A new theory of how people see proposes that past perceptions, as well as evolution, influence the recognition of visual stimuli, perhaps replacing the accepted theory that perceptions correspond exactly to the real world.
Dr. Dale Purves, chair and George B. Geller professor of neurobiology, co-authored Why We See What We Do: An Empirical Theory of Vision with Beau Lotto-a faculty member of the Institute of Ophthalmology at University College London. The project is a culmination of the research the two have conducted for the past eight years.
Purves' and Lotto's theory is based on probability distribution--a statistical theory that helps state that, although one object can produce many stimuli, some stimuli are more likely to occur based on the past perceptions one has experienced. The perception of vision, according to the new research, is a reflex response to the summation of all previously experienced scenes.
"You see in terms of the statistics of things you've seen before... and unusual scenes are interpreted to the best of your ability," Purves said. He added that, in a sense, everything one sees is "new" because no two encountered scenes are ever similar.
If one does experience a "radically novel thing," the brain's best response will be to interpret it based on the most related past experiences, however removed they may be, Purves said.
This idea is the latest theory advanced in an effort to address the mystery of how light falls unidimensionally and ambiguously on the eye's retina but is nevertheless translated into a three-dimensional picture.
Purves' team studied the way people perceived objects in relation to what was really "out there," which was determined by using an objective measuring device. The best way to validate this theory, he said, is to obtain information about the world by using technology--through brain-imaging studies.
Jim Voyvodic, assistant research professor in the Department of Radiology's Brain Imaging and Analysis Center, is collaborating with Purves in an effort to prove his theory.
Voyvodic said that Purves is not interested in where brain activity is occurring, but rather what is happening. "It's likely to mean that, as the brain accumulates information through experiences, it's modifying the 'this is here and that's there' theory, although it's not contradicting it," Voyvodic said.
Voyvodic and Purves are, in essence, attempting to elucidate how the brain makes sense of three-dimensional objects because the brain encodes the source of the vision, not its properties.
"What we view is the net of activation over a lot of regions," Voyvodic said, emphasizing again that the "where" was not as important because it is difficult to know exactly the activation over a bundle of areas.
As this research continues, Purves is diversifying his own research to extend it to the auditory system, studying how people respond to music. The implications of this new avenue of research extend out to many other fields, including artificial intelligence. An explanation of the way in which vision actually works--which could aid computer scientists in constructing robots to see the way humans do--hinges on the successes of brain-imaging studies.
"My hope is that the probability distribution and the three-dimensional perception are very closely related," Voyvodic said.
Get The Chronicle straight to your inbox
Signup for our weekly newsletter. Cancel at any time.