Abstract
Purpose. Perceived depth curvature may readily be interpolated across sparse samples of the specifying disparity cues. Location of monocular shapes defined by luminance shading also survives sparse sampling, suggesting the involvement of an interpolation process, which might operate at the level of generic depth representation rather than being specific to disparity processing. To test this hypothesis, we combined luminance and disparity curvature in a sampled position discrimination task (Kontsevich & Tyler, 1997, VR) to measure the accuracy of centering of Gaussian disparity/luminance profiles. Methods. Stimulus samples were vertical 1′ bars separated by uniform gaps of 15′, on a dim background. The depth profile, with a width at half-height of 1 deg, was carried either by the disparity or the luminance of the bars. Position thresholds were measured relative to a reference whose disparity was offset from the peak of the Gaussian by a randomly varying amount (to avoid depth matching), using the Psi Bayesian staircase procedure. Results. Localization was poor (5–10′) for Gaussians specified by luminance cues alone, but improved by as much as a log unit solely by the addition of disparity cues to the identical Gaussian profile. Perceived depth from the luminance cues biased the position sensitivity for disparity so as to degrade performance around the point of the perceived flatness (with large individual differences). Conclusion. Localization of broad luminance profiles is enhanced in rough proportion to their perceived depth, suggesting that depth processing plays an unsuspected role in the position coding of sampled stimuli. Localization could be degraded in luminance stimuli by nulling their perceived depth with a disparity profile. We suggest that profile interpolation may be restricted to the depth-processing system, and is available to support the localization of luminance stimuli only to the extent that they evoke a depth percept. See www.ski.org/cwt for details.