In the Calls of Bonobos, Scientists Hear Hints of Language

After listening to hundreds of hours of ape calls, a team of scientists say they have detected a hallmark of human language: the ability to put together strings of sounds to create new meanings.

The provocative finding, published Thursday in the journal Science, drew praise from some scholars and skepticism from others.

Federica Amici, a primatologist at the University of Leipzig in Germany, said that the study helped place the roots of language even further back in time, to millions of years before the emergence of our species. “Differences between humans and other primates, including in communication, are far less distinct and well-defined than we have long assumed,” Dr. Amici said.

But other researchers said that the study, which had been conducted on bonobos, close relatives of chimpanzees, had little to reveal about how we use words. “The present findings don’t tell us anything about the evolution of language,” said Johan Bolhuis, a neurobiologist at Utrecht University in the Netherlands.

Many species can communicate with sounds. But when an animal makes a sound, it typically means just one thing. Monkeys, for instance, can make one warning call in reference to a leopard and a different one for an incoming eagle flying.

In contrast, we humans can string words together in ways that combine their individual meanings into something new. Suppose I say, “I am a bad dancer.” When I combine the words “bad” and “dancer,” I no longer mean them independently; I’m not saying, “I am a bad person who also happens to dance.” Instead, I mean that I don’t dance well.

Linguists call this compositionality, and have long considered it an essential ingredient of language. “It’s the force behind language’s creativity and productivity,” said Simon Townsend, a comparative psychologist at the University of Zurich in Switzerland. “Theoretically, you can come up with any phrase that has never been uttered before.”

For decades, scientists found no clear sign of compositionality in other species. But a few years ago Dr. Townsend and his colleagues discovered a hint of it in chimpanzees.

In a Ugandan forest, Dr. Townsend’s team recorded more than 330 hours of chimpanzees going about their daily lives, and identified a dozen distinct calls. To the untrained ear, the recordings might sound like a random cacophony. But Dr. Townsend and his colleagues noticed that certain calls followed others more than would be expected by chance alone. All told, they identified 15 distinctive pairs of calls.

The scientists wondered if a pair of calls carried a meaning greater than that of two individual calls on their own. To test that hypothesis, they spent two years studying one pair in particular: a call known as “waa-bark,” followed by another known as “alarm-huu.”

Chimpanzees make the waa-bark call as a way to bring other chimpanzees to them. An ape will make the call during a hunt, for instance, or to summon allies during a fight. They make the alarm-huu call when frightened or surprised — in response to an earthquake, perhaps, or the unexpected sight of a scientist’s raincoat.

Dr. Townsend and his colleagues wondered if “alarm-huu” when it was followed by “waa-bark” meant something else. They noticed two occasions in which a chimpanzee paired the calls when it encountered a snake while other chimpanzees were within earshot. Perhaps, the scientists thought, the two calls together meant something like, “Get over here and help me deal with this snake!”

Experiments followed. In one, the researchers pulled a fake snake across a trail as chimpanzees passed by. The apes, as predicted, often responded with “alarm-huu” followed by “waa-bark.”

The researchers then played the pair of calls through loudspeakers and watched how chimpanzees reacted. The apes tended to look at the loudspeaker for a long time; almost a minute. If it played only “alarm-huu” or “waa-bark” on their own, the chimpanzees glanced over for just a few seconds.

An additional clue suggested that the two calls combined to form a snake alarm: When some chimpanzees heard the paired calls, they leaped into a tree, a typical response (among apes) when snakes are around.

As intriguing as these ideas were, testing them was slow going. To broaden the research, and speed it along, Dr. Townsend began collaborating with Martin Surbeck, a behavioral ecologist at Harvard who studies bonobos, a species of ape that split off from chimpanzees two million years ago. Dr. Surbeck and his colleagues have spent years following apes in the Kokolopori Bonobo Reserve in the Democratic Republic of Congo.

In 2022, Melissa Berthet, a postdoctoral researcher in Dr. Townsend’s lab, joined with them to eavesdrop on the apes. She made 400 hours of recordings, capturing 567 single calls and 425 pairs. Dr. Berthet also made a note of what had happened just before the bonobos made their calls. Did a tree fall? Was the ape making a nest for the night, or grooming a friend? Dr. Berthet filled out a 336-item checklist for every call.

Shane Steinert-Threlkeld, a computational linguist at the University of Washington who was not involved in the study, said that the scale of the collected data was unmatched in this line of research. “For that reason, I’m very excited about it,” he said.

Back in Zurich, Dr. Berthet listened to the recordings and classified the calls into a dozen different types. To analyze the meaning of the calls, she analyzed her checklist. She and her colleagues then used some of the mathematical techniques that artificial intelligence systems like ChatGPT use to learn how words are related to each other. This analysis allowed the scientists to map the bonobo calls visually; the closer the calls appeared to each other on the map, the more similar their meanings.

The researchers also found that the bonobos frequently employed 16 specific pairs of calls, and that most pairs showed up on the map in the same neighborhood as the two individual sounds comprising them. This suggested that their combination conveyed no special meaning.

But four pairs of calls stood out. These landed on the map far from the placement of their two individual calls; together, it seemed, they carried a meaning unlike either call alone. One such pair, for instance, combined two calls: a high hoot, generally made when a bonobo is trying to draw the attention of others far away, and a low hoot, made when the bonobo is excited by some emotion.

Combined, the two calls seem to express something more, perhaps a rescue plea to distant bonobos when under attack. “It would be like, ‘Pay attention to me because I am in distress,’” Dr. Berthet said.

Dr. Berthet said that the new results should address any skepticism about Dr. Townsend’s earlier study on chimpanzees. “Linguists would always say, ‘Yeah, OK, but it’s just one combination — what does it really tell us?’” she said. “Here we show actually bonobos have several compositional structures, and they use them a lot.”

Together, the two studies on bonobos and chimpanzees suggest that our common ancestor with these apes also possessed compositionality, the researchers argue.

But Dr. Bolhuis questioned whether the new study could actually detect compositionality in bonobos. “Compositionality is not just about combining two words,” he said: It’s also about following rules of syntax to assemble words into phrases and bigger units of meaning.

Dr. Townsend countered that perhaps the act of pairing calls was a first step toward a full-blown compositionality that had emerged later, in early humans.

A next step, Dr. Steinert-Threlkeld said, would be for researchers to analyze the bonobo data with more sophisticated methods, to see if these results hold up. Maybe a computer could be trained to learn the meanings of individual calls, then tested to see if it could predict the meanings of pairs of calls it had never heard before.

“It’s imperfect,” he said of the new study. “But it’s a good first step.”