Toward an objective measure for a "stream segregation" task.

A procedure to estimate the relative contribution of "A" and "B" tones for a stream-segregation task is described. Listeners detected a delay in the penultimate A tone in an A-B-A-B sequence of tones. For small A-B frequency separations, for most listeners, classification models based on both the A and B tones were superior to models based on just the A tones. For large frequency separations, models based on just the A tones were superior, indicating the A and B tones were segregated. The results also revealed individual differences in the strategies adopted to complete the task.