Stephanie M Shields¹ (); ¹The University of Texas at Austin

The visual system uses information in the two 2D retinal images to estimate the 3D structure of the environment, including the 3D orientation of surfaces. The information available depends on projective geometry, the mapping of the 3D environment onto the two retinae. One could expect that, as a result, projective geometry has shaped how the system processes 3D scene structure and can be used to at least partially explain patterns in perceptual performance. To test whether that prediction holds true for the perception of 3D orientation, the present experiments measured human performance on a two-interval forced choice slant discrimination task. Within an experiment, stimulus slant and viewing distance varied, and across experiments, the slant cues in stimuli varied. Observers likely combine information from multiple cues when estimating slant in everyday scenarios, so the first experiment used a cue-rich stimulus, specifically a square plane with texture that approximated pink noise. To assess the impact of various sources of information on performance, the cues available to observers were reduced in the three subsequent experiments, either by removing binocular disparity, texture, or global shape from the cue-rich stimulus. Results confirmed that observers weighted cues differently. Removing texture perturbed performance the least, suggesting that monocular and binocular shape cues most strongly influenced performance. A notable source of shape cues was the projection of the stimulus's vertical borders, so the geometry of that projection was used to predict performance patterns. In line with those predictions, observers generally performed best at high magnitude slants (beyond approximately +/- 70 degrees), and their performance typically worsened as slant approached frontoparallel (0 degrees) and as viewing distance increased. Thus, the results of these experiments both further understanding of slant perception and support the theory that projective geometry can explain patterns in perceptual performance.

Acknowledgements: National Science Foundation Graduate Research Fellowship Program