Why We See What We Do
A probabilistic strategy based on past experience explains the remarkable difference between what we see and physical reality
If this general explanation is correct, then the same perceptual effect should be elicited by any stimulus in which target territories of the same luminance have typically turned out to be differently reflective objects in different amounts of light. A particularly interesting challenge is the percept generated by a more complex stimulus called the Cornsweet edge, which is named after Tom Cornsweet, the psychologist who described this effect in the late 1960s.
In the Cornsweet effect, opposing luminance gradients that meet at an edge make physically identical adjoining regions look differently bright. Specifically, the region adjacent to the lighter gradient appears brighter than the region next to the darker gradient. Because this perceptual effect is the ooposite of the effect of standard simultaneous brightness contrast, the Cornsweet stimulus provides yet another example of why explanations based on local contrast relations do not work.
Despite its seemingly complicated structure, the Cornsweet-edge effect can also be explained in empirical terms. The common denominator of the Cornsweet stimulus and conventional simultaneous-brightness-contrast stimuli is that the percepts can both be understood in terms of the possible sources of the physically identical target territories. Thus, the equiluminant regions bordering the gradients that comprise a Cornsweet edge could have been generated by similarly reflective surfaces under the same illuminant—including painted gradients on the surface of a piece of paper on which light falls uniformly—or differently reflective surfaces under different intensities of illumination—including a cube with rounded edges placed so that one side is in light and another in shadow. Since both scenarios, and a host of others, are real possibilities, the percept elicited by the stimulus will, according to a wholly probabilistic theory of vision, take all the possible sources into account in proportion to their occurrence in the past. Given that the stimulus will often have been generated by differently reflective surfaces in different illumination, such as the cube scenario, the target territories will look differently bright.
If this statistical explanation based on past experience has merit, then the perceptual effect of the Cornsweet edge should be increased, decreased or even abolished simply by altering the relative probabilities of the possible sources of the stimulus without changing the stimulus as such. As various experiments show, this is the case.