Abstract
Tracking of multiple moving targets has been found to be largely unaffected by the presence of narrow occluders, implying that the object tracking system is immune to brief occlusion events (e.g, Scholl & Pylyshyn, 1998). Object tracking across occlusion was robust when occlusion occurred as an accretion/deletion along a fixed contour relative to the object's path of motion. Movements were random except for the stipulation that object paths would never cross. The objects therefore influenced each other's direction of motion.
The purpose of the present study was to examine whether the tracking system treats all occlusion events the same, with two new conditions. First, we used a shadow occluder whose edges were semi-transparent gradients instead of fixed contours. The total amount of occlusion area remained the same, but time to become occluded was prolonged, relative to the accretion/deletion event. We reasoned that the tracking system might have difficulty following objects whose occlusion time was extended. Second, motion paths were independent, and objects occasionally occluded each other. We reasoned that extra occlusion events might pose additional challenges to the tracking system, resulting in a decrement of performance.
When no occluders were present, subjects accurately identified 89% of targets, replicating the original Scholl and Pylyshyn results. Performance in the shadow event was similar to our other occlusion events (luminance and virtual borders), implying that the tracking system is robust to shadow occlusions as well as accretion/deletion occlusions. However, performance was far worse (between 78 and 81%) than no-occluder events, and worse than results reported by Scholl and Pylyshyn. Tracking through occlusion is thus more difficult when objects can also occlude each other, but only when static occluders are also present. The tracking system may be unable to individuate objects effectively when occlusion events are made more ecologically valid.
Supported by NSF grant BCS-0094814