Sterile worker castes

Why do worker bees raise a queen's offspring instead of having their own?

Selection is supposed to be an accounting of descendants. A trait that leaves fewer of them should disappear. Yet a honeybee colony contains tens of thousands of females who never reproduce at all, and who spend their entire working lives feeding a sister's offspring. Darwin himself called this "one special difficulty, which at first appeared to me insuperable, and actually fatal to my whole theory."

The problem is not that helping is odd. It is that a gene for never breeding has no obvious route into the next generation. So what is the route?

The first move is to stop counting bodies and start counting copies of genes. If a worker's labour causes extra sisters' offspring to exist, copies of her genes go forward through them. Hamilton put this as a threshold: helping is favoured when the relatedness between helper and beneficiary, multiplied by the benefit conferred, exceeds the cost to the helper's own reproduction.

Now the famous move — and this is where the popular account goes wrong, so it is worth doing slowly. Bees, ants and wasps are haplodiploid: females come from fertilised eggs and have two sets of chromosomes, males from unfertilised eggs and have one. A male therefore passes his single, identical set to every daughter. So full sisters share all of their father's contribution and, on average, half of their mother's — three quarters overall, against the one half a female would share with a daughter of her own. The conclusion writes itself: a worker propagates her genes better by raising sisters than by breeding.

Except it does not hold up. She also rears brothers, to whom she is related by only a quarter. Weight the two by how many of each the colony makes, and if investment is split evenly between the sexes the average relatedness to the brood she raises comes back to one half — exactly what she would have with her own offspring. The three-quarters figure is real but it is half of a pair, and the other half cancels it.

So haplodiploidy alone explains nothing. To make it pay, workers would have to bias the colony's investment toward sisters — which Trivers and Hare argued in the 1970s they in fact do, since workers control the feeding. That is a live and much-argued finding rather than a settled correction.

And the deeper trouble is that the correlation the haplodiploidy story predicts is simply absent. Termites are ordinary diploids and fully eusocial. So are naked mole-rats, which are mammals. Meanwhile most haplodiploid insects are solitary. Whatever explains sterile castes must work without haplodiploidy, and must fail to work in the many haplodiploid species that never evolved it.

Return to Hamilton's threshold and look at the other two terms, which the popular story neglects. Relatedness does not have to exceed one half; it only has to reach it, provided helping is more productive than breeding alone. And it reaches one half exactly when a mother mates with a single male for life — lifetime monogamy makes full siblings precisely as related as offspring. At that point the helper is indifferent on relatedness grounds, and any efficiency advantage at all tips the balance. Boomsma's observation that every independent origin of obligate sterile castes appears to trace back to a monogamous ancestor reframes the question: monogamy does not create an incentive to help, it removes the disincentive.

Then the ecology supplies the rest. What makes helping more productive than leaving? A nest that is expensive to build, defensible and reusable; a brood needing continuous care; high mortality for a lone female founding a colony. Each raises the benefit of staying or the cost of leaving — and each turns up repeatedly at the origins of social life in bees, wasps, termites and mole-rats.

Three refinements, and one honest admission.

Workers are mostly not sterile in the strict sense. Honeybee workers retain ovaries and will lay eggs — unfertilised ones, which become males — when the queen is lost, and at low rates even with a queen present. The colony suppresses this: other workers eat the worker-laid eggs, a behaviour called worker policing, favoured precisely because a worker is more related to the queen's sons than to another worker's. The caste is enforced from inside, not merely accepted.

Second, "for the good of the colony" describes the outcome, not the mechanism. The colony functions as a unit because its members' interests largely coincide, and where they do not — over sex ratios, over who lays the males — there is measurable conflict. Third, kin selection is not an alternative to ordinary natural selection but a way of doing its bookkeeping, counting a gene's effects wherever they land.

The admission: the field is not at peace. In 2010 Nowak, Tarnita and Wilson argued that inclusive fitness theory is unnecessary and sometimes misleading, and that standard population genetics plus ecology does the work; the response from over a hundred researchers was fierce, and the argument continues. Much of it concerns which mathematics better describes the same process rather than what happens in nests — but it would be dishonest to present the kin-selection account as uncontested.

• Why worker policing is predicted to be stronger in species where the queen mates many times.
• Whether the same threshold explains cooperative breeding in birds and mammals, where helpers usually do eventually reproduce.

Every term the collection defines is gathered in the glossary.