At Meadowmount jagged bursts of notes are stretched into whale sounds. One teacher has a rule of thumb: if a passerby can recognize the song being played, it's not being practiced correctly. When camp director Owen Carman teaches a class, he spends three hours covering a single page of music. New students are surprised at the seemingly glacial pace—it's three or five times slower than they've ever gone. But when they're finished, they have learned to play the page perfectly; such a Clarissa-like feat would otherwise take them a week or two of shallower practice.

Why does slowing down work so well? The myelin model offers two reasons. First, going slow allows you to attend more closely to errors, creating a higher degree of precision with each firing—and when it comes to growing myelin, precision is everything. As football coach Tom Martinez likes to say, "It's not how fast you can do it. It's how slow you can do it correctly." Second, going slow helps the practicer to develop something even more important: a working perception of the skill's internal blueprints—the shape and rhythm of the interlocking skill circuits.

For most of the last century, many educational psychologists believed that the learning process was governed by fixed factors like IQ and developmental stages. Barry Zimmerman, a professor of psychology at City University of New York, has never been one of them. Instead, he's fascinated by the kind of learning that goes on when people observe, judge, and strategize their own performance—when they, in essence, coach themselves. Zimmerman's interest in this type of learning, known as self­regulation, led him in 2001 to undertake an experiment that sounds more like a street-magic stunt than regular science. Working with Anastasia Kitsantas of George

Mason University, Zimmerman posed a question: Is it possible to judge ability solely by the way people describe the way they practice? To take, for instance, a roomful of ballerinas of varying ability, query them about demi-plies, and then accurately pick out the best dancer, second-best dancer, third-best dancer, and so on, based not on their performance but solely on how they talked about practicing those demi-plies?

The skill Zimmerman and Kitsantas chose was a volleyball serve. They gathered a range of expert players, club players, and novices, and asked them how they approached the serve: their goals, planning, strategy choices, self-monitoring, and adaptation—twelve measures in all. Using the answers, they predicted the players' relative skill levels, then had the players execute their serve to test the accuracy of their predictions. The result? Ninety percent of the variation in skill could be accounted for by the players' answers.

"Our predictions were extremely accurate," Zimmerman said. "This showed that experts practice differently and far more strategically. When they fail, they don't blame it on luck or themselves. They have a strategy they can fix."

In other words, the volleyball experts are like de Groot's T. Rex chess players. Through practice, they had developed something more important than mere skill; they'd grown a detailed conceptual understanding that allowed them to control and adapt their performance, to fix problems, and to customize their circuits to new situations. They were thinking in chunks and had built those chunks into a private language of skill.