Abstract
The classic aperture problem describes the ambiguity inherent to the motion of a frontoparallel (2D) contour (such as a line or an edge) viewed through a circular aperture. Despite a continuum of 2D velocities consistent with the apertured view, observers consistently perceive the direction of motion as orthogonal to the contour. Here we present an analogous 3D version where observers judged the 3D direction of motion of a slanted planar surface defined by a moving random dot stereogram presented behind a circular aperture. If the surface is specified by single frame dot lifetimes, the only potential factors influencing the perceived motion direction of the surface are the change in binocular disparity across time and 3D surface orientation. Provided observers use a similar heuristic in the 2D and 3D cases, such a surface should be perceived as traveling normal to its 3D orientation.
In separate sessions, observers judged either the perceived surface slant or direction of motion of the surface using a bird's-eye-view matching paradigm. We varied the surface slant, the lifetime of individual dots, and the 3D motion direction specified by the dots.
Slant judgments were close to veridical in all conditions. When dot lifetimes were more than one frame, and thus unambiguously specified surface motion, motion judgments were consistent with previously reported biases in the perception of 3D motion, and relatively close to veridical.
However, when the surface was specified by single frame dot lifetimes, motion was always perceived as moving directly towards or away from the observer. Thus, in the 3D version of the aperture problem, the perception of surface motion was heavily biased as moving along the line of sight, and not towards the perceived surface normal. These results suggest that the visual system might resolve perceptual ambiguity distinctly in 2D and 3D motion processing.