In music, you have scales. In Jiu Jitsu, it’s drilling. Most of us just call it practice. Whatever you label it, many believe that greatness, heck even mere competency, requires training a skill well past proficiency. It's continuing to practice your free throw even after you've nailed every shot. It's playing through that song one more time even though you've made no mistakes. Scientists call this training past the point of improvement ‘overlearning.’ And a recent study in Nature Neuroscience suggests that it might improve performance by altering chemicals in the brain that “lock” in training.

To understand how overlearning affects our ability to obtain a new skill, researchers exposed two groups to a series of visual perception learning exercises—basically orienting lines on a screen known as Gabor patches.

In the first group learners stopped practicing as soon as they stopped getting better. This happened usually around the eighth block of training. They then took a 30-minute break. After the break, they trained on another distinct, but similar visual learning exercise. The next day they took a post-test. In the post-test, subjects performed well on the second task—the one they learned more recently. They tanked the first task. Their results were the same as if they had never trained at all.

“In the usual situation in which you stop training on a new skill immediately after you’ve mastered it, the area of the brain related to the skill is still plastic,” said Takeo Watanabe, a professor of Cognitive, Linguistic and Psychological Sciences at Brown University and an author on the study.

Brains are flexible—they are adept at learning new tasks. What Watanabe’s research suggests is that if you stop training a skill right after you've acquired it the brain stays in its ready-to-learn state. If you then train on a second similar task while your brain is still in a plastic state, it overwrites the first skill. It becomes as though you haven't studied the first skill at all.

“This is called retrograde interference,” said Watanabe going on to note that this problem has been recognized for many years.

But in the same study a second group of subjects "overlearned." This group continued practicing past the point of competency, for eight more blocks, or sixteen blocks total. Like the previous group, after a 30-minute break they trained on a second task, and the following day they took a post-test.

Subjects who overlearned performed much better on the first trained task than those who didn’t overlearn. It turns out that spending as little as 20 extra minutes practicing a task you've already perfected leads to lasting improvements. The benefit is that the second learning doesn’t interfere with the first learning. But this benefit comes with a cost.

“The other part of the story, is that you learn that second task less well,” said Robert Goldstone. Goldstone is a distinguished professor for psychological and brain sciences at Indiana University and was not a part of this study.

The first group, the one that didn’t overlearn, performed better on the second task. But the sum of their improvements on both tasks was below that of the overlearning group. In other words, the overlearning group learned the first task much better, and learned the second task roughly half as well as the first group. The first group, though it trained on both tasks, basically only learned the second task.

To understand why Watanabe and his colleagues turned to Magnetic Resonance Spectroscopy (MRS). When it comes to brain scans, functional MRI or fMRI machines are better known. But fMRI machines measure brain function by tracking oxygen in the brain. Areas of the brain that are working hard use more oxygen, so it’s possible to connect brain function to oxygen usage. MRS machines, in contrast, track chemicals like carbon and nitrogen that are present in the brain’s neurotransmitters. They allow researchers to assess which neurotransmitters are present in the brain.

Using the MRS machine, researchers repeated the earlier experiments but with two changes. First, the two groups trained on either the usual (eight-block) condition, or the overlearning (sixteen-block) condition, but without a second training. And, before the first test and training researchers scanned the subjects’ brains in the MRS machine. Researchers also scanned the subjects’ brains 30 minutes after the training, and again 3.5 hours after the training. The post-test was again held on the second day.

What Watanabe found is that if you don’t overlearn, the brain has higher amounts of glutamate-dominate excitatory. Glutamate is a chemical that makes your brain plastic or more adept at learning. But overlearning decreases the amount of glutamate, and increases the amount of GABA, a chemical that stabilizes the brain.

“If you overlearn the skill, your brain state changes very rapidly from being plastic to being stable,” said Watanabe. Which in turn means that your brain has more time to lock in the skill preventing it from being overwritten.

If you’re a teacher and you want to make sure that your students get a foundation in a basic topic before moving onto a more complex, related topic, it may make sense to overlearn the first topic before tackling the second with the goal of revisiting the latter at a later date.

And Goldstone cautions not to put all one's eggs in the basket. Some studies suggest that the benefits gained by overlearning can erode over as little as four weeks. It might be that we have to pair overlearning with other learning techniques. For example, researchers know that spacing out learning, and mixing up topics also helps.

"I’m on board with the idea that overlearning can add fluency to your processing. It can make you respond faster, can make you respond in cases where you have dual task interference—that is when you’re doing something else, you’re tired, you’re cognitively limited," said Goldstone.

So, if you're a surgeon who performs under life-or-death pressure, or a quarterback who has to land that pass when he’s exhausted, overlearning may make sense. For the rest of us, it may be useful but most likely in conjunction with other learning techniques.