Abstract
Biases in perceived 2D orientation are asymmetric in the frontal plane: Deviations from horizontal are exaggerated such that a line of 37° from horizontal appears to be 45° (Durgin & Li, 2011). How flexible is the reference frame within which these biases occur? Last year (VSS2012) we showed that these biases stayed with perceived external horizontal/vertical axes in tilted observers. Here we show that, once depth compression (under-constancy) is taken into account for obliquely-viewed planes, the asymmetrical biases remain. Participants (N = 56) made orientation estimates in four virtual display types. In the first and the last displays, oriented lines appeared on frontal walls. In the two middle displays, line orientations were judged relative to implicit planar axes on a textured vertical surface (side wall) viewed to the left of the observer at a 30° observation angle and on a textured horizontal surface (ground plane) viewed downward at a 30° angle. Orientation estimates were made either with respect to the depth-aligned axis (horizontal on the wall plane; "vertical" on the ground plane) or the frontal axis (vertical on the wall plane, "horizontal" on the ground plane). Once depth compression (factor of 0.73) was taken into account, orientation estimates in both planes showed asymmetric patterns of bias with exaggerated deviations from horizontal axes. In other words, the perceptual orientation bias exists with respect to post-constancy representations of surface orientation. Note that the asymmetrical spatial orientation bias depended on the horizontal/vertical axes of the respective planes rather than the reference axis defined as zero. As in the body-tilt paradigm, the results indicate that the bias is not in the retinotopic coding of orientation, but at a later, post-constancy, processing stage in which orientation is coded in extrinsic coordinates imposed by the visual system.
Meeting abstract presented at VSS 2013