If you were looking for an animal in which to study the evolution of language, the zebra finch—a native of Australia that weighs about an ounce and has a brain the size of a grape—might not seem an obvious choice. Until recently, in fact, few researchers even considered the possibility that birds, with an evolutionary history that diverged from that of mammals 300 million years ago, have much to teach us about intelligence. But, says Duke University's Erich Jarvis, zebra finches have one truly outstanding quality—"vocal learning," the ability to remember and reproduce the sounds they hear, an ability shared, as far as we know, by only a handful of animal families. Besides human beings, these are elephants, bats, whales, dolphins and three orders of birds: parrots, hummingbirds and songbirds.

Pigeons coo and chickens squawk; wolves howl and monkeys chatter. But those sounds are innate. Dogs and horses can learn to respond to vocal commands, and some apes can be taught a primitive sign language. But none of those animals can learn to imitate new sounds, which appears to be one of the skills that made human culture possible. So in a basement a few flights down from Jarvis' laboratory, several hundred orange-beaked, gray-backed zebra finches fill the room with a soft chirruping, like so many tiny, squeaking fan belts. Their fate is to have their brains analyzed for the presence of proteins that signify the activation of certain genes during singing. "What's the most complex thing a brain can do? Language," says Jarvis, a 41-year-old neurobiologist. "I am starting from the naive position that if you can decipher the most complex thing about how the brain works, everything else will fall into place."

Oddly, perhaps, Jarvis began his career by impressing the zoologist Fernando Nottebohm as "the most disorganized and chaotic member of my laboratory." That was in 1989, when Jarvis, a product of the New York City public schools and Hunter College, arrived at Rockefeller University to study under Nottebohm, renowned for discovering that songbird brains actually grow new nerve cells throughout the bird's life. Jarvis is tall, lean and loose-limbed, with a short, curly beard; a former dancer who considered a career in ballet, he walks quickly and speaks rapidly in a soft, even voice. Growing up, he lived with his mother in Harlem after his parents' divorce, and with grandparents in Brooklyn and the Bronx. He attended the High School of the Performing Arts, and a bird, in his early experience, was something you cooked for Sunday dinner. In 1998, having completed his doctorate at Rockefeller, Jarvis moved to Duke, where four years later he received the Alan T. Waterman Award, the highest honor given by the National Science Foundation to a young researcher.

One great thing about what Jarvis studies is that there's funding for it, because he can plausibly claim to be doing work that might lead to a cure for human disease. Tracing the pathways of vocal learning in birds might help researchers working to restore speech in stroke victims, Jarvis says, adding: "We now know the avian brain is a lot more like the mammalian brain than people realized." The possibility of new treatments for brain injuries led to a $2.5 million "Pioneer" grant to Jarvis from the National Institutes of Health. The award, for high-risk, groundbreaking research, makes him an object of envy to other researchers, including even his old mentor Nottebohm, who wryly notes that Jarvis "is getting these days much more grant support than I get."

Jarvis is well known in the field for his speculation about the evolution of vocal learning—or, rather, its failure to evolve more often in nature. In songbirds, where as a rule the males sing, the trait plays a role in mating. "The more variable syntax you produce, the more likely a mate will choose you," Jarvis says. "They call them sexy songs." (It is, he adds, not all that different in another vocal-learning species, the one that produces poetry and improvisational jazz.) Jarvis believes that the neural architecture for vocal learning is basic to the vertebrate brain—specifically, that it is related to the motor-learning system that enables an animal, say, to walk. In that case, Jarvis wondered, why has such a useful trait not evolved more often?

His proposed answer (along with that of Japanese researcher Kazuo Okanoya, who came up with the same idea independently) represents a conceptual leap that his colleagues regard as somewhere between audacious and preposterous. He suggests that getting listened to is, for many animals, just another way to get eaten. Predators that become accustomed to the routine sounds of their prey take notice when someone whistles a different tune. Thus, evolution—which favors vocal learning by way of sexual selection—also works against it, through predation. Is it just coincidence that most of the species capable of vocal learning can also fly? Jarvis thinks not. When he talks about the idea at scientific meetings, he says, "half the room will be denouncing it...which doesn't bother me. At least until someone comes along to disprove it."

As a scientist, Jarvis has come a long way since his first years at Rockefeller, when he was also dancing with a small company called Ballet Afrique and nearly washed out of the lab. "I really didn't see that he was going to survive," recalls David Clayton, a Rockefeller colleague now at the University of Illinois at Urbana-Champaign. "It's interesting now to see him flourish. He's a very strong personality." Nottebohm says, "Erich was very insecure. He told me he thought he was being 'dissed.' I didn't even know what the word meant. I told him, you're right, they're discriminating against you, because you're a lousy scientist. And I think that really struck him. He became not only the most hard-working member of the lab, but he started doing things that were creative and well thought through."

It's not unusual for the name of a scientist at a major university like Duke to show up in the New York Times, as Jarvis' has half a dozen times. Yet the first Times reference to him is a 1983 review of his high-school graduation performance. ("Lisa Arrington and Erich Jarvis had the audience cheering with their Soviet-style lifts in a War and Discord pas de deux," the reviewer wrote.) Six years later, while in his first year at Rockefeller, he turns up again, in a long article under the headline "Random Death Claims a Man Who Struggled to Regain Life." The man was Jarvis' father, James, who had been homeless for years, living in caves in New York City parks. James Jarvis had graduated from the High School of Music and Art at 15 and majored in chemistry at City College. He had been a mentally ill drug addict for much of his later life, but in the year or so before his murder—an apparently motiveless crime that has not been solved—he had made efforts to reenter society. Erich Jarvis had been close to his father, when he could find him. "I don't know how many homeless people out there are scientists," he told the reporter. "But my father was a scientist. He was an explorer. He was studying survival techniques. He wanted to find the universal law of all laws."

Jarvis has some theories about how cultural biases have entered into his own, highly abstract field of science. For example, he suspects that the role of "housekeeping" genes, which perform routine functions such as regulating metabolism or maintaining the integrity of the cell wall, have been overlooked by many scientists because "housekeeping" is regarded as uninteresting: "Who invented this term? Why are [those genes] uninteresting? Because in the past, I believe the [male researchers] were thinking about their wives. In the family I come from—in most African-American families—housekeeping is highly respected." Recently his lab discovered that actin, a protein produced by one of the so-called "housekeeping" genes, appears to play a role in vocal learning.

When he's not traveling, Jarvis is in the laboratory six or seven days a week. He long ago gave up performing, although he manages to go out dancing once in a while with his wife, Miriam Rivas, a molecular biologist whom he met at Hunter College and is now a member of his lab. (One of her conditions for marrying Jarvis was that he know how to Latin dance. One of their concerns in moving to Durham, North Carolina, was whether they could find a suitable salsa club. They did.) He thinks it's not entirely a coincidence that someone whose life was steeped in music ended up studying bird song. (His colleague at Duke, neurobiologist Richard Mooney, who also works with songbirds, is a classical guitarist.) Jarvis' most recent paper was on the release of dopamine, a neurotransmitter that controls the brain's "reward" circuitry, in the bird brain during song. "We don't have the data to prove this," Jarvis says, "but the implication is that the bird gets pleasure out of singing." So much of life, for human and finch alike, comes down to music—and that singular gift of listening and learning.

Jerry Adler is a senior editor at Newsweek specializing in science. Greg Foster, a freelancer based in Atlanta, took the photographs in our story last month about cricket in America.