Surface perception and motion integration

Purpose. We investigated the rules by which the visual system générales percepts of global object motion from locally ambiguous image-motion signals. Method. We constructed a stimulus similar to Anstis' "chopsticks illusion." A vertical bar and a horizontal bar each oscillated sinusoidally, 90 degrees out of phase, so that the cross-point executed a circle. When the bars cohered and appeared to move circularly as a cross, subjects could correctly judge whether the motion was clockwise or counterclockwise; without coherence performance was poor. We then manipulated the surfaces surrounding the bars, varying the endpoinls1 tendency to be seen as intrinsic or extrinsic. Results. In accord with the prior literature, coherence improved when the addition of a frame around the bars converted the previously intrinsic bar endpoints to extrinsic T-junctions (the motions of which were presumably attributed to occlusion and ignored - left figure). We then constructed an illusory occluding frame and found that it also improved coherence, even though the local structure of the endpoints was unchanged (middle figure). More interesting, coherence also occurred when the occlusion relationships between the frame and the bars were reversed such that the cross was reported to occlude the frame (right figure). We also used frame motion and transparency to vary whether frames were seen as occluded or occluding. Our results suggest that coherence occurs whenever contours are locally aligned with the bar endpoints. This is true even when the global depth ordering is inconsistent with an interpretation of the endooint motion as extrinsic. Conclusion. The motion intégration system tanned bv this stimulus appears to use a simple contourbased heuristic to segment motion signals, even when the assignment is inconsistent with perceived borderownership.