Abstract
Aim: Correspondence noise is a major factor limiting threshold coherence for detecting motion in Random-dot Kinematograms (RDKs) (Barlow and Tripathy, 1997, Journal of Neuroscience, 17, 7954–66). Computer modelling suggests that correspondence noise could also limit Dmax (Tripathy & Barlow, 2001, Perception (Suppl.), 30, p.32]. This model assumes that the radii of the catchment area of the Reichardt-type local detectors scale directly with the displacement that the detectors are tuned to detect. We directly test this assumption psychophysically: if detectors tuned to larger displacements are associated with larger catchment areas, then they should be less susceptible to positional jitter of the dots.
Methods: We generated 2-frame RDKs (15 ×10 , 3000 dots) with different levels of coherence and with positional jitter. On each trial the stimulus moved either right or left and the observer reported the direction of motion. For the different dot-displacements used (10–50 arcmin), the coherence levels were adjusted to equate performance without saturation (82–90% correct responses) for the unjittered stimuli. To generate jitter, each dot was displaced to lie on the circumference of an imaginary circle centred on the dot's unjittered position with radius proportional to the amount of jitter. Within a block the average dot-displacement was kept fixed and the ratio: r =(radius of jitter / average displacement size) was varied to obtain a complete psychophysical function (r=0−1.5). Across blocks the average dot-displacement was varied.
Results: Consistent with our predictions, the different psychometric functions were similar, suggesting that catchment regions of local detectors scale with the size of the displacement they are tuned to detect, and that correspondence noise could indeed limit threshold coherence and Dmax.