Generalization of repetition-based segregation with novel sounds and nonsense speech.

Repetition helps listeners identify an unfamiliar target sound in the presence of different competing sounds and later recognize it in quiet. To further characterize this perceptual learning process, we tested whether recognition remains robust when the target is paired with a novel distractor sound. Exposure to the repeating target with either a single repeating distractor or in quiet offered limited benefit for detecting the target with a new distractor. In contrast, exposure with varying distractors led to strong recognition, suggesting that listeners adapt their expectations and learn to isolate the target only when distractor variability is present.