A robot taught these birds a long-lost birdsong
The song of the chingolo can be heard across South America. But young songbirds were no longer learning the tunes of their elders—until scientists stepped in.
The song of the male chingolo, or rufous-collared sparrow (Zonotrichia capensis), is one of the most distinctive sounds in South America. In just a few seconds, this small bird—light brown or whitish with black spots—delivers its melody, composed of two to four introductory notes and a final trill. Each family of chingolos has its unique song, which they sing throughout their lives. But first, they must learn it.
This behavior is deeply rooted in the interaction between young and adult male chingolos, who transmit the patterns of their song. But what happens when this learning is interrupted—when habitat loss, population decline, or the disappearance of adult tutors break the chain of transmission?
Between 2020 and 2023, researchers from the Faculty of Exact and Natural Sciences at the University of Buenos Aires (UBA) took on a bold task: reintroducing a song that had vanished from the wild—known only from a musical notation made in the 1960s—to a population of chingolos. Using an innovative approach that combined advanced technology, scientists developed a “robotic tutor,” a device capable of emitting the forgotten melody so that young birds could learn from it. As a result, chingolos living in Pereyra Iraola Park, a natural reserve of about 25 acres in Buenos Aires, learned the synthetic song, incorporated it into their repertoire, and now sing it proudly.
“We always think of preserving biodiversity as a genetic issue, but it also implies a cultural issue,” explains Gabriel Mindlin, director of the Institute of Interdisciplinary and Applied Physics (INFINA) at UBA and co-author—together with Ana Amador and Roberto Bistel—of the resulting 2024 study published in Physica D: Nonlinear Phenomena. “Here, we made an extinct song fashionable, and it is a case of how an entire culture can be reintroduced if necessary.”
Learning a lost song to uphold tradition
The chingolo is one of the 4,000 known species of songbirds in the world. While some aspects of its song are genetically conditioned, in the case of the male chingolo, it’s learned from a tutor, most often the father. The young bird imitates and learns to vocalize a song that combines the family song with the population’s song. (For example, chingolos from the Río de la Plata do not sing the same final trill as those from other parts of the country.) This process takes approximately three months after birth. At first, its muscular control is somewhat imprecise, but eventually, it produces a highly refined acoustic pattern.
Limited-Edition Classic Collection
(Songbirds are being snatched from Miami’s forests.)
Each individual usually sings one song, although some can sing two or even three. The melody lasts only two seconds and is repeated from dawn until the sun reaches its zenith.
“It is a distinctive song—like a fingerprint, but learned,” says Amador. “It serves to attract the female and protect its territory. It’s its way of saying, 'This is me, and here I am.’”
How an extinct melody was revived
Thanks to handwritten musical notes made by Argentine ornithologist Fernando Nottebohm, professor emeritus at Rockefeller University in New York, the researchers knew the most popular songs heard in Pereyra Iraola Park in the 1960s. “If he noticed a rising frequency, he would mark it with an upward sign,” says Mindlin. “If it went down, becoming deeper, he would draw a downward sign. The trill was a set of small lines, and that’s how he represented the song.”
The question that triggered the investigation was this: Which of the hits from the '60s are still being sung? The team went to the park, recorded the melodies, then trained a neural network to verify which of the old tunes were still popular and which had disappeared. They discovered that only three classic songs were still being sung in the area. One possible reason for this is the urbanization of the area surrounding the park, which has displaced much of the local chingolo population; another is that over the years, the park has been invaded by other singing bird species that have taken over the chingolos' territory.
After deciding to revive the tunes that had vanished, the scientists analyzed the parameters of each real song—the start and end frequency of each note, its duration, and so on—and generated a mathematical model capable of producing synthetic copies of the bird’s song based on the physics of its phonation. To achieve this, they simulated the bird’s vocal tract, including the trachea, oropharyngeal cavity, glottis, and beak. The team even observed the activation of the syrinx muscles, the bird's vocal organ.
This synthetic song was introduced to the chingolos in the same park during the species’ critical sensory period, from October to February, when juveniles learn by imitating an acoustic model. The sessions took place during peak singing hours—which occur in the early morning—and were limited to a maximum of eight hours. The intervals between the synthetic songs were set to a speed slightly higher than the natural intervals in order to stimulate vocal responses.
“The interval between each song was random so that the birds perceived the playback from the three devices placed in the area as a real exchange, as if they were responding to each other,” explains Bistel.
From February to July, chingolos enter a period of complete silence due to low temperatures, which lasts until September. They then resume vocal practice and fine-tune their song based on the learned model. Surprisingly, the robotic tutor had been chosen by some young chingolos, who learned the extinct song and incorporated it into their vocal repertoire.
The song acquired by the juveniles alternated between syllables with increasing and decreasing frequencies, in line with the synthetic theme. However, the last descending syllable had a wider frequency range than the synthetic version—a characteristic shared by all adults recorded in the park. This variation seems to be the distinctive mark of the “dialect” of this chingolo population, possibly acquired from other real tutors or as an expression of the species' genetic programming.
“It’s like rescuing an ancient and forgotten language,” Mindlin says, emphasizing how cultural traits—whether in birds or humans—can be revived through learning and transmission across generations.
(What are animals saying? AI may help decode their languages.)
The future of vocal preservation
Mindlin, Amador, and Bistel assert that their tool holds promise for preserving the vocal repertoire of wild birds.
“There are genetic banks with large freezers. But what about cultural preservation? Shouldn’t we have song recording banks?” Amador says. “Our work paves the way for thinking about species preservation in a more holistic manner.”
The INFINA team is already working with chingolos and other species from southern Argentina, with a near-term goal: designing a neural network to aid in the automatic recognition of each individual’s song. The next step is to study vocal transmission in a population of ringed birds.
“The loss of individuals means not only the reduction of genetic diversity but also the loss of true cultures,” Mindlin says. “With mathematical equations that we transform into sound waves, we can ensure that this culture is not lost.”
