Depth Interval Perception: Comparing Binocular Stereopsis With Motion Parallax In “Action Space”

Motion parallax is widely regarded as providing metric depth information that is equal or superior to that obtained from binocular stereopsis — in part because it is assumed to be fairly robust with viewing distance. But excellent depth-recovery from motion parallax may be limited to very near “personal” space, where most motion parallax experiments are done. For example, Ono et al. (1986) found that although motion parallax produced rigid depth percepts in very near space, simulated objects appeared to become non-rigid as viewing distance was increased past a meter or so. We sought to investigate the perception of empty intervals between objects. To ensure there were no artifacts of temporal lag, we created a “cloud chamber” in which fibrous clouds could be suspended on invisible wires in front of a back-lit matte blue screen. A volume of space 2 – 4 m from the observers was used — the near part of “action space” (Cutting & Vishton, 1995). Naive subjects viewed pairs of clouds that were offset in depth (and offset slightly horizontally, to avoid occlusion), and estimated the depth separation between them. Static binocular viewing was compared with dynamic monocular viewing where the extent of lateral head motion (12.5 cm) was twice the typical interocular distance. Despite the theoretical advantages accorded by doubled parallax, the gain of perceived depth separations from motion parallax was only about half that from binocular stereopsis. This inferiority of motion parallax corresponds to our own subjective experiences: At these distances, monocular motion parallax from lateral sway produces the impression of relative motion, not depth, between objects. Whether object-directed action would show greater evidence of depth sensitivity remains to be seen