Abstract
Correspondence noise is a major factor limiting the detection of motion in random-dot kinematograms, whether detection performance is measured using threshold coherence or Dmax [Barlow & Tripathy, 1997 Journal of Neuroscience 17 7954–7966; Tripathy & Barlow, 2001 Perception(Suppl.) 30 32]. The previous studies provide a unified theoretical approach for studying the factors limiting threshold coherence and Dmax. Can this unified theory provide an explanation for Dmin, the smallest detectable coherent displacement for the dots of a kinematogram?
This study assumes that for kinematogram stimuli, motion is detected by global motion detectors that sum the outputs from arrays of local detectors, each having bi-local catchment regions. For such motion detecting systems, several factors could contribute towards the existence of a Dmin:
There might exist a physiological lower limit to the spatial separation between the two catchment regions of the local detector.
The local detectors could undersample the stimulus plane for small displacements; the smaller the local detectors, the greater their required number for tiling the stimulus plane.
The areas of the two catchment regions for each local detector could limit the smallest detectable displacement; these areas must be small so as to reduce correspondence noise but large enough for the detection of non-rigid motion in the real world.
The theory from Tripathy & Barlow (2001) has been extended to incorporate the above constraints. Theoretical predictions for the changes in Dmin produced by changing dot density or the proportion of coherently moving dots compare favourably with results from computer simulations for two-frame motion. Psychophysical experiments are in progress to test which of the above three constraints limit performance. This study brings us closer towards the goal of providing a unified theoretical framework for studying Dmin, Dmax and threshold coherence.