Abstract
Thresholds for the discrimination of disparity-defined structure in random-dot stereograms (RDS) are elevated when the images presented to each eye are of different contrasts, or when a proportion of the signal dots in a stimulus are replaced with uncorrelated noise dots. Such threshold elevations reveal limiting factors in the visual system's computation of binocular disparity. In the present study, we examine how these factors of inter-ocular contrast difference and decorrelation noise combine to elevate thresholds in a disparity discrimination task. Observers were presented with a series of RDS containing disparity modulated gratings, oriented at ±20 degrees, and asked to decide whether each grating was clockwise, or counter-clockwise rotated. Grating amplitude was varied to obtain thresholds at which disparity-defined orientation could be reliably discriminated. Thresholds were obtained across a range of grating spatial frequencies, for different inter-ocular contrast differences, and different proportions of decorrelation noise. As expected, discrimination thresholds were elevated by an increase in inter-ocular contrast difference, and by an increase in the proportion of uncorrelated dots present in the stimulus. However, the effect of these factors in combination was not simply additive. Instead, increased decorrelation noise led to proportionally greater threshold elevations when inter-ocular contrast differences were also high. This suggests that inter-ocular contrast differences affect disparity computation in a manner that exacerbates the correspondence noise problem produced by decorrelation. We compare the observed effects of inter-ocular contrast difference and decorrelation to other forms of noise in stereoscopic stimuli.