Looking through Ames' window

Ames' rotating trapezoidal window demonstrated that planar trapezoidal shapes are often interpreted by the visual system as resulting from rectangular shapes slanted differently in depth, even in the face of contradictory information from stimulus motion. In Ames' demonstration, surface markings and silhouette were both compatible with the same slanted rectangle. We have noticed that when surface texture simulates a slant opposite to that conveyed by the silhouette of a trapezoidal shape, a bistable percept results: with attention to the outline of the shape, the texture is captured by the silhouette, and appears painted on the surface and distorted; with central fixation, the surface texture appears slanted in depth receding away from a window formed by the outline. We used moving textures viewed through trapezoidal windows to examine combination and separation of layers defined by texture and silhouette cues. Textures consisted of sinusoidal gratings oriented at 0, 45, 90 and 135 degrees, and plaids constructed from orthogonal grating pairs. Texture patterns were distorted to simulate perspective projections of planar surfaces slanted left, right or frontoparallel drifted horizontally on a monitor, and viewed monocularly through a trapezoidal aperture, which was oriented with the parallel sides vertical. The silhouette and texture cues to slant were thus either compatible or incompatible. Drifting vertical and oblique gratings were perceived as moving in the same slanted depth plane as the window (3D Aperture Effect). Drifting horizontal gratings were perceived as slowly receding and advancing in depth. Compatibly slanted plaids adhered to the window, but incompatibly slanted plaids separated from the window and were immune to the 3D aperture effect. Trapezoidal apertures create 3D percepts for moving 1D patterns, separation into layers and immunity from the aperture effect for 2D patterns cannot be explained in terms of their 1D components.