Definition
Kin selection is the evolutionary process by which selection favors behavior that helps genetic relatives, even at a cost to the helper, because aiding kin propagates copies of the helper's own genes through an indirect route. Formalized by W. D. Hamilton in 1964, it is captured in the inequality rB > C — when relatedness times benefit to the recipient exceeds the cost to the actor, a gene for that helping behavior spreads.
Kin selection is not an alternative to natural selection; it is natural selection, viewed from the level of the gene rather than the individual. The wider accounting it produces — direct reproductive output plus indirect contribution to relatives' reproduction, each weighted by relatedness — is called inclusive fitness. Together these ideas dissolve the long-standing puzzle of biological altruism: a gene that builds bodies which sacrifice for kin can outcompete a gene that builds bodies which never do.
Why it matters
How it works
Hamilton's rule, in plain arithmetic
Imagine a gene that disposes its bearer to perform an altruistic act costing C units of reproductive output and yielding B units to a recipient. The recipient carries that same gene with probability r. So the expected gain in copies of the gene flowing through the recipient is rB; the expected loss through the actor is C. If rB > C, the gene spreads; if rB < C, it doesn't. The arithmetic is bookkeeping at the level of the replicator, and it explains why a parent will sacrifice substantially for a child (r = 1/2) but rarely for a stranger (r ~ 0).
The rule does not require animals to perform any calculation. It says, instead, that the genes which built bodies behaving as if rB > C were the genes that, on average, left more copies in the next generation. The behavior is the rule's downstream signature, not its conscious application.
Relatedness, exactly
Hamilton's r is a precise quantity: the probability that an allele in the actor is identical-by-descent with the allele at the same locus in the recipient — that is, the probability they share the gene because both inherited it from a common ancestor. For diploid sexual organisms the canonical values are r = 1 for self and identical twins, 1/2 for parent-child and full siblings, 1/4 for half-siblings and nieces/nephews, 1/8 for first cousins, and 1/32 for second cousins. These are averages over many pairings; any particular pair may share more or less, but selection cares about averages because gene-pool dynamics are averages.
The framework cleanly predicts the gradient observed across thousands of species: organisms favor close relatives over distant ones, and the strength of the favoritism scales with r. Sapolsky (Behave, ch. 11) treats this gradient as one of the most powerful explanatory tools in all of social behavior — it accounts for hereditary ranking systems, cousin-mating preferences, competitive infanticide, and the family politics of dozens of primate species.
Inclusive fitness — the right generalization
Classical Darwinian fitness counted only an individual's own offspring. Hamilton's inclusive fitness counts an individual's direct reproductive output plus the indirect contribution to relatives' reproductive output, each weighted by r. A successful gene is one that maximizes the inclusive fitness of the bodies it builds. Once fitness is measured inclusively, the apparent contradiction between "selfish gene" and "altruism toward kin" evaporates — they are two sides of the same accounting.
This is the technical reframing of natural selection that underwrites Dawkins's gene-centric view. In The Selfish Gene, kin selection is not a special case to be tacked on at the end; it is the first big consequence of treating the gene, rather than the organism, as the unit of selection. From the gene's-eye view, a body is a temporary vehicle, and helping the bodies of relatives is just one route by which a replicator secures copies of itself.
Haldane's quip and the worked examples
The topic retells J. B. S. Haldane's reply when asked whether he would lay down his life for his brother: "No, but I would for two brothers or eight cousins." Two brothers at r = 1/2 sum to r = 1, the same value as Haldane's own life. Eight cousins at r = 1/8 also sum to 1. The quip is not advice on actual behavior; it is a way of seeing why the math works. The same arithmetic explains aunts who babysit nieces and nephews, alarm calls that warn nest-mates at the caller's expense, and the food-sharing patterns of many social mammals.
When you encounter biological altruism — alarm calls, food sharing, parental care, social grooming — estimating r, B, and C is the right first move. B and C are usually proxies (calories, time, survival probability); r is the relatedness coefficient. If rB > C makes the trait look advantageous, kin selection is the null hypothesis. If rB < C, you need another explanation — reciprocal altruism, manipulation, or mismeasured B and C.
