Geometric context influences ambiguous apparent motion

Abstract

How sensitive is the matching process in long-range apparent motion to contextual influences? The M points in N frames computational models make little accommodation for more general scene analysis constraints. Alternatively, if apparent motion perception is to be understood as generating a model of a scene transformation due to changing object/object and object/observer relationships, then other factors might affect one's interpretation. We examined the ability of contours to influence the interpretation of dynamic dot quartet motion. In the first two experiments, we measured the frequency of seeing horizontal vs. vertical quartet motion in the presence of flanking real or Kanizsa rectangles arranged outside the sides or top/bottom of the quartet. Most observers exhibited a significant change in quartet interpretation with the motion more frequently appearing to be along the contours. The biasing effect was greater for real contours than for cognitive contours. In some control conditions, the Pacman disks were oriented so that no virtual figure was seen, and no biasing effect was observed. In these experiments the observed along-the-contour bias was very sensitive to proximity. In contrast, a central “blocking” rectangle is also effective in biasing quartet motion, but in a way that is not proximity-dependent. A third experiment measured the influence of a central virtual blocker crossing either horizontal or vertical quartet paths. Virtual blockers were effective in all experienced observers, but only in some naive observers. Therefore the effect may depend on the observer being actively aware of the virtual contours and objects.