Animals communicate in ways that are far more structured, specific, and flexible than most people realize. Some species use distinct sounds to label different predators, others combine calls in rule-governed sequences, and a few even address each other by something resembling a name. Whether any of this counts as “language” depends on where you draw the line, and that line has become blurrier with each new discovery. The honest answer is that no animal communication system matches human language in its full complexity, but several come startlingly close on individual features that were once considered uniquely human.
Alarm Calls That Label the World
One of the clearest examples of meaning in animal calls comes from vervet monkeys in East Africa. When a vervet spots a leopard, it produces one type of alarm call; when it spots an eagle, it produces a different one; and when it sees a snake, it produces a third. These are not just generic shrieks of fear. In a landmark study at Amboseli, Kenya, researchers played back recordings of each alarm type when no predator was present and filmed the monkeys’ reactions. Leopard alarms sent them scrambling into trees, eagle alarms made them look up, and snake alarms made them look down. The length or loudness of the call did not change the type of response, and the surrounding context was not a systematic factor either. The monkeys were responding to what the call meant, not how alarmed the caller sounded.1Animal Behaviour. Vervet monkey alarm calls: Semantic communication in a free-ranging primate More recent quantitative analysis confirmed that these three call types are acoustically distinct enough to be classified correctly over 90% of the time for both males and females.2Scientific Reports. Vervets revisited: A quantitative analysis of alarm call structure and context specificity
Prairie dogs push this even further. Gunnison’s prairie dogs produce alarm calls that differ depending on not just the type of threat but also its physical characteristics. Research has shown that the acoustic structure of their calls varies with the size and shape of an approaching silhouette.3Ethology. Communication of Stimulus Size and Shape in Alarm Calls of Gunnison’s Prairie Dogs In a separate experiment, prairie dogs produced distinguishable alarm calls for humans wearing different-colored shirts, with the calls for blue and yellow being significantly different from each other. The colors they could encode corresponded to what their visual system could actually perceive.4Animal Cognition. Prairie dog alarm calls encode labels about predator colors In other words, these rodents are not just saying “danger.” They are describing the danger with something like adjectives.
Combining Calls With Rules
One feature long considered unique to human language is syntax: the idea that the order of elements changes the meaning. You understand “the dog bit the man” differently from “the man bit the dog” because English has word-order rules. For decades, no animal was known to do anything comparable. Then researchers found it in a small bird.
Japanese tits produce different note types that they string together into sequences. Two of the key building blocks are what researchers labeled “ABC” notes and “D” notes. When tits hear ABC calls alone, they scan their surroundings for danger. When they hear D calls alone, they approach the sound source, a recruitment behavior. When the two are combined in the specific order ABC-D, the birds do both: they scan and then approach. But when researchers reversed the order to D-ABC, the combined response disappeared. The meaning depended on the sequence, not just the presence of the component parts.5Nature Communications. Experimental evidence for compositional syntax in bird calls This was recognized as the first firm evidence of compositional syntax in a nonhuman animal.6PLOS Biology. Call combinations in birds and the evolution of compositional syntax
Even more striking, the tits could apply this ordering rule to entirely novel sequences. When researchers replaced the D call with a functionally similar recruitment call from a different species, the birds still responded with the combined behavior, but only when the foreign call came after the ABC notes, not before. They were not just memorizing fixed sequences. They were applying a rule about what order means.7Current Biology. Wild Birds Use an Ordering Rule to Decode Novel Call Sequences
Calling Each Other by Name
Bottlenose dolphins develop individually distinctive signature whistles early in life and use them throughout adulthood. These are not just voices that happen to sound different from one dolphin to the next. Researchers demonstrated that dolphins extract identity information from signature whistles even when all voice features, such as pitch and timbre, have been digitally removed. The dolphins recognized who the whistle belonged to based purely on its melodic shape. This made dolphins the only animals besides humans known to transmit identity information independent of the caller’s voice or location.8PubMed Central. Signature whistle shape conveys identity information to bottlenose dolphins
What makes this especially interesting is that dolphins copy each other’s signature whistles, apparently to address specific individuals. In a study of wild bottlenose dolphins, animals responded to hearing a copy of their own signature whistle by calling back, but did not respond to whistles that belonged to other dolphins. The copied whistle functions as a label, much like using someone’s name to get their attention.9PubMed Central. Bottlenose dolphins can use learned vocal labels to address each other This is a far cry from a generic contact call. It implies that dolphins operate in a social world where individuals are tracked, referred to, and specifically addressed.
