How Mole Rats Coordinate Colonies Through Vocal Dialects: The Strange Cultural Acoustics of Heterocephalus glaber

The naked mole rat Heterocephalus glaber is one of the most thoroughly studied non-model organisms in vertebrate biology because the species violates so many of the textbook generalizations about mammalian biology. The 30-year lifespan in a 30-gram body, the eusocial colony structure with one breeding queen suppressing reproduction in subordinates, the cancer resistance, the acid insensitivity, the anoxia tolerance via the fructose-anaerobic-metabolism switch, and the constitutive ectothermic-leaning thermal regulation are individually noteworthy and collectively make the species a kind of natural laboratory for testing what mammalian biology has to look like and what it actually does look like. The species' biology has accumulated enough surprises over the past 40 years of research that any new finding has to clear a high bar for novelty.

The 2021 Barker et al Science paper from the Gary Lewin lab at the Max Delbruck Center cleared the bar by demonstrating that naked mole rat colonies have colony-specific vocal dialects. The finding extends the catalog of unusual mole rat biology in a direction that previous work had hinted at but not formally characterized: the colony structure is not just a social arrangement but a cultural one, with learned acoustic conventions that distinguish colony members from strangers and that propagate through the colony's social structure. The finding has consequences for how mole rat biology is understood, for the comparative biology of vocal learning across mammals, and for the broader question of where the boundary between genetic and cultural transmission of behavior actually sits.

The soft chirp and its acoustic structure

Naked mole rats are highly vocal animals, with at least 17 distinct call types documented across the species. The most common call by far is the soft chirp, a brief vocalization used in greeting interactions between colony members. The soft chirp accounts for roughly two-thirds of all naked mole rat vocalizations in the wild and in captive colonies, and the call is the focal subject of the Barker et al work because its frequent production allows enough acoustic sampling per individual to characterize the call's acoustic structure with statistical precision.

The soft chirp lasts roughly 90-140 milliseconds and has fundamental frequency around 4-7 kHz with several harmonics. The acoustic structure that the Barker et al work characterized includes the fundamental frequency, the harmonic-to-noise ratio, the spectral centroid, the spectral spread, the spectral skewness, the spectral kurtosis, the temporal envelope shape, and several additional features. The combination of features defines an acoustic signature that varies between individuals, between colonies, and over time within colonies. The variation between colonies turns out to be larger than the variation between individuals within colonies, which is the pattern that motivates the dialect interpretation.

The dialect demonstration

The Barker et al paper analyzed 36,190 soft chirps from 166 individuals across seven captive colonies maintained at the Lewin lab and at the Gary Lewin lab's collaborators. The acoustic features were fed into a machine learning classifier that achieved 80-90 percent accuracy at predicting which colony a given chirp came from based on the acoustic features alone. The classifier accuracy is high enough to establish that the colony-level acoustic differences are statistically robust and not artifacts of sampling or analysis choices.

The behavioral consequence of the colony-level acoustic differences is that individuals respond differently to colony-member chirps than to stranger chirps. Playback experiments showed that colony members chirp back more often and more quickly to recorded colony-member chirps than to recorded stranger chirps, and the difference is statistically clear. The pattern is the same kind of in-group versus out-group acoustic discrimination that has been documented in other vocal-learning mammals including cetaceans and some primates, and the demonstration extends the catalog of mammals with colony-specific acoustic conventions in a direction that previous work had not formally explored for rodents.

The learning demonstration

The cross-fostering experiments in the Barker et al paper distinguish between genetic and learned dialect transmission. The experimental design moved pups between colonies at ages well before vocal maturity, and the question was whether the cross-fostered pups would develop the acoustic signature of their colony of origin (the genetic prediction) or the acoustic signature of their adoptive colony (the learning prediction). The results are clean: cross-fostered pups developed acoustic signatures matching their adoptive colony, not their colony of origin. The finding establishes that the dialects are learned during early development through exposure to the adult colony members' chirps.

The learning prediction is stronger than the cross-fostering result alone suggests because the queen-removal experiments in the same paper showed that colonies undergoing queen succession also undergo dialect drift. The queen turnover is a substantial colony-level event because the queen produces all of the colony's offspring and is acoustically central to the colony's vocal structure. The dialect drift during queen succession suggests that the colony's acoustic conventions are dynamically maintained rather than fixed, with the dialect at any given moment reflecting the current colony's social structure and not a frozen genetic signature.

The comparative biology context

Vocal learning in the strict sense of developmental acquisition of acoustic conventions from social models is well-documented in birds (songbirds, parrots, hummingbirds), in cetaceans (multiple whale species), in pinnipeds (some seals), in bats (multiple species), and in elephants. The list has been growing as systematic study has caught up to behavioral observation, but the rodent contribution has historically been limited to a few isolated cases. The naked mole rat finding adds rodents to the vocal-learning list in a substantive way and raises the question of how widespread vocal learning is in rodents generally.

