Abstract
The human visual system has a highly developed sensitivity to relative binocular disparity, as shown by stereoacuities of a few arcseconds. Relative disparity requires a pair of stimuli with disparities that can be differenced. What sort of pairing of stimuli is required to compute a relative disparity?
To find the spatial and temporal requirements for optimal disparity discrimination, disparity thresholds were measured for a target stimulus as a function of reference stimulus parameters. In one experiment, the target (a gabor annulus) surrounded the reference (a gabor patch), much like a foveal stimulus surrounds a fixation point. In another experiment, the reference gabor (shown for 400 ms) was overwritten by the target gabor (150 ms) of the same spatial extent, in the manner of an apparent-motion display but with no spatial displacement. In these experiments the reference has zero absolute disparity and is more salient spatially or temporally—more foveal or longer in duration—than the target.
Invariances in psychophysical performance often reveal the underlying computation. Here the invariance in the target thresholds is found with respect to reference parameters. Target thresholds vary with the spatial frequency of the reference stimulus, so as to maintain a near-constant reference phase angle. This reference constancy holds at least to moderate spatial frequencies (∼4 c/d). At the highest of these reference frequencies, thresholds are about as low as observed when the most effective broad-bandwidth fixation point plays the reference role.
A reference stimulus is not just a fixation point. It sets the resolution for detecting relative disparities. In spatial and temporal configurations that are balanced between the two stimuli, as in the bipartite gabor patches used in a third experiment, the stimulus taken as ‘reference’ is not arbitrary; the one with lower frequency determines the resolution.