December 11, 2013 | Salute to Scholars, The University
By Richard Firstman
MOST ADVANCES in science these days tend to come out of laboratories with the very latest and most sophisticated equipment. And then there is the groundbreaking science emerging from Ofer Tchernichovski’s lab at Hunter College.
You won’t find much in the way of high-tech, big-ticket instruments in his Laboratory of Vocal Learning. What you’ll see are walls lined with Igloo coolers — 50 of them, stacked in rows and turned on their sides, covers facing out but rarely opened. The coolers have been repurposed from mobile beer fridges to miniature behavioral labs. Each one is soundproof, climate-controlled and equipped with lighting that simulates day and night.
And each is occupied by a single bird — a young zebra finch — and its live-in lab partner of sorts: a plastic version of a similar bird outfitted with a tiny speaker that pipes in a repertoire of chirps recorded from adults of the species. From the day a young bird arrives at four weeks of age, the faux finch is its personal singing teacher. And for the next 50 days, 24 hours a day, every chirp of the student’s performance will be documented. By the end there will be a million chirps, sometimes twice that many – all of them recorded, analyzed, classified and entered into a database by software Tchernichovski developed.
For nearly 20 years, Tchernichovski, a research professor in the Hunter Department of Psychology, has been studying how songbirds learn songs. But it’s not just about the birds. By a quirk of nature, their brains and ours have a peculiar thing in common that makes newborn songbirds a stand-in for human babies in one of the more challenging frontiers of child-development research. Humans and songbirds are two of the few species that are “vocal learners,” meaning they develop the ability to imitate and modify a range of sounds, and sound combinations, by listening to adults of their species.
It doesn’t work that way with dogs or horses or cows: Calves come out of the womb knowing how to moo — and a moo is pretty much a moo. Puppies have to be trained to wait until they’re outside to do the most natural thing, but that other most natural thing is inborn and fixed: You can’t teach an old dog a new bark. Even primates, our closest evolutionary and genetic relatives, arrive at their calls innately. But there is something about the brains of human babies and young songbirds that makes their acquisition of sounds a process of listening to those around them and repeating what they hear.
“Songbirds even have local dialects,” Tchernichovski says. “Birds of the same species but come from different places don’t make the same sounds. Go right here to Inwood Forest and listen to the cowbirds, then go to Poughkeepsie, and it’s a completely different song. They change the song very locally. And the females only like the local guys.”
Human language, of course, is a far more complex form of vocal learning than that of songbirds and the few other species that learn sounds by listening to them — dolphins, whales and, some research suggests, elephants and bats. It’s partly that complexity that makes childhood language and speech development — and impairment — a hard thing to study. That and this: “You can’t put babies in a cage,” as Tchernichovski’s colleague Dina Lipkind puts it. But you can put zebra finches in Igloo coolers and control what they hear, record and analyze what they repeat and compare it to a database of infant babbling sounds.
That’s what Tchernichovski, Lipkind and their colleagues did, and this spring they published a major study in the journal Nature suggesting that the key to vocal development in babies — how they convert babbling to speaking — isn’t merely learning different sounds. It’s learning to make the difficult connections, or transitions, between those sounds. The researchers, including collaborators at New York University and the Riken Brain Science Institute in Japan, showed that babies seem to use the same step-by-step learning process as songbirds.
It’s been known since the 1970s that the ability of songbirds to sing comes from a specific area of their brains that is similar in some respects to the part of the human brain that’s responsible for speech. The new research takes the parallel a leap forward, identifying for the first time the elaborate process at the heart of vocal learning. The findings, and the inventive method Lipkind devised to reach them, have gotten the attention of language-development researchers because they could lead to new understanding of developmental disorders in children, and perhaps even to treating speech deficits caused by strokes in adults.
Tchernichovski never expected his studies of songbirds to cross over into the science of human-language development. He is a zoologist and veterinarian by training and began his CUNY career in the biology department of City College. Until four years ago, his focus was on the cagey way that young songbirds learn complex sounds by imitating adults during a critical period of development. “If you expose a three-week-old nightingale to 50 different songs, a year later the bird will start singing back each one of those 50 songs,” Tchernichovski says. “And it all happens within a few weeks of babbling, without any further input during that whole year, like a miracle.”
That much has been known since early in the last century. The mystery has been how they do it: What happens in the bird’s brain that allows it to master so many different combinations of sounds, seemingly out of nowhere? To explore that big question, Lipkind, then a postdoctoral student in Tchernichovski’s lab, had come up with a way of simulating the lessons young songbirds get from adults in the wild, but in a tightly controlled way that might allow the researchers to isolate and observe the components of the learning process with extraordinary precision.
