In the Japanese city of Sendai, in Miyagi Prefecture in northern Honshu, carrion crows have been seen dropping walnuts into the path of moving cars, waiting for the tires to do the cracking, and collecting the meat when the traffic stops. The behavior was first noticed at a driving school in 1975 and later documented at nearby intersections. It remains one of the most cited examples of urban animal learning in the literature.
The most detailed account comes from the zoologist Yoshiaki Nihei, then at Tohoku University, who studied the Sendai crows and published his observations in the Japanese Journal of Ornithology in 1995. Across 21 months of observation he recorded 43 occurrences in detail, and concluded that the birds did not simply drop nuts and hope. They positioned them, adjusted them, and timed their retrieval around the traffic cycle.
It is a strange thing to sit with. A wild bird using a signalized intersection the way a person uses a nutcracker.
Where the story actually begins
The origin point, as best anyone has reconstructed it, is the Kadan driving school in Sendai, where instructors run learner drivers in slow loops around a paved course. In 1975, the local carrion crows (Corvus corone) there discovered how to use cars as nutcrackers, dropping Japanese walnuts (Juglans ailantifolia) into the path of the slow-moving vehicles. The walnuts are smaller than the commercial variety and stubbornly hard. A crow’s beak cannot open them. Even a drop from a rooftop often fails.
The driving school offered something better than gravity: a predictable, low-speed wheel. From there, as the evolutionary biologist Menno Schilthuizen describes in that Aeon essay — he sat on the school’s crumbling wall himself, waiting for the birds — the behavior spread to other parts of the city where cars moved slowly and reliably: sharp bends, parking lot entrances, and intersections with traffic lights.
The spread was slow, and that slowness is the interesting part. Nihei returned to the question with Hiroyoshi Higuchi in 2001 (Tohoku Psychologica Folia 60: 93–97), combining a questionnaire survey of local residents with their own observations. The car-using behavior originated at the driving school in the 1970s and took thirteen years to reach the next recorded site, an intersection roughly a kilometre away, before radiating out to about a five-kilometre radius over some two decades. A later review of urban corvids describes the same shape: the behavior spread slowly from the place where it was first recorded, which is what social learning looks like rather than sudden collective insight.
Nihei’s observations went further, describing crows waiting near a red light, walking into the crosswalk to place a nut, then hopping back to the curb until the light changed again. The birds were recorded nudging a walnut a few centimeters if the wheels were missing it. In one case, a crow walked into the path of an oncoming car, forced it to brake, and tossed a nut in front of its wheels.
The behavior reached a global audience through the BBC. Footage shot for The Life of Birds shows the birds hopping in front of the cars to place walnuts picked from the adjoining trees, then returning alongside the pedestrians to collect them.
What the observations do, and do not, show
A careful reading of the source material matters here, because the Sendai crows have become a set piece in popular science writing, and the popular version has drifted from the primary record.
What is well documented: carrion crows in Sendai drop walnuts onto roads and recover them after cars have passed. The behavior was first noted in 1975 at one location. By 1995 it had been recorded at intersections in a way Nihei judged systematic rather than accidental.
What has been contested: whether the birds understand what the tire is for. In 1997, David Cristol and colleagues published a rebuttal in The Auk — titled, bluntly, “Crows do not use automobiles as nutcrackers” — after watching American crows in Davis, California, and finding no evidence that the birds positioned nuts where wheels would hit them. Their argument was that car-cracking can be an incidental by-product of ordinary food-dropping onto hard surfaces. The Sendai record is considerably stronger than the Davis record, but it is a record of behavior, not of intention.
What holds up on the geography: the traffic-signal version does appear to be a Sendai speciality. The only published comparison comes from Hakodate, on Hokkaido, where Kanami Ara and Katsura Mikami watched carrion crows through the autumn and winter of 2016. The behavior had previously only been studied in Sendai City, and the Hakodate birds did not reproduce it: they dropped walnuts from electric wires onto roads in a much less targeted way, which the authors judged less sophisticated than the Sendai version. No other Japanese population has been documented placing nuts at a signalized crossing. No national survey has been published either, so this is an absence in the record rather than a proven absence in the country’s crows.
