Abstract
When a pair of stereo pictures were interchanged, all disparities are reversed, but inverted depth is never seen with familiar objects such as human faces (reversed disparity illusion). In contrast, Georgeson(1979) reported that concave face can be seen when a face is depicted as random-dot stereograms. Thus, the illusion for regular pictures is often related to top-down effects of familiarity, and existence of visible face seems the key. In this study, we examined depth appearance for conditions between these two extreme cases to clarify the effects of familiarity by superposing random-dot stereograms and regular stereo pictures of human face, and manipulating relative strength. Stimuli were stereo pictures of a female face. Pictures were taken with two digital cameras with inter-camera distance (disparity) varied for 4 steps. For random-dot stimuli, similar pictures were taken with random-dots projected on the face by a projector. Dot images were extracted from these pictures by image processing. Then the contrast of regular picture and random-dots were normalized. In Exp. 1, intensity ratio between regular picture and random-dots were varied in 5 steps, while keeping the total contrast constant. The regular face appeared convex regardless of disparity, but rate of concavity perception increased rapidly as intensity of dots was increased. It reached 100% for 50% intensity ratio. In Exp. 2, contrast of regular pictures was halved and the random-dots with varied intensities were simply added. For these stimuli, although the face is visible in all the conditions, rate of seeing concave face increased as dot intensity increased. Another experiment with negative pictures (Exp. 3) produced similar results, but the rate of seeing concave face increased more rapidly. These results indicate importance of clearly defined depth cues such as random-dots as well as strong cognitive effects for the reversed disparity illusion.