Whoever said that we learn from our mistakes made a mistake. Albert Einstein said it, Winston Churchill said it, and they got it wrong. It turns out that we are better at learning after doing something right rather than after doing something wrong.
Researchers from MIT have shown for the first time that the brain learns more after a success than a failure. This study indicates, contrary to previous research, that neurons in the brain are able to keep a memory of recent success and failures during learning and performed better after doing it right than after doing it wrong. The researchers found that after monkeys did something correctly, there were prolonged neural signals, which continued to fire until the next action, therefore affecting subsequent neural response. But after a wrong action there was less neural activity and no improvement in further attempts.
“The findings suggest that learning may not require the changing of connections of neurons on each trial, as several other studies have suggested, and instead suggest that information about outcomes on each trial are held in a sort of buffer for guidance in the next attempt,” says Professor Howard Eichenbaum from the Centre for Memory and Brain at Boston University.
So, it turns out our history, piano and tennis teachers had it right all along, practice does indeed makes perfect. If you can get something right repeatedly, you're likely to keep getting it more right. In other words, “perfect practice makes perfect,” says Mark Histed, one of the co-authors on the paper published in Neuron.
Although scientists have long identified that the prefrontal cortex in the brain is associated with learning, researchers have had very little idea about what are the exact neural mechanisms that convert environmental feedback, our successes and failures, into the brains ability to learn and respond to experience.
Even though this study was done in monkeys, Earl Miller, the lead investigator on the paper and professor of neuroscience at MIT, points out that “the brains of humans and animals are essentially identical in these fundamental functions, so it is very likely that the exact same thing is going on in human brains.”
In their study, monkeys were rewarded after shifting their gaze to the right or left after being shown specific pictures on a computer screen. The monkeys were able to use trial and error to work out which images were correctly associated with which movements. The investigators found that when the monkeys made the correct movement in accordance with the correct picture some neurons started to fire for several seconds until the next action was made. “We found a set of neurons that maintains a memory of whether what the animals did in the past was correct or not,” Histed. These prolonged signals carried information into the next action influencing subsequent neural responses.
“Normally rewards following a ‘successful’ action indicating it was correct occur a short time after it, but a long time before the next,” said Professor John Stein at Oxford University. “These results show, contrary to previous ideas, that single nerve cells in the caudate nucleus and prefrontal cortex, with which the caudate is intimately connected, remain active for several seconds after the reward, probably contributing to medium-term learning from experience. These longer lasting responses could also increase synaptic strengths and thus underlie even longer term learning.”
Our study found that “it is reward, rather than its absence, that is driving learning,” said Miller, and it is also clear that there are different types of learning. Stein indicates that these findings helps us understand why repetitive learning is so successful in humans, and honing in on which parameters favour this type of learning could be beneficial in terms of teaching, training, and education.
The prefrontal cortex and basal ganglia are the areas of the brain that are typically impaired in diseases such as schizophrenia, Parkinson’s disease, and autism. “Understanding how the brain operates normally is a key step to finding therapies for these disorders,” Miller says. The authors hope that this study will further better understanding the role of dopamine, a neurotransmitter associated with reward and pleasure, in learning, and incidentally is also associated with gambling and drug addiction.
“Gradual trial-and-error learning is common across all species, and is fundamental to many types of learning that confront animals and humans, including skill learning and perceptual classification,” Stein said. Miller thinks that this mechanism may have evolved because success is more informative than a failure, “so the brain has taken advantage of that.”
This article was originally published by WIRED UK
