Abstract
How do we represent where we have been during the course of visual search? Although several studies have shown memory for rejected distractors during oculomotor search, this memory might also take the form of knowledge of the spatial path generated to search a display. If memory is distractor-based (arising from either IOR or inhibitory object tags), we would expect it to be limited to the four or five most recently inspected items. However, if we remember the spatial path transcribed by our search, we might expect memory for distractors viewed even far back in our search sequence. Our general approach was to: (1) briefly present a search display, (2) replace the search items with a placeholder display, then (3) present a probe at either a previously fixated (on-probe) or non-fixated (off-probe) distractor location. A gaze-based n-back methodology was used to present on-probes at distractor locations fixated either early or late in the search path. The observers' task was to rate (on a 7-point scale) their confidence that a target had appeared at the probed location. The target in Exp 1 was an O among Qs; observers in Exp 2 searched Where's Waldo displays. We analyzed confidence ratings as a function of (1) intervening distractors fixated during search (in on-probe trials), (2) the distance between the probe and the nearest fixated distractor (in off-probe trials), and (3) whether an off-probe appeared in the search path (i.e., between two fixations) or off of this path. In addition to the expected high confidence ratings for the most recently fixated locations, we found evidence for an extremely high-capacity distractor memory extending practically the entire length of the search path. We interpret this evidence as suggesting two forms of memory during search. One takes the form of capacity-limited inhibition or tags attached to individual distractors or their locations; another takes the form of high-level knowledge for where we have been during search.
This research was supported by Army Research Office grant DAAD19-03-1-0039 and NIMH grant R01 MH63748.