Despite the seemingly commonsensical belief that everyday experience with visual perceptions corresponds precisely with the characteristics of the 'real world,' in "Why We See What We Do: An Empirical Theory of Vision," (Sinauer Associates, 2003), Dale Purves and Beau Lotto provide detailed scientific evidence to the contrary.
Purves is the George B. Geller Professor of Neurobiology at Duke University Medical Center, and Lotto is a faculty member in the Institute of Ophthalmology at University College London.
Basically, say Purves and Lotto, their evidence demonstrates that what humans and other mammals see is a reflex response to the accumulation of possible sources that a given stimulus has turned out to be in past experience. This way of generating vision explains why visual perceptions are often at odds with physical measurements of the underlying objects -- the angles and line lengths of a simple geometrical figure, for instance.
According to Purves and Lotto, their empirical theory of vision explains a wealth of striking visual illusions of brightness, color, form, depth and motion that have puzzled vision scientists for decades, sometimes centuries. (See http://www.purveslab.net for demonstrations of these effects and their explanation in terms of the new theory.).
The authors emphasize that any successful theory of vision must address a conundrum recognized more than a century ago that visual stimuli are inevitably ambiguous.
"The physicist Hermann von Helmholtz was the first to clearly state the fundamental problem in vision, namely that there is no way to directly specify objects and conditions in the world by means of the information conveyed to eye by light," Purves said in an interview. "Even though it doesn't seem that way to us, the information carried by the light that falls on the retina is inevitably ambiguous. A particular stimulus can have many different physical sources. Since the goal of any visual animal is to react appropriately to the sources of visual stimuli, this fact presents a major problem."
For example, he said, it is impossible to know whether a given amount of light reaching the retina signifies a highly reflective surface in weak illumination, or a weakly reflective one in strong illumination. Since the amount of light and therefore the effect on the retina is the same in either case, the viewer's perception cannot be a simple "report" of the amount of light reaching the eye from the surface in question. "What we see has to correspond to what is really out there -- the shiny surface or the dull one," says Purves. "If that weren't the case, we would all be in deep and continual trouble with respect to the responses we make to retinal stimuli."
Ironically, Purves explained, this central problem identified more than a century ago has taken a back seat in the twentieth century, primarily because of rapid progress in anatomical and neural recording techniques that have proven enormously successful in determining how the nerve cells in the visual system are wired. Despite the scientific success from using these techniques, said Purves, the successes have not kept the implicit promise that understanding the detailed wiring of relevant parts of the brain will lead to a general theory of vision.
"The problem is that the quite wonderful knowledge vision scientists now have about nerve cells and the cells' connections has not been able to explain what we actually see," he said. "Especially puzzling have been the obvious discrepancies that exist between physical measurements of the real world with photometers, rulers and the like, and the corresponding perceptions."
Thus, Purves, Lotto and their colleagues set out to collect the data that would, explain the range of striking visual illusions that has long fascinated and perplexed neurobiologists, psychologists, philosophers and others curious about vision.
"The work that we began in the mid-nineties was motivated by our conviction that the problem of visual stimulus ambiguity shouldn't continue to be relegated to the back burner," said Purves. "Of course, we're not the only people since Helmholtz to have wrestled with this problem, but I think we've done a pretty good job of showing how the visual system goes about solving it, at least in general terms."
The premise of the book -- and of the many scientific papers on which the theory is based -- is that the only way the problem of stimulus ambiguity can be solved is to generate perceptions on the basis of what a given image on the retina has, in statistical terms, turned out to be in the past.
"The tutoring in this process comes from the feedback of the results of visual behavior in response to visual stimuli," said Purves. "If you make a visual mistake, you suffer the consequences," said Purves. "The connectivity of the visual brain has thus been shaped gradually by this trial-and-error process over the eons of human evolution and the decades of individual development," said Purves. "As a result, what we see at any moment is, quite literally, always predicated on the probability distribution of the possible sources of the retinal stimulus." The obvious implication of this fact is that the nuts and bolts of the visual system are going to have to be understood in these statistical terms, said Purves.
Purves also pointed out that this strategy of vision has implications for other aspects of brain function, such as the generation of auditory perceptions. As with the light reaching the retina, the sources of auditory stimuli at the ear are ambiguous. The sound pressure waves at the ear, like the light waves affecting the retina, cannot specify the physical source of the sound. Thus, understanding the sources of sound stimuli in these probabilistic terms may be just as useful in understanding what we hear as it has been in understanding what we see.
"If we are to make real progress in understanding how the human brain makes sense of the physical world," said Purves, "then visual behavior and its underlying physiology will have to be understood in terms of this empirical theory of neural function."