“What Ofer had found was that if you put a bird alone in a soundproof chamber and just let him hear the sounds of an adult bird, he will learn that song,” Lipkind says. “What I did was design a way to teach the birds a second song. The first one has a certain order of sounds. For example, A-B-C, A-B-C. Then I give them the same sounds but in a different order: A-C-B, A-C-B. And after a while the bird will change to the new song. What’s important is that the birds don’t have to relearn any sounds. They just have to learn a new order of the sounds. The question is how does this ability develop?”
In 2009, Tchernichovski published a paper in Nature describing how his lab’s method of raising young zebra finches in the isolation of the coolers demonstrated that both nature and nurture were at work in the birds’ “culture” of vocal learning. Among the readers most fascinated by the paper was Gary Marcus, a prominent research psychologist who studies how infants learn language and directs NYU’s Center for Language and Music. Marcus wrote Tchernichovski “a fan letter,” he says, and it led to an invitation to the songbird lab — and to a collaboration. They set out to use Lipkind’s method to break down the elements of vocal learning and see, perhaps, if birds and babies use a similar process.
The first step was for Tchernichovski and Lipkind to try to tease out precisely how birds do it. They knew it was an arduous process — that it took many thousands of chirps for a zebra finch to go from A-B-C to A-C-B. But was it a sudden advance, like a button being pushed, or did it happen incrementally, in stages? Even more intriguing, was it a matter of random trial-and-error, or was there an identifiable pattern — a system?
What emerged from the computer analysis of those many millions of chirps was striking. The birds didn’t go directly from singing the first song to suddenly one day singing the second. There was a bottleneck in the progression, and it was the surprising difficulty they had in learning to switch the order of just two syllables. “If they can sing A-B, you would think it should be easy to sing B-A,” Tchernichovski says. “But it’s not. It takes them a long time of training themselves to learn the transition. It’s not that they need to keep hearing it again and again. They heard the sequence only 20 times a day. But they had to practice it thousands of times before they mastered it. The transition was happening in their minds.”
The breakthrough discovery was that learning a new arrangement of three notes was always a three-step process, even when the first note stayed the same. “To change their song from A-B-C, A-B-C to A-C-B, A-C-B, they first say A-C, then C-B and finally B-A,” says Tchernichovski. “Only then can they switch to the new song. It’s a stepwise process, where new transitions appear one by one, with gaps of days or weeks between them.” And in many cases, the birds could learn the individual transitions but not the entire set.
The findings were exciting to the researchers, but not, initially, to the peer reviewers of Nature. They were wary of applying the results with the zebra finches too generally. So Tchernichovski extended the collaboration to a birdsong researcher he knew at the Riken Institute in Japan. This colleague, Kazuo Okanoya, worked with Bengalese finches, which use more complex combinations in their songs than zebra finches do. Okanoya used Lipkind’s method and gathered data demonstrating the same learning pattern in the smarter birds.
Finally came time to test the idea that human babies might use essentially the same process. It was a head-on challenge to the widely accepted thought that humans are born with an ability that allows them to make an almost seamless transition from babbling to speaking starting around one year of age. “The idea is that we have this language-learning machine,” Tchernichovski says. “But maybe it’s not like that. Maybe babies are like birds.”
Lipkind and a graduate student in Marcus’ lab at NYU analyzed a database of infant language, called CHILDES, that goes back 30 years and has been cited in more than 3,000 studies. “These are huge data sets of recordings of babbling babies, taken every two weeks, that someone listened to and transcribed,” Tchernichovski said. “We took the database of nine American babies and looked at the connection of each syllable, like ba, gu or di, and how it developed over time.
“The data look very much like song development in birds. Every time an infant babbles he learns to produce new syllables, but connecting them together seems much more challenging. For example, it takes the infant 20 to 30 weeks to connect a new syllable type to other syllable types.”
The study was published in May by Nature with Lipkind as the lead author. It amounts to the discovery of “a previously unidentified component” of how we learn to speak, she says. Tchernichovski adds, “Now we see that this notion that humans are born with the capacity to rearrange vocal elements is incorrect. It develops very slowly, either by maturation or by learning, or both. The ability to rearrange elements is not the starting point of vocal development, it’s a laboriously achieved end point.”
Marcus, who has published several popular books on the human mind and is a contributor to The New Yorker, wrote on the magazine’s website: “Nobody had ever really explained why babbling took so many months; our birdsong data has finally yielded a clue.”
The researchers hope the findings might be a foundation for new understanding of speech and language disorders in children and adults. “Can we predict developmental disorders in human infants based on the development of their combinatorial abilities?” Tchernichovski asks. “Can we improve treatment of aphasia — speech and language — after stroke? The similarities between song development and speech development suggest a shared, primitive mechanism that we and others can now explore.”