Each colony of naked mole rats speaks its own unique dialect of chirps, a social feature that appears to be mediated by the colony’s queen and learned during adolescence, according to a study published January 29 in Science. This ability to modify and adapt language may be the first example in a rodent of production learning—the type of language learning that humans use to create new sounds in response to our experiences—opening up a new model system for studying the joint evolution of vocal and social complexity.
“It’s surprising to find a new model of a social animal that has dialects,” says Thomas Park, a neurobiologist at the University of Illinois at Chicago who has...
Naked mole rats (Heterocephalus glaber) have puzzled scientists since the 1970s. Beyond their hairless and wrinkly bodies, these rodents are highly tolerant of pain, have an unusual resilience to many cancers, and live for decades. Moreover, despite being functionally blind and nearly deaf, they are among the most social mammals, structuring their underground communities like those of ants or termites: a single queen oversees a colony of hundreds of workers and is the only one to reproduce, choosing a handful of consorts from among her subordinates to mate with.
In adapting to a subterranean lifestyle, naked mole rats have done away with many of the physical abilities that humans most associate with communication. “They’ve basically abandoned everything that they don’t need, but surprisingly, they’ve expanded their vocal repertoire,” says Alison Barker, a neurobiologist at the Max Delbrück Center for Molecular Medicine in Germany and a coauthor on the new study. “As soon as I met them, I thought ‘Wow, this is a perfect organism to study . . . how social units get organized.’”
Barker chose to study social information contained within the most common of the mole rats’ 17 unique vocalizations: the soft chirp, used as a greeting call. To determine how the call might differ among colonies or individuals, the team trained a machine learning algorithm using more than 36,000 individual vocalizations across seven colonies. The algorithm parsed each call into eight acoustic features, including its pitch, peak frequency, and asymmetry, and identified key patterns. “The logic was that if a machine can learn this, then there’s some features of the data that a mole rat brain could also learn,” Barker tells The Scientist.
The algorithm was able to distinguish between colonies with an accuracy of 74 percent on average, and within a colony, it could identify individual mole rats with an accuracy of between 45 percent and 84 percent. Among the eight variables analyzed in each call, the vocalization’s peak frequency and its asymmetry—or the difference in frequency between the beginning of the call and the end—were the most predictive.
Next, the team tested whether mole rats recognize social information conveyed through the chirps and modify their behavior in response. They placed mole rats in a chamber between two tubes and played calls through the tubes from animals in their own and other colonies. The mole rats responded to all of the calls with their own return chirp, but responded more strongly to calls made in their native dialect. This finding held even when the researchers generated artificial calls that followed native patterns of peak frequency and asymmetry to show that the mole rats were responding to the dialect and not simply a voice they recognized.
This use of machine learning allowed Barker and her team to generate and harness a volume of data that would have required thousands of hours of human effort to analyze even a few years ago, says Sonja Pyott, a molecular neurobiologist at the University Medical Center Groningen in the Netherlands who has published a book chapter with Barker but was not involved in the current study.
A challenge to working with this type of sound data lies in the fact that naked mole rat colonies are constantly abuzz with chatter and noise, making it difficult to parse out individual calls. To get around this, researchers have previously separated mole rats and recorded their calls individually, a practice that doesn’t preserve the social context of the communication. By feeding a colony’s collective vocalizations into the algorithm, which could then identify distinct chirps, the team was able to study single calls while keeping their social context intact. “They really did such a nice, careful analysis using really modern approaches like machine learning,” Pyott tells The Scientist.
To further determine whether these dialects are genetic or learned, Barker carried out a pair of pup transplant experiments in which three young pups were fostered in new colonies before they learned to speak. She and her colleagues then compared their chirps with those of pups raised in their native colonies. All three of the foster pups learned to speak the dialect of their adoptive families, suggesting that these unique vocalizations are learned and not encoded in the genome.
The queen controls a lot of aspects of the colony, and we don’t know how this works.—Alison Barker, Max Delbrück Center for Molecular Medicine
Over the course of the experiment, one colony also underwent a series of coups during which two consecutive queens were killed and replaced by new females. These events gave the team a chance to study how different aspects of the colony, such as the stability of the group dialect, are maintained by the presence of a queen. “This is the million dollar question in naked mole rat research,” Barker says. “The queen controls a lot of aspects of the colony, and we don’t know how this works.”
When the colony was without a queen—a period the authors dubbed “anarchy”—its dialect became more variable, with the predictive power of the algorithm for that colony dropping from 75 percent to 39 percent. This “tantalizing” finding, Barker tells The Scientist, suggests that the queen may be using some sort of hormonal process to suppress superfluous vocalizations—keeping all the workers chirping in the same way—although the exact mechanism remains a mystery.
One big hint in support of the hormone hypothesis came when the researchers looked at individual variations in soft chirps. Certain naked mole rats seemed to have a free pass to be more vocal, and they were almost always breeding males. “We think that the queen is allowing this because the males sing to her during the courtship exchange, and she likes that,” Barker says.
Another burning question to emerge from the study is whether naked mole rats are true production learners that are able to make up new sounds in response to experience, or if they are usage learners, meaning that they mold a rote set of vocalizations to new circumstances, as cats and certain nonhuman primates such as rhesus monkeys do. The study’s findings, while in need of further confirmation, provide some evidence that naked mole rats are modifying their vocalizations—changing the pitch and asymmetry of their calls to conform to their queen—in ways that mirror how humans form speech.
Moving forward, Barker plans to further probe whether naked mole rats truly rely on production learning and how their communications are processed in the brain. “I think the naked mole rats are going to be somewhere in the middle” between true production learning and usage learning, she says. “We know that they have a pretty set vocabulary, but the fact that this one standard call seems to already have within it some frequency modulation is a bit more than we see in some of these other species that are purely usage learners.” Unraveling mole rat communication may go a long way toward understanding how language evolved alongside social complexity.
A. Barker et al., “Cultural transmission of vocal dialect in the naked mole-rat,” Science, doi:10.1126/science.abc6588, 2021.