A pattern of luminances equivalent to that of a traditional simultaneous lightness display (two equal gray squares, one on a white background and the other on an adjacent black background) was presented to observers under two conditions, and matches were obtained for both perceived reflectance and perceived illumination level of the squares and their backgrounds. In one condition, the edge dividing the two backgrounds was made to appear as the boundary between a white and a black surface, as in the traditional pattern. The squares then were perceived as almost the same shade of middle gray. In the other condition, a context was supplied that made the edge between the backgrounds appear as the boundary between two illumination levels, causing one square to appear black and the other white. These results were interpreted as a problem for local ratio theories, local edge theories, and lateral inhibition explanations of lightness constancy, but as support for the concepts of edge classification, edge integration, and the retinal image as a dual image.Theories that appear different or even opposite may nonetheless share a common starting assumption, and the preoccupation with such apparent differences may serve to obscure flaws in the common assumption. Believing this to be true for theories of lightness perception, we have chosen an alternative, and somewhat radical, starting point for the experiments we report here. This approach will be more easily grasped if we begin by reviewing the central role of luminance information in current theories of lightness perception.The Photometer Metaphor: Absolute Lumiuauce Levels as the Basic Iuput Implicit in most theory and research on the perception of surface lightness is a conception we would call the photometer metaphor. This term refers to the assumption that, fundamentally, the visual system measures the intensity of light reflected by each point in a visual scene. Helmholtz and Hering, although their views were usually in opposition, both assumed that absolute intensity of reflected light (luminance) was available as an input to the visual system. Of course, neither their theories of lightness perception nor the more modem theories are

Observers looked into a miniature room in which everything was painted matte white, or--in another room--matte black. They made both reflectance and illumination judgments for eight test spots. The test spots (which varied in luminance) were perceived as approximately equal in reflectance--not different, as conventional contrast theories would seem to require. The illumination matches made to the same points, however, closely paralleled the pattern of actual illumination levels, and this result is discussed as evidence that edges are classified as changes in either reflectance or illumination. The white room was correctly perceived as white, and the black room was perceived as middle gray; similar results were obtained even when the luminances in the black room were higher (owing to higher illumination) than the corresponding luminances in the white room. An explanation in terms of differences in gradient patterns is presented and supported with luminance profiles.

A pattern of five gray squares ranging from white to black was presented to observers at four levels of illumination, spanning a range of six log units. This replicated an earlier experiment by Jameson and Hurvich (1961) in which a 1.1-log-unit range was used. Three measures of perception were used: (1) a "lightness" measure consisting of a square of variable luminance surrounded by a bright white field (after Jameson & Hurvich), (2) a Munsell chart, and (3) a "brightness" measure consisting of a square of variable luminance surrounded by complete darkness (after Heinemann, 1955; Leibowitz, Mote, & Thurlow, 1953;and Leibowitz, Myers, & Chinetti, 1955). The first two measures yielded the same results-a very high degree of constancy over the entire range. No diverging or negative functions were found. The brightness measure yielded almost no constancy, but did yield approximate luminance matching. It is argued that these results, together with those of three other published studies, indicate that the concept of intensity dependence is not valid. It is also suggested that the term "brightness constancy" is a misnomer, since brightness varies with illumination.In 1948, Wallach published a now classic set of experiments in which he showed that the perceived shade of gray of a surface depends not on its luminance (the absolute amount of light it reflects), but on the ratio between its luminance and the luminance of the surrounding region. He showed that a disk of constant luminance could be made to appear as any shade between white and black simply by varying the luminance of a surrounding annulus. In addition, he offered quantitative results showing that when observers are presented with two such disk! annulus displays on opposite sides of a darkened room, and asked to adjust the luminance of one disk (the comparison) until it appears the same shade of gray as the other (standard) disk, they will actually set the luminance of the comparison disk to almost the same ratio with the luminance of its annulus as that of the standard disk in relation to its annulus, even though this may require that the luminance of the comparison disk be set as much as eight times that of the standard.These experiments were important because they offered a simple explanation of lightness constancy under changing illumination that bypassed Helmholtz's (1867/1962) less operational cognitive account. They further suggested the radical possibility that the visual system might have no need at all for absolute luminance levels-that lightness perception might be determined rather simply by relative amounts of light.

An early experiment by Hess and Pretori (1894) was replicated and modified in an attempt to determine why they failed to find the ratio principle later discovered by Wallach (1948). Separating the two surround-infield patterns by darkness made very little difference. However, allowing the observer to adjust the infield luminance (as in Wallach) rather than the surround luminance (as in Hess & Pretori) revealed some startling effects. At surround:infield luminance ratios greater than approximately 100:1, there is no ratio effect; all infields appear equal and totally dark. Converging evidence is presented that Hess and Pretori's data in this region actually represent surround-matching by the observers. Nor are ratio effects found with increments (infield brighter than surround). When free to match either infield luminances or ratios (by controlling infield luminance), observers match luminances. For decrements with ratios between 1:1 and approximately 100:1, lightness constancy and the ratio principle hold.One of the most extensive experiments in simultaneous contrast was published by Hering's students Hess and Pretori (1894/1970). In this early study, which calls to mind the more modem work by Wallach (1948), observers were presented with two adjacent infield-andsurround patterns, each pattern consisting of a 1.1 0 infield square surrounded by a 11.40 square background region. Each of the four regions was a white surface oriented at a 45 0 angle to the observer's line of sight so as to reflect light from a petroleum lamp that moved along a blackened tunnel at right angles to the observer's line of sight (see diagram in Evans, 1948, p. 165). This elegant method allowed good and independent control of, and a valid determination of, the four luminances at a time that predated photometers. The actual slanted position of each of the four regions was not perceived, since each was visible only through a mask containing an aperture that projected a rectangular region to the viewpiont of the observer, with the surround region serving as the mask for the infield.The infield-surround pattern on the left side was used primarily as a standard, and, for a given series of matches, the luminances of the left-hand infield and the left-hand surround were held constant. Within such a series, the right-hand infield was set at different luminances, and, for each, the observer was required to adjust the righthand surround luminance until the two infields appeared equal in brightness.Correspondence may be addressed to A. Gilchrist at the Department of Psychology, Rutgers University, Newark, NJ 07102. 7In all, the Hess and Pretori study used luminances ranging from 0 to 5,000 units (one unit was equal to 0.12 Hefner-Altenech units, or 0.01 tL), and tested luminance ratios from less than 1:5,000 to 256:1.Hess and Pretori's results did not show, as Wallach showed later under somewhat different conditions, that a given luminance ratio on the standard side was matched by the same ratio on the comparison side. What they did show, however, ...