As the stage lights went down on his graduation dance performance at New York's High School of the Performing Arts, Erich Jarvis decided on his future. He'd been struggling to choose between a career in science or in dance. With a scholarship to Hunter College waiting for him, he decided he could make a bigger impact on the world as a scientist. "My mother influenced me to think that science was one of the ways to change the world," Jarvis recalls. "It just clicked: I want to be a scientist."

At Hunter, Jarvis worked closely with molecular biologist Rivka Rudner. He threw himself into research, Rudner recalls, coauthoring six papers as an undergraduate. "You had to kick him out of the lab so he'd go home and sleep," says Rudner.

Jarvis earned a PhD in molecular neurobiology and animal behavior at Rockefeller University. He's now an associate professor of neurobiology at Duke University. In 2002, he was honored with the Alan T. Waterman Award, the National Science Foundation's highest award for young investigators.

"Science turned out to be quite like being a dancer," Jarvis says now. "You must practice a lot; sometimes experiments fail and you try again and again. You're always being creative, and you have to be very disciplined."

Unsatisfied with the "stimulus-and-response" behavioral research going on in many labs at the time, Jarvis joined a Rockefeller lab studying avian brains. There he felt he could apply his molecular biology experience to look at the brain in new ways. He began studying vocal-learning birds, those capable of learning and imitating sound. Only three orders of birds - songbirds, hummingbirds, and parrots - are capable of vocal learning.

Jarvis discovered that the three bird groups have evolved similar neural pathways and seven similar brain nuclei that aren't present in common ancestors or close relatives. "Pigeons and chickens don't have them," he says. With an estimated 65 million years of evolutionary history between them, a parrot and a songbird are as distantly related as a human and a dolphin. Jarvis believes that the brain structures evolved independently in each of the three orders.

If he's right, it implies that there are certain constraints on the evolution of the vertebrate brain. "I think [vocal learning] evolved out of a preexisting system," he says. Besides birds, only humans, elephants, bats, and cetaceans are capable of imitating sounds. Jarvis believes the brains of those mammals likely share similarities with vocal-learning birds. If so, unraveling the neural pathways could be a key to unlocking the origins of language. "You've got everyone searching for the language gene," he says. "I think they should be searching for the language neural network."

Jarvis is doing just that at his Duke laboratory, which includes three other African-Americans, as well as individuals from Brazil, China, Africa, Japan, Puerto Rico, and Slovakia. "I haven't made efforts to purposely create a diverse lab, but it's happened that way," he says. "I want the best people out there." The stew of cultures might have something to do with the innovative discoveries his team has made. "Sometimes the science you do is based on your culture and experience," he says. "Diversity in the lab leads to greater scientific discovery."

Jarvis' former mentor has a different idea for the perfect lab: "If I could only have cloned a few Erichs so that after he left I'd still have a few to work with me in my lab," says Rudner, "but there was only one."