Abstract
A person depicted in a portrait painting does not become slanted even when observers move around. The gaze is also fixed to the observer. This effect is called the Mona Lisa effect. In addition, the face appears thinner when the Mona Lisa effect occurs. In this study, we examined this face narrowing to explore relationship between the Mona lisa effect and depth perception. The face narrowing is a break-down of shape constancy and suggests the information about the physical orientation of the painting/photography or the face/body is not obtained by the observer. If the slant of the portrait is obtained by the observer, the face width should be perceived wider than that implied by the width of the retinal image because of shape constancy. In experiment 1, we presented 2D portraits generated from CG 3D model as test stimuli and a line-drawing oval as references. Rotation angles of −30 deg to + 30 deg were used. Binocular disparity was given to the 2D portrait to represent the slant and presented with mirror stereoscope. Observers were asked to compare the face and flat oval and to judge which appeared wider. The results showed that facial width was perceived narrower with increasing rotation angle and indicated that the Mona Lisa effect occurred. Moreover, intensity of the effect was stronger when edges of background were blurred by Gaussian mask to reduce binocular disparity cues. Following these experiments, we investigated whether the Mona Lisa effect is specific for human face by using inverse and negative/positive portraits. The results showed that changing of facial width became smaller than upright portrait. However, intensity of positive portrait also became smaller. These results indicate that the Mona Lisa effect occurs from conflict between face specific cues (occlusion) and binocular disparity cues.