Cues and Pseudocues in Texture and Shape Perception

In estimating properties of the world, we often use multiple sources of information. For example, in estimating the three-dimensional (3D) layout of a scene, there are many sources of information or “cues” available for the estimation of depth and shape (Kaufman, 1974). These include binocular cues (disparity, vergence), motion cues (motion parallax, the kinetic depth effect), pictorial cues (texture, linear perspective, occlusion, etc.), and more. Human observers often combine these cues in a near-optimal fashion so as to maximize the precision of their estimates of scene layout (see Chapter 1, also Landy, Maloney, Johnston, & Young, 1995). Some cues are acquired after birth. In infancy, themotion and binocular cues appear to develop around 3–4 months of age, with use of the pictorial cues beginning some 3 months later (Kellman & Arterberry, 1998, 2006). The onset of responses to binocular disparity appears to be based on maturation of the required vergence control and cortical architecture (binocular, disparity-tuned cells). On the other hand, it is not known whether the onset of the response to pictorial depth cues is the result of maturation or learning. If a new depth cue is to be learned, the visual system has to note a correlation between values of that cue and other indicators of the values of the environmental variable being estimated. For example, consider the pattern of texture as a cue to slant. For observers to learn how to use texture, they must associate larger texture gradients (rapid changes in the size and density of texture elements across the image) or larger values of foreshortening (in which circles on the surface appear as eccentric ellipses in the retinal image) with larger values of surface slant. An observer could do so by noting the correlation of these image features with previously learned cues to surface slant such as the gradient of binocular disparities, or by using haptic cues as the observer handles the surface manually. Backus and colleagues (see Chapter 6; Backus & Haijiang, 2007; Haijiang, Saunders, Stone, & Backus, 2006) investigated the visual system’s ability to learn new depth cues. They did so by artificially pairing different values of a new, arbitrary “cue” with depth cues on which the visual system already relies. In their experiment, observers viewed Necker cubes that rotated about a vertical axis (Fig. 14.1). ANecker cube (Necker, 1832) is a picture of a transparent cube in which only the edges are drawn. It is an ambiguous figure; every few seconds the perception of the cube “switches” so that the face that was previously seen as behind is now perceived to be in front, and vice versa. When such a figure is rotated, the perceived direction of rotation reverses along with perceived depth.