In 1880, a naturalist in Vermont noted that a particular crow seemed to use one call when a hawk appeared and a different call when a fox trotted through the meadow. He mentioned this observation in a letter. Nobody followed up on it for about a century.
Today, a crow at the University of Tubingen sits before a screen, watches a colored numeral flash, and produces exactly three calls. Not two, not four. Three. Then it pecks a button to say: I’m done. The acoustic signature of its very first call already contains the number it intends to reach, as though the entire sequence was planned before the beak opened.
Between that Vermont meadow and that German laboratory, something enormous happened. We started listening.
The Voice Box That Sits in the Wrong Place
Every mammal on Earth produces sound with its larynx, a structure positioned at the top of the trachea that uses vocal folds vibrated by exhaled air. Birds took a different path. The avian vocal organ, the syrinx, sits at the bottom of the trachea, right where it forks into the two bronchi leading to the lungs. This positioning isn’t random. A 2019 study by Tobias Riede and colleagues, published in PLOS Biology, showed through physical and computational modeling that a sound-producing organ at the syringeal position is significantly more efficient than one at the laryngeal position, because of the dynamics between inertance and tracheal length.
But the real trick is structural. Because the syrinx straddles the bronchial fork, it contains two independent sound sources, one in each bronchus, each controlled separately by the brain. Some birds can produce two different sounds simultaneously. Songbird syringeal muscles are among the fastest vertebrate muscles in any species, capable of modulations too rapid for the human ear to parse.
Corvids, the family that includes crows, ravens, jays, magpies, and about 135 species total, put this machinery to extraordinary use. A 2025 study in Animal Cognition found that 39 species of corvids across 13 genera are known to vocally mimic other species, representing about 30% of all corvid species. Ravens mimic human speech. Crows imitate car alarms and cell phone ringtones. And critically, corvids are open-ended vocal learners: unlike many songbirds that must finalize their repertoire during a juvenile sensitive period, corvids can acquire new vocalizations throughout their entire lives.
The syrinx does not explain what they choose to say with it. That requires a different kind of hardware.
More Than Caw
Kevin McGowan of the Cornell Lab of Ornithology has spent decades cataloging American crow vocalizations. His assessment is blunt: the familiar “caw-caw-caw” is a fraction of the repertoire. American crows produce more than twenty distinct call types, including rattles, clicks, clear bell-like notes, and a quiet, rambling “subsong” given at rest that mixes hoarse coos, caws, and grating noises in long improvised sequences.
The known categories form a system. Short, sharp, explosive barks serve as alarm calls in response to hawks, owls, or cats. Loud, rapid cawing recruits other crows to mob a predator. Softer, more varied calls bring others to a food source. Distress calls, produced under physical capture, are harsh and unmistakable. Contact calls between paired birds are quiet and short. Territorial calls run longer and lower-pitched at boundary disputes.
Ravens take this further. Studies have documented between 15 and 33 vocalization categories for the common raven, with one spectrographic analysis distinguishing 79 call types. Beyond the deep, resonant “kronk” that distinguishes them from crows, ravens produce knocking sounds, gurgles, high-pitched trills, and metallic “tok” sounds that no crow can manage. Bernd Heinrich, in his book Ravens in Winter (1989), documented how ravens produce specific food recruitment calls, far-reaching “yells” that bring other ravens to a carcass they cannot monopolize alone.
But quantity isn’t the point. What matters is whether these calls carry meaning. And the evidence says they do.
They Know Who’s Talking
In 2015, a team led by Elizabeth Mates published a study in Bioacoustics that analyzed calls from 18 wild, marked American crows using 23 pitch-related and spectral parameters. The result: individual crows can be identified from their calls alone, with 24% classification accuracy across all caw-type calls, against a 6% chance baseline. This means a crow’s voice carries its identity even when you don’t know what it’s doing.
Sex matters too. A 2006 study by Yorzinski and colleagues in The Condor found that female caws tend to have higher pitch, shorter duration, and more peaked contours than males. Using 25 acoustic variables, their analysis achieved 65% accuracy in distinguishing individual crows. One subordinate male clustered acoustically with the females, raising the possibility that social status partially shapes vocal characteristics.
