Abstract
Visual information is often noisy and ambiguous. To overcome this, the visual system exploits multiple sources of information to resolve ambiguity. Here we investigate the use of different kinds of motion information to recover the shape of a displaced object from the noisy background of a random dot cinematogram (RDC). The ability to perceive shape-from-motion in an RDC depends on the ability to segregate the coherent motion of the object from incoherent background motion. The stimuli in Experiment 1 were two-frame RDCs; a rectangular figure is displaced and the background updated randomly. For half the trials, the contrast is reversed during the second frame, creating reverse-phi motion. Subjects indicated the perceived motion direction (left or right) and shape (tall or wide). Judgments could be based on either 3rd-order counterchange-specified motion or 1st-order motion energy when contrast was not reversed, but only on motion energy when it was reversed. Experiment 2 was identical except that triangular figures were displaced; shape discrimination was based on the triangles’ orientation. Following Sato (1989), shape was easily discernible for small displacements in the non-reverse condition, but not in the reverse-contrast condition, for which reverse-phi motion is perceived in the direction opposite to the displacement. Differences in shape were weakly inferred from differences in reverse-phi motion strength between the flat and upright rectangles; this was eliminated when reverse-phi motion strength was balanced for the triangular shapes in Experiment 2. The results suggest that motion energy provides information that can inferentially aid simple shape discrimination, but is distinct from the non-reverse condition, for which a segregated figure can be perceived. We conclude that 3rd-order, counterchange-specified motion is the basis for perceiving shape-from-motion in the RDC. Its absence for motion in the reverse-phi direction is consistent with the "objectless" quality of 1st-order motion energy.
Meeting abstract presented at VSS 2012