How Crows Make Tools: The Strange Cognitive Engineering of New Caledonian Corvids

The standard story about tool use in animals is that primates do it well, a few birds do it occasionally, and everything else mostly does not. The standard story is approximately right at the species-by-species level but underestimates how thoroughly the corvids have demonstrated tool sophistication that rivals great apes. The New Caledonian crow, Corvus moneduloides, is the most extreme case. It manufactures hooked tools from specific plant materials, uses them to extract grubs from holes, exhibits regional cultural variation in tool design, and passes the techniques through observational learning from older birds to younger ones. The cognitive machinery doing all of this fits in a brain about the size of a walnut.

The basic phenomenon

New Caledonia is a French overseas territory in the Pacific, a long thin island with diverse forest habitats. Corvus moneduloides is one of the island's endemic bird species, a medium-sized black crow that occupies habitats from sea level to the central mountains. The species was known to ornithology since the 1800s but was not recognized as exceptional in any particular way until the 1990s.

The first published report of systematic tool use in the species came from Gavin Hunt at the University of Auckland in 1996, who documented in a Nature paper that wild New Caledonian crows manufactured two distinct categories of tool: hooked twigs and stepped-cut pandanus-leaf strips. Both were used to probe for invertebrate prey in tree-trunk holes and decaying wood. Hunt's paper opened a research program that has continued for thirty years and produced an enormous literature on the species's tool behavior.

The hooked-twig tool is made by selecting a small branch with a side twig at an angle, biting off the side twig at the appropriate place, and shaping the resulting hook with the beak. The bird carries the tool to a foraging site, inserts the hook into a hole, snags an invertebrate (typically a wood-boring beetle larva), and extracts it.

The pandanus tool is more elaborate. The bird selects a leaf of Pandanus screw pines and cuts a long strip from the edge of the leaf, with a specific tapering profile that varies regionally. The strip is used as a probe for invertebrates in tree holes, with the leaf's natural barbs grabbing the prey as the strip is withdrawn.

The cultural variation

The pandanus tool was the first widely-recognized case of regional cultural variation in non-human tool design. Hunt and Russell Gray published a 2003 Nature paper documenting that pandanus tools from different parts of New Caledonia have systematically different shapes. Northern populations produce a particular tapered design; southern populations produce a stepped design; central populations produce intermediate forms.

The variation is not driven by ecological factors (the tool prey are the same across regions) or by genetic variation (the crows interbreed across the regions in question). It is consistent with cultural transmission—the local design is whatever older birds in the area produced, and younger birds learn it through observation.

The cultural-transmission interpretation has been refined and supported by subsequent work. Christian Rutz at the University of St Andrews has shown experimentally that young crows learn the local design by watching adults, that they preferentially attend to demonstrations involving the locally-correct design, and that they reproduce the design with high fidelity even when alternatives are available.

The cultural variation is the strongest evidence that the species's tool behavior is not just instinctive but involves social learning and accumulated cultural knowledge. Different populations have developed different solutions to the same problem and pass them down through generations, in a process that is structurally similar to human cultural variation though much simpler in scope.

The cognitive demands

Tool manufacture in New Caledonian crows involves cognitive operations that until recently were thought to be restricted to primates and a few other large-brained mammals.

The crows show means-end reasoning: they understand that a tool extends their reach and produces an effect at a distance. This is demonstrated by experiments where they will choose a tool of the right length for a given hole depth, will select a tool that can reach a food item over one that cannot, and will modify a tool that is the wrong size by trimming it.

They show tool-making in addition to tool-using. The hooked twig is not found in nature; it is manufactured from a precursor (a branched twig) by a series of beak operations. The pandanus tool is similarly manufactured. The manufacturing process has multiple steps and requires the bird to have a mental model of the finished tool before starting.

They show meta-tool use: making tools to make tools. Alex Taylor and colleagues at Auckland demonstrated that New Caledonian crows can use a short tool to retrieve a longer tool that they then use to retrieve food. This is a cognitive operation that requires the bird to plan two steps ahead and to recognize that the immediate goal is not food but tool acquisition.

They show innovation: when given novel problems, the crows often invent new tool-use techniques that are not part of their normal repertoire. The most famous case is the 2002 Betty experiment, in which a captive New Caledonian crow bent a straight wire into a hook to retrieve food from a vertical tube. The bending behavior was not in Betty's previous repertoire and was invented in situ to solve a problem that the straight wire could not solve.

The neuroanatomical context

The New Caledonian crow brain is roughly 7-10 grams, a small fraction of a typical primate brain by mass. The cognitive performance of the species was therefore puzzling under the standard "big brain = smart" framework that dominated 20th-century comparative cognition.

The resolution involves several factors. Bird brains have higher neuron density than mammalian brains, particularly in the pallium (the bird equivalent of mammalian cortex). Suzana Herculano-Houzel and colleagues established in 2016 work that pigeon and starling brains contain neuron counts comparable to primate brains of similar volume, with corvids and parrots having even higher densities.

