Kin Selection and Altruism

A worker who never breeds does not have zero descendants — they are booked under someone else's name

First, block the sentence that comes most easily: the worker is not sacrificing herself "for the good of the colony". That explains nothing — if the payoff were tallied at the level of the group, then a variant that quietly laid its own eggs and did nothing else should spread through such a group fast.

Change the arithmetic and it works. Any stretch of genetic material in you has a one-in-two chance of sitting in your full brother as the same copy, inherited from the same ancestor. That number is called relatedness: one half between parent and child too, a quarter for a nephew or niece, an eighth for a cousin.

So the books can be kept this way: you do something, you end up with fewer offspring, and some relative ends up with more. As long as "the offspring your relative gains" times "how related you are to them" beats "what you gave up", the behaviour spreads generation after generation. In 1964 Hamilton wrote it as a single line:

r is relatedness, b is the offspring the other side gains, c is what you lose. This is the only formula this site will ever spell out — because it is the protagonist of the whole explanation. The combined tally, what you pass on directly plus what you pass on through relatives you helped, is called inclusive fitness.

The best-travelled line here is Haldane's: "I would not lay down my life for one brother, but I would for two brothers or eight cousins" — each set exactly cancelling his own share. Take it with a pinch of salt: there is no written source for it, and it comes to us through Maynard Smith's later recollection. And block the teleology once more: nobody is doing this arithmetic. All that happens is that individuals carrying a "more likely to help" variant pass on more copies of it on average, so its share of the population drifts up.

The cleanest field evidence comes from a ground squirrel in the California mountains. In 1977 Sherman published three summers and three thousand hours of observation: every animal in the population was individually marked, and maternal kinship was on record. When a coyote or a weasel shows up on the ground, some squirrels stand and scream. That scream is not free — the records show that the caller is more likely to be targeted. Who calls? Mostly females who happen to have mothers, sisters or daughters living nearby; females with no relatives around call noticeably less. Males leave home as adults, have no close kin nearby, and barely call at all.

What does the work is not kinship itself, but a cue that can be fooled

For that line to run, something has to be true: the one being helped really is more likely to be carrying the same copy. The trouble is that animals cannot check a pedigree.

One workaround is not needing to. In many species the young do not travel far and settle near where they were born — so "the one next door" is a relative with fairly high probability by default. Just by not moving, relatedness in that inequality is already above zero, with no recognition ability required at all.

Another is to learn a cue. Long-tailed tits are a small European bird; adults whose own nest has failed go and feed chicks in someone else's nest, and clearly favour a relative's. How do they tell? In 2005 a group answered: by a contact call known as the churr, which sounds alike among birds from the same nest. The decisive step was what came next — swap eggs into another nest, and when the chicks grow up their calls resemble their foster parents', not their genetic ones. The cue is learned in the nest from whoever feeds you, not born with you.

These birds did not grow an organ for measuring kinship. What they have is a coarse, cheap rule of thumb: whatever you heard in the nest as a chick, treat as family later. It is good enough under normal conditions, because the nest really is family — but it can obviously be fooled, if something else is put in the nest.

There is a stranger route as well. In 1998 a report came out of the red fire ant: workers carrying one variant at a particular locus execute queens who lack it — what gets recognised is not kinship but a specific marker. Such things are called green beards: one and the same thing both grows the green beard and favours those who wear one. Whether this counts as a clean green beard is disputed (the genetic structure in that region is far more complicated than a single gene), but it makes one thing plain: relatedness in the inequality is really "the probability of sharing the same copy", and common ancestry is just the most common way of raising that probability — not the only way.

Why ants and bees keep arriving here: a beautiful answer, cut in half by its own field's evidence

Back to the worker who does not breed. Bees, ants and wasps determine sex in an unusual way: fertilised eggs become females, unfertilised eggs become males directly — so a male has only one set of chromosomes, and only that one set to pass to a daughter.

Hamilton noticed a consequence. If a queen mates with just one male in her life, her daughters get an identical full set from their father and half-matching material from their mother, which adds up to three quarters between sisters — higher than the one half between a mother and her own daughter. So "stay home and help mother make more sisters" beats "go out and make daughters". The inference was too beautiful not to enter textbooks for decades.

Then its own field cut the argument down. Termites are ordinary diploids, have nothing to do with this sex-determination system, and evolved sterile workers anyway. Three quarters holds only when the queen mates once, and plenty of ant and bee queens mate with dozens of males; add in that a worker is related to a brother by only a quarter, and most of the remaining edge cancels out.

The version more people now accept is plainer: what matters is not three quarters, it is the single mating itself. Under strict monogamy your younger siblings are related to you exactly as much as your own children would be — one half — so helping at home only has to be slightly more efficient than going it alone for the inequality to hold. In 2008 a group compared 267 eusocial bee, ant and wasp species and reconstructed their ancestral states: in all eight lineages that independently took this path, the ancestor was monogamous without exception, and multiple mating always appeared later.

The strength of that evidence should be stated plainly: it is a correlation across species rather than an experiment, reconstructing ancestral states carries its own uncertainty, and a recent paper argues the case is "inconclusive"; in 2010 others proposed replacing the whole inclusive-fitness framework. The argument has not fully settled, though most researchers still work with Hamilton's line.