October 2000

The fact that humans learn to talk by listening to and imitating other people seems obvious. If speech were built in, as breathing and swallowing are, we would all speak the same language, free of dialect, and life might be a lot simpler (think foreign travel or peace negotiations). While children pick up language with enviable ease, the process of vocal learning is actually quite complex, and most linguists agree that there is a critical time when it should occur. But given the proper teachers and timing, most humans can learn to produce a seemingly infinite number of sounds and sound combinations.

Such is not the case with most animals. In fact, only whales, dolphins, bats, and some birds are known to have the ability to learn vocalizations. Separate a kitten from its mother or other tutor, and its mew will be basically the same as that of its littermates—a pattern that holds true for the howl of a wolf, the grunt of a gorilla, and the whinny of a horse.

Vocal learning has been repeatedly demonstrated in two bird orders, Passeriformes (specifically the oscine songbirds) and Psittaciformes (parrots), and is believed to occur in a third, the Trochiliformes (hummingbirds). By comparing brain structures in these three bird orders, which are widely separated from one another on the avian family tree, Rockefeller University biologist Claudio Mello and his colleague Erich Jarvis, of Duke University, have shown that the same areas that control song learning and production in songbirds and parrots are also present in hummingbirds, a finding that strengthens the case for vocal learning in the latter.

Most people are surprised to learn that hummingbirds even have songs. "The songs aren’t particularly loud and you sort of have to know what to listen for," explains Mello. "They are higher pitched than those of songbirds, but the songs are amazingly rich, and in some species they can be quite complex."

The outcomes suggested that the males were imitating each other’s vocalizations—evidence that hummingbirds learn their songs.

In the 1950s biologists began to investigate the processes by which birds imitate the sounds they hear and incorporate them into songs. Appropriately enough, that work began with songbirds, a suborder that includes almost half the nearly 8,500 living species of birds. W. H. Thorpe, of the University of Cambridge, was the first to demonstrate learning in birds by performing what is now considered to be a classic experiment, involving the isolation of male chaffinches (European songbirds) in soundproof chambers equipped with speakers. Young chaffinches that heard recorded chaffinch songs were able to imitate these songs, while birds deprived of the recordings developed abnormally simple songs.

Isolation experiments are exceedingly difficult to do with hummingbirds, however. Because of their extraordinarily fast metabolism, baby hummers must be fed every ten minutes around the clock. But in 1990, this type of experiment was conducted on one species of trochilid, the Anna’s hummingbird (Calypte anna). The late Luis Felipe Baptista, of the California Academy of Sciences, and Karl Schumann, at the Zoologisches Forschungsinstitut in Bonn, Germany, found that a male Anna’s hummingbird raised in isolation produced a much simpler song than did wild males. The song was also very different from that of three males hand-raised together. The outcomes suggested that the males were imitating each other’s vocalizations—evidence that hummingbirds learn their songs. Other research showed that the nearest neighbors of hermit hummingbirds (Phaethornis longuemareus) sing more similar songs than nonneighbors of the same species; and the songs of green hermit hummingbirds (P. guy) living in Costa Rica are different from those of the same species living in Trinidad.

In the 1970s and 1980s, Fernando Nottebohm, of Rockefeller University, and several colleagues set about mapping the parts of the brain involved in the singing process. The researchers identified six anatomically distinct areas—clusters of cells called nuclei—in the forebrain of songbirds. These nuclei are organized into two distinct paths: the posterior pathway, which controls song production, and the anterior pathway, which controls song learning. Together these pathways form a song control system that must be intact if birds are to sing the songs they’ve learned.

Forebrain nuclei similar in structure and location have also been found in the budgerigar (an Australian parakeet). No such nuclei have been found in the birds most closely related to songbirds, the suboscines (woodcreepers, ovenbirds, antbirds), or in other nonlearners of songs such as pigeons and doves (order Columbiformes) and chickens, turkeys, and quails (order Galliformes). And before Mello and Jarvis, no one had bothered to look for these nuclei in hummingbirds.

Working with songbirds in the 1990s, bird researchers added a novel tool to their toolbox, a gene called ZENK, that would make the search for nuclei in hummingbirds much easier. Nottebohm, Mello, and Jarvis noticed that the number of activated ZENK genes in certain areas of the brain was very low when the songbirds were quiet. When the birds sang or heard songs, however, ZENK activity increased. By measuring the levels of activated ZENK in specific locations, the researchers were able to see the previously identified nuclei "in action." ZENK gave the researchers a window into the brain, enabling them to see how certain behaviors set into motion the molecular activity of cells in specific brain areas.

Did hummingbirds, too, have the brain structures and pathways necessary for learning?

