Signal Pooling across ON and OFF Motion Detectors

In motion studies with random-dot patterns or random-pixel-arrays (RPAs) and a signal-to-noise ratio (SNR) paradigm, detection thresholds have been found to depend upon size, form, and duration of the stimuli. This suggests a pooling of information from motion detectors in local space-time neighbourhood regions. Spatial pooling changed the threshold in inverse proportion to the square root of the number of pooled detectors, while temporal summation could be described as signal build-up in a leaky integrator. We investigated the size of these space-time neighbourhoods and their contrast-sign specificity (if any). Is information from equal-sign and opposite-sign motion detectors combined separately or also mutually, and if so, how?

Equal-area RPA-patches [256 square pixels, from 256 × 1 (height × width) to 4 × 64] were presented on a larger background RPA (256 × 256 pixels of equal size, C=70%, L=50 cd m⁻²), and displaced relative to the background (velocity 1.41 deg s⁻¹). With the use of a staircase procedure, SNR thresholds for direction detection were determined in a 2AFC design. In a second experiment, lines containing 64 pixels (either all dark or all bright or both randomly distributed) were divided in 1 to 64 segments, oriented vertically or horizontally. Minimal thresholds were obtained when temporal and spatial summations were possible in about equal proportion. Sensitivity to fractionated arrays decreased gradually as a function of the number of fractions. Whereas subjects showed equal sensitivity to the bright and dark segments, fractionated RPAs were far less effective stimuli. A preliminary model explaining these results is presented.