Abstract
The gaze and face orientation of a person depicted in a portrait painting appears to follow observers when they move around. This is called the Mona Lisa effect (MLE). The binocular gradient of disparity on the picture surface indicates the slant of picture surface, but pictorial cues from the facial parts (relative positions and occlusions between them) indicate the face is facing straight to the observer. Morita et al. (2020) showed that the existence of pictorial cues suppresses the functioning of binocular disparities when the MLE occurs. In this study, we examined the effect of disparities given to the background, the external frame, and the face on the occurrence of MLE by using the perceived facial width of slanted portrait as an index as for our previous study. In Experiment 1, we presented 2D portraits rotated ±30 degrees around z-axis as test stimuli and a line-drawing oval as the reference. We used lowpass-filtered random-dot texture as the background to enhance the disparity cues. Observers judged which, the face or the reference, appears wider. The results showed that the MLE intensity was almost equal regardless of the existence of background. In Experiment 2, we measured the MLE intensity for portraits containing disparity only in the external frame or only in the facial part, and compare the results with those obtained from portraits containing disparities in both components. It was found that the intensity was almost equal between the three conditions. In sum, the present results demonstrated two intriguing aspects of the MLE; (1) the disparity gradient given to the picture background does not affect the intensity of MLE, and (2) the disparity given to either the picture frame or the facial part, or both generate the MLE of about the same intensity.