rich Jarvis nestles himself into the thick green foliage of the Brazilian forest, just as the morning sun begins to drive vapor off the broad wet leaves. He has arrived just in time for the most important daily event in his scientific fieldwork--the "dawn chorus."

Songcatcher: Jarvis and his parabolic microphone, used for recording birdsong. photo:Chris Hildreth

After a long, frustrating wait and too many missed opportunities, a prime target appears near his wire-cage trap. He aims a parabolic microphone at his small, quick quarry as his fellow scientist tracks it through binoculars. It takes four attentive eyes to track this lightning-fast creature--an absolute necessity, because if the scientists lose sight of it even for an instant, they cannot claim that they have gathered data on the same bird and the creature would be lost as a candidate for valid scientific study.

Jarvis barely manages to keep his flitting target framed in the video camera's sights. Then, to his utter gratification, the creature abruptly launches the performance he hoped for--emitting a raspy squeak so soft and high-pitched that Jarvis, when he first arrived in the forest, had taken for an insect's buzz. Again and again the creature sings, finally attaining the thirty-minute duration required for Jarvis' experiments. After zipping about, almost seeming to tease, his quarry flits into the trap to imbibe the sugar-water bait. Jarvis springs the trap and the tiny hummingbird is captured. Now he can measure the telltale changes in its brain produced by its obliging dawn chorus.

The hummingbird's half-hour of squeaky twitters has switched on a specific marker gene that the scientists know signals activity of the neural machinery underlying production of "learned vocalizations." Studying vocal learning is important because such learning of meaningful sounds is different from, say, the barking of a dog. Even though dog-lovers know that their pooch's bark is a form of communication, it is not learned but inborn. In contrast, vocal learning is imitative and rare in nature. It is found only in a few birds--songbirds, parrots, hummingbirds--and in just as few mammals, including whales, dolphins, bats, and, of course, humans.

Jarvis and his colleagues had already used their marker gene, with the faintly comical name of ZENK, to map the brain regions involved in vocal learning in songbirds and parrots, discovering that basically the same seven regions are involved. And once back at Duke, their analysis of the brains of the trapped hummingbirds resulted in a perfect scientific trifecta. As reported in August 2000 in the journal Nature, Jarvis and his colleagues discovered that hummingbirds also used the same seven brain structures to generate their song.

What's startling to biologists, says Jarvis, is that these three bird species occupy distant twigs on the bird family tree, each having closer relatives that are vocal nonlearners. The finding that these birds evolved to use a very similar set of structures for vocal learning offers a doorway to study a fascinating mystery of evolution. "One hypothesis is that the three orders evolved these structures independently over the last 65 million years in almost identical ways," says Jarvis.

This suggests that nature exerts some kind of external constraints other than genetics--called "epigenetic constraints"--on how complex brain structures evolve for complex behavior. "It's like the independent evolution of wings in both bats and birds. It is thought that the environment dictated that the wings had to be on either side of the body at the center of gravity for the least energy use, not one on the head and one on the foot."

Among the less likely evolutionary possibilities for the birds' song, says Jarvis, are that all modern birds had a common ancestor with vocal learning, and that many intervening orders lost the trait multiple, independent times. "If you apply that argument to mammals, then maybe some ancient mammal had the capability of speech, but it was lost multiple times at least between that ancestor and dolphins and humans." In both birds and mammals, such a multiple disappearance and then reappearance of this trait in one of many primates, humans, seems highly improbable, he says. Or all birds might have rudimentary, nonworking brain structures for vocal learning that are simply amplified in songbirds, parrots, and hummingbirds.

If such rudimentary brain structures exist, asks Jarvis, then why not have them in reptiles and dinosaurs? He and his colleagues can use modern techniques of genetic and molecular analysis to study the brains of "nonlearning" birds to answer such evolutionary questions. But more broadly, he says, his birds challenge all of us to rethink outmoded concepts of evolution, including our own place in it.

"Throughout our education, we have this concept of linear evolution instilled in us," he says. "We're told that there's this hierarchy of evolution, in which vertebrates evolved from some worm-like creature to fish, amphibians, reptiles, birds, mammals, and so forth, and that living vertebrates represent these stages in both body plan and brain intelligence. And once there were mammals, they evolved to primates, then humans, being last at the top of the hierarchy. But this concept of lower and higher in the vertebrate lineage is just completely false. Actually, evolution is a process in which each group of animals develops independent and diverse specializations. Songbirds have them, and we--as humans, mammals, and primates--have them."

What will surprise many, and even surprised Jarvis, is how little scientists really know about the language-fostering structures in our own brains. "After publication of the Nature paper on the hummingbirds, I started looking into the scientific literature on human speech more deeply than I ever had before," he says, waving a hand across an office floor piled with 800 or so scientific papers he has scrutinized. "I looked for any kind of brain differences or similarities with the bird vocal-learning and human-language literature. And the first thing I found is the human literature is a mess."

Of course, he notes, ethical constraints have prevented scientists from using the same rigorous techniques of dissection and experiment to study the human brain that they have with animals. It's understandable, then, that scientific understanding remains fragmentary of the language-related brain structures that sit between our very own ears. "Even the information in textbooks is not correct, because much of it doesn't come from studying human brains. Rather, it comes from extrapolating to humans studies of rats and nonhuman primates such as macaques."

Since neither of these species has been known to utter an intelligible learned vocalization, says Jarvis, "what's in textbooks is not reality, even though it's what students are learning as a mantra, and what professors are passing from generation to generation." In birds, vocal learners have different brain vocal pathways than vocal nonlearners. So it is impossible to extrapolate one to the other.

If Jarvis sounds like a man set to rattle some scientific cages, he is. Next year, he will help bring together at Duke a group of comparative neurobiologists, who will negotiate a renaming of the structures of bird brains from names given to them nearly a hundred years ago. The problem now, says Jarvis, is that names of all bird forebrain structures reflect an outmoded analogy to a basal structure in mammals called the "striatum," which scientists believed was a region involved in primitive, instinctual, and savage behaviors. Such invidious comparison doesn't reflect the reality that the striatum is involved in complex learned behaviors, and that a much larger proportion of the bird forebrain is instead analogous to the human cortex--an outer forebrain division that also manages complex behaviors. Although humans are taught to be inordinately proud of our cerebral cortex, it is actually the interaction of different parts of the forebrain, including the striatum, that manages complex processes as thought. Likewise, says Jarvis, birds have comparable brain divisions that manage their complex behavioral processes, among them several of the seven vocal learning nuclei in each of these parts.

• continues on page two.