News of the WeekNeurobiology

Birds May Refine Their Songs While Sleeping

Science  18 Dec 1998:
Vol. 282, Issue 5397, pp. 2163-2165
DOI: 10.1126/science.282.5397.2163

Like novice tenors learning an aria, young male songbirds first learn their species' courtship songs by copying the melodies sung by other males; later, each bird adds flourishes that make his rendition unique. Some researchers think that happens “on line,” with the birds correcting errors and improving their technique as they sing. But on page 2250, Daniel Margoliash and his colleagues at the University of Chicago argue that at least some song learning and refinement may occur while the birds sleep.

Margoliash's team based that conclusion on measurements comparing the activity of song-specific neurons in the brains of waking and sleeping zebra finches. The researchers found that in sleeping birds, auditory signals triggered by a recording of each bird's own song flowed freely between the brain areas that govern singing. But when the birds woke up, it was as if a gate came down to block that flow. Margoliash suggests that during sleep the wide-open gate allows the birds' brains to refine the neural firing patterns that produce the song, an “off-line” learning similar to the memory strengthening that some neuroscientists think may occur during sleep when rats learn mazes and humans learn motor tasks (see Science, 29 July 1994, p. 603).

Other birdsong researchers praise the new work, noting that it's the first time anyone has done such a study on naturally sleeping birds. Birdsong pioneer Fernando Nottebohm of Rockefeller University's Field Research Center in Millbrook, New York, calls it “novel and intriguing,” and Richard Mooney, who studies bird song learning at Duke University Medical Center, adds that it may provide new clues to human language learning. But both Mooney and Nottebohm say it fails to prove that song refinement takes place during sleep. At this point, Nottebohm says, “there are really no grounds to suggest that anything like ‘off-line learning’ is taking place” while the birds sleep.

Just as the human brain contains special areas that control speech, birds have brain areas devoted to producing song. Neurons in an area called HVc send signals to a second region, RA, which connects to motor neurons that directly control the singing muscles. Because researchers have found activity in HVc and RA not only when birds sing, but also when they hear their own song played back, some suggested that the neurons self-correct while the bird is singing, modifying their activity to improve the song.

But those results came from anesthetized birds, and Margoliash's team saw a different picture when they recorded from individual HVc and RA neurons while the birds were awake. When those birds heard recordings of their own songs, team member Albert Yu found that HVc neurons responded, but those in RA did not, instead firing in a monotonous pattern. But when the birds naturally drifted off to sleep, the firing patterns in response to the recorded songs shifted to resemble those in the anesthetized animals. At that point, team member Amish Dave found, the RA neurons came alive and began to fire in response to signals from HVc. When the birds awoke, RA returned to its monotonous firing pattern.

The team fingered a molecule that may help cause the blockade: norepinephrine, a neurohormone whose levels fall during sleep and rise with waking. When the team boosted the norepinephrine level in anesthetized birds, the RA responses dropped. Margoliash notes that other as yet untested signaling molecules, such as dopamine, may contribute to the effect as well.

To Margoliash, the wide-open communication between HVc and RA during sleep suggests that that's when the birds learn to refine their songs. He speculates that even though a sleeping bird doesn't normally hear its own song, as the birds did in the experiment, its HVc neurons might spontaneously fire in the same pattern that is induced by the song while the bird is awake. That information would pass freely to RA neurons, which could use it to fine-tune the commands they give to the singing muscles the next time the bird sings. His team, he says, is now studying HVc firing patterns during sleep to see whether they do mimic the song response in awake birds.

Without such evidence, Mooney argues, the wide-open circuitry during sleep may be a “red herring,” the result of the fact that the brain has little else to attend to. What makes the new work “profoundly important” in his view are the results obtained with birds that are awake, in which, he points out, RA's response to the HVc activity elicited by the song recordings is “throttled down,” but “not shut down entirely.” That “in-between state,” he says, makes the circuits sensitive to modulating influences such as attention, which could regulate the information channels to control when song learning can occur.

Mooney finds the results tantalizing for another reason as well. They may provide a clue to a well-known human phenomenon: the loss of ability to learn new languages fluently at puberty. At puberty, bird songs become less responsive to auditory feedback. Mooney notes that sex hormones affect the turnover rates of norepinephrine in ways that could locally increase its levels, and he speculates that increases of sex hormones at puberty could reduce the bird's ability to self-correct its song. If so, he adds, it would “not be a big leap” to consider that a similar mechanism may be responsible for the problems humans have learning to speak a language like a native after puberty.

Those ideas remain to be tested, but to Nottebohm, that's another benefit of the new results. “What opportunities for future work,” he enthuses. Indeed, just as the tenor and the zebra finch use feedback to fine-tune their songs, song researchers will likely be tweaking their hypotheses in response to these results and the new experiments they are bound to inspire.


Navigate This Article