Eusocial insects — the crown jewel
Hymenopteran societies (ants, bees, wasps) take kin altruism to its logical extreme: most individuals never reproduce, devoting their lives to raising their queen's offspring. Hamilton showed this becomes evolutionarily sensible under haplodiploidy, the genetic system in which females have two parents but males have only one. Under haplodiploidy, full sisters share r = 3/4 on average through their shared father — more than they would share with their own potential offspring (r = 1/2). A worker can therefore propagate more copies of her genes by helping her queen mother produce more sisters than by trying to produce daughters of her own. The architecture of bee, ant, and wasp colonies — sterile worker castes, daughter-rearing devotion, suicidal hive defense — is Hamilton's rule made flesh, with no hand-waving required.
Alarm calls and other classic field cases
A Belding's ground squirrel that gives an alarm call when a hawk appears raises its own risk of being eaten but lowers the risk to nearby relatives. Sherman's classic field work showed that callers are disproportionately females with nearby kin — exactly the pattern Hamilton's rule predicts. Vampire bats regurgitate blood meals to roost-mates that failed to feed, but the sharing is heavily biased toward genetic kin and stable reciprocal partners, again consistent with kin selection (and, for unrelated pairs, with reciprocal altruism). Sapolsky uses examples like these to illustrate how a once-baffling class of behavior dissolves into routine prediction once the gene-level accounting is applied.
How animals recognize kin
Hamilton's rule tells us what gene-level logic favors kin altruism. It does not tell us how an actual animal recognizes its kin. The Selfish Gene surveys the mechanisms: spatial co-occurrence (the chicks in my nest are probably mine), olfactory matching (the smell of this lamb matches my own smell), phenotype matching (the bee that smells like me is my hive-mate), and the more speculative "green beard" effect — a gene that recognizes copies of itself by an external marker. Green-beard genes are rare in pure form but have been observed in social amoebae and in fire-ant queen-recognition systems. Kin-recognition mechanisms are themselves under selection: a mechanism that misidentifies kin costs the gene; one that fails to identify true kin also costs the gene. The result is good enough recognition — better than random, calibrated to the structure of the ancestral environment.
Group selection is not the answer
Sapolsky and Dawkins are emphatic that kin selection replaces, rather than supplements, the older "for the good of the species" story. The discredited rival — group selection — held that animals voluntarily restrain themselves to avoid over-population, or sacrifice for the herd. Sapolsky's example is sharp: the old wildebeest did not nobly sacrifice itself for the herd; look closely and it was old and weak, and the others pushed it in. Animals behave to maximize copies of their genes; what looks like species-level altruism is almost always kin-level accounting, reciprocal exchange, or coercion. Dawkins (Family Planning, ch. 7 of TSG) demolishes Wynne-Edwards's group-selectionist demography on the same grounds: clutch sizes optimize individual gene transmission, not collective restraint.
Roots of morality without religion
Dawkins (The God Delusion, ch. 6) deploys kin selection — together with reciprocal altruism, reputational concerns, and "misfiring" of these mechanisms in modern contexts — to ground human morality in evolution rather than scripture. Goodness is not incompatible with the selfish gene; the gene is selfish, but the organisms it builds are often programmed to be kind, because kindness toward kin (and toward likely reciprocators) is exactly how a gene secures copies of itself. Once a moral psychology evolves for genetic reasons, it can fire promiscuously in environments very different from the small-band Pleistocene where it was shaped — which is why we feel pity for strangers on the news, generosity toward unrelated friends, and tribal warmth toward fellow citizens. The same machinery that built family loyalty now powers loyalty to nation, faith, and football team.
Pseudo-kinship: hijacking the system
Limits and complications
Kin selection is the null hypothesis for biological altruism, not the only answer. When rB < C, other mechanisms must be doing the work — typically reciprocal altruism (delayed pay-back across repeated encounters), reputation in transparent communities, or manipulation. Hamilton's framework also interacts with parent-offspring conflict (Trivers, ch. 8 of TSG): a parent is r = 1/2 to each child, but each child is r = 1 to itself and only r = 1/2 to siblings, so each child evolutionarily "wants" twice as much parental investment as the parent's own genes "want" to give. Weaning conflict, sibling rivalry, and even the prenatal tug-of-war over blood sugar between fetus and mother are predicted, not pathological. The family is not a refuge from selfish-gene dynamics; it is the most-studied arena of them.