Then there’s the recognition itself. In 2012, Kondo, Izawa, and Watanabe published a study in Proceedings of the Royal Society B that demonstrated cross-modal recognition in large-billed crows. The experimental design was elegant: a crow saw a group member (visual input), then heard a contact call (auditory input). When the voice didn’t match the face, crows looked more rapidly and for a longer duration, the same expectancy violation response that human infants show. But this worked only for group members. Unfamiliar crows triggered no such response. The birds had built mental representations linking specific voices to specific faces, but only for individuals that mattered socially.
This isn’t just hearing. It’s knowing.
Gestures, Grammar, and the Question of Syntax
In 2011, Simone Pika and Thomas Bugnyar published a paper in Nature Communications that changed the conversation about animal communication. Over three consecutive field seasons in the Austrian Alps, they recorded 38 social interactions between 7 raven dyads. What they documented were referential gestures: ravens showing and offering non-food objects (moss, twigs, small stones) to other ravens, predominantly to partners of the opposite sex. Recipients responded by orienting to both the object and the signaler, and interactions frequently led to affiliative behaviors, sitting close, touching.
Before this study, referential gestures (the pointing and showing that human infants develop around 9 to 12 months, before speech) had been documented only in great apes. Ravens were the first non-primate species to demonstrate them.
The structure of calls themselves also follows rules. In 2025, Claudia Wascher and Sean Youngblood published the first evidence of Menzerath’s law in corvid vocal communication. Menzerath’s law is a linguistic principle: in longer sequences, constituent elements tend to be shorter. (Longer sentences have shorter words; longer words have shorter syllables.) They studied carrion crows, hooded crows, and hybrids, and found that crow call sequences adhere to this pattern. Shorter calls occur in longer sequences. The effect was stronger in females and younger individuals, suggesting individual traits shape vocal efficiency.
The deeper question is whether corvids combine calls to create new meanings, true compositional syntax. A 2016 study by Suzuki, Wheatcroft, and Griesser demonstrated compositional syntax in Japanese great tits: “ABC” notes mean “scan for danger,” “D” notes mean “approach the caller,” and the combination “ABC-D” produces a compound meaning (scan AND approach) that disappears when the order is reversed. If a small passerine can do this, corvids, with larger brains, more complex social lives, and richer vocal repertoires, almost certainly have more sophisticated combinatorial systems. But formally demonstrating it remains one of the great open challenges in animal cognition research.
The Crow That Counted
In May 2024, Diana Liao, Katharina Brecht, Lena Veit, and Andreas Nieder at the University of Tubingen published a study in Science that no one had anticipated. Three carrion crows had been trained to produce a specific number of vocalizations, one to four, in response to arbitrary visual and auditory cues. After producing the correct count, each crow pecked a button to signal completion.
The results were striking on three levels. First, the crows could do it at all. Producing a planned number of vocalizations requires numerical cognition (understanding quantity), vocal motor control (producing the right number of sounds), and working memory (keeping track while executing). No non-human animal had previously demonstrated this degree of voluntary numerical vocal control.
Second, the acoustic features of the very first call in each sequence predicted how many total calls the crow intended to produce. The count was planned before the first sound left the beak.
Third, errors were detectable from the acoustic signature. A call that should have been “second” but had the spectral characteristics of a “third” indicated the crow was losing track, the same kind of counting error a toddler makes.
The same lab had published an equally startling finding two years earlier. In November 2022, Liao and Nieder showed in Science Advances that carrion crows could generate center-embedded recursive sequences: patterns like {[()]}. Recursion, the ability to embed structures within structures of the same type, is what Noam Chomsky proposed as possibly the one feature unique to human language. The crows selected recursive structures in approximately 40% of trials, on par with human children and outperforming macaque monkeys on the same task.
A 2016 PNAS study by Olkowicz and colleagues found that corvid brains contain far more neurons per gram than mammalian brains of similar size. A raven brain weighing about 14 grams packs roughly 1.2 billion neurons into the pallium alone, comparable to some primates. And the corvid brain uses a fundamentally different architecture, a nuclear (clustered) pallial organization rather than the layered mammalian neocortex. This is convergent cognitive evolution: two separate lineages arriving at similar cognitive abilities through entirely different neural structures.
