December 1964
Further Consideration of a Paradox in the Visual Perception of Translatory Motion
J. J. Gibson, Cornell University
The World Wide Web distribution of James Gibson’s “Purple Perils” is for scholarly use with the understanding that Gibson did not intend them for publication. References to these essays must cite them explicitly as unpublished manuscripts. Copies may be circulated if this statement is included on each copy.
In 1958, a number of paradoxes were described that arise when one tries to make experimental studies of the visual perception of motion (Second Symposium on Physiological Psychology, ONR, p. 165-175). The classical apparatus used in this research is a moving endless belt behind a window, with an observer stationed in front. One of these paradoxes is that the phenomenal velocity of the belt is substantially equivalent whether the subject observes with pursuit-fixations of the moving surface or with a stationary fixation of the edge of the window (Gibson et al., AJP, 1957, 70, 64-68). In one case the retinal stimulus is a stationary inner image with a moving surrounding and in the other case it is a moving window-image with a stationary surrounding. The sensations, of course, are quite different, and there is said to be a constant error in matching velocities across the two modes of observation (the Aubert-Fleischl paradox which, however, is not easy to measure). But the perceptions of motion are essentially the same, that is, there is an invariance of perception despite varying sensations (Useful Dimensions of Sensitivity, Amer. Psychol., 1963, 18, 1-15). How can this be? Can we doubt that the stimulus for the perception of motion is retinal motion?
One resolution of this paradox has been to suggest that when a psychologist speaks of motion he should always mean relative motion displacement with respect to the “frame of reference”. Duncker’s experiments on induced motion are in line with this explanation (Psychol. Forsch., 1929, 12, 180-259). The implications is that the brain somehow responds to relative motion, not to absolute motions of elements in the field of view. This formula brings up the whole unresolved problem of the perception of relations among sensory elements and how this could be achieved.
Another resolution of the paradox, more fundamental, is to suggest that when the eyes shift from pursuit fixation to stationary fixation there actually exists at the retina some stimulus that remains invariant, i.e., constant, and that this is the information for the equivalent perceptions. This stimulus would have to be common to the two cases of window-moving-relative-to-the-surround and surround-moving-relative-to-the-window. What might this invariant be? It could be something occurring in the transformations of optical structures that correspond to the edges of the window.
One thing that occurs at an optical margin corresponding to the physical edge separating two surfaces undergoing relative motion (if the surfaces are patterned) is frequency. This optical event is not flicker precisely, but it is more like flicker than it is motion. It is invariant under size-transformations, that is, constant with different distances of the window apparatus from the eye. Smith and Sherlock (AJP, 1957, 70, 102-105) thought of it as frequency of “passing” and they used it to explain J. F. Brown’s velocity-transformation phenomenon without any reference to velocity whatsoever.
What are the optical transformations at the edges of windows (which have a surrounding front ground) and at the edges of objects (which have a surrounding background)? In figurative language, I have called the transformations at the leading and trailing edges “winking” or “flickering” (Smith’s “passing”) and I call the transformations at the lateral edges “shearing”. Consider the former. The optical events at leading and trailing edges can now be better understood since Michotte has published his experimental studies of “screening effects” (Studia Psychologia, Louvain, 1964). Considered as transformations of texture there is a sort of wiping-out of elements (which then appear to be covered). Michotte considers the forms as such in his experiments which, although screened or occluded from view, continue their phenomenal existence. This is true even though there are no sensory data to support the continued perceptions of form.
The family of transformations connected with occlusion, more exactly, those accompanying a change in occlusion deserve much more study. These kinetic edge phenomena normally result, of course, from the interception of pencils of rays in the network of pencils filling an illuminated space (MS of Perceptual Systems, ch. 9). They occur whenever the of the environment changes and whenever the observer’s viewpoint changes. These transformations normally correspond to edge-depth in the world, unlike the family to transformations that correspond to slant-depth in the world (Flock, 1964, on slant, and Gibson et al, AJP, 1955, 68, 372-385). The transformations that correspond to edge-depth seem to involve a kind of mathematical discontinuity whereas those that correspond to slant depth (perspective transformations) do not.
Next consider, for example, the stimulus that occurs when an individual moves his head from one side to the other in the presence of an “optical cliff” (Gibson and Walk, Scient. Amer., 1960, 202, 64-71). It is essentially a lateral shearing of the optical texture, a slicing into two parts along a horizontal line. This stimulus is invariant. Considered as a slippage it is unchanged despite variations in speed or direction of the upper and lower slices of texture in the optic array. It is not itself a motion, if a motion is something defined by a definite speed in a definite direction. The flow can be to right or to left, fast or slow, but the invariant continues to specify the difference in depth between the near and the far surface. Moreover the slippage of one half of the retinal image relative to the other remains the same whether the observer fixates on a detail above the edge or one below the edge and thereby shifts the retinal motion from one half of the retina to the other.
It seems to be implied, therefore, that kinetic edge effects are essential for the perception of the motion of an object in the world (including the motion of an endless belt behind a window) and are important, if not essential, for the perception of the depth at stationary edges in the world. The puzzles and anomalies that arise when one tries to establish a psychophysical correspondence between retinal notions and sensations of motion are understandable in the light of this conclusion.