Despite their miniature brains, insects exhibit substantial variation in brain size. Although the functional significance of this variation is increasingly recognized, research on whether differences in insect brain sizes are mainly the result of constraints or selective pressures has hardly been performed. Here, we address this gap by combining prospective and retrospective phylogenetic-based analyses of brain size for a major insect group, bees (superfamily Apoidea). Using a brain dataset of 93 species from North America and Europe, we found that body size was the single best predictor of brain size in bees. However, the analyses also revealed that substantial variation in brain size remained even when adjusting for body size. We consequently asked whether such variation in relative brain size might be explained by adaptive hypotheses. We found that ecologically specialized species with single generations have larger brains —relative to their body size— than generalist or multi-generation species, but we did not find an effect of sociality on relative brain size. Phylogenetic reconstruction further supported the existence of different adaptive optima for relative brain size in lineages differing in feeding specialisation and reproductive strategy. Our findings shed new light on the evolution of the insect brain, highlighting the importance of ecological pressures over social factors and suggesting that these pressures are different from those previously found to influence brain evolution in other taxa.

We measured brain and body size ofrom 93 species of bees (Apoidea). For these species, we also collected data on diet specialization, sociality and life-history from literature sources (Supplementary Data 1). For a subsample of 12 species, we measured the volume of the mushroom bodies and optic lobes and the remaining brain (Supplementary Data 2). Finally, we built a phylogenetic tree for the species included in the study, by constraining genera and subgenera to be monophyletic (Supplementary Data 3) and generating a sample of 100 trees to account for phylogenetic uncertainty when running all the comparative analyses (Supplementary Data 4).