Division of labor, the specialization of workers on different tasks, largely contributes to the ecological success of social insects [1, 2]. Morphological, genotypic, and age variations among workers, as well as their social interactions, all shape division of labor [1-12]. In addition, individual experience has been suggested to influence workers in their decision to execute a task [13-18], but its potential impact on the organization of insect societies has yet to be demonstrated [19, 20]. Here we show that, all else being equal, ant workers engaged in distinct functions in accordance with their previous experience. When individuals were experimentally led to discover prey at each of their foraging attempts, they showed a high propensity for food exploration. Conversely, foraging activity progressively decreased for individuals who always failed in the same situation. One month later, workers that previously found prey kept on exploring for food, whereas those who always failed specialized in brood care. It thus appears that individual experience can strongly channel the behavioral ontogeny of ants to generate a lasting division of labor. This self-organized task-attribution system, based on an individual learning process, is particularly robust and might play an important role in colony efficiency.

In social insect colonies, recognition of nestmates, kinship, caste and reproductive status is crucial both for individuals and for the colony. The recognition cues used are thought to be chemical, with the hydrocarbons found on the cuticle of insects often cited as being particularly important. However, in honeybees (Apis mellifera) the role of cuticular hydrocarbons in nestmate recognition is controversial. Here we use the proboscis extension response (PER) conditioning paradigm to determine how well honeybees learn long-chain linear alkanes and (Z)-alkenes present on the cuticle of worker bees, and also how well they can discriminate between them. We found large differences both in learning and discrimination abilities with the different cuticular hydrocarbons. Thus, the tested hydrocarbons could be classified into those which the bees learnt and discriminated well (mostly alkenes) and those which they did not (alkanes and some alkenes). These well-learnt alkenes may constitute important compounds used as cues in the social recognition processes.

In insect societies, worker policing controls genetic conflicts between individuals and increases colony efficiency. However, disentangling relatedness from colony-level effects is usually impossible. We studied policing in the parthenogenetic ant Cerapachys biroi, where genetic conflicts are absent due to clonality and reproduction is synchronized through stereotyped colony cycles. We show that larval cues regulate the cycles by suppressing ovarian activity and that individuals that fail to respond to these cues are policed and executed by their nestmates. These individuals are genetically identical to other colony members, confirming the absence of intracolonial genetic conflicts. At the same time, they bear distinct cuticular hydrocarbon profiles, which could serve as proximate recognition cues for policing. Policing in C. biroi keeps uncontrolled reproduction at bay and thereby maintains the colony-level phenotype. This study shows that policing can enforce adaptive colony-level phenotypes in societies with minimal or no potential genetic conflicts. In analogy to immunosurveillance on cancer cells in genetically homogeneous multicellular organisms, colony efficiency is improved via the control of individuals that do not respond properly to regulatory signals and compromise the functioning of the higher-level unit.