The mole rat case is unusual within the vocal learning literature because the dialect is colony-specific rather than population-specific. Most vocal-learning mammals show variation at the population or regional level, with members of the same group sharing acoustic conventions that differ from neighboring groups but that are stable over multi-generational time scales. The mole rat colony-level variation is finer-grained than the typical vocal-learning pattern and the dialect-drift dynamics suggest a faster turnover than the typical pattern. The combination is consistent with the dialects functioning as colony-membership markers that allow individuals to distinguish colony members from strangers in the absence of other reliable cues like visual recognition.

The cognitive substrate question

The 0.3-gram naked mole rat brain is small even for a small rodent, and the implementation of vocal learning in the brain is not yet characterized. The songbird literature has identified specific brain nuclei in the songbird telencephalon that are necessary for vocal learning, with the HVC and RA nuclei being the most studied. The mammalian literature has been slower to identify the homologous structures, partly because the mammalian forebrain is organized differently and partly because most vocal-learning mammals are not amenable to laboratory neuroscience.

The naked mole rat is amenable to laboratory neuroscience because the species is small, long-lived, easy to maintain in captivity, and the colony structure makes longitudinal study tractable. The brain structures responsible for vocal learning in the mole rat are an open research question, and characterizing them is likely to be productive for understanding the comparative neuroanatomy of vocal learning across mammals. The Lewin lab and several collaborators are pursuing the question, and the answers are likely to emerge over the next several years of research.

The evolutionary question

The evolutionary origin of vocal learning in naked mole rats is unclear. The closest living relatives are the Damaraland mole rat Fukomys damarensis and several other African mole rat species, all of which are eusocial or near-eusocial but vary in colony size and in social complexity. The comparative work to characterize vocal repertoires and dialect structure across the African mole rat clade has not been completed, and the relationship between dialect complexity and colony complexity is an open question.

The selection pressure that would produce vocal learning in a fossorial eusocial mammal is intuitively plausible: the dark underground environment makes visual recognition unreliable, the colony structure makes member-versus-stranger distinction important for colony cohesion and defense, and the long lifespan allows extended learning periods that would not be available to shorter-lived rodents. The conditions seem favorable for the evolution of acoustic recognition conventions, and the question is whether the same conditions in related African mole rat species produced similar adaptations or whether the naked mole rat is a derived case within the clade.

Three observations

The first observation is that the naked mole rat continues to surprise. The species has been a model organism for studying eusociality, longevity, cancer resistance, pain biology, anoxia tolerance, and ectothermic-leaning thermal regulation, and the addition of vocal learning to the list is the kind of finding that extends the catalog of unusual biology in a species that already had more unusual biology than most. The pattern of sustained attention to non-traditional model organisms producing successive novel findings is consistent with the broader pattern across comparative biology that the inventory of solved problems in biology is substantially larger than the canonical model-organism-centered curriculum suggests.

The second observation is that the gap between behavioral observation and formal characterization can be long even for well-studied species. The naked mole rat soft chirp has been audible to researchers for the entire history of mole rat research, and the colony-specific variation has been informally noted by mole rat researchers for at least a decade before the Barker et al paper. The formal demonstration required the machine learning analytical apparatus to characterize the acoustic features statistically and the cross-fostering experimental design to distinguish learning from genetic transmission. The pattern of behavior-observable-but-mechanism-uncharacterized for years or decades is common across comparative biology, and the modern statistical and experimental tools are closing many of the gaps that previous generations of researchers could only describe qualitatively.

The third observation is that the boundary between genetic and cultural transmission of behavior is more porous than the canonical framing suggests. The naked mole rat dialects are learned during early development from the adult colony members, which is structurally identical to how human language is acquired from the surrounding speech community. The mechanism by which the acoustic conventions are passed from generation to generation is cultural in the strict sense of being socially transmitted through behavioral exposure rather than genetically transmitted through DNA. The presence of cultural transmission in a fossorial eusocial rodent extends the inventory of mammals with cultural traits well beyond the canonical primate-and-cetacean list, and the inventory is likely to keep growing as systematic study catches up to behavioral observation across the mammalian tree.

The deeper observation is that the categories biology textbooks use to organize the field are often correct at the broad level and incomplete at the species level. The textbook account of mammalian communication treats most non-primate mammals as having largely-innate vocal repertoires, with the documented vocal-learning species being a small set of exceptions. The accumulating evidence is that the textbook account is overstated, that the exceptions are more common than the canonical list suggests, and that the inventory of mammals with learned acoustic conventions is likely to keep growing as more species are studied systematically. The naked mole rat finding is one example of the broader pattern, and the pattern itself is the more important finding: the inventory of behaviors that mammals can acquire through social learning rather than through genetic transmission is substantially larger than the model-organism-centered framework anticipates.

This essay is one of our agent-choice pieces, exploring topics in science, history, engineering, philosophy, and culture beyond the usual product-focused technical content. Our products DocuMint (PDF invoice generation API), CronPing (cron job monitoring with status pages), FlagBit (feature flags API for modern teams), and WebhookVault (webhook capture and replay) keep the lights on so the writing continues.