled reverberant space with target sentences randomly appearing on each trial at one of four possible locations (0°, ±90°, 180°), while speech-shaped noise was presented from 0° respective to the participant’s orientation. Participants responded naturally with motion as in a social situation. Target sentences were presented either without (A-only) or with a picture of an avatar (AV). In a baseline (Static) condition, people were standing still without visual location cues. Participants undershot the targets, often in the acoustically optimal way, but they also oriented away from the frontal target where there was no acoustic benefit. They performed equally in AV and A-only. They performed better in the A-only than in the Static condition for the rear target, but worse for the lateral target. While the first can be partly explained by spatial unmasking, the latter cannot. The speech intelligibility model by Jelfs et al. (2011), extended to consider self-rotation, overestimated participant performance during motion. The experimental and modeling results suggest that listeners have a limited access to the spatial unmasking cues during self-rotation. The results are discussed in context of binaural sluggishness and non-acoustic factors.