Abstract
When we look at an object, multiple sources of information define its 3D shape, with binocular disparity being the most prominent. However, the elements on a surface that yield these binocular disparities are often sparse. Previous work found evidence of a cooperative interaction between shading and disparity information for surface interpolation. Here, we tested whether luminance gradients that vary smoothly across the retinal image are able to ‘fill in’ the missing values of the disparity field. If this is the case, we can predict a paradoxical outcome: a shading gradient may warp a disparity field. Observers were asked to perform a probe placement task where they adjusted the binocular disparities of a small dot until it appeared to rest on a presented surface in depth. We tested surfaces defined by a smooth disparity dot field with monocular shading information that either corresponded to the surface (cue-consistent), or was generated by a surface perturbed by random bumps (cue-inconsistent). The disparity and shading were combined with a method of back-projection such that the disparity field was completely consistent with a smooth surface. Corresponding probe positions were selected on each surface such that, in the cue-inconsistent trials, the probes intersected with controlled bumps that were either convex or concave. Importantly, during the cue-inconsistent trials, observers exhibited a strong depth bias in the direction of the shading gradient when placing the probes on the controlled bumps. The depth bias was highly reliable given the degree of precision across conditions. This novel result suggests that the bumpy shading was able to warp the smooth disparity field by constraining the interpolation between the defined points, despite the fact that the shape specified by shading was inconsistent with the smooth disparity field.