Abstract
The visual system is often faced with small discrepancies between consecutive views of a scene (introduced, e.g., by motor inaccuracies in eye movements), yet it typically does not attribute such events to changes in the world. In this study, we introduced a series of small discrepancies to a scene image by rotating it in 1° intervals in a flicker paradigm. Two stimuli were used: a living room scene that rotated in depth around the viewing position, and a top-down view of a desk that rotated in the picture plane. First, we considered whether, through the incremental addition of small rotations, participants would come to consider a significantly different view as an unchanged continuation of the original view. This was indeed the case. Participants required an average of 31° and 50° of rotation before detecting the change to the living room and desk scenes, respectively. Control experiments, in which participants reliably detected much smaller total rotations made in a single step, eliminated the possibility that this insensitivity derived from failure to represent the scenes in memory or to compare consecutive views. Second, we considered the relationship between explicit change detection and visual memory. Participants' extended insensitivity to the 1° rotations suggests that, despite the absence of explicit change detection, visual memory was incrementally updated with each change to reflect the most recently available scene information. Thus, comparison processes from one view to the next operated over representations of consecutive views, even though the current and immediately prior views were often significantly different from the original view. These results demonstrate an important dissociation between explicit change detection and visual memory. Following a change, visual memory is updated to reflect current perceptual information, even if the change was not detected.
Supported by NSF grants SBR 9617274 and ECS 9873531.