Iroshini Gunasekera^1,2 (iroshini@yorku.ca), Domenic Au^1,3, Abtin Zaker³, Robert S. Allison^1,4, Laurie M. Wilcox^1,2; ¹Centre for Vision Research, York University, ²Department of Psychology, York University, ³Department of Biology, York University, ⁴Department of Electrical Engineering and Computer Science, York University

Shape constancy allows the brain to interpret the 3D structure of an object despite changes in its retinal projection. For example, when a planar surface is slanted in depth, its retinal projection becomes foreshortened, and the visual system must account for slant to recover its shape. If depth and slant are estimated accurately, then the shape should be accurately perceived. If not, it would be distorted. While such distortions have been documented, it is not clear which sources of depth information best support shape constancy. Here, we systematically assessed the impact of monocular and binocular depth cues on shape constancy for planar surfaces rendered in perspective. Renderings of rectangular table-top surfaces were presented sequentially using a mirror stereoscope; one was positioned so that it receded in depth along its long x-axis (reference surface), and the other was aligned horizontally. Using a method of adjustment, participants (N = 18) modified the length of the horizontal surface to match the perceived length of the reference surface. Four conditions with varying amounts of depth information were tested in random order in both perspective (2D) and stereoscopic 3D: untextured rectangles, dropped shadows, texture, and contextual shape cues (table legs and plates). In all viewing conditions the stimuli were rendered from a 25-degree angle at 0.45 m. Results showed that shape estimation improved most under stereoscopic 3D viewing, with the contextual cue condition producing the most accurate estimation in both 2D and 3D. Although shape estimation did not significantly differ between the untextured rectangles, dropped shadow and texture conditions, participants showed consistently higher shape estimates when stereopsis was available. These findings highlight the strong impact of contextual cues on shape constancy and demonstrate that even small reference shapes (table legs and plates) greatly improve the visual system’s ability to recover shape from slanted projections.

Acknowledgements: This research was supported by grants RGPIN-2019-06694 to LW and ALLRP 570802–21 to RA and LW from the Natural Science and Engineering Council (Canada), and CF-REF Connected Minds funding to I. Gunasekera.