Abstract
It has frequently been reported that stereoscopically-defined, slanted surfaces have greater perceived slant when the rotation axis is horizontal (tilt = 90 deg) than when it is vertical (tilt = 0 deg). It has also been observed in cue-conflict experiments that more weight is given to the slant specified by monocular cues when tilt = 90 deg than when tilt = 0 deg. We asked whether these two effects - which have been called slant anisotropy - are caused by a reduction in the precision of stereoscopic slant estimates for tilt=90 deg compared to tilt=0 deg. We measured slant discrimination thresholds for disparity alone and for texture alone for a variety of base slants, and for tilts of 0 and 90 deg. Discrimination thresholds for disparity alone were slightly lower when tilt = 90 deg and thresholds for texture alone did not vary consistently with tilt. From those thresholds, we can determine the variances of the disparity and texture estimators at different slants and tilts. We next examined how disparity and texture are weighted when both cues are relevant to the task. From the single-cue variances (measured in the single-cue discrimination experiment), we can predict the relative weights that should be given to disparity and texture in a cue-combination experiment. The agreement between predicted and observed weights was quite reasonable. More weight was given to texture as base slant increased due to a concomitant increase in the reliability of slant-from-texture. Slightly more weight was given to disparity for tilt = 90 deg than for tilt = 0 deg. However, the small tilt variation we observed in stereoscopic precision was insufficient to explain the large variations in perceived slant that has been previously reported.
Original language | English (US) |
---|---|
Pages (from-to) | 80a |
Journal | Journal of vision |
Volume | 3 |
Issue number | 9 |
DOIs | |
State | Published - 2003 |
ASJC Scopus subject areas
- Ophthalmology
- Sensory Systems