Gestures and Intentional Messaging in Apes
Great apes are not especially impressive vocalists by animal standards. Their vocal repertoires are relatively small and largely innate. Where apes shine is gesture. Wild chimpanzees in the Budongo Forest of Uganda were observed producing 66 distinct gesture types over the course of a long-term study, and the repertoire appeared close to the total number the species uses. Chimpanzees deploy these gestures flexibly across different social contexts and adjust them depending on whether the recipient is paying attention: they use silent visual gestures when the other chimp is watching and physical contact gestures when the other chimp is looking away.10PubMed. The gestural repertoire of the wild chimpanzee
This sensitivity to the audience is one marker of what researchers call intentional communication: the signaler is not just emitting behavior reflexively but is directing it at a specific recipient with a specific goal in mind. Studies of chimpanzee infants have tracked the developmental arc of these markers, including audience checking, persistence toward a goal when the first attempt fails, and adjusting signals based on whether the recipient is attending. All three markers increased with age within individual chimps.11PubMed. The ontogeny of intentional communication in chimpanzees in the wild Both bodily and manual gestures were used intentionally to achieve goals, though manual gestures showed a qualitatively different form of flexibility, with signalers more likely to persist and try alternative gestures when the first attempt failed.12PubMed. The repertoire and intentionality of gestural communication in wild chimpanzees
Ape language acquisition studies have also demonstrated that apes raised with human-designed symbol systems can learn arbitrary associations between symbols and concepts. But their production of multi-symbol combinations remains quite limited compared to human language.13PubMed Central. Evidence of Grammatical Knowledge in Apes: An Analysis of Kanzi’s Performance on Reversible Sentences This gap is revealing: apes clearly have the cognitive architecture for flexible, goal-directed communication, but their natural systems and even their trained ones do not build up the kind of open-ended, recursive sentence structures that human children acquire effortlessly.
Songs That Travel Between Populations
Male humpback whales produce long, elaborate vocal displays known as songs. Within any one population, all the males converge on the same version of the song at any given time. But the songs are not static. They evolve progressively, with themes being added, dropped, and modified. Across the western and central South Pacific, researchers documented a striking pattern over an 11-year period: new song types spread repeatedly in one direction, eastward, rippling through multiple populations like a cultural wave. This was the first documentation of such large-scale, repeated cultural transmission in any nonhuman species.14PubMed. Dynamic horizontal cultural transmission of humpback whale song at the ocean basin scale
The fidelity of this cultural transmission is itself noteworthy. When a song type moved from the east Australian population to New Caledonia, its complexity did not degrade during the transfer. Simple songs stayed simple and complex songs stayed complex, suggesting that the whales were copying the structural features of the songs with considerable accuracy.15Scientific Reports. Song complexity is maintained during inter-population cultural transmission of humpback whale songs Killer whales show a parallel phenomenon. Different matrilines maintain their own dialects, and the structure of their calls drifts over time. There is evidence that modifications can spread horizontally between groups rather than only being passed from mother to offspring, implying that killer whale vocal traditions involve active learning from peers.16Animal Behaviour. Dialect change in resident killer whales: implications for vocal learning and cultural transmission
Communication Beyond Sound
Not all animal communication systems use sound. Honey bees communicate the location of food, water, and nest sites through their waggle dance, which encodes distance as the duration of a waggle run and direction as the angle of the dancer’s body relative to vertical on the comb.17PubMed. Incorporating variability in honey bee waggle dance decoding improves the mapping of communicated resource locations The dance integrates information about celestial cues, optic flow, and food quality into a physical performance that other bees can decode. Research has shown that this is partly a socially learned skill: bees that grow up without experienced dancers to learn from produce less precise dances.18PubMed. Social signal learning of the waggle dance in honey bees
Cephalopods like cuttlefish and octopuses communicate visually through rapid changes in skin color and texture. Their chromatophores, tiny pigment-containing cells controlled by the nervous system, can be coordinated to produce pulsing, wave-like patterns across the body. In some species, these dynamic displays include rhythmic “flashing” or “strobing” signals, with moving bands of contrast flowing across the mantle in coordinated sequences.19Frontiers in Physiology. Dynamic Skin Patterns in Cephalopods The speed and complexity of these displays are astonishing, and they serve functions ranging from camouflage to courtship to aggression, all without producing a single sound.
Elephants, meanwhile, rely on a multimodal approach. They combine vocalizations with gestures such as ear-flapping, trunk-reaching, and body postures, and the specific combinations they use vary depending on the individual and social context. A recent study of greeting behavior in African savannah elephants found that these vocal and gestural signals are not deployed randomly but are integrated into meaningful multicomponent combinations.20Communications Biology. Multimodal communication and audience directedness in the greeting behaviour of semi-captive African savannah elephants Elephants also produce low-frequency rumbles near the bottom of the human hearing range, and these infrasonic calls can carry over considerable distances, allowing communication between individuals that cannot see each other.21PubMed. Long-distance, low-frequency elephant communication However, the identity-carrying features of these calls may become distorted over long distances, placing limits on social recognition at range.22Animal Behaviour. Long-distance communication of acoustic cues to social identity in African elephants
A Shared Genetic Foundation for Vocal Learning
One reason some species can learn new vocalizations while others cannot traces back, at least partly, to shared genetics. The gene FoxP2 is involved in brain circuits that underlie sensory-guided motor learning. In humans, mutations in this gene cause severe speech and language disorders. Expression of FoxP2 in the basal ganglia is necessary for normal vocal learning and production in both humans and songbirds.23PubMed. Using FoxP2 to Distinguish Direct and Indirect Basal Ganglia Pathways for Vocal Learning in Songbirds Research across primates, rodents, and birds suggests that FoxP2 and other language-related genes participate in neural networks underlying not just vocal learning but also symbolic understanding and social cognition.24PubMed Central. Genes and vocal learning
The FoxP2 story matters because it suggests that vocal learning, one of the prerequisites for anything like language, has deep evolutionary roots. Learned vocal behavior evolved independently in songbirds, parrots, hummingbirds, bats, whales, and a handful of other groups. The gene network shaping the brain circuits for this ability shows similarities across these otherwise very different lineages.25PubMed Central. Evo-devo, deep homology and FoxP2: implications for the evolution of speech and language This does not mean FoxP2 is a “language gene” in any simple sense, but it does mean that the biological toolkit for learning complex vocalizations is not unique to humans. It has been assembled, in partly overlapping ways, multiple times across the animal kingdom.
