Abstract
Objects that shrink and expand at occluding surface boundaries in multiple object tracking are more difficult to follow than objects that delete and accrete (Scholl & Pylyshyn, 1999). Here we ask whether this difficulty is due to the lack of a top-down naturalistic explanation for the shrinking or whether it is due to the bottom-up optical transformation of shrinking itself. If a naturalistic visual context for the shrinking disks was available, would observers exhibit higher tracking capacity? Observers tracked four of eight disks that could pass behind an occluding surface in the middle of the display. The disks moved for 10 seconds and then observers attempted to identify the four disks they were tracking by clicking on them. In the occluded condition, disks deleted/accreted at the occlusion boundary, while in the shrinking condition, disks shrank/expanded at the boundary. In the critical falling condition, disks also shrank/expanded at the occlusion boundary but a background image with a steep slope made the disks appear to fall into and then emerge from a ravine below the occluding surface. The low level optical transformations in the shrinking and falling conditions were identical " only the background image varied. Observers had the highest tracking capacity in the occluded condition, followed by the shrinking condition. Contrary to the naturalistic optical transformation hypothesis, observers exhibited the lowest tracking capacity in the falling condition. We interpret this result to indicate that bottom-up optical transformations per se caused the decrease in tracking capacity. If an object shrinks at an occlusion boundary, the visual system seems to stop tracking it, even if there is a naturalistic visual explanation for the shrinking.