A crucial stage in hominin evolution was the development of metatool use -- the ability to use one tool on another [1, 2]. Although the great apes can solve metatool tasks [3, 4], monkeys have been less successful [5-7]. Here we provide experimental evidence that New Caledonian crows can spontaneously solve a demanding metatool task in which a short tool is used to extract a longer tool that can then be used to obtain meat. Six out of the seven crows initially attempted to extract the long tool with the short tool. Four successfully obtained meat on the first trial. The experiments revealed that the crows did not solve the metatool task by trial-and-error learning during the task or through a previously learned rule. The sophisticated physical cognition shown appears to have been based on analogical reasoning. The ability to reason analogically may explain the exceptional tool-manufacturing skills of New Caledonian crows.

The ability of some bird species to pull up meat hung on a string is a famous example of spontaneous animal problem solving. The “insight” hypothesis claims that this complex behaviour is based on cognitive abilities such as mental scenario building and imagination. An operant conditioning account, in contrast, would claim that this spontaneity is due to each action in string pulling being reinforced by the meat moving closer and remaining closer to the bird on the perch. We presented experienced and naïve New Caledonian crows with a novel, visually restricted string-pulling problem that reduced the quality of visual feedback during string pulling. Experienced crows solved this problem with reduced efficiency and increased errors compared to their performance in standard string pulling. Naïve crows either failed or solved the problem by trial and error learning. However, when visual feedback was available via a mirror mounted next to the apparatus, two naïve crows were able to perform at the same level as the experienced group. Our results raise the possibility that spontaneous string pulling in New Caledonian crows may not be based on insight but on operant conditioning mediated by a perceptual-motor feedback cycle.

No abstract

New Caledonian (NC) crows are the most sophisticated tool manufacturers other than humans. The diversification and geographical distribution of their three Pandanus tool designs that differ in complexity, as well as the lack of ecological correlates, suggest that cumulative technological change has taken place. To investigate the p possibility that high-fidelity social transmission mediated this putative ratchet-like process, we studied the ontogeny of Pandanus tool manufacture and social organization in free-living NC crows. We found that juvenile crows took more than 1 year to reach adult proficiency in their Pandanus tool skills. Although trial-and-error learning is clearly important, juveniles have ample opportunity to learn about Pandanus tool manufacture by both observing their parents and interacting with artifactual material. The crows' social system seems likely to promote the faithful social transmission of local tool designs by both favoring the vertical transmission of tool information and minimizing horizontal transmission. We suggest that NC crows develop their Pandanus tool skills in a highly scaffolded learning environment that facilitates the cumulative technological evolution of tool designs.

New Caledonian crows are the most proficient non-hominin tool manufacturers but the cognition behind their remarkable skills remains largely unknown. Here we investigate if they attend to the functional properties of the tools that they routinely use in the wild. Pandanus tools have natural barbs along one edge that enable them to function as hooking implements when the barbs face backwards from the working tip. In experiment 1 we presented eight crows with either a non-functional ('upside-down') or a functional pandanus tool in a baited hole. Four of the crows never flipped the tools. The behaviour of the four flipping birds suggested that they had a strategy of flipping a tool when it was not working. Observations of two of the eight crows picking up pandanus tools at feeding tables in the wild supported the lack of attention to barb direction. In experiment 2 we gave six of the eight crows a choice of either a barbed or a barbless pandanus tool. Five of the crows chose tools at random, which further supported the findings in experiment 1 that the crows paid little or no attention to the barbs. In contrast, a third experiment found that seven out of eight crows flipped non-functional stick tools significantly more than functional ones. Our findings indicate that the crows do not consistently attend to the presence or orientation of barbs on pandanus tools. Successful pandanus tool use in the wild seems to rely on behavioural strategies formed through associative learning, including procedural knowledge about the sequence of operations required to make a successful pandanus tool.