Tools, and What They Tell Us About the Mind Behind the Calls
The connection between tool use and communication isn’t obvious, but it’s deep. Both require the same cognitive foundations: planning, causal reasoning, means-end thinking, and mental representation of goal states.
In 1996, Gavin Hunt published a paper in Nature documenting something unprecedented in a wild, free-living non-human animal. New Caledonian crows manufactured and used tools with a high degree of standardization, distinct discrete tool types, and the use of hooks. Their stepped-cut pandanus leaf tools, wide at the base and tapering to a fine working tip through a planned series of rips and cuts, showed what Hunt called features that “only first appeared in the stone and bone tool-using cultures of early humans after the Lower Palaeolithic.”
In 2002, a captive New Caledonian crow named Betty at the University of Oxford spontaneously bent a straight piece of garden wire into a hook and used it to retrieve a food-baited bucket from a vertical tube. Out of ten successful retrievals, Betty bent wire into a hook nine times. This was the first observation of spontaneous tool manufacture from a novel material by any animal.
In 2018, Auguste von Bayern and colleagues showed that New Caledonian crows could combine non-functional elements into compound tools. One crow constructed three- and four-piece compound tools when the task required it. Previously, this had been observed only in humans and great apes.
The Aesop’s fable test brought things full circle. In 2014, Jelbert, Taylor, Cheke, Clayton, and Gray published a study in PLoS ONE showing that New Caledonian crows dropped stones into water-filled tubes to raise the water level and reach a floating reward. They chose sinking objects over floating ones, solid objects over hollow ones, and tubes with higher water levels. Their performance rivaled that of 5- to 7-year-old children.
A mind that can plan a four-piece compound tool, count to four out loud, and recognize recursive patterns is not a mind producing random noise when it opens its beak. The episodic-like memory demonstrated in scrub jays, who remember what they cached, where, when, and who was watching, points to the same conclusion: corvid cognition operates at a level that demands we take their vocalizations seriously as communication, not dismiss them as instinct.
A Funeral That Teaches
When a crow discovers a dead crow, something happens that looks, to a human observer, unmistakably like mourning. The discoverer gives loud alarm calls. Other crows arrive rapidly. Dozens of birds gather, calling loudly, circling above or near the body. The cacophonous aggregation can last from several minutes to over an hour.
Kaeli Swift, working under John Marzluff at the University of Washington, set out to determine what was actually happening. Her 2015 study in Animal Behaviour used a clean experimental design: wild crows were presented with a dead crow, a dead song sparrow, a taxidermied red-tailed hawk holding a dead crow, and a taxidermied hawk alone. The dead sparrow triggered nothing. The dead crow triggered mobbing and decreased foraging in the area. The hawk-with-dead-crow combination produced the strongest response of all. And here was the kicker: crows learned to associate the specific human who presented the dead crow with danger, scolding that person for up to six weeks after a single exposure.
Swift’s follow-up brain imaging study (2020, Behavioural Brain Research) used FDG-PET scans and found something surprising. Crows viewing dead conspecifics did not show heightened activity in brain regions associated with emotion or fear learning. Instead, the nidopallium caudolaterale, the corvid equivalent of the mammalian prefrontal cortex, showed the most difference. The crow brain treats a dead crow the way it treats other known threats: as a problem to be analyzed, not an emotional event.
The funeral is a classroom.
The Face They Never Forget
In February 2006, John Marzluff and students at the University of Washington put on grotesque caveman masks and captured and banded seven crows on the Seattle campus. A Dick Cheney mask served as the neutral control. The variable was the face, not the body: different-sized people of different ages and genders wore the same masks.
Before trapping, fewer than 5% of crows scolded the person in the dangerous mask. After trapping, up to two-thirds reacted with scolding, mobbing, and dive-bombing.
