Abstract
To determine the global (overall) direction of image motion the human visual system must integrate local motion signals across space and time. However, the spatial extent over which this integration process operates in central vision is uncertain. Previous studies that have measured direction discrimination for random-dot-kinematograms (RDKs), suggest that the spatial integration limit for global direction is at least as large as 63 deg2 (e.g. Watamaniuk & Sekuler, 1992; Vision Research 32: 2341 2347). However the total area of retinal stimulation and the dot number (to maintain dot density) both co-varied with changes in RDK size. Consequently, regional variations in motion sensitivity across space and the availability of local motion samples potentially confound these estimates. To address this issue we developed a novel display that enabled us to measure the global motion integration area without changing the overall stimulus dimensions, dot number and density. Global motion coherence thresholds (79% correct) for direction discrimination were measured binocularly for 5 observers using RDKs in which the signal dots (all moving either up or down at 3.91 deg/s) were confined to a central, circular region of the display. The surrounding annular region only contained randomly moving noise dots. The spatial extent of the signal-dot-region varied from 1.95 to 249.14 deg2, but overall RDK size (249.14 deg2) and dot number (1024) remained constant. Coherence thresholds should be constant when all the signal dots are within the spatial integration area for global motion, but rise proportionally once this is exceeded. Results showed that the integration area for global motion was, on average, 12 deg2 and was unaffected by a log-unit change in dot contrast (100 to 10 %). The size of this integration area in humans is consistent with physiological measurements of foveal-centred receptive fields in primate area MT.
Wellcome Trust (PM) Chief Scientist Office Scotland (AS).