The problem of how the visual system matches corresponding inputs from one instant to the next to produce the perception of motion has been experimentally examined. The specific concern was whether this correspondence problem is solved prior to the interpretation of three-dimensional distance. Observers judged the degree of apparent motion between pairs of lights in a conflicting motion display. Spatial separation of the lights was varied in two and three dimensions in order to assess whether retinal distance, actual depth, or some combination of these provided critical information for correspondence. The results support Ullman's contention that only two-dimensional (retinal) distances are used in establishing correspondence in motion perception.