Amodal completion does not require attention

Abstract

When one object occludes another, occluded objects appear to continue invisibly behind the occluder. Does this “amodal completion” require attention? Hulleman (VSS07) showed that efficient search for vertical bars among horizontal bars was disrupted by occluding diagonal bars. This suggests that the amodal completion that created oriented occluded items did not occur prior to search. However, Hulleman's oriented occluders might have disrupted orientation search. Here, four experiments manipulated orientation information in the occluder. Participants searched for a vertical bar among horizontal bars or vice versa. Bars were either unoccluded, occluded or “gapped” (i.e., occluded by an object of the same color as the background). In all experiments, search for unoccluded objects was efficient (0 – 2 ms/item), while search for gapped objects was inefficient (12 – 49 ms/item; except for one condition). In the first two experiments occluders were 45° tilted bars as in Hulleman's experiment. In Experiment 1, each bar had its own occluder, whereas in Experiment 2 the occluder covered three bars at once. Both experiments replicated Hulleman's finding that search for occluded bars was inefficient ([[gt]] 19 ms/item). In Experiment 3 we used an unoriented square occluder. Here we found efficient search for occluded bars (0 – 3 ms/item). However, this may have been due either to the oriented convex hull of each item or to simple connectedness, rather than amodal completion. Experiment 4 controlled for this by adding a condition where the “gapped” objects were connected and had the same convex hull as the occluded object. Search for occluded objects was again efficient (1 ms/item), but search for the connected control objects was not (29 ms/item). These experiments suggest that the orientation signals created by amodal completion are weak, easily masked by orientation noise in the display. However, amodal completion appears to be a preattentive process.