December 1967
TO: Participants in the first Rochester-Cornell Conference on Vision (12/9/67)
FROM: James J. Gibson, Cornell University
SUBJECT: Ideas Worth Thinking About
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 References to these essays must cite them explicitly as unpublished manuscripts.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.
A considerable part of the discussion centered on the nature of stimulating light and on the problem of the information in light. Are the concepts of physical optics adequate for “ecological” optics (in future if not at present) and what new conceptions, if any, are required for a theory of visual perception?
1. First, how do we go from the concept of radiant or propagated energy in empty space to the concept of reverberating (multiply reflected) energy in an enclosed, or semiclosed, space? The latter clearly requires the notion of a surface layout.
a. An illuminated space might be described as a dense network of photon paths or rays after the reverberating energy has reached a steady state (one should consider the interior of an integrating sphere, then the space between two facing mirrors, and then other layout arrangements).
b. Or at another extreme one might describe the steady state network of projections of the faces and facets of the surface layout (e.g. a terrestrial layout). This would consist of solid angles, not rays.
2. Second, it was generally agreed that we need the concept of an optic array at a station point (what Gibson calls an “ambient” optic array). But it seems that this concept could take two forms.
a. The potential stimulus at a point. Boynton described what might be called the intensity array, that is, the distribution of radiant flux per unit solid angle coming to a point from each direction. The different intensities in different directions could be measured in radiometric units along the meridians of the array. These could be converted into photometric units if one wished to speak of the potential sensations of “brightness” at this station point. Can we call this a “Boynton point?”
b. The perspective projections at a point. This form of the concept assumes discontinuities of the intensity distribution in the array without any measurement of the intensities. It might be called astructural array, emphasizing the potential stimulus information at a station point instead of the potential stimulation. This kind of array can be described at various levels of decreasing simplification.
– The perspective projections to a station point of objects, faces and backgrounds, ignoring the variables of illumination, surface-texture, surface reflectance, and shadows. This is the level of perspective used in architectural drawing (outline perspective) and it is even more simplified or abstracted in the disciplines of perspective geometry and projective geometry.
– Perspective projections with consideration of the surface texture of objects, faces, and backgrounds. This is the kind of perspective exemplified by gradients of texture-density in relation to surface-slant. It ignores the amount of illumination (although the direction of illumination must be considered in dealing with faceted or rough surfaces).
– Perspective projections with further consideration of the reflectance (“color”) of the parts of the layout, and of the attached shadows. This gets very complicated, as painters and students of color constancy know.
– Perspective projections with still further consideration of cast shadows in the surface layout, specifying the direction of illumination.
Note that all these levels of perspective are more or less invariant with changes in amount of illumination.
3. Third, it was suggested that we need the concept of the ambient optic array at a moving station point with a stationary layout of surface.
a. For the array as a moving “Boynton point,” there will occur a changing overall distribution of intensities in different directions. How can one go from this analytic method of describing an array to a description of the overall progressive transformation of the structure of the array? Can the “cues for depth” be derived from it? The participants seemed to agree that all the possible stimulusinformation in an array is somehow mathematically implicit in the stimulus array of intensities and their changes, but how can the complex functions ever be determined? Is it better to resort to a non-metric or wholly projective mathematical method?
b. For the structural array at a moving station point, there will occur an overall pattern of perspective transformations of solid angle forms (the projected faces and backgrounds of objects), and a set of “accretions and deletions” of optical texture (where some surfaces occlude others). It is specific to the path of locomotion. And it contains projective invariants that are specific to the layout of surfaces. (It was suggested that a bee, for example, can directly pick up both his path of locomotion and the layout of his environment from such optical information.)
4. Fourth, we also need the concept of a stationary optic array with moving parts corresponding to the motions of objects.
a. For the intensitive array at a “Boynton point,” can the information to specify the motion of an object against its background be derived? (Note that the analytical description of the absolute displacements of certain points in the array will not suffice to specify this fact, although it would be sufficient to specify various motions of a star cluster.)
b. For the structural array, the progressive deletions (and accretions) of the elements of an optical texture on one side of a contour will specify both the background that is behind the object (superposition) and the motion of the object relative to the background. (This hypothesis was illustrated with a film.)
It was agreed by some (if not all) participants that the physiological and psychological problems of the activity of the visual system are separable from these problems of optics, i.e., of potential stimulation and information. Experiments on actual stimulation and information pickup will be discussed in a future conference.