Definition
Reciprocal altruism is cooperation between non-relatives sustained by the expectation of a return — one individual bears a cost to benefit another in the reliable expectation that the favor will be repaid. The mechanism was formalized by Robert Trivers in 1971 and, alongside individual selection and kin selection, became one of the three foundational answers to the puzzle of biological altruism.
Unlike kin selection, reciprocal altruism does not require shared genes. It requires repeated interaction, individual recognition, and a way to withdraw help from those who fail to return it. Where those conditions hold — in cleaner fish on a reef, in vampire bats in a roost, in primate troops, in human villages — cooperation between strangers can be evolutionarily stable. Where they fail, defection dominates.
Why it matters
How it works
Trivers's mechanism and the three conditions
Robert Trivers's 1971 formalization of reciprocal altruism asked a precise question: under what circumstances can a gene that disposes its bearer to help an unrelated individual spread? The answer is a small set of necessary conditions. First, the interaction must be repeated — if you and I will meet only once, my help to you costs me C and gains me nothing, so I should defect. Second, individuals must recognize each other — if we cannot distinguish past partners from strangers, kindness cannot be specifically rewarded and cheating cannot be specifically punished. Third, the relationship must include a credible withdrawal mechanism — if I help you and you defect, I must be able to stop helping next time, or the gene for naive helping is invaded by the gene for exploitation.
Where all three conditions hold, reciprocal altruism stabilizes. Where any one fails — anonymous large groups, one-off encounters, no memory — cooperation collapses back to the ceiling that kin selection alone can support. The framework gives a sharp prediction: cooperation between unrelated individuals should appear precisely where these structural conditions are met, and nowhere else.
Dawkins on the gene's-eye view of cooperation
In The Selfish Gene, Dawkins frames reciprocal altruism as the third pillar of biological altruism alongside individual selection and kin selection. The apparent paradox of the title dissolves once you see the level of analysis correctly: a gene that is selfish in the sense that it tends to spread copies of itself can build bodies that are altruistic, when altruism is exactly the strategy that secures more copies. Topic 10 — You Scratch My Back, I'll Ride on Yours — surveys cooperation between unrelated individuals and even between species: bees and flowers, cleaner wrasse and their client fish, lichens (an alga plus a fungus), gut bacteria and their human hosts.
The cleaner fish case shows all three conditions in operation. Each cleaner holds a station on the reef; each client returns to the same station; both recognize one another; both benefit repeatedly. Cheaters — mimic species that gain admission and then bite — are remembered and avoided, exactly as Trivers's theory predicts. The line between mutualism and parasitism is thin and shifts with circumstance: gut bacteria are mutualistic in normal conditions but can turn pathogenic in stressed ones. The topic also notes that long-lived, recognizing species accumulate the most reciprocity — a clue to why humans, with long lives and strong memory, are so heavily cooperative.
Axelrod's tournament and the winning shape
The 1989 second edition of The Selfish Gene added Topic 12 — Nice Guys Finish First — to walk through Robert Axelrod's 1980 iterated-Prisoner's-Dilemma computer tournament. In a one-shot Prisoner's Dilemma, defection dominates: whatever the other player does, you do better by defecting, even though mutual cooperation would beat mutual defection. Axelrod asked what happens when the same two players meet again and again, remembering each other's past moves — the iterated version, where the shadow of the future enters every choice.
Game theorists submitted strategies ranging from elaborate (long memory, statistical modeling of opponents) to trivial. The winner was the simplest entry, Anatol Rapoport's four-line tit-for-tat: cooperate on the first move; thereafter, do whatever the opponent did on the previous move. That was the entire algorithm, and it won. Axelrod ran a second tournament with entrants knowing the first result and trying to design strategies that exploit tit-for-tat. Tit-for-tat won again. Analyzing the winners, Axelrod identified four shared properties: niceness (never defect first), retaliation (punish defection immediately), forgiveness (return to cooperation as soon as the partner does), and clarity (be transparent enough that the partner can learn your rule). The most counterintuitive of the four is clarity — in repeated games, being predictable is an asset, because it lets the other side learn how to work with you.
Limits and noise
Tit-for-tat is not invincible. Under noisy perception — where a cooperative move can be mistaken for a defection — two tit-for-tat players can lock into a punishing defection cycle from a single mis-read move. Variants like tit-for-two-tats or generous tit-for-tat tolerate the occasional mistake. Against an unconditional defector, tit-for-tat loses on the first move and breaks even thereafter, never inflicting net damage. Grim trigger (defect forever after one defection) punishes more harshly but cannot recover from noise. And a cunning opponent can test tit-for-tat by defecting once, absorbing the retaliation, and then settling back to cooperation, having taken a small one-off advantage.
Even with these caveats, the broader pattern holds across tournaments, simulations, and real biological systems: the winning shape of an iterated-cooperation strategy is nice, retaliatory, forgiving, and clear. Dawkins lists biological cases that appear to implement something close to it — vampire bats sharing regurgitated blood and remembering non-sharers, stickleback fish pairing scouts and avoiding the ones that hang back, cleaner fish maintaining stations and recognizing mimics.
Sapolsky on cooperation among strangers
In Behave, Sapolsky presents reciprocal altruism in Topic 11 as one of the three foundations of behavioral evolution, alongside individual selection and kin selection. Together they account for behaviors that would otherwise be baffling — competitive infanticide, hereditary ranking systems, cousin-mating preferences, vampire bats feeding unrelated young. Reciprocity, specifically, is the evolutionary route by which cooperation becomes possible outside the circle of genetic relatives.
Sapolsky's contribution is to follow the consequences forward into psychology. Once selection favors conditional cooperation, it also favors the psychological machinery that polices it: gratitude rewards cooperators and motivates return; guilt restrains the tempted from defecting; trust allows the relationship to extend in time; moral outrage punishes cheaters and signals to bystanders. The result is that fairness, a sense of cheating, and even the felt indignation of being wronged are not cultural inventions laid on top of an indifferent nature — they are evolved adaptations for keeping reciprocal exchange stable. Sapolsky frames this as the soil from which much of human social emotion grows.
Dawkins on morality without religion
In Topic 6 of The God Delusion, Dawkins uses reciprocal altruism — together with kin selection — to argue that human moral instincts predate religion. The reasoning is direct: if natural selection on genes builds organisms that cooperate with relatives (because relatives share genes) and with reliable partners (because the favor will be returned), then dispositions like kindness, fairness, gratitude, and indignation have a Darwinian explanation that requires no scripture and no deity. Goodness is not incompatible with the selfish gene; in the right circumstances, kindness is exactly how a gene secures copies of itself.
The polemical point is that this dissolves a load-bearing pillar of the case that society needs religion. The descriptive point is more interesting: morality, on this view, is older and more universal than any particular tradition, available to believer and non-believer alike, grounded in the same evolutionary mechanics that built cleaner fish and vampire bats. Reciprocal altruism becomes part of the secular case for an evolved ethics.