Abstract Group living is favorable to pathogen spread due to the increased risk of disease transmission among individuals. Similar to individual immune defenses, social immunity, that is antiparasite defenses mounted for the benefit of individuals other than the actor, is predicted to be altered in social groups. The eusocial honey bee (Apis mellifera) secretes glucose oxidase (GOX), an antiseptic enzyme, throughout its colony, thereby providing immune protection to other individuals in the hive. We conducted a laboratory experiment to investigate the effects of group density on social immunity, specifically GOX activity, body mass and feeding behavior in caged honey bees. Individual honeybees caged in a low group density displayed increased GOX activity relative to those kept at a high group density. In addition, we provided evidence for a trade‐off between GOX activity and body mass: Individuals caged in the low group density had a lower body mass, despite consuming more food overall. Our results provide the first experimental evidence that group density affects a social immune response in a eusocial insect. Moreover, we showed that the previously reported trade‐off between immunity and body mass extends to social immunity. GOX production appears to be costly for individuals, and potentially the colony, given that low body mass is correlated with small foraging ranges in bees. At high group densities, individuals can invest less in social immunity than at low densities, while presumably gaining shared protection from infection. Thus, there is evidence that trade‐offs at the individual level (GOX vs. body mass) can affect colony‐level fitness.

Animals living in large colonies are especially vulnerable to infectious pathogens and may therefore have evolved additional defences. Eusocial insects supplement their physiological immune systems with ‘social immunity’, a set of adaptations that impedes the entrance, establishment, and spread of pathogens in the colony. We here find that honey bee workers (Apis mellifera) that had been experimentally immune-challenged with bacterial lipopolysaccharide (LPS) often exited the hive and subsequently died; some individuals were dragged out by other workers, while others appeared to leave voluntarily. In a second experiment, we found that healthy workers treated with surface chemicals from LPS-treated bees were evicted from the hive more often than controls, indicating that immune-challenged bees produce chemical cues or signals that elicit their eviction. Thirdly, we observed pairs of bees under lab conditions, and found that pairs spent more time apart when one member of the pair had received LPS, relative to controls. Our findings suggest that immune-challenged bees altruistically banish themselves, and that workers evict sick individuals which they identify using olfactory cues, putatively because of (kin) selection to limit the spread of pathogens within colonies. Just as in humans, animals living in large groups must contend with infectious diseases. Social insects such as honey bees have evolved a range of behavioural and organisational defences against disease, collectively termed ‘social immunity’. Here, we describe experiments in which we introduced immune-stimulated bee workers into hives to study social immunity. We find that bees that were wounded or immune-challenged were more likely to leave the hive—resulting in their death—compared to healthy controls. Some of the bees leaving the hive were ejected by other workers, while some left the hive seemingly by choice: we thus find evidence for both ‘banishment’ of immune-challenged bees and self-imposed exile. Furthermore, using experiments transferring chemical signals between healthy and immune stimulated bees, we establish that the latter are identified for banishment by the chemicals present on their body surface.