In kindergarten, several of my friends and I were very serious about learning to tie our shoes. I remember sitting on the edge of the playground, looping laces into bunny ears and twisting them into a knot over and over again until I had it just right. A few years later, whistling became my new challenge. On the car ride to school or walking between classes, I puckered my lips and blew, shifting my tongue like rudder to direct the air. Finally, after weeks of nothing but tuneless wooshing, I whistled my first note.

Although I had no inkling of it at the time, my persistence rewired my brain. Just about everything we do modifies connections between brain cells—learning and memory are dependent on this flexibility. When we improve a skill through practice, we strengthen connections between neurons involved in that skill. In a recent study, scientists peeked into the brains of living mice as the rodents learned some new tricks. Mice who repeated the same task day after day grew more clusters of mushroomlike appendages on their neurons than mice who divided their attention among different tasks. In essence, the scientists observed a physical trace of practice in the brain.

Yi Zuo of the University of California, Santa Cruz, and her colleagues studied how neurons changed in the brains of three groups of mice that learned different kinds of behaviors over four days, as well as a fourth group of mice that went about business as usual, learning nothing new. Of the three learning groups, the first practiced the same task each day, learning how to stretch their paws through gaps in a Plexiglass cage to get a tasty seed just within reach. The second group practiced two tasks: reaching for a seed and learning how to eat slippery bits of capellini, a very thin pasta. Each day mice in the third group played in a cage outfitted with a different set of toys, such as ropes, ladders and mesh on which to scamper and climb.

After each day's training, Zuo and her colleagues took the mice aside, anesthetized them and used a dental drill to cut a tiny window in their skulls, through which they could see a piece of the motor cortex—a band of brain tissue that regulates movement. The mice in the experiment had been genetically engineered to express a protein that glows yellow in infrared light. Zuo and her teammates examined the radiant neurons with an incredibly powerful two-photon microscope, searching specifically for tiny cellular structures called dendritic spines. Neurons receive signals from neighbors through a crown of thin branches called dendrites, off of which sprout even tinier dendritic spines, which are often mushroom-shaped with bulbous heads and thin necks. Dendritic spines are highly dynamic, springing up or disappearing to strengthen or weaken the connections between neurons.

Both the mice that learned new tricks and those that went about business as usual grew new spines on their neurons, but mice that were learning something grew far more clusters of spines made of two or more neighboring offshoots. Mice that repeatedly practiced reaching for a seed during the four-day period grew more clusters of dendritic spines than the mice who practiced two different tasks or those that played in the variously equipped cages. Further, spines that grew in clusters had a much higher chance of surviving, even four months later, than new spines that appeared on their own.

With their microscope, Zuo and her colleagues often observed one spine pop out of a dendrite on the first day of training and another spine pop up near it a few days later. In more than half the clusters, the first spine grew on the first training day and the second joined it by the fourth, and nearly all of the clusters in all the learning mice grew between the first and fourth days. These observations suggest that the clusters are one example of how practice physically manifests itself in the brain. The findings appear in the March issue of Nature. (Scientific American is part of Nature Publishing Group.)

"I think it is a very active process," Zuo says. "The neurons work very hard to form clusters, to place spines close to one another. Even after a short training period on the first day, a mouse makes of a lot of new spines—they might make double what they make in an ordinary day, but these spines are not clustered. Only after repeated training are they clustered." Clusters of dendritic spines on a single neuron could strengthen the connection between two neurons involved in the practiced task, but the researchers have not experimentally shown this yet—that is what they want to do next.

Zuo says such specific clustering has not often been observed or well studied, but this is not the first time that scientists have observed neurons growing new dendritic spines as part of learning—although in vivo studies, like this one, are much rarer than studies of brain cells in petri dishes. In earlier work Zuo and her colleagues have shown that dendritic spines can pop into existence incredibly quickly—within an hour of a training session. Researchers have debated whether it is the number of new spines that matters or the size of the spines—their mushroom heads seem to enlarge the more an animal practices, and larger spines are stabler than smaller ones. Studying the behavior of dendritic spines is part of the larger challenge of figuring out how the brain stores memories. Memory depends on the plasticity of neural connections—that much is known—but scientists are still discovering the precise ways in which neurons make and break their numerous links.