The real discovery was what happened next. A 2011 study by Cornell, Marzluff, and Pecoraro in Proceedings of the Royal Society B documented three types of learning. Individual learning: previously captured crows scolded the mask immediately. Horizontal social learning: crows that were never captured but witnessed the trapping mob later scolded the mask independently. Vertical social learning: young crows whose parents had been conditioned to scold also scolded, though they never witnessed the original event. Scolding doubled in frequency and spread at least 1.2 kilometers from the trapping site over five years. Marzluff later recounted that during one campus walk, he was scolded by 47 of 53 crows encountered while wearing the dangerous mask.
The peer-reviewed documentation shows recognition lasting at least 2.7 years. But the experiment has essentially continued across multiple crow generations: by 2023, seventeen years after the original trapping, crows on the UW campus were still being studied for their responses. Since wild crows live 7 to 8 years on average (up to about 17 in the wild, 20 to 30 in captivity), some of the scolding crows in recent years were never alive when the trapping occurred.
Brain imaging (Marzluff et al. 2012, PNAS) showed that threatening faces activate the crow’s amygdalar regions, with the nucleus taeniae showing 6.5% increased activation. Threatening faces triggered predominantly right-hemisphere activity. Caring faces triggered left-hemisphere activity. The same lateralization pattern exists in humans.
Every Culture Heard Them Speak
The scientific findings are modern. The recognition that something is happening behind those black eyes is ancient.
In the Grimnismal (stanza 20 of the Poetic Edda), Odin sends two ravens out at dawn to fly over the entire world and return at breakfast with news. Their names are Huginn (from Old Norse hugr, “thought”) and Muninn (from munr, encompassing thought, desire, and emotion, though conventionally translated as “memory”). Odin fears for Huginn, but fears more for Muninn. Losing memory is more terrifying than losing thought. The Prose Edda, compiled by Snorri Sturluson around 1220, adds that Odin is called Hrafnagud, the raven-god, because of them.
In Irish mythology, the Morrigan, the “phantom queen,” is a triple goddess whose aspect Badb (literally “crow” in Old Irish) takes the form of a screaming raven on battlefields. In the Aided Con Culainn (“The Death of Cu Chulainn”), as the hero stands dying tied to a standing stone, the Morrigan settles on his shoulder in the form of a crow, signaling his end to all who watch.
In Ovid’s Metamorphoses (Book 2), Apollo appoints a white raven to guard his lover Coronis. The raven discovers her infidelity and reports back. Apollo kills Coronis, rescues their unborn child Asclepius from her funeral pyre, and turns the raven’s plumage from silver-white to black as punishment. Not for lying. For telling the truth. The moral is about the penalties of being a tattletale.
Pliny the Elder, in Natural History (Book 10, Chapter 60), tells of a raven hatched on the Temple of Castor and Pollux during Tiberius’s reign. It flew to a cobbler’s shop, learned to talk, and every morning greeted Tiberius, Germanicus, and Drusus Caesar by name in the Forum. A neighboring shopkeeper killed the bird. The murderer was lynched by an angry mob. The raven received a full funeral procession with bearers and wreaths. Pliny notes it received more honor in death than many of Rome’s leading citizens.
In the Quran (Surah Al-Ma’idah 5:31), after Cain kills Abel and does not know what to do with the body, Allah sends a raven that scratches the earth to show Cain how to bury his brother. The raven becomes the teacher of the first burial in human history.
Among the Tlingit and Haida of the Pacific Northwest, Raven is the trickster-creator who steals the sun from a greedy chief who hoards light in cedar boxes. His creative acts are accidental, driven by selfishness and curiosity rather than benevolence. He transforms into a hemlock needle, is swallowed by the chief’s daughter, is “born” as a human child, begs for the boxes of light, then transforms back into a bird and flies out through the smoke hole with the sun in his beak. The world is born in that moment.
In Hindu tradition, during Pitru Paksha (the “Fortnight of the Ancestors,” a 16-day period in the lunar calendar), crows are considered messengers of Yamraj, the god of death. Food offered to crows during the Shraddha ritual is believed to reach the ancestors. If crows accept the food, the ancestors are pleased. If they refuse, the ancestors are dissatisfied. This is a living tradition, practiced across India today.
