Abstract
Previous work suggested that objects in search displays are preattentively completed behind occluders, but these studies were confounded in that occluded targets were identical across entire blocks—amodal completion may have occurred for the target representation used to guide search rather than preattentively for objects in the search display. To explore this possibility, we created four-item search arrays consisting of real-world objects, and systematically varied whether the search target was occluded or not at preview and in the search displays. We found that target verification times (the time between fixating the target and making a present/absent button press) were longest when the preview was unoccluded but the target in the search display was occluded (p <.001). This is consistent with a verification process that checks for specific features of the target, half of which may be hidden by an occluder. Guidance, measured as the percentage of initial saccades directed to the target, was strong when the target was occluded at preview but not in the search display (59%), which did not differ from a baseline condition having unoccluded targets during preview and search (63%; p= .87)—occluding a target at preview did not significantly weaken guidance. However, guidance was weaker whenever the target was occluded in the search display, regardless of whether it was occluded during preview (40%) or not (37%). If amodal completion occurs preattentively, we should not have found this ~20% guidance difference (p < .001) between occluded and unoccluded targets in the search display. In two control experiments we determined that these effects were not specific to our occluders or stimuli. Contrary to previous reports, these findings suggest that participants complete the target during preview, but that amodal completion does not occur preattentively during search, at least for the photorealistic objects used in this study.
Meeting abstract presented at VSS 2012