Julie Elie has spent years tuning in to the small, everyday sounds of zebra finches, not the songs that usually get the attention.

Elie, a researcher at the University of California, Berkeley, studied the birds’ calls and chirps, the vocalizations they use to communicate with each other. Using data collected over years of observation, she identified 11 core calls in the zebra finch vocabulary, including calls for distress, hunger and saying hello.

She found the birds announce who they are and what they are doing, and that they use individual signatures that let other finches recognize them.

Elie and her colleagues also tested the birds directly. In one set of experiments, the birds had to discriminate and categorize calls by their meaning. The team first tested if the birds could recognize other individuals from a specific call type, the distance call.

“And then I said, ‘okay, let’s export that to the other call types and see whether they can identify each other throughout the repertoire,’ and surely, they would, they were able to do it,” Elie says.

The birds sometimes made mistakes, but “still, they were always above chance level,” she says.

In another test, the birds were asked to validate the call classifications Elie and her team had developed. The results backed the team’s system.

“This is comforting to me, and so, yeah, cool. I’ve not been hallucinating for all these years. They agree with my organization,” she says.

The birds consistently grouped calls by perceived meaning, not by sound. They sometimes confused calls with similar meanings, such as aggression and distress, but not calls that sounded alike and had very different meanings.

The work has earned Elie the 2026 Coller-Dolittle Prize, a $100,000 reward for progress toward interspecies communication, specifically enabling humans to talk to animals, and for animals to talk back in a way humans can understand. The competition also offers a $10 million grand prize for solving the problem in full.

Elie used machine learning to help parse a large set of observational data and match zebra finch calls to behavior patterns.

“I think the zebra finch is just right level of complexity,” she says.

She developed an algorithm that could classify calls from sound alone, but she said it could not always tell some calls apart, including distress and aggressive calls.

“You want to have machine learning, you want to have artificial intelligence that helps you to capture acoustic differences between things,” she says. “But communication is not only about that, and having information about the behavior of the animal, like the context of the condition, is what really also puts some more light onto the language of the species you’re studying.”

Elie said zebra finch calls are complex enough to encode meaning and easy to observe in a lab, while doing similar work with other vocal animals, such as dolphins, would be much harder.

“But I have hope that by constructing level by level, we’ll be able to climb up there,” Elie says. “The aim of this challenge is to be able to establish a communication with the animal that goes both ways. It’s not only the human understanding what the animal says, but it’s also the human communicating to the animal, and the animal understanding it. And this, I think, is achievable.”

Read more from Scientific American.