PERCEPTION 1625 



not much beyond 6 m from the eyes) ; it may be more 

 important in enhancing accuracy of distance judg- 

 ments than in providing us with elementary im- 

 pressions of depth. The role of these mechanisms in 

 aiding perceptual constancy, especially of size, is 

 another matter (see below). 



depth from gradients. A refreshingly new approach 

 to the study of depth perception has been introduced 

 by Gibson (154). Rejecting the emphasis on manifold 

 but separate clues or cues, Gibson has pointed out 

 that a convincing two-dimensional picture of a 

 surface slanting into the third dimension is obtained 

 by presenting 'gradients of texture'; an array of 

 patterns (e.g. circles) set up in rows and columns 

 will give the impression of a receding slanting plane, 

 if successive rows, from the bottom of the picture to 

 the top, diminish in size and increase in density. The 

 same effect can be obtained with converging lines, 

 as shown in figure 21. Gibson believes that such 

 texture gradients provide a psychophysical correlate of 

 monocularly obtained depth and that main of the 

 traditional cues can be subsumed under if 1 is gradient 

 notion. 



These gradients in a scene become even more 

 effective when the observer (or his eyes) move relative 

 to the environment. As Gibson points out, there is 



FINE 



TEXTURE 



COARSE 

 TEXTURE ' 



TOP OF 

 RETINA 



~n r 



CORNER BOTTOM OF 



TOP OF 



RETINA 



BOTTOM OF 



RETINA 



fig. 21. Depth from gradients. Left, a change of gradient 

 corresponding to a corner; right, a jump between two gradients 

 corresponding to an edge. [From Gibson (154).] 





fig. 22. Flow patterns (continuous perspective trans- 

 formations) in the visual field of a pilot in level Bight < • 

 and during a landing glide {below). [From Gibson (154).] 



more than classical motion parallax involved in the 

 resultinc; transformations. During the landing glide of 

 an aircraft, for instance, complex but characteristic 

 flow patterns deform the ground below, and the 

 clouded sk\ above, in a continuous sequence of 

 'perspective transformations.' The point on the 

 landing strip toward which the pilot moves forms the 

 center of an expanding pattern; everything around 

 it moves radially away from this center (fig. 22). 



In addition, Gibson suggests that there is usually 

 enough in the stimulus pattern, at least in natural 

 surroundings, to enable a perceiving organism to 

 distinguish his own movement through space from 

 the motion of objects in the scene relative to the 

 perceiver; self-movement results in characteristic 

 perspective transformations of the entire scene (156). 

 It is for that very reason that movement of the entire 

 scene, or most of it, relative to the observer, tends to 

 be ambiguous; the observer readily believes himself 

 to be moving, in direct confirmation of Duncker's 

 experimental results (no) on induced motion (that 



