Abstract
Perception of dynamically occluded objects is highly accurate when occluders contain many apertures. With only two misaligned apertures, however, it is subject to a strong illusion such that an aligned rod appears misaligned in the direction of the apertures (the Aperture-Capture Illusion [ACI], VSS 2001, 2002; Hecht, 1924). In 9 experiments, we used a nulling procedure to assess the magnitude of the illusion under a range of object velocities, fixation conditions, and aperture arrangements. We will present a quantitative model of the ACI that largely explains these data using two assumptions: 1) Underestimation of object velocity after occlusion, and 2) A short binding time to integrate visible and occluded regions of a moving shape into a single perceptual unit. We estimated occlusion velocity by calculating the amount of time the rod was occluded, along with the resulting illusion in the 9 experiments. This model makes an unintuitive prediction: If two parts of a moving object appear together perfectly aligned, and then one part becomes occluded, there will be a strong illusion of misalignment. Results from 2 experiments confirm this prediction, and suggest that it takes approximately 40ms to bind visible and occluded regions of an image. Our model may account not only for the ACI but also for a class of illusions from the anorthoscopic literature. When a circle moves behind a vertical slit, the circle appears as an oval compressed along its axis of motion. Additionally, when a square is seen moving behind a thin, C-shaped slit, the sides of the square appear bent in the same direction as the slit. In general, when a moving object is seen through a deformed slit, the perceived shape appears compressed and takes on the form of the slit (Anstis & Atkinson, 1967). This class of illusions may be explained by a perceptual process that integrates occluded and visible regions of a moving shape over time, but underestimates the velocity of occluded regions of the shape.