Disentangling the "many-eyes", "dilution effect", "selfish herd", and "distracted prey" hypotheses in shaping alert and flight initiation distance in a colonial seabird.

Group living is thought to have important antipredator benefits for animals, owing to the mechanisms of shared vigilance ("many-eyes" hypothesis), risk dilution ("dilution effect" hypothesis), and relative safety in the center of the group ("selfish herd" hypothesis). However, it can also incur costs since social stimuli, such as conspecific aggression, may distract individuals from anti-predator behavior ("distracted prey" hypothesis). We simultaneously evaluated how these four different hypotheses shape anti-predator behaviors of breeding king penguins (Aptenodytes patagonicus), which aggregate into large colonies, experience frequent aggressive social interactions, and are regularly exposed to predation by giant petrels (Macronectes sp.) and brown skuas (Catharacta loonbergi) when breeding on land. We approached 200 incubating penguins at four different periods of the breeding season across a range of overall increasing colony densities. We measured the distance at which focal birds detected the approaching threat (alert distance: AD), whether birds decided to flee or not, and the distance of flight initiation (flight initiation distance: FID, viz. the bird attempting to walk away with its egg on its feet). We quantified relative local neighbor density, centrality within the colony (rank), and the number of aggressions the focal bird emitted towards neighbors during the approach. We found that birds engaged in aggressive conflicts with neighbors were less likely to flee, and that increasing relative local neighbor density at low and medium overall colony density resulted in a decrease in bird AD, both supporting the "distracted prey" hypothesis. However, at maximal overall colony density, increasing relative local neighbor density resulted in longer AD, supporting the "many-eyes" hypothesis. We found no support for the "dilution effect" and "selfish herd" hypotheses, and no effects of any hypothesis on FID.