The Surprising Intelligence of Crows

For most of the twentieth century, the study of animal intelligence concentrated on great apes. The choice was natural: apes are our closest evolutionary relatives, share a recognizably similar neural architecture, and produce results that translate readily to human terms. By the 1990s, however, a small group of researchers studying corvids (the family that includes crows, ravens, jays, magpies, and rooks) had begun to publish results that were uncomfortable for the standard narrative. Birds, with brains the size of walnuts and a neural architecture quite unlike a mammal's, were solving problems at the level of chimpanzees and sometimes beyond.

The interesting question is no longer "are corvids intelligent." That was settled. The interesting questions are about what their intelligence is like, what it implies about how minds work, and why it appears to have evolved at all in such a different lineage.

The face-recognition study

The most cited modern crow study is John Marzluff's 2008 work at the University of Washington. Marzluff and his team trapped, banded, and released wild American crows on the Seattle campus, while the trappers wore distinctive caveman masks. They then walked around campus periodically, sometimes in the caveman mask, sometimes in a control mask (Dick Cheney, for what that is worth, became the control), and recorded crow scolding behavior.

The crows recognized the caveman mask immediately and scolded any researcher wearing it. They did not scold the Cheney mask. They did not scold the same researcher's bare face. The recognition persisted across years; crows that had not personally been trapped scolded the mask too, having apparently learned it from older birds.

Marzluff's follow-up work with brain imaging found that the crows' recognition of the threatening face activated the avian equivalent of the amygdala (fear processing) and the avian equivalent of the medial striatum (memory), an architecture functionally analogous to mammalian face-recognition networks. The structures evolved independently; they converged on a similar division of labor.

The tool manufacture sequence

New Caledonian crows are the most studied tool-using birds. They manufacture hooked tools from twigs by stripping bark and trimming the end into a barb, then use the tool to extract grubs from holes in dead wood. The behavior is regional: New Caledonian crow populations on different islands make tools of slightly different shapes, suggesting cultural transmission.

The Auckland researcher Russell Gray's lab has run a long-running set of experiments testing how far this behavior generalizes. In one sequence, captive crows were given a problem they had never seen: a piece of food on a string, dangling from a perch. The crows pulled the string up using a foot-and-beak coordination, holding the gathered slack with the foot.

The harder version of the problem requires compound tool use: the food is reachable only by inserting a small stick into a tube that releases a longer stick, which can then be used to reach the food. Some crows solved this on the first attempt without training. The cognitive prerequisite (planning a sequence of subgoals where the immediate action does not produce the immediate reward) had previously been documented only in great apes and a few human children.

The mental time travel question

A more philosophically loaded question is whether crows are capable of mental time travel: imagining a future state and acting now to influence it, beyond simple instinctual caching. Nicola Clayton at Cambridge has spent decades on this with scrub jays.

The classic experiment cycles jays through three rooms with breakfast available in only one of them on a given day. After a few cycles, the jays start caching food in the room where breakfast will not be available, the night before. They are not responding to current hunger; they are anticipating future hunger and pre-positioning food for it. Clayton's terminology is "future-oriented behavior"; the philosophical hedge is to avoid claiming subjective experience.

Clayton's group has also shown that scrub jays remember the perishability of cached food. They retrieve worms (perishable) before they retrieve nuts (non-perishable) when both are available. They re-cache food they observed being watched by another jay. They cache food less in front of dominant jays and more in front of subordinate ones. The pattern is hard to explain without something resembling a model of other minds.

The brain that should not work

The puzzle is anatomical. A crow brain weighs about ten grams. A chimpanzee brain weighs about 400 grams. The crow nonetheless performs many cognitive tasks at chimpanzee-or-better levels. How?

The answer, established by Suzana Herculano-Houzel's neuron-counting research at Vanderbilt in the 2010s, is that bird brains have much higher neuron density than mammal brains. A crow's pallium (the bird-brain analog of the cerebral cortex) packs roughly twice as many neurons per gram as a primate's neocortex. The total neuron count of a crow's pallium is comparable to that of a small monkey's neocortex, despite the difference in physical brain size.

The deeper architectural difference is that birds evolved their cognitive structures in parallel with mammals, from a common ancestor that lived 320 million years ago and had only the most rudimentary forebrain. The result is convergent evolution at the algorithm level: the bird and mammal brains arrived at structurally different solutions to the same computational problems. The avian "nidopallium caudolaterale" performs the executive functions that the mammalian prefrontal cortex performs, but the connectivity, layering, and developmental origin are quite distinct.

This matters because it suggests that intelligence is not a property of any specific brain architecture. It is a property of certain computational organizations, and there is more than one way to instantiate them in tissue. The implications for thinking about intelligence elsewhere (in machines, in octopuses, in possible aliens) are loose but not negligible.

The grudges

One detail in the Marzluff face-recognition work that gets less academic attention but has stayed with me: the grudge persisted longer than the original group of crows. Birds that were never themselves trapped, who had no personal cause for complaint with the caveman mask, scolded it as adults. The behavior had been transmitted, plausibly through observation of older birds' alarm responses, into a sort of cultural memory of who was dangerous.

It is a strange thought. Wild birds in Seattle, today, almost two decades after the original study, may still be teaching their young that the caveman face means trouble. There are very few non-human species we can demonstrably show holding multi-generational grudges against specific human visual patterns. The crows are doing it on a university campus, in plain sight, while we mostly ignore them.

What it changes

The general conclusion, after thirty years of corvid cognition research, is that the inheritance of human-like cognitive capabilities is much wider than a strict primate-centric view would suggest. Tool use, planning, deception, theory of mind, face recognition, episodic-like memory, cultural transmission, multi-generational grudges. These appear, in some form, in animals separated from us by 320 million years of independent evolution.

The reasonable response is to treat any reasonably brain-bearing animal we encounter with more curiosity and a little more respect. The crows on the lawn are watching. They know your face. They are deciding whether you are a threat and they are telling each other. The cognitive distance between them and us is smaller than it looks.