Abstract
In images of textured surfaces, orientation flows formed by perspective convergence invariably convey 3-D shape. We examine whether second-order orientation flows convey 3-D shape, whether they induce 3-D shape aftereffects, and whether these aftereffects are invariant to how the flows are defined. Three sets of surface textures were used: a full contrast 2.6 cpd luminance-modulated horizontal-vertical plaid (LM); a 2.6 cpd vertical full contrast grating contrast-modulated by an equal-frequency horizontal grating envelope (CM); and a horizontal abutting grating illusory contour stimulus (IC). These textures were mapped onto carved sinusoidal concavities and convexities that varied in depth as a function of horizontal position. For these surface shapes, orientation flows are formed by horizontal components of the surface texture. Baseline measurements show that the perceived flat point for each of the three sets of stimuli was close to physical flatness, indicating that CM and IC orientation flows convey 3-D shape. For all three sets of stimuli, adapting to curved stimuli caused 3-D shape aftereffects on subsequently presented test stimuli, quantified by a shift of the perceived flat point away from the baseline. Adapting to LM orientation flows induced robust 3-D shape aftereffects on CM and IC tests. Aftereffects using CM and IC adapting stimuli on LM tests were substantially weaker, but were robust when adapting/test stimuli were both CM or both IC. These results can be explained by the adaptation of 3-D shape-selective neurons in extra-striate regions that invariantly extract first- and second-order orientation flows from striate and extra-striate signals. The asymmetry of aftereffect strengths can be explained by stronger input to these shape-selective neurons from neurons selective for orientation defined by first-order differences.