The Perception of Surface Layout: A Classification of Types
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 References to these essays must cite them explicitly as unpublished manuscripts.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.
It was suggested in 1950 (Visual World, p. 8) that the fundamental types of space perception were those of “surface” and “edge.” But the term “edge” has several meanings, and they need to be distinguished. A “surface,” moreover, can approach the state of transparency and become invisible, or it can be mirror-like so as to yield “virtual” or illusory perception. I meant an opaque, scatter-reflecting surface which is visible as such.
We do not yet fully understand the conditions for the visibility of surfaces; the implications of the Ganzfeld experiments are not agreed upon. Nevertheless it is useful to theorize that space perception is not one problem, how we see the third dimension, but a set of problems connected with the perception of surface-layout. In order to disentangle the problems from one another we need a classification of the types or forms that layout may take. It might seem that solid geometry would provide such a classification, but the familiar extension of plane geometry into a third dimension will not serve, for it does not take account of opaque planes. We shall have to use an opaque solid geometry.
The problem of perceiving the layout of opaque surfaces is different from that of perceiving the layout of transparent surfaces inasmuch as the former conceal or hide whatever is “behind” them. This fact will be called occlusion. The occluding of one surface by another depends, of course, on the point of observation. At any given point of observation, some opaque environmental surfaces will be projected and others will be unprojected, that is, some will be “uncovered” and other “covered.” This fact is characteristic of ecological space as distinguished from abstract geometrical space.
It is assumed that the perception of layout in the space of a picture presents a special subset of problems in layout perception. A picture is a frozen sample of an optic array, it usually has a “frame,” and its own surface is usually visible, so that the problem of depth perception is complicated by ambiguities and discrepancies of information. Pictures are not here considered.
The question is how can the fundamental forms of environmental layout be classified? One might make a classification without reference to a point of observation and then another with reference to a point of observation. I shall attempt both. What follows is tentative.
I. Geometrical Layouts
What are the combining forms of ecological space, the perception of which needs to be understood? We have too long been preoccupied with the points, lines, and planes plus the “forms” (planar and solid) of abstract geometry.
1. The planar, the convex, and the concave surface. A surface may depart from planarity by a positive or a negative dihedral angle (convex or concave). Either type has an apex, that is, a “corner,” or “an edge.” A convex dihedral angle that is acute is called a “sharp” edge; if obtuse it is a “dull” edge.
A surface may also depart from planarity by a positive or negative curvature. A curve has no apex. Curvature on one dimension only yields a “ridge” or a “valley.” Curvature in both dimensions yields a “protuberance” or an ” indentation.” Opposite curvature on the two dimensions yields a “saddle.”
2. The closed continuous surface. This, I suggest, should be called an “object.” Its surfaces may have any of the above forms of layout. If composed of dihedral angles, the planes may be called “faces,” and the object is then a polyhedral.
3. The elongated object. When a closed surface is elongated in two dimensions relative to the third it tends to become a “sheet.” When it is elongated in one dimension relative to the other two it tends to become a “rod,” “stick,” “wire,” or “fiber.” The abstract plane is a sort of vanishing ghost of a sheet and the abstract line is a ghost of a wire.
4. The junction of two surfaces. When two surfaces are immediately adjacent there will exist a “crack.” If separated, there will be a “gap.” Note that a “crack” will cause a line in an optic array, and that this is not the same as a contour.
II. Layout with reference to a point of observation
We are now prepared to bring in the fact of occlusion. Except for the bare interior of a windowless enclosure, or perhaps the extended plane of a featureless earth (but even there a horizon will exist) some parts of the total surface layout of the environment will hide other parts, and this will depend on the point of observation. The “back” of a convexity and the “front” of a convexity may be hidden; the “back” of an object is always hidden and a detached object always hides the “background.” But note that there is always some point of observation at which any piece of surface is unhidden. For describing the forms, I shall borrow the term “edge” from opaque solid geometry. For occlusion there must always be an occluding edge. But note that this need not be the apex of a dihedral angle. What are the combining forms of layout with respect to possible observation?
1. The apical occluding edge. Examples of this are the corner of a corridor, the overhang of a roof, and the brink of a cliff. In the diagram below, the occluding surface is represented by a dashed line and the non-occluded surface by a solid line. The point of observation is indicated, but it stands for a set of such points. The diagram may have any size or orientation.
2. The curved occluding edge. An example is the “brow” of a hill. Note that the curved occluding edge moves along the surface as the point of observation moves since, in the diagram, it is the tangent to the curve. In both kinds of occluding edge there is an abrupt contrast in the optic array, a contour, separating the optical textures of the hidden surface from that of the unhidden surface, but in the latter kind there is no geometrical edge in the world.
3. The front-back occluding edge. This is the separation between what we call the “front side” of an object and the “back side.” The edge is apical for a polyhedral object but curved for a curved object. The fact is that the back of an object is perceived in some meaning fo the term, and this is a paradox for sensation-based theories of perception.
4. The limitiing case intermediate between an occluding edge and a dihedral angle. In the special case diagrammed below, there is depth at the contour in the array (one surface is farther than the other) but it is ambiguous whether or not one surface is occluded or extends behind the other).
5. The occluding of a background by an object. We now introduce two forms of layout that involve a continuous background surface. The commonest type is an object that occludes this background, which results in a closed contour in the optic array. This yields the familiar figure-ground phenomenon. Phenomenally, the ground seems to extend continuously behind the figure. there is an increase of apparent depth outside the contour and the contour is said to “belond” to the inside, not the outside. These phenomenal facts can be noticed even in a frozen picture, and much has been made of them. But the figure – ground phenomenon is not the unique prototype of visual perception that Gestalt theory has assumed.
6. The occluding of background in the case of a window (aperture). The background may be occluded by a surface with a hole in it, as well as by an object, although less commonly. This also results in a closed contour in the optic array. But the occluding edge of a hole is the opposite of the occluding edge of an object. When the edge is detected, there is an increase of apparent depth inside the contour. The background seems to extend continuously outside of the contour, and the contour “belongs” to the outside, not the inside.
The implication is that the direction of depth at a contour is univocally determined not by “closure” of the contour but by the stimulus information for occlusion, that is, the information to specify which side of the contour is an occluding edge. This information is more or less equivocal when the optic array comes from a drawing, or is frozen in time. The best information comes from progressive change of occlusion, as evidenced by the motion picture studies of this phenomenon (Gibson et al., 1969; Kaplan, 1969). Perhaps there is equally good information from binocular disparity of occlusion at an edge, but this is a new kind of binocular disparity and it has not yet been studied. However, the experiments of Julesz might be interpreted in this way. Evidently, the term “depth” is vague and misleading; there are many kinds of depth perception that need to be sorted out. Even the term “edge” is ambiguous, and the effort has here been made to distinguish several kinds of edges that are perceptible and to distinguish these, in turn, from “contour” and “line.” There is little use in postulating “edge – detectors” in visual perception unless we know what there is to be detected. It should be kept in mind that these “forms” of the occluding edge, together with their surfaces may enter into a limitless set of combinations. All surfaces, I assume, have physical texture and this is layout on a microscopic level. Are these the “fundamental” forms of layout?