Researchers use brain scans to predict behavior

By peering into the minds of volunteers preparing to play a brief visual game, neuroscientists at Washington University School of Medicine in St. Louis have found they can predict whether the volunteers will succeed or fail at the game.

"Before we present the task, we can use brain activity to predict with about 70 percent accuracy whether the subject will give a correct or an incorrect response," says lead author Ayelet Sapir, Ph.D., a postdoctoral research associate in neurology. Eleven seconds before volunteers played the game -- discriminating the direction of a field of moving dots -- scientists showed them a hint: an arrow pointing to where the moving dots were likely to appear. The dots were visible only for one-fifth of a second and therefore were easy to miss if a subject was not paying attention to the right area.

After the hint and prior to the appearance of the moving dots, researchers scanned the volunteers with functional brain imaging, which reveals increases in blood flow to different brain areas indicative of increased activity in those regions. Based on brain activity patterns that reflected whether the subjects used the hint or not, scientists found they could frequently predict whether a volunteer's response would be right or wrong before the volunteers even had a chance to try to see the dots.

Results are published online in the Proceedings of the National Academy of the Sciences and will appear in the journal's print edition on Dec. 6.

Sapir and her colleagues concluded that volunteers don't use the hint the same way every trial. One speculation was that some of the brain signals they detected might be signs of the brain's struggle to cope with an ambiguity built into the test: the volunteers knew the hint was only accurate 80 percent of the time.

"Whether the hint is accurate or not was determined by the computer's random number generator, and the volunteers were not going to be able to beat that," says coauthor Giovanni d'Avossa, M.D., an instructor in neurology. "But regardless of how hopeless it was to try to outguess the computer, some of our data suggest that the brain may still have been trying to do just that: to figure out a formula or a rule based upon which it could predict whether a hint was valid and should be trusted."

Researchers based this speculation on a spike in brain activity partially found in the rewards system in the frontal lobes. "The rewards system is involved in regulating behavior based on previous experiences of rewards and punishments," d'Avossa says. "It also may help us build up predictions of what the world should be like and how certain events go together. When it works well, the world makes sense to you."

Sapir noted that the reward systems' predictive abilities may be damaged or missing in some patients with mental illness, causing these patients to perceive the world as alien and unpredictable.

Other areas involved in prediction included regions in the visual cortex involved in the analysis of the motion display and regions involved in the control of visual attention.

"These activations may reflect the degree to which subjects variably directed attention on each trial to the location of the stimulus prior to its presentation," says Maurizio Corbetta, M.D., the Norman J. Stupp Professor of Neurology and senior author of the study. Regardless of how the results are interpreted, Corbetta notes, the study clearly showed that visual perception not only depends on the quality of sensory signals but also on the variability of internal signals.

Sapir A, d'Avossa G, McAvoy M, Shulman GL, Corbetta M. Brain signals for spatial attention predict performance in a motion discrimination task. Proceedings of the National Academy of the Sciences, Dec. 6, 2005.

Source : Washington University School of Medicine