Spatial summation properties of directionally selective mechanisms in human vision

Our goal in this paper was to measure psychophysically the receptive-field size of motion units in human vision. To this aim, length and width spatial summation functions were measured for drifting (8-Hz) sinusoidal gratings of spatial frequencies 0.1, 1.0, and 10.0 cycles per degree (c/deg) with two threshold criteria: direction discrimination and simple detection. For each spatial frequency, contrast sensitivity for detection of the direction of drift increased with increasing stimulus size (length or width), at first rapidly (slope greater than or equal to 1.0) and then more gradually (slope 0.29). For most stimuli, the detection and direction-discrimination contrast thresholds were nearly the same. However, for stimuli severely curtailed in width, significantly more contrast was required for direction discrimination than for detection. These results were predicted with a summation model, which incorporated three-dimensional (space-space-time) linear input filters, and probability summation over space and among different filter types. The fit of the model gave an estimate of both the receptive-field length and width of motion-detector units in human vision. At each spatial frequency, the estimates of receptive-field width and length were similar, indicating that the receptive fields of motion-detector units are as long as they are wide at all spatial scales. Receptive-field size varied from approximately 0.12 cycle at 0.1 c/deg to 0.52 cycle at 10.0 c/deg.