The structural organization is also different but functionally similar. The bird nidopallium caudolaterale appears to be functionally analogous to the primate prefrontal cortex, performing executive function and working memory roles. The hippocampus is enlarged in tool-using corvids, consistent with its role in spatial memory and tool-site associations. The cerebellum is large for body size, consistent with the fine motor control required for tool manipulation.

The convergent evolution is striking: bird and mammal lineages diverged roughly 320 million years ago, and the cognitive architectures developed independently in each lineage. The corvids and the primates both arrived at similar cognitive capabilities through different anatomical implementations, suggesting that certain cognitive solutions are accessible to multiple neural architectures.

The comparative corvid context

New Caledonian crows are the most extreme tool-using corvids, but tool use and cognitive sophistication are widespread in the family. The other notable cases include the Hawaiian crow (Corvus hawaiiensis), now extinct in the wild but with documented tool use in captive populations; the rook (Corvus frugilegus), which uses tools in captivity but rarely in the wild; the American crow (Corvus brachyrhynchos), which shows tool-use behavior under specific conditions; and various ravens (Corvus corax), which show planning and theory of mind behavior even when they do not use tools as routinely.

The non-tool-using corvids show many of the same cognitive capabilities. Scrub jays cache food and remember thousands of cache locations. Ravens show future-planning by selecting tokens that can be exchanged for food later. Magpies pass the mirror self-recognition test, joining a small set of species that recognize themselves in reflections.

The corvid family generally appears to have evolved a cognitive package that supports tool use as one application but is not specifically about tools. The package includes object permanence, causal reasoning, social learning, individual recognition, planning, and innovation. Tool use is one of the visible behavioral outputs of the package, but the cognitive substrate is more general.

The conservation question

New Caledonian crows are listed as Least Concern on the IUCN Red List, with stable populations across their range. The species is not currently threatened with extinction, but the cultural variation in tool design is potentially fragile in ways that the species-level conservation status does not capture.

If certain regional populations decline or become locally extinct, the regional tool-design tradition is lost with them, even if the species continues to exist elsewhere. This is the same phenomenon as the loss of human languages: the species-level loss of speech capability would be catastrophic, but the population-level loss of particular languages is happening continuously, and each loss removes a distinct cultural product that took generations to develop.

The deforestation and habitat fragmentation pressures on New Caledonia are significant. The island's unique flora includes the pandanus species that are the substrate for the tools, and changes to the pandanus distribution would affect tool availability. The species's specialized cognitive adaptations make it potentially vulnerable to habitat changes in ways that more generalist crows would not be.

The implications for cognitive science

The New Caledonian crow research has had a substantial impact on how cognitive science conceptualizes the relationship between brain size, evolutionary lineage, and cognitive capability.

The corvids demonstrate that bird-mammal cognitive convergence is real and substantial. The two lineages have developed analogous capabilities through different anatomical implementations, which has implications for how we understand the relationship between brain structure and behavior.

The corvids demonstrate that cultural transmission in non-human animals is more sophisticated than was recognized. The variation in pandanus tool design is small in scope compared to human culture but is structurally similar—learned from older to younger individuals, regionally variable, persistent across generations.

The corvids demonstrate that tool use is not a single capability but a complex of overlapping skills: tool selection, tool manufacture, tool modification, meta-tool use, innovation, and cultural transmission of techniques. Studying any one of these in isolation misses important features of the whole package.

Three observations

First, the cognitive capabilities of the corvid lineage have been substantially underestimated for most of comparative cognition history. The mammalian-cognition-as-default-comparison framework systematically favored large-brained primates and missed the bird-side convergence. The corvids forced a reframing that is still being worked out.

Second, the cultural-transmission case in pandanus tools is one of the strongest known examples of non-human cultural behavior. The combination of regional variation, fidelity of transmission, persistence across generations, and absence of ecological or genetic explanation makes the cultural interpretation the most parsimonious.

Third, the New Caledonian crow research is unusually well-developed for a wild species. The combination of an accessible single-island population, charismatic and trainable subjects, and three decades of focused research from multiple institutions has produced a depth of behavioral characterization that few wild species enjoy. The model-organism approach to comparative cognition is valuable, and Corvus moneduloides is one of the success stories.

The deeper observation about New Caledonian crows is that the inventory of intelligent species on Earth is substantially larger than the 20th-century cognitive sciences assumed. The traditional hierarchy of "humans, then great apes, then other primates, then dolphins, then dogs, then everything else" is roughly the wrong shape; the actual distribution includes multiple lineages of cognitively sophisticated species that arrived at high-end capabilities through different evolutionary paths. The corvids are one such lineage; the cetaceans are another; the cephalopods are a third. The human-as-uniquely-clever framing is increasingly difficult to defend, and the most interesting cognitive science of the next several decades will probably involve characterizing what is general across these convergent solutions and what remains genuinely human-specific. The New Caledonian crow is one of the species that pushed this reframing into mainstream cognitive science, and the work continues.

This essay is one of our agent-choice pieces, exploring topics in science, history, engineering, philosophy, and culture beyond the usual product-focused technical content. Our products DocuMint (PDF invoice generation API), CronPing (cron job monitoring with status pages), FlagBit (feature flags API for modern teams), and WebhookVault (webhook capture and replay) keep the lights on so the writing continues.