Perception Of Color Contrast, Constancy Depends On Neural "Reflexes," Says Vision Theory

In an article in the Nov. 7 issue of Proceedings of the National Academy of Sciences, neurobiologists Beau Lotto and Dale Purves present experiments showing that the visual system perceives color based not on the light that actually reaches the eye, but on the reflectances and illuminances that usually would have generated the stimulus in the past.

The study reported in the PNAS is based on the well-known illusion that the color of a surface can appear quite different, depending on its context.

"The perception of color has fascinated scientists from Isaac Newton to Edwin Land [inventor of the Polaroid camera," said Purves. "A basic problem, recognized early on in studies of color, is that the physical properties of light stimuli do not always give rise to expected color sensations. Thus, a patch that would look purplish on a white background looks distinctly bluish on a purple background; this illusion is called color contrast. In the same vein, different stimuli can elicit the same color when viewed in different circumstances, an illusion called color constancy."

In experiments that explore the basis of color contrast and constancy, Lotto and Purves presented volunteer subjects with images displayed on a computer screen in which pairs of colored targets were shown in the context of two differently colored surroundings. The subjects were then asked to use onscreen "buttons" to adjust the apparent color of one target until it matched the perceived color of the other target.

By measuring the subjects' adjustments, the researchers could determine how the colors of the surroundings affected perceptions of the target color. They found that by changing the context so that the identical targets were more likely to have been generated by differently reflective objects under different illuminants, the perceived color difference between the physically identical targets was increased. However, if the context was changed such that the targets were more likely to signify similar objects under similar illumination, the perceived color difference between the targets decreased.

"This finding is what one would expect if the illusion of color contrast represents the experience of the visual system with the physical laws that govern how reflectance and illumination combine in generating the ambiguous light that hits the eye," Purves said.

The rationale for this kind of explanation, according to Lotto and Purves, is that the light that enters the eye does not carry unambiguous data about the visual world.

"Since photons don't carry any information about their history, there's no direct way to disentangle what has actually given rise to the light falling on the retina," Purves said. "The light that reaches the eye is always a product of both the quality of the object's surface and the quality of its illumination. So it's impossible to know the extent to which the stimulus coming from an object is determined by the object's reflectance properties or the conditions of its illumination.

"Therefore, the basic problem in vision is that the meaning of a light stimulus is inevitably ambiguous," he said.

According to the argument presented in the PNAS paper, the only way observers can sort out the "meaning" – or more appropriately, the behavioral significance – of such ambiguous stimuli is to use trial and error to indicate what they should perceive. Thus, humans gradually evolved the reflexive visual circuitry that enables them to see, not the actual properties of the light falling on the retina, but the sources in the physical world that typically would have generated that type of stimulus in the past.

Despite its simplicity, this wholly empirical theory remains highly controversial, Purves said. "The idea of empirical influences on vision has been around for a long time, but to a large extent has been ignored by people doing neurobiology, or considered as a way of modulating basic visual processing machinery under the rubric of ‘top-down' influences on ‘bottom-up' mechanisms," said Purves. "The reason is that neurobiologists have made great progress using conventional techniques of anatomy and physiology to map the brain's visual circuitry, and this work has proven enormously valuable.

"Given that kind of success, you really don't need to think about whether some arcane aspects of visual perception provide a better framework for understanding vision," he said. "The peculiar phenomena we have been concerned with can easily be ignored as anomalies that don't have much importance.

"The problem is that, after 50 years of work, neurobiologists still can't explain in terms of visual circuitry any aspect of visual perception, no matter how simple," Purves said. "This evidence about how color is perceived, together with what we have recently discovered about the perception of luminance [black and white stimuli], orientation and motion argue pretty strongly that all these visual percepts arise on the same empirical footing. This way of thinking rationalizes a long list of phenomena that are otherwise very hard to explain."

In related papers published over the past three years, Purves and his colleagues have reported other experiments using visual illusions to explore perception of brightness, shading, color and geometry (See Web site at http://www.adm.duke.edu/alumni/purves for examples).

According to the neurobiologists, the empirical theory of vision, if supported by further experiments, could yield a far more productive approach to understanding the brain, as well as practical applications such as computers that more realistically employ the strategies used by the brain. "The visual part of the brain – and presumably the rest of the brain as well – seems to work like a computer that doesn't ‘understand' the rules of the game it's playing," said Purves. "Nevertheless, it's already clear a computer can develop a pretty good game of, say, checkers simply by changing its connectivity according to feedback about the moves that worked well in the past. Getting computers to see things, however, has been a difficult problem that remains largely unsolved,"said Purves.

"Humans have had millions of years as a species, and a fair number of years as individuals, to perfect the neural networks triggered by visual stimuli. This very long process of shaping connections by trial and error is apparently what has made us so good at visualizing the world we live in, even though we never really see what's there."