Interposition, minimal depth, and depth-from-disparity

In the context of the renewed interest for the combination of ordinal and metric cues to depth (Burge et al. JOV 2005; Burge et al. ECVP 2007), we presented the bar-in-the-box effect (Gerbino & Fantoni ECVP 2007). Consider a pictorial display involving only interposition and perspective, in which a bar crosses an open box, so that the front face of the box occludes the bar, which in turn occludes the back face of the box. Despite the lack of optical specification, most observers perceive the bar just behind the occluding surface in the foreground, consistent with a tendency towards minimal depth. We evaluated the strength of such a tendency in two experiments in which the box and the bar were both specified by binocular disparity. Observers judged whether a textured bar of variable disparity was closer to the front vs. back face of an untextured box. Were used different boxes, keeping their height and inclination constant, but varying their width/height ratio to change their binocular structure: the hole between front and back faces was present in both eyes, in one eye only, or absent in both eyes. In Experiment 1 the bar height co-varied with disparity, simulating a bar aligned with the medial axis of the box, while in Experiment 2 the bar was aligned with the line of sight. The perceived depth of the bar was smaller than the depth-by-disparity, showing a strong minimal-depth bias in binocular hole and no-hole conditions; while minimal depth did not overcome disparity in the monocular-hole condition. Configural factors - like the tendency to minimize depth in ambiguous displays and the presence of monocular features - affect perceived depth beyond what one might expect on the basis of the metric information provided by disparity, demonstrating their relevance in unambiguous conditions.