Change Perception in Complex Auditory Scenes

Article excerpt

In four experiments, we examined the role of auditory transients and auditory short-term memory in perceiving changes in a complex auditory scene comprising multiple auditory objects. Participants were presented pairs of complex auditory scenes that were composed of a maximum of four animal calls delivered in free field; participants were instructed to decide whether the two scenes were the same or different (Experiments 1, 2, and 4). Changes to the second scene consisted of either the addition or the deletion of one animal call. Contrary to intuitive predictions based on results from the visual change blindness literature, substantial deafness to the change emerged without regard to whether the scenes were separated by 500 msec of masking white noise or by 500 msec of silence (Experiment 1). In fact, change deafness was not even modulated by having the two scenes presented contiguously (i.e., 0-msec interval) or separated by 500 msec of silence (Experiments 2 and 4). This result suggests that change-related auditory transients played little or no role in change detection in complex auditory scenes. Instead, the main determinant of auditory change perception (and auditory change deafness) appears to have been the capacity of auditory short-term memory (Experiments 3 and 4). Taken together, these findings indicate that the intuitive parallels between visual and auditory change perception should be reconsidered.

Change perception in the visual modality-and particularly its counterpart, change blindness-have received much interest in the past decade (Rensink, 2002; Simons & Levin, 1997). In particular, change blindness studies have shown that, contrary to our impression of seeing everything around us, human observers often fail to see even large changes in the scene if attention is not directed to the object or location of change (O'Regan, Rensink, & Clark, 1999; Rensink, O'Regan, & Clark, 1997). Change blindness does not directly address the conscious perception of objects per se, but addresses only the awareness of change (Rensink, 2000); yet it supports the view that attention might be crucial for the contents of our conscious visual perception, or, put differently, that we consciously see what we attend to (Mack & Rock, 1998). In the present study, we examined an auditory analogue of change blindness-namely, change deafness-in the context of complex auditory scenes.

In the visual modality, the mechanisms allowing correct change perception and those leading to change blindness are relatively well understood (O'Regan, Deubel, Clark, & Rensink, 2000). Any sudden change in the environment that can be registered by the visual system gives rise to a local visual transient. When such a local transient is detected, attention is automatically attracted to the change location, and the novel aspect of the visual scene is reported (Rensink et al., 1997; Turatto & Bridgeman, 2005). However, when the local visual transient is masked or made ineffective in capturing visual attention, the change location remains unattended, and change blindness occurs. Although different visual events can mask the change-related transient (Rensink, 2002), this has typically been achieved by interposing a blank screen between two consecutive images of the same scene differing in only one particular (Rensink et al., 1997). The appearance of the second scene after the blank creates a global visual transient (i.e., all locations in the second scene change with respect to the previous blank image), so that detection of the local visual transient that is associated with the changed object is no longer available to summon attention. Under these conditions, change blindness is often observed, and the change can be detected only through a time-consuming serial scanning of the scenes. According to this account, selective attention is the mechanism that permits information from the two scenes to be stored in visual short-term memory (vSTM; see Luck & Vogel, 1997; Phillips, 1974) for comparisons across views that would then allow conscious change detection. …