Why honey bees produce these extra queens is not fully clear, perhaps as insurance against one of them dying. But they do limit the conflict to a few individuals by controlling queen production through the special feeding they require. This limitation is extremely common in social insects with queen and worker castes. Queens generally require more food, offering the opportunity to control queen production (Wilson, 1971). It is instructive to see what happens in the unusual cases in which this constraint does not hold.

In one genus of stingless bees, Melipona, caste is not determined by differential feeding. Instead, workers and queens develop in cells of the same size, provisioned with the same amounts of food. This presumably leaves the choice of being queen up to each developing female larva. As a consequence, a significant fraction of females (5–20%) develop into queens, with small heads, large abdomens, and lacking the pollen baskets required to be effective workers (Wenseleers and Ratnieks, 2004). Because stingless bees reproduce by colony division, this amount of queen production is far more than the single queen they need, every once in awhile, to head a new colony. The excess queens, useless for work and a potential threat to the old queen, are slaughtered by workers (Wenseleers et al., 2004). The 5–20% reduction in worker production must constitute a significant cost to the colony and clearly shows that cross-purpose can remain important in advanced social insects.

Stingless bees other than Melipona determine caste by the usual means of feeding some larvae more, but this does not entirely solve the problem. In some species, in which brood cells are adjacent, a larvae that is supposed to be a worker can tunnel from its own cell into its neighbor’s, consume the food stores intended for its neighbor, and develop as a queen (Engels and Imperatriz-Fonseca, 1990). In other species, some individuals with worker-sized food allotments will nevertheless develop into morphological queens (Wenseleers et al., 2005). In Schwarziana quadripunctata, these dwarf queens make up only 0.6% of all individuals reared in worker-sized cells but 86% of all queens reared. These queens are less successful than normal-sized queens in attaining reproductive status and are executed more readily by the workers. Still, the strategy appears to be successful often enough to pay, inasmuch as 22% of all reproductive females are dwarf queens (Wenseleers et al., 2005). Some ant genera, such as Myrmica and Solenopsis, also have some small queens (called microgynes), which may be the result of individual larvae determining their own caste fate in colonies that are initiated by budding (Bourke and Franks, 1995; McInnes and Tschinkel, 1995).

The threat of other queens may lie behind another colony-level design flaw that is usually not obvious but is present in many species: the lack of a backup queen. Consider the fungus-growing ants of the genus Atta. A