In Serbian epic poetry, the formulaic phrase “dva vrana gavrana” (two black ravens) in a poem’s opening guarantees that the poem will continue in a tragic direction. The phrase appears across multiple cycles of heroic ballads collected by Vuk Karadzic, including the Kosovo Cycle commemorating the defeat of the Serbian Empire at the Battle of Kosovo in 1389.
The convergence across these unrelated traditions is not cultural diffusion. Norse Scandinavia had no contact with Tlingit Alaska. Hindu India developed its corvid symbolism independently of Celtic Ireland. What drove the convergence is the bird itself. Black plumage suggests death. Carrion feeding places them on battlefields. Mimicry of human speech suggests prophecy. Visible problem-solving suggests wisdom. Claude Levi-Strauss identified the raven as a mediating figure between opposing categories (life and death, herbivore and carnivore, the scavenger that eats meat but does not kill), functioning as psychopomps, soul guides, across multiple unrelated cultures: to Valhalla in Norse tradition, to Yamalok in Hindu, to Navia in Slavic.
Every culture that watched these birds closely arrived at the same symbolic conclusions independently. The mythology is convergent because the observable behavior is real.
The Black Bird on the Alchemist’s Floor
The alchemical tradition took the raven’s death associations and made them the foundation of its entire transformative system.
The first stage of the Great Work, the Magnum Opus that aimed to produce the Philosopher’s Stone, is called the nigredo, the blackening. It represents putrefaction, decomposition, the necessary death of old form before anything new can emerge. The four stages progress: nigredo (blackening) to albedo (whitening) to citrinitas (yellowing) to rubedo (reddening). The bird associated with each stage follows the same chromatic logic: the raven (black) gives way to the swan (white), and finally to the phoenix (red).
The nigredo is also called corvus, and the specific symbol is the caput corvi, the Head of the Raven, associated with mortificatio, the death of the ego. The operations of this stage, putrefaction, calcination, repetition, are described in the alchemical texts as slow, difficult, desiccating, severe. The Rosarium Philosophorum (1550) tells the practitioner: “When you see your matter going black, rejoice, you are at the beginning of the work.”
Saturn governs the nigredo. Lead is its metal. Melancholia is its temperament. The alchemical lead must be transmuted into gold. The raven stands at the beginning of this transformation as the saturnine bird, the creature of necessary darkness. In the Ripley Scroll, named after the 15th-century English alchemist George Ripley, the opening illustration shows Hermes Trismegistus presenting a book about alchemy. The progression from black bird to golden completion mirrors, in symbolic language, what the Tubingen crows demonstrate in empirical terms: that something complex and structured emerges from what initially looks like mere noise.
Two Readings
The materialist reading is straightforward. Crows are highly encephalized birds with dense neural architecture, strong social selection pressures, and millions of years of vocal learning capacity. Their calls encode information because natural selection favors communication systems that transmit useful signals. Face recognition, funeral learning, tool manufacture, counting, and recursive pattern generation are cognitive adaptations shaped by ecological and social pressures. The mythology is projection: humans saw smart birds doing smart things and dressed them in divine clothes.
The other reading notes the pattern. Here is a creature that counts, remembers faces across generations, teaches its children who to fear, holds assemblies around its dead, bends novel materials into tools, and produces vocalizations structured by the same mathematical laws that govern human language. Every civilization that encountered this creature, independently and without contact, placed it at the boundary between the living and the dead, between thought and speech, between chaos and creation. The Tlingit made it the world’s architect. The Norse made it the whisperer in wisdom’s ear. The alchemists made it the first stage of transformation. The Hindus made it the messenger who carries your offerings to the other side.
Is the mythology just projection? Or did these cultures, watching the same bird, recognize something that the Tubingen counting experiments, the Seattle mask studies, and the Vienna gesture recordings are now confirming through different means?
A crow at the University of Tubingen produces exactly three calls, then signals completion. Its first sound already contains the plan for the whole sequence.
That’s worth paying attention to. The ancient world certainly thought so. Whether science and myth are describing the same phenomenon from different angles, or whether one is romance and the other fact, is a question we present without answering. The bird, as always, has the last word.
