Abstract
Two experiments were performed to examine the perception of surface slant from texture gradients and binocular disparity. In Experiment 1, observers matched the apparent slant of a planar surface defined by monocular texture gradients or binocular disparity by adjusting the orientation of a line on a separate monitor. In Experiment 2, observers matched the apparent slant of a planar surface defined by monocular texture gradients with an adjustable stereoscopic planar surface. The results of both experiments reveal that observers' judgments of slant from texture are systematically biased, such that stereoscopic surfaces appear significantly more slanted than those that are defined by monocular texture gradients. In addition, the variance of observers' judgments for both sources of information increases with the slant of the depicted surfaces. These results are fundamentally incompatible with data obtained from slant discrimination studies (e.g., Knill & Saunders, Vis Res, 2003) that have consistently shown that the reliability of observers' slant estimates is higher for large slants than for small slants. In order to determine the reliability of slant estimates from discrimination thresholds, it is typically assumed that observers‘ judgments are veridical except for internal random noise. This assumption is necessary because thresholds can also be affected by any systematic biases in perceived slant or the presence of 2D cues. Our results indicate that the assumptions employed for interpreting the results of prior discrimination studies may not be valid, and that the use of discrimination procedures for investigating the reliability of perceived slant is questionable.