Figure 1 / The Cornsweet edge effect. A) Diagram of the spinning disk used by Cornsweet to demonstrate that when two equiluminant regions are separated by an edge comprising a pair of oppositely disposed luminance gradients, the adjoining territories are 'filled-in' by different lightness/brightness values. B) Standard presentation of the Cornsweet stimulus, shown here as a blowup of a portion of the rotating disk (for the sake of simplicity, the edge in the blowup has been straightened). C) Comparison of the photometric and perceptual profiles of the stimulus in (B). Despite the equal luminances of the territories adjoining the two gradients, the territory (1) to the left of the dark gradient (2) looks darker than the territory (4) to the right of the light gradient (3). D) This effect is abolished by covering up the opposing luminance gradients, as indicated here. (After Purves et al., 1999)

Figure 2 / Possible sources of the Cornsweet stimulus. The luminance gradients in the standard Cornsweet stimulus (or anyplace else) could arise from gradual changes in surface reflectance adjacent to territories having the same reflectance properties and observed under the same illuminant (top panel), or from gradual changes in illumination of two surfaces that have different reflectance properties and are under different illuminants (bottom panel), or any combination of these possibilities. (Although the illuminated side of the darker cube and the shadowed side of the lighter one in the bottom panel look differently bright, they are actually equiluminant; the perceptual effects under discussion here can't be avoided, even in a didactic illustration like this). The statistical significance of these different possible sources of the Cornsweet stimulus is that equiluminant territories adjoining gradients of reflectance will typically have been similarly reflective surfaces under the same illuminant, whereas territories adjoining gradients of illumination will typically have been differently reflective surfaces under different amounts of illumination. (After Purves et al., 1999)

Figure 3 / Diminishing the Cornsweet effect by information that increases the probability that the Cornsweet gradients signify variations in reflectance. Panel (A) shows the standard stimulus presented such that the equiluminant territories adjoining the gradients now extend around the Cornsweet stimulus as such (see dotted outline in [B]). In this case, observers see the territories adjoining the Cornsweet edge as having about the same brightness. Panel (C) shows a perspective view to indicate the source that is made more likely by this presentation of the stimulus (i.e., a flat surface with the territories adjoining the Cornsweet edge receiving the same amount of illumination). (After Purves et al., 1999)

Figure 4 / Enhancing the Cornsweet effect by information that increases the probability that the Cornsweet gradients signify variations in illumination. A) The stimulus here is presented in perspective, i.e., the way an object extending away from the observer would appear as a result of the 3-D to 2-D transformation required by projection onto the retina. The addition of perspective enhances the difference in brightness between the two territories compared to the standard presentation. Inset shows the source of the stimulus that is made more likely by this presentation. As in all the other manipulations shown here, the addition of perspective only alters the probability distribution of the possible stimulus sources, since the two territories could still lie in the same plane if their shapes implied a diminution of size with distance only coincidentally. B and C) Given that illumination usually comes from above, a similar enhancement occurs when the Cornsweet stimulus is rotated so that the dark gradient is above the light gradient, as in (B). Conversely, placing the stimulus in the opposite configuration makes it more likely that the source is similarly reflective surfaces under the same illuminant. As a result, the Cornsweet effect in (C) is diminished in comparison with the effect in (B), even though the physical structure of the stimuli are identical. (After Purves et al., 1999)