|Budget Amount *help
¥8,600,000 (Direct Cost : ¥8,600,000)
Fiscal Year 1997 : ¥500,000 (Direct Cost : ¥500,000)
Fiscal Year 1996 : ¥8,100,000 (Direct Cost : ¥8,100,000)
When the left and right pictures of a stereo pair are interchanged, binocular disparity is completely reversed, but we tend to perceive a normal convex rather than a concave face, similar to the hollowface illusion with 3D shapes. We examined this illusion with stereo pictures by manipulating disparity and familiarity factors. Method. Stereo photographs with normal, zero, and reversed disparity were presented on a CRT screen and viewed through a mirror stereoscope. Subjects were asked to identify the disparity type by a 3AFC method. Results, In Expt 1, stereo photographs depicting human faces, statues, and buildings were presented for an unlimited duration to untrained subjects. Contrary to our expectations, reversed disparity was detected fairly easily. The average correct rate was 60% (chance=33%), but less for facial stimuli. In Expt 2, only facial stimuli were presented for 500ms. Disparity and contrast polarity were also manipulated. With this short duration, detection of reversed disparity was at chance level improving as disparity became exaggerated. With reversed luminance contrast (photographic negatives), however, detection was dramatically improved. We also examined the effect of orientation, in a separate experiment, by presenting inverted facial stimuli. Performance improved, but the effect was modest compared to contrast reversal. Conclusion. The weakened illusion with reversed-contrast or upside-down facial stimuli suggests a contribution of familiarity factors, but such factors could be overridden by a low-level disparity signal, indicating complex dynamics among different depth cue systems. The results also indicate that, contrary to van den Enden & Spekreijse's (1989) conjecture, texture gradient cannot be a determining factor since it is kept constant in either contrast-reversed or inverted stimuli.