Learning a new cue to depth

Can learning transform a non-informative 2D shape into a new cue for 3D depth perception? To answer this question, we determined (i) whether the visual system can associate, through learning, a specific depth-order to the features of an ambiguous 2D shape, and (ii) whether the learned association can successively affect the perceived depth-order, when a 3D depth-map is attributed, through disparity information, to the same 2D shape. The experiment comprised two phases. During the learning phase, observers repeatedly viewed a 2D shape coupled with a depth-map specified by disparity information. In the test phase, two versions of the 2D shape used in the learning phase were presented. For the target stimulus, disparity information specified a (near threshold) depth map, having either the same sign as in the learning phase, or the opposite one; for the comparison stimulus, disparity specified a 2D shape. Observers were asked to choose the 3D stimulus. Performance improved across sessions when the depth-order of the learning phase was preserved in the test phase; performance decreased across sessions when the depth order of learning and test phases had opposite signs. Our results indicate that observers associated, through learning, a specific depth-order to an ambiguous 2D shape. Such results imply that a short-term associative process can transform an uninformative 2D signal into a novel cue to depth.