None of this diminishes the finding. It just means the Sendai crows are a well-observed local phenomenon rather than a nationwide experiment in avian traffic engineering.
Why a red light is a useful thing to a crow
The reason the story keeps drawing attention is that a signalized intersection is one of the most tightly engineered pieces of infrastructure in a city. Signal phase, pedestrian cycle, vehicle throughput: every second of it is designed around human movement. For a bird to slot itself into that rhythm, and to use it as a nut-processing station, is genuinely striking.
Carrion crows are members of the corvid family, alongside ravens, jackdaws, magpies, and jays. Corvids have been the subject of a long run of comparative-cognition work showing capacities for planning, tool use, and social learning once assumed to belong to primates alone. New Caledonian crows, for example, have been documented assembling compound tools from separate parts to reach food: in a 2018 Scientific Reports study, four crows spontaneously combined elements to make functional tools, and did so conditionally on the position of the food.
The wider pattern is not confined to corvids. A 2025 paper in Frontiers in Ethology, prompted by a city-wintering Cooper’s hawk, puts the general case plainly: moving vehicles are a feature of the urban environment, and birds learn not just to avoid the danger they represent, but also to exploit them. The Sendai crows are the most famous case, not an isolated one.
How a behavior like this might travel
The interesting question is not whether one crow is clever. It is how a trick gets from one clever bird to a neighborhood of birds, and then to the next generation.
The mechanism most often invoked is social learning: young crows watching adults, copying, refining. Corvids are unusually well suited to this. They live in extended family groups, they pay close attention to what other crows are doing, and juveniles stay near their parents long enough to acquire complicated foraging routines.
The clearest parallel is not in birds at all. Carel van Schaik and colleagues, writing in Science in 2003, documented geographic variation in orangutan behavior and found a correlation between the abundance of opportunities for social learning and the size of the local cultural repertoire — evidence, they argued, that great-ape cultures exist. The point generalizes. When an environment contains a resource that is hard to access, and when the animals living in that environment can watch each other closely, a workable trick can travel further than any individual could invent it twice.
That framing sits behind the Sendai story. One crow at one driving school in 1975 does not explain fifty years of behavior. A social learning pathway does.
The urban animal beat, taken seriously
The Sendai crows are part of a longer pattern in urban biology: animals reading human infrastructure and using it in ways engineers never anticipated. Some of it is charming. Some of it is a conservation puzzle. Most of it looks obvious only in retrospect.
Science Blog has covered several of these cases. In Pennsylvania, engineers bolted ultrasonic speakers to wind turbine nacelles to jam bat echolocation; in the first year of the trial, an estimated 21–51% fewer bats were killed per treated turbine than per control turbine, though the second year’s results were far less conclusive. In Florida, a nuclear plant south of Miami has become a major American crocodile nursery, entirely by accident of its cooling canals.
The Sendai story fits the same shape. A built environment produces an opportunity nobody designed for, and a species that is paying attention takes it.
What the next study should do
The honest gap in the record is a modern, systematic follow-up. Nihei’s detailed work is three decades old. The BBC footage is nearly as old. Sendai’s traffic patterns, its walnut trees, and its crow population have all shifted in the intervening decades. A new survey covering multiple intersections across multiple cities, with video and GPS-tagged birds, would establish whether the behavior is stable, whether it has spread, and whether it really is absent from other Japanese urban areas.
The wider question, the one the corvid literature keeps circling, is how much of what gets called animal intelligence is really animal culture. A single crow that once dropped a walnut in front of a slow car does not explain what happens at a Sendai intersection today. Generations of crows watching each other do.
The Long View has written before about how long a brain can persist after death. The Sendai crows raise the mirror question. How long can a piece of learning persist inside a population of brains that keep replacing themselves?
Autumn is walnut season in Sendai. If the pattern still holds, there is a bird somewhere near a crosswalk this week, watching the light change.
