Abstract
The perception of disparity-defined cyclopean form requires both the successful measurement of disparity and the linkage of multiple disparity measurements across the image. Performance limitations in cyclopean tasks may therefore depend on both difficulties arising at the disparity measurement stage and at later stages, where surface structure is determined. To assess the contribution of these factors in cyclopean tasks, participants were presented with disparity-defined random-dot sinusoidal corrugations in depth and asked to perform an orientation discrimination task. Corrugations were presented embedded in random-dot masks, which were structured as either random disparity distributions, or anti-phase disparity sinusoids. Multiple mask types were used in order to assess the contributions of both random and structured disparity signals in disrupting cyclopean performance. Thresholds were obtained for the number of masking dots required to reduce orientation discrimination performance to 75% correct levels. For random disparity distributions, mask-to-surface-dot ratios increased with increasing numbers of surface dots, ranging from 0.91 for 60-dot corrugations to 3.5 for 200-dot corrugations. For anti-phase masks with a 120-dot target corrugation, thresholds were around a mask-to-surface-dot ratio of 0.5, falling to chance performance by a ratio of 1. These results show that anti-phase mask structures are more effective at disrupting cyclopean performance than random disparity distributions. At a ratio of 1, where the numbers of mask and surface dots are equal, performance in the cyclopean orientation discrimination task improved, to above threshold performance, with increasing disparity amplitude of the corrugation, decreasing cyclopean frequency, or an increase in the total number of stimulus dots. Using a local cross-correlation model of disparity measurement, we show that psychophysical performance cannot be fully accounted for by a decrease in cross-correlation measures at target disparities, but must instead result from ambiguities arising in processes involved with the linkage of multiple disparity measurements.
Meeting abstract presented at VSS 2016