August 18, 2026
Crows memorize facial features of dangerous humans and pass the information along to their friends, remembering the person for up to 17 years.
Editor
Follow Earth on GoogleWild crows learned to treat a rubber mask as a threat after watching people wearing it trap and band their neighbors. Birds that felt the net themselves got the identification right almost every time.
Birds that only picked up the warning from other crows were far less certain. That gap is the most striking number in the study.
The work tracked American crows (Corvus brachyrhynchos) at five sites for five years and captured something researchers rarely document outside a laboratory.
Information about a specific dangerous human moved through a wild population. It moved sideways between adults, downward from parents to chicks, and outward over roughly three-quarters of a mile.
It also degraded along the way.
A mask that meant trouble
Heather Cornell, John Marzluff, and Shannon Pecoraro of the University of Washington set up the experiment around a simple pairing. Two trappers wore the same full-face mask while they fired a net launcher, banded birds, and released them. Every capture happened at the start of the study. Once trapping stopped, no crow could learn about the mask by being caught in it again.
Between 7 and 15 crows were taken at each site, for a total of 40. Handling lasted 10 to 30 minutes. At four of the five sites, a mob of 15 to 40 uncaptured crows circled overhead and scolded for as long as 10 minutes before breaking up.
Afterward, observers walked survey routes of 1.2 to 2.4 miles (2 to 3.8 kilometers). They wore either the trapping mask or one of several neutral masks, in randomized order.
Many of the observers had no idea what the study was testing. They logged whether each crow they passed scolded, which the team defined as a harsh, repeated “kaw” delivered with agitated wing and tail-flicking.
Firsthand fear beats secondhand fear
Ten of the trapped crows were later encountered wearing bands on 41 separate occasions. They scolded the trapping mask in 21 of 23 encounters, a hit rate of 91.3 percent. They scolded the neutral masks twice out of 18 encounters.
Crows that had never been caught behaved differently. They scolded the trapping mask in 137 of 421 encounters (33 percent) and scolded neutral masks in 11.8 percent of encounters. The signal was real. It was also much noisier.
That contrast is the reason animals do not rely on social information alone. Learning by getting caught is expensive and sometimes fatal, but it produces a sharp memory.
Learning by watching costs almost nothing and produces a blurry one. Crows appear to use both, which fits a broader pattern in how corvids solve novel problems by mixing trial and error with observation.
How the warning traveled
The University of Washington campus served as the long-term site. Before any trapping in 2005, only 3 percent of 167 crows scolded a person in the mask.
Two weeks after trapping, 26 percent did. At 1.25 years, the figure was 30.4 percent. By autumn 2008, roughly 2.7 years after the last capture, 66 percent of crows encountered were scolding. The average mob had grown from about seven birds to 19.
The team then searched outside the usual route. Four of 65 crows in that wider sweep scolded the mask. They were sitting at three spots 1,775 feet, 2,188 feet, and 4,094 feet from the trap sites. The farthest was nearly 0.8 miles (1,248 meters) out.
The shape of the increase mattered as much as the size. A linear model proved to be the best fit for the data. An exponential rise to a plateau did not.
Had the birds been conditioned by the original trapping spectacle, scolding should have jumped and then leveled off. Instead, it kept climbing for years, suggesting crows continued to learn from each other long after the traps were gone.
Mobs did the teaching
Scolding rarely stayed private. In 82 percent of 126 scolding events, the noise pulled in other crows. Mobs averaged 6.4 birds and reached 21 at the high end.
They held a fairly consistent distance of about 57 feet (17.4 meters) from the masked person. Birds perched around him and moved when he did. Escorts ran 33 to 330 feet (10 to 100 meters) before the mob lost interest.
The mob locations left a trail the researchers could read. At spots along the routes where a mob had previously formed, 27.8 percent of 54 lone unbanded crows later scolded the mask on their own.
At spots where no mob had ever formed, only 3.3 percent of 181 lone crows did. Attending a mob, in other words, changed what a bird did when it was by itself.
Discrimination also sharpened over time. During the second half of the trials at four sites, lone unbanded crows scolded the dangerous mask more often than they had earlier. They scolded neutral masks less often.
Parents teaching fledglings
The vertical route showed up in the next generation. Researchers followed 24 young fledglings from 10 pairs on the Washington campus in 2007 and banded and radio-tagged fledglings from four pairs at other sites in 2008. None of the chicks witnessed a capture. All watched their parents scold a masked visitor.
Later, tested without their parents nearby, fledglings from seven of 11 families scolded the person in the mask. Only one of the nine families scolded the same person with the mask off.
Young crows scolding alone never managed to recruit adults to join them, which the authors attribute to underdeveloped calls rather than indifference.
That is cultural transmission in the strict sense. A specific human face became a piece of shared knowledge among birds who had no direct evidence of it.
Why crow rumors matter
Marzluff has argued that this kind of learning is what allows a species to live alongside people. “The behavior of individual people towards animals is often changing,” he told CBC News.
Some humans put out peanuts. Three crows were shot at one of the study sites during the research. Telling those two categories apart and warning the neighbors has obvious value.
Juvenile dispersal could carry a learned face several miles per generation, which suggests the reputations crows build can outlast the person who earned them. Earlier work by the same lab found that crows hold onto these associations for years.
One caveat sits at the center of the design. Every captured bird was released unharmed. Learning that follows an actual killing may run through different channels and may place greater weight on direct observation. The authors flag the result as a likely underestimate rather than a flaw.
Studies of bolder behavior in urban animals and of rapid change in city bird populations have since made the same general point from other angles. Crows just happen to be loud about it.
The full study was published in the journal Proceedings of the Royal Society B.
—–
Like what you read? Subscribe to our newsletter for engaging articles, exclusive content, and the latest updates.
Check us out on EarthSnap, a free app brought to you by Eric Ralls and Earth.com.
Exposing the Big Game 
