Abstract
It is widely believed that motion processing can be split into two stages, a local directionally selective stage followed by a global velocity sensitive stage. Here we investigate the phenomenon of motion repulsion and ask at which of these stages in the motion processing hierarchy repulsion occurs. Our stimulus consisted of two translating, superimposed planes of Laplacian of Gaussian dots whose directions of motion differed by 60 degs. Using an adaptive method of constants procedure, we measured the strength of the direction repulsion of a target (speed 2.5 deg/sec) as a function of distractor speed (0.625–15.0 deg/sec) and found an inverted-U function peaking at about 5 deg/sec. To distinguish between local and global theories of motion repulsion we investigated the target repulsions induced by distractors containing pairs of velocities (2.5 & 10 deg/sec, 1.25 & 12.5 deg/sec, 0.625 & 15 deg/sec). Each constituent dot was assigned one of the two speeds (with equal probability) for the duration of the stimulus. We define a local repulsion model as one in which the repulsion of the target is a weighted sum of local repulsion measures. Our inverted-U tuning function gives the local repulsion for each distractor speed. The model cannot produce a magnitude of motion repulsion that is greater than that given by the more efficacious of the distractor velocities. For both subjects, the magnitude of repulsion was consistently (and significantly) greater than the maximum repulsion predicted by the local model. In fact, the magnitude of repulsion would be better predicted by the mean of the distractor set (our global prediction). This pattern of results occurs in spite of the fact that the distractors themselves appear to contain two transparent motions. Our findings argue strongly against the local model of direction repulsion. We propose that motion repulsion occurs after global motion extraction.