Abstract
Although the position in depth of monocularly occluded regions (half-occlusions) is ambiguous, there is considerable evidence that the visual system capitalizes on this information to enhance the speed and accuracy of binocular depth perception (Gillam & Borsting, 1988). To date, depth percepts from half-occlusions has been studied almost exclusively in configurations where a surface occludes part of the background in one eye. However, monocular regions also result from self-occlusion and in such cases, because the monocular region forms part of the foreground object, they could influence its perceived shape. Here we evaluated how monocularly occluded regions resulting from self-occlusion influence the perception of depth in 3D volumetric objects. Stimuli were greyscale textured truncated half-ellipses rendered stereoscopically using Blender and viewed in a mirror stereoscope. In a series of experiments, we manipulated the degree to which half-occluded regions were consistent with 3D viewing geometry. We also assessed perceived object coherence using a method of constants (yes/no), and perceived depth magnitude for these objects under geometrically valid and invalid viewing conditions. Our results show that there is surprising tolerance to invalid occlusion arrangements when monocular regions are perceived as part of an occluding object. Furthermore, unlike conventional surface/background half-occlusions, we find that quantitative depth percepts from self-occlusions do not follow model predictions based on binocular viewing geometry even when the occluded region is consistent with viewing geometry. Taken together our experiments show that the interpretation half-occlusions depends critically on their context; heuristics used to extract depth information from such monocular regions are contingent on foreground/background segmentation.