Segregation of computations underlying perception of motion discontinuity and coherence.

Motion is one of the most important cues for detecting discontinuities in images. The major dichotomy among theories of motion-defined discontinuity concerns whether the computations related to the extraction of discontinuity and large scale integration of motion signals are organized hierarchically or occur simultaneously in the brain. In this study we investigated the hierarchical nature of these computations using data from two patients with unilateral brain lesions on two psychophysical tasks: one requiring motion for spatial integration of direction in a stochastic motion field, and the other requiring motion to extract discontinuities in the same type of stimuli. The results showed a surprising double dissociation of deficits on these motion tasks which suggests that models for discontinuity detection requiring a single neural substrate for computing coherence and discontinuity are unlikely to be applicable to the human visual system. We discuss the computational implications of these results. Using morphometric three-dimensional reconstructions of the lesions from the magnetic resonance imaging data we suggest possible anatomical sites mediating these computations.