The ZENK gene soon led the researchers to discover another nucleus in the songbird vocal system, bringing the total to seven. With ZENK they were also able to define more clearly the roles of the anterior and posterior pathways. The anterior pathway (which, they knew, needed to be intact for learning to occur) showed signs of activity only when the birds heard songs by members of their own species. And after they had heard the song a number of times, the ZENK response began to fade. "With ZENK," says Jarvis "we now had a technique that essentially lit up the learning center."

Jarvis and Mello went on to look at budgerigars’ brains. "The brains were much more similar to songbirds’ brains than people thought," says Jarvis, "We found many of the same structures, and their placement in the brain was identical." One question remained: Did hummingbirds, too, have the brain structures and pathways necessary for learning?

To answer that question, Mello and his colleagues have made several trips to his native Brazil, to an area with one of the world’s highest densities and species diversity of hummingbirds. The researchers set up a hummingbird feeder on the veranda of the Museu de Biologia Mello Leitão, a small museum and biological research station on the outskirts of the town of Santa Teresa, nestled high in the Atlantic tropical forest of Brazil’s east coast. Part of the research involved learning to identify individual birds—no easy feat.

Family tree: Did vocal learning capacity evolve independently in parrots, hummers, and songbirds, or did a common ancestor possess the trait? AMNH; after Jarvis, et al.

For the uninitiated, it is difficult to tell apart the thirty or so species of bejia-flores (Portuguese for “flower-kissers,” the creatures we call hummingbirds) that live in this swath of forest, and it is mind-numbing to try following the activity of individual birds. They move so fast that keeping your eye on one is like trying to follow a single bee in a swarm. Banding—the field technique commonly used for telling birds apart—doesn’t work: hummingbirds’ legs are so tiny that there is little space for a band, and the birds weigh so little that attaching a band could significantly affect their ability to fly.

"After a couple of days, you start to see individual variation, a messy feather here, or a slightly darker color," says Linda Wilbrecht, a graduate student at Rockefeller. Working at Mello’s field site in Brazil, Wilbrecht and her fellow graduate student Sidarta Ribeiro observed the behavior of individuals of two species, sombre hummingbirds (Aphantochroa cirrhochloris) and rufous-breasted hermits (Glaucis hirsuta). Wilbrecht and Ribeiro followed birds that were doing one of three things: singing in territorial defense, listening to recorded songs of conspecific birds, or perching quietly. These species were chosen, in part, because their songs are very different. G. hirsuta’s is complex, while A. cirrhochloris’s is fairly simple. A complex song is itself evidence that a bird learned its song: the innate vocalizations of birds (those produced by pigeons, for example) tend to be simple, with few syllables and a narrow range of notes. Because of the differences between the two species, Jarvis and Mello reasoned, one might be a learner and the other not—a possibility that would make their results quite intriguing.

Back at Rockefeller University, the researchers hoped to learn whether the behavior they had observed in the field translated into observable differences in the ZENK expression within the birds’ brains. In the laboratory, the researchers confirmed that seven nuclei are active in singing hummingbirds, and that these structures are strikingly similar to the seven forebrain structures involved in vocal learning and production in songbirds and parrots. It also turned out that Aphantochroa’s song is not so simple after all. While these birds do sing relatively few notes, each note is in itself more like a chord—more complex than notes in a Glaucis song. As for their brains, there is little difference in their structure. For the first time scientists can say definitively that hummingbirds have the necessary brain circuitry for vocal learning.

And what of the evolutionary path that brought hummingbirds to this point? Of the twenty-three bird orders, the three in which song learning occurs are very distantly related; on the avian family tree they are separated by several orders in which learning doesn’t occur. This suggests two possible scenarios: First, all the birds shared a common ancestor that had the necessary brain structures, but only three orders retained them, developing into learners, while the intervening orders did not. But that would mean that some birds with the structures necessary to learn, and maybe even the ability to learn, have lost them over time. To Jarvis, this seems an unlikely scenario. The second possibility is that the three orders independently evolved the same learning structures in their brains. “Maybe complex behaviors can evolve more than once. What we’re saying is that there is an overarching factor that says if vocal learning is going to occur, then Mother Nature is only going to let it happen this way,” Jarvis says.

The next step is to use ZENK to look at the brains of the intervening orders to see if birds that don’t learn songs share the same brain structures and pathways with their educated (or educable) cousins. The researchers will start with doves, and if the learning structures aren’t found, the second hypothesis becomes even more likely. Jarvis doesn’t think learning structures will be found in the other birds, "but if in a million years pigeons develop learning, I say I can predict where in their brains those structures will be."

Click here to hear hummingbird songs

Copyright © 2000, 2001 American Museum of Natural History

Return to Web Site Archive