Why Researchers Hesitate to Call It “Language”
If animals can name predators, follow ordering rules, address each other individually, and pass vocal traditions across generations, why do most scientists still stop short of calling these systems languages? The standard framework for comparing animal communication with human language has historically been a checklist of design features proposed by the linguist Charles Hockett in the mid-twentieth century. These features include things like arbitrariness (the signal bears no physical resemblance to its meaning), displacement (referring to things not present), and productivity (creating novel messages). Human language checks all the boxes; animal systems each check some but not others.
However, the Hockett framework itself has come under criticism. It was developed as a way to describe how human language works and then used, somewhat awkwardly, to judge animal systems by that standard. Modern researchers in language evolution have argued that while Hockett’s features remain useful for description, they are of limited value as a theoretical framework for understanding how communication evolves.26PubMed Central. Language Evolution: Why Hockett’s Design Features are a Non-Starter The problem is not that the features are wrong but that treating them as a pass/fail test obscures how much animals actually do. A vervet monkey’s alarm call system exhibits something very much like arbitrariness and semantic reference, but it lacks productivity: vervets cannot invent new calls for new predators. Dolphin signature whistles show something like naming, but there is no evidence dolphins combine those names into sentences. Each system has pieces of what we recognize as language, assembled in a different configuration.
The more productive way to think about it, and the direction the field is moving, is to ask what specific capacities different species share with human language rather than whether any one species has “the whole package.” When the question shifts from “do they have language?” to “what language-like abilities have evolved, and how?”, the answers become genuinely interesting rather than just a list of failures to be human.
Machine Learning and the Race to Decode Animal Communication
Artificial intelligence is opening a new chapter. Sperm whales communicate using patterns of clicks called codas, and the structure of these exchanges has long been difficult for human researchers to parse. A recent study trained a neural sequence model on sperm whale vocal exchanges and found that the whales’ vocalizations show order dependence and long-range dependencies stretching back up to eight codas in a conversation. The model also found predictable turn-taking patterns. When trained to predict behavior from vocal exchanges, it could identify the current behavioral context with about 72% accuracy and predict the whales’ next actions with about 86% accuracy.27bioRxiv. WhaleLM: Finding Structure and Information in Sperm Whale Vocalizations and Behavior with Machine Learning These findings provide the first evidence that sperm whale vocalizations contain enough structured information to plausibly coordinate group behavior, something that had long been suspected but never demonstrated.
The broader ambition behind projects like the Cetacean Translation Initiative is to apply the same machine learning tools that have been so successful with human language to animal communication. Researchers have proposed using large-scale collection of bioacoustic, behavioral, and environmental data, processed by models trained to find patterns that human ears and eyes miss.28arXiv. Cetacean Translation Initiative: a roadmap to deciphering the communication of sperm whales The technology is still in its early stages, and “decoding” animal communication does not mean building an interspecies Google Translate. What it does mean is that researchers can now detect statistical regularities in animal vocal streams that would have been invisible to previous methods. Whether those regularities turn out to be language-like or something altogether different is one of the more genuinely open questions in biology right now.
Bats, Context, and the Information Hiding in Sequences
One underappreciated finding involves bats. Egyptian fruit bats produce vocal sequences made up of different syllable types, and a recent study used machine learning to test whether the order of syllables within those sequences mattered. It turned out that context, meaning whether the bats were feeding, disputing space, or engaged in mating interactions, could be identified from the vocal sequences well above chance. Sequences of seven syllables allowed the model to identify the social context with roughly twice the accuracy of a random guess.29PubMed Central. Bat vocal sequences enhance contextual information independently of syllable order
What made this finding unusual was that the contextual information seemed to build up across the sequence in a way that did not depend strictly on syllable order. The accumulation of information was more like adding brushstrokes to a painting than following a grammatical rule. This sits in an interesting middle ground: it is more than random noise, clearly, but it also is not syntax in the way the Japanese tit example is. It is a reminder that animal communication systems do not all have to work the same way to carry real information. Some may organize meaning through sequence order, others through cumulative statistical properties, and others through channels humans have barely begun to investigate.