Abstract

Accurately timing acoustic events in dynamic scenes is fundamental to scene analysis. To detect events in busy scenes, listeners must often identify a change in the pattern of ongoing fluctuation, resulting in many ubiquitous events being detected later than when they occurred. This raises the question of how delayed detection time affects the manner in which such events are perceived relative to other events in the environment. To model these situations, we use sequences of tone-pips with a time-frequency pattern that changes from regular to random ('REG-RAND') or vice versa ('RAND-REG'). REG-RAND transitions are detected rapidly, but the emergence of regularity cannot be established immediately, and thus RAND-REG transitions take significantly longer to detect. Using a temporal order judgment task, and a light-flash as a temporal marker, we demonstrate that listeners do not perceive the onset of RAND-REG transitions at the point of detection (∼530 ms post transition), but automatically re-adjust their estimate ∼300 ms closer to the nominal transition. These results demonstrate that the auditory system possesses mechanisms that survey the proximal history of an ongoing stimulus and automatically adjust perception to compensate for prolonged detection time, allowing listeners to build meaningful representations of the environment.