Abstract
Human hypercooperativity and the emergence of division of labor enables us to solve problems not only effectively within a group but also collectively. Collective problem-solving occurs when groups perform better than the additive performance of separate individuals. Currently, it is unknown whether this is unique to humans. To investigate the evolutionary origin of collective problem-solving and potential precursors, we propose a continuum of group effects on problem-solving, from simple to complex ones, eventually culminating in collective problem-solving. We tested captive common marmosets with a series of problem-solving tasks, either alone or in a group. To test whether the performance of a group was more than the sum of its parts, we compared real groups to virtual groups (pooled scores of animals tested alone). Marmosets in real groups were both more likely to solve problems than marmosets within the virtual groups and to do so faster. Although individuals within real groups approached the problem faster, a reduction in neophobia was not sufficient to explain the greater success. Success within real groups arose because animals showed higher perseverance, especially after a fellow group member had found the solution in complex tasks. These results are consistent with the idea that group problem-solving evolved alongside a continuum, with performance improving beyond baseline as societies move from social tolerance to opportunities for diffusion of information to active exchange of information. We suggest that increasing interdependence and the adoption of cooperative breeding pushed our ancestors up this scale.