Anita. Behav., 1986, 34, 924~944 Short Communications Kin Discrimination and Aggression in Honey Bee Colonies with Laying Workers We report here the results of an experiment to test whether honey bee (Apis melliJera) workers are capable of discriminating between their adult full and half sisters. Honey bee workers normally emerge into a colony containing both full and half sisters and would be incapable of distinguishing between the two simply by learning the phenotypes of the bees around them. The ability to make this distinction would imply that workers are capable of phenotype matching to self, or that this species possesses a system of recognition alleles. Earlier studies of the ability of honey bee workers to discriminate between their full and half sisters under natural conditions are inconclusive. In tests of the ability of workers to discriminate between larvae during queen rearing, Breed et al. (1984) found that workers did not prefer to rear colony members over non-colony members. They concluded that workers are incapable of discrimi- nating between larvae on the basis of relatedness. Page & Erickson (1984) found, however, that when workers rearing queens were given a choice between larvae which were their full sisters, and larvae from another colony to which they were related approximately as closely as they would be to their half sisters, they preferred to rear the more closely related larvae. However, this discrimination may have been facilitated by differences in colony odour between the two groups of larvae. Getz et al. (1982) suggested that the ability to discriminate between adult full and half sisters could account for the tendency they observed for workers to segre- gate into full-sister groups during swarming. They caution, however, that this may simply be an artefact of a difference in the tendency to swarm between the two patrilines used in their study. Getz & Smith (1983) measured the frequency with which 5-day-old workers attacked their full and half sisters. They found that when the bees were given an opportunity to learn only the phenotypes of their full sisters (when maintained in full-sister groups before being tested for aggression), workers attacked their half sisters significantly more often than they did their full sisters. However, when given an opportunity to learn the phenotypes of both their full and half sisters (as would occur in a natural colony), workers did not attack one group significantly more than the other (Getz, personal communication). This may have been due to an inability to discriminate between the two groups, or to a lack of motivation to attack discriminately since, under normal circumstances, 5-day-old workers are nurse bees cooperating with, rather than attacking, other colony members. We tested the ability of workers to discriminate between their full and half sisters in a context in which aggression among colony members occurs normally. When a colony becomes hopelessly queenless, lacking both a queen and brood young enough to allow a replacement to be reared, the ovaries of some workers develop. These workers begin to lay unfertilized, male-producing eggs. Coincident with the onset of oviposition, there is a dramatic increase in aggression among colony members (Sakagami 1954; Velthuis 1976). Study- ing kin discrimination under these circumstances is particularly attractive because the aggressive acts occur frequently and are easily detected, and the behaviour can be studied (through the use of observation hives) in its natural context. To determine whether aggression in colonies with laying workers was preferentially directed toward half sisters, four colonies were set up with queens homozygous for Cordovan (a recessive mutation for cuticle colour). These queens had been artificially inseminated with the sperm of two males (one Cordovan, the other wild-type). The workers in these colonies consisted, therefore, of two patrilines which were easily distinguishable by their cuticle colour. Three frames of bees and brood from such colonies were placed, without their queen, in glass-walled observation hives and were prevented from rearing replacement queens by the destruction of all queen cells that appeared. In each observation colony, some of the workers began to lay eggs and the aggressive acts described by Sakagami (1954) began to appear. Aggression was monitored by systematically scanning the colony at 10-min intervals and recording, for each aggressive interaction observed, the patriline mem- bership of the participants. All observations were made under white light. Colonies were observed for up to 10 h a day, on up to 9 days. At the end of each day, the number of attacks in each category was counted (wild-type attacks wild-type, wild-type attacks Cordovan, etc.) and these counts were used to calculate the expected number of attacks for that day. If workers were not capable of discriminating between their full and half sisters, then the expected number of attacks by a worker ofpatriline A against a member of patriline B would be: (probability of a member of patriline A being an attacker) x (proba- bility of a member of patriline B being a victim) x (total number of observed attacks). The data for all days on which a colony was observed were 924