In 2021, Berlin teams studied 166 naked mole-rats and found they recognised colony-specific vocal dialects; cross-fostered pups learned the dialect of their new colony

A team based at Berlin’s Max Delbrück Center for Molecular Medicine (MDC), working with colleagues at the University of Pretoria in South Africa, set out to answer a basic question about one of the animal kingdom’s stranger societies: do naked mole‑rats’ constant chirps carry real information, or are they just noise, as described in a press release from the Max Delbrück Center. Over two years, the researchers recorded 36,190 individual ‘soft chirps,’ the greeting call naked mole‑rats use most often, from 166 animals across seven colonies housed in labs in Berlin and Pretoria, according to the Max Delbrück Center’s description of the study and the original Science paper. The findings appeared in the study 'Cultural transmission of vocal dialect in the naked mole‑rat', published in Science on January 14, 2021, which suggests vocal dialects can be learned.

Naked mole‑rats live in large, multigenerational colonies ruled by a single breeding queen, more closely resembling an ant or termite nest than a typical rodent burrow, as described in a review of their eusocial biology published in PMC. With food scarce underground and colonies fiercely territorial toward outsiders, the researchers wanted to know how individual animals keep track of who belongs and who doesn't.

A machine-learning classifier linked chirp features to individual and colony identity


In order to categorize tens of thousands of chirps, the research team used a random forest classifier, a form of machine learning, which was trained using eight acoustic characteristics of each soft chirp, three from the original waveform and five from its spectrogram, as detailed in the study's supplementary materials. After training, the classifier was capable of recognizing which specific mole rat the chirp came from. When the analysis was limited to just four colonies, three from Berlin and one from Pretoria, the same classifier could sort chirps by colony rather than by individual, relying most heavily on two of its eight acoustic features: asymmetry and peak frequency.

Mole-rats responded far more often to playbacks of their own colony's dialect

The individual mole-rats were put in a box with two chambers and the researchers played recordings of chirps made either by members of their own colony or members of a different colony. Across at least 36 individual playback trials per animal, more than 5,900 trials in total across the 166 mole-rats, the researchers found that the animals chirped in response to the chirps of the home colony much more frequently than those of the alien colony, a difference they report as highly statistically significant. Later, Alison Barker, the study's lead author, said in interviews that the animals respond to their own colony's dialect nearly every time but largely ignore other dialects.

To ensure that the animals were not responding to a specific voice rather than a common dialect, the researchers used computer‑generated ‘mock chirps’ which consisted only of the characteristic asymmetry and peak frequency of a colony, without matching any individual animal's voice. The animals responded more strongly to mock chirps that contained features of the home colony despite having bedding of alien colonies placed inside the box.

Cross-fostered pups adopted the dialect of their new colony within months

To see whether the dialects were learned or innate, the researchers conducted a cross-fostering experiment, placing orphaned pups in different colonies right after they were born. When these pups grew up, the researchers found that they had indeed picked up the dialect of their foster colony, with the rate of accuracy ranging from 59% to 95% in three successful experiments. The younger the pups when placed in the foster colony, the closer they were to their new colony’s dialect.

Colony dialects became less consistent after the queen died

There was one particular colony that provided an unintentional experiment for testing how dialect cohesion works. During the study, the colony lost its queen, then lost her replacement as well, a sequence described in news coverage of the work that notes dialect cohesiveness fell during these ‘anarchy’ periods and recovered only after a new queen took charge. According to the study mentioned above, the researchers discovered that in this period when there were no queens in the colony, individual chirps became more variable, and the accuracy of classification of colonies decreased, improving only when a new queen appeared.

A companion analysis placed the finding among a small list of vocal-learning mammals

In addition to the original research article, Science magazine featured a commentary on this discovery in the same January 2021 issue, called ‘Colony-specific dialects of naked mole-rats,’ by Rochelle Buffenstein of Calico Life Sciences. In this commentary, Buffenstein places this discovery within the context of the relatively small group of mammals that are capable of learning their sounds, as opposed to inheriting their sounds, which, aside from naked mole-rats, includes only bats, cetaceans, elephants, and primates. Moreover, this particular capability appears to be found in naked mole-rats among mole-rats, as a number of other mole-rat species apparently depend more on scents than sounds to distinguish colony members.