Bottom-up and top-down influences on spatial unmasking

The ability to detect and understand a source of interest (a "target") in the presence of a competing source (a "masker") is better when the sources are spatially separated than when they are at the same location, an effect known as "spatial unmasking". Many models account for spatial unmasking by predicting reduction of within-frequency-band masking; however, recent studies report significant spatial unmasking even when withinband "energetic masking" is minimal. The current study examines whether spatial unmasking depends on the veracity and kinds of spatial cues present when the target and masker are similar and processed to have little spectral overlap. For the tested stimuli, traditional within-band models predict (at most) a modest amount of spatial unmasking that varies with condition. Instead, we observe large spatial unmasking effects. Moreover, after accounting for the broadband target and masker intensities at the acoustically better ear, the amount of spatial unmasking is essentially independent of the kind of spatial cues that cause the target and masker the be perceived at different locations. Only at the lowest target-to-masker ratios (when within-band masking becomes significant) does the amount of spatial unmasking depend on the interaural phase differences in target and masker. These results emphasize that the relative overall intensities of the masker and target are critical for predicting how much perceptual interference the masker causes, even when "energetic masking" is minimal. We believe that in everyday settings, both traditional bottom-up factors and a higher- level mechanism depending on spatial perception contribute to spatial unmasking. We argue that the latter mechanism is a form of spatial attention, critical for meditating competition between similar, simultaneous sources in everyday settings.