The brain is more likely to learn when the ‘unexpected’ happens
A daily commute to and from work can be a monotonous experience. Little ever changes, and most long drives down the interstate are largely forgettable.
Until, of course, the day you get in a wreck.
Not much is known about the role the human brain’s sensitivity to unexpected outcomes plays in our ability to learn. But now, for the first time, a Penn study has shown the brain is more likely to learn when the unexpected happens than when things are predictable.
“When you are driving, the brain is not going to encode the presence of the road over and over again—you learn to expect the road will be there,” explains lead author Kareem Zaghloul, a postdoctoral fellow in the Department of Neurosurgery at Penn Medicine. “But if a deer jumps out, the brain is wired to encode that new information.”
The study, published in the journal Science on March 13, is the first to directly record neural activity underlying this learning process in humans and has given scientists a greater understanding of fundamental cognitive activity.
To attempt to understand when observed feedback differed from expected feedback, researchers used microelectrode recordings to monitor brain-wave activity in 10 Parkinson’s patients (six men and four women) undergoing deep brain stimulation (DBS) surgery. Specifically, researchers set out to follow neuronal activity in the substantia nigra, a midbrain structure that plays an important role in movement, as well as reward and addiction.
For the experiment, participants were asked to select between a red and blue deck of cards presented on a computer screen, one of which carried a higher probability of a virtual reward, or payout. If the participant made a lucky pick, a stack of gold coins was displayed along with an audible ring of a cash register and a counter showing the accumulated earnings. If the draw did not yield a reward, or if no choice was made within the five-second time limit, the screen turned blank and participants heard a buzz. The idea was for participants to learn which deck had the better odds of paying out.
Data collected showed neurons in the SN structure “exhibited significantly higher firing rates after unexpected gains than unexpected losses.” But no such differences were seen after expected wins or losses. The findings confirm previous animal research that has shown this area of the brain performs a critical function in reinforcement learning.
“Our findings should serve as a point of validation for animal models of reward learning,” the authors write.
The study brought together researchers specializing in the fields of psychology, neurosurgery and bioengineering. Neurosurgeons provided the opportunity to record neural activity in the inner brain. Michael J. Kahana, senior author and professor of psychology at Penn, and Gordon Baltuch, associate professor of neurosurgery, helped construct the experiment and analyze the data.
“In order to get at these complicated questions, you need these different disciplines” working together, Zaghloul says.
Supported by grants from the National Institutes of Health, the Conte Center and the Dana Foundation, the study could lead to more research to better understand the mechanisms that regulate reward learning.
“If these structures are involved in other psychiatric disorders, potentially we could begin to think of new therapies,” Zaghloul says.
Originally published on March 26, 2009
