Brain may make mistakes on purpose so we can learn more

To study mistakes, we first had to get people to make mistakes repeatedly in the lab. To do this, we had participants put on headphones and make a series of difficult perceptual judgments about rapidly presented sounds.

Each judgment involved listening to 20 clicks presented in the space of a second. Some of the clicks occurred in the left ear and some in the right, and the participant’s job was to decide which ear received the most clicks.

Because the clicks come quickly, answering correctly is surprisingly hard and people made mistakes about 20 percent of the time. This high frequency of mistakes allows us to look for physiological changes associated with right and wrong answers and lets us probe the neural correlates of mistakes.

Our initial studies have looked at the how pupil size correlates with making mistakes. Pupil size fluctuates from moment to moment and reflects the level of the norepinephrine, a neurotransmitter that modulates arousal, in our brains.

We found that if a person’s pupils were large at the start of the clicks, they were more likely to make a mistake than when their pupils were small.

This suggests that norepinephrine modulates the frequency of mistakes, possibly by amplifying the effects of random “noise” in the brain, corrupting the decision process and leading to an incorrect answer.

Thus, one reason we make mistakes may be that the level of norepinephrine in our brains is too high. This finding suggests that by controlling norepinephrine, our brains can control how many mistakes we make. But if that is the case, why do we make any mistakes at all? One possibility is that we make mistakes in order to learn from them.

To test the idea that people make mistakes in order to learn, we had participants make a series of gambling decisions in the lab. Specifically, people chose between virtual slot machines whose payoff probabilities were unknown and that they had to learn over time.

As participants played the slot machines, they gradually learned which one paid out the most, enabling them to maximize their reward over time.

Intriguingly, even when they knew which machine was best, they would occasionally make a mistake and choose an option with lower value.

Moreover, they made more of these mistakes when they had more time left to play, suggesting that the “mistakes” were not mistakes at all, but exploratory actions that enabled them to learn more about the payoffs from all the slot machines, rather than the one that currently appeared to be best.

Consistent with this idea, we found that mistakes such as these actually improved performance in the long run, and that people that made some (but not too many) mistakes earned more rewards overall.

So the next time you flunk out on a test or take a wrong turn on the way into work, try not to sweat it. It may just be norepinephrine in your brain telling you that it wants you to learn.