Reciprocal altruism

Reciprocal altruism is cooperation in which an individual helps another at some cost, with the possibility that the favor will be returned later. Its evolutionary problem is the delay: an individual that accepts help can benefit by refusing to repay it. The book explores how repeated encounters, recognition, and conditional responses to cheating can make delayed cooperation possible, and how the same logic illuminates human institutions and changes in the apparent stakes of a game.

The problem of deferred repayment

When favors and returns occur at the same time, mutual benefit is relatively easy to imagine. A delay changes the calculation. The first recipient may take the benefit and withhold repayment when the roles are reversed, leaving the original helper with the cost and no return. The account therefore treats reciprocal altruism not as an automatic consequence of a mutually useful exchange, but as a problem of incentives across time.

A hypothetical case makes the difficulty concrete. A bird can remove parasites from most of its own body by preening, but cannot reach the top of its head. A companion can remove them there, and later receive the same service when it is the one in need. Mutual grooming is common among birds and mammals, but the intuitive appeal of this exchange does not explain how it evolves: genes have no foresight with which to plan a mutually beneficial arrangement.

The account associates the possibility of delayed repayment with the ability to recognize and remember individuals. It notes that G. C. Williams discussed the issue in 1966 and that R. L. Trivers developed it further in 1971, connecting the problem to the Prisoner’s Dilemma. Repeated interaction matters because the current act can affect what happens when the same partners meet again; a one-off favor offers no such future consequence.

Sucker, Cheat, and Grudger

The hypothetical grooming population begins with two behavior programs. Suckers groom anyone who needs help, while Cheats accept grooming but never return it. These are not conscious plans but inherited patterns of behavior. If parasites are sufficiently common, Suckers in a population of Suckers receive help about as often as they give it, so the exchange benefits both sides. But a lone Cheat can receive assistance from everyone while bearing none of the cost. Its relative advantage allows Cheat genes to spread, even though the population as a whole would do better if everyone remained a Sucker.

The trouble persists at mixed frequencies. Even when Cheats make up half the population, they still do better than Suckers by taking the benefits without paying them back. At 90 percent Cheats, the average return for all individuals becomes very low, and many may die from the infection carried by the parasites; nevertheless, Cheats still outperform Suckers. If those two strategies were the only possibilities, the Suckers would be driven out and the population might also decline toward extinction. The example separates individual advantage from collective benefit: a strategy can spread even as the conditions shared by the population deteriorate.

Grudger introduces conditional cooperation. Grudgers help strangers and those who have helped them before, but remember a Cheat and refuse to help that individual in future. In a population of Suckers and Grudgers, the two are indistinguishable in their conduct toward others and earn the same high average return. Against a population dominated by Cheats, however, a rare Grudger initially does poorly: it spends energy helping others before it has accumulated enough experience to identify Cheats, while receiving little help in return.

Once Grudgers become common enough, their chance of meeting one another offsets the cost of helping Cheats early on. Their average return then rises above that of Cheats, which are driven toward extinction. Cheats may remain as a minority for a long time: a rare Cheat is unlikely to meet the same Grudger twice, so relatively few individuals will have learned to withhold help from it. A computer simulation confirmed that the outcome was not as intuitively obvious as the initial story might suggest. Grudger is stable against invasion by Sucker or Cheat when Grudgers predominate, but Cheat can also be stable when Cheats predominate; depending on the assumed payoffs, either state may have a larger zone of attraction.

Repeated games and the structure of cooperation

The Prisoner’s Dilemma gives this problem a general strategic form. In its human version, a banker pays two players according to cards they select simultaneously and reveal only afterward. Each player chooses either to cooperate or defect, and the outcome depends on both choices. The game matters because mutual cooperation can benefit both players, while each may also face an incentive to defect; it captures the tension between a shared gain and the temptation to take an immediate advantage.

A crucial condition for reciprocal strategies is that the interaction be repeated. If players know that a game will end after exactly 100 rounds, the final round becomes a one-off Prisoner’s Dilemma. Working backward, the 99th round then becomes effectively one-off as well, and the same reasoning applies successively to earlier rounds. The book therefore treats the end date as important: for repeated cooperation to be viable, the players must not know when the interaction will finish. Even when the exact number of rounds is uncertain, signs that the end may be near can affect behavior by making defection seem more attractive.

Tit for Tat illustrates a cooperative strategy adapted to repetition. It is “nice” because it never defects first, “forgiving” because it has a short memory for past misdeeds, and “not envious” because it can accept an equal return when both players receive a high payoff. It does not outscore its partner in any particular game, since it defects only in retaliation; its advantage is the possibility of sustaining a high shared score. In this setting, calling the other participant an opponent can be misleading: the players may gain more by cooperating with each other than by trying to do one another down.

The book contrasts this with zero-sum games, where one player’s gain is another’s loss. Chess is the example of a zero-sum contest; the Prisoner’s Dilemma is nonzero-sum because a banker can provide an outcome in which both players benefit. This distinction clarifies the point of reciprocal altruism: cooperation need not mean that participants cease to pursue their own returns. It can be favored when the rules and repeated structure of the interaction allow both parties to do better through cooperation than through mutual defection.

Human arrangements and changing payoffs

The book applies the distinction between zero-sum and nonzero-sum situations to divorce. A marriage is described as a nonzero-sum game sustained by cooperation, and a couple may still benefit from cooperation after the relationship breaks down. Yet, in England and, until recently, in all fifty states of the United States, lawyers’ professional rules required each lawyer to represent only one member of the couple. A partner seeking advice from the first lawyer could be turned away and forced to seek a separate one.

The resulting process can make a potentially cooperative dispute adversarial. Each lawyer may make a proposal both sides expect to be rejected, prompting another unacceptable counterproposal. The letters and calls add to the bill, and the process can continue for months or years. The two lawyers need not coordinate consciously: their professional separation can itself structure a kind of cooperation between them, while the couple pays the costs. The example shows how rules can shape incentives and turn one set of participants’ shared interests into another set of participants’ mutual returns. The account also notes that some lawyers take the contrary role, persuading clients who want a zero-sum fight that an out-of-court settlement would serve them better.

A football match shows how external information can alter a game’s payoff structure in the middle of play. On May 18, 1977, Sunderland, Bristol, and Coventry were contesting the final relegation place in the First Division. Sunderland were playing elsewhere, while Bristol and Coventry were facing each other in a match running five minutes late. If Sunderland lost, a draw between Bristol and Coventry would keep both teams in the First Division; if Sunderland won, one of the two would be relegated depending on their match’s result.

The Bristol–Coventry game was 2–2 in the 80th minute. Two minutes before the end, news arrived that Sunderland had lost, and the Coventry manager had it displayed on the stadium’s electronic message board. Both sides now needed only a draw to avoid relegation. Trying to score became a risk, since taking players away from defence could cause a loss, so the teams shifted from striving to defeat one another to securing the shared result. Supporters who had been rivals joined in celebration, while the players moved the ball with little challenge. The outside result acted like a banker, transforming what had been a zero-sum contest into a nonzero-sum situation for the two clubs.

Evidence and limits of the account

Blood sharing by vampire bats provides a biological case in which the proposed conditions for repeated cooperation can be examined. The bats feed at night; some return to the roost without food, while others may have more than they need. On a later night, luck may reverse. G. S. Wilkinson observed successful bats regurgitating blood for less fortunate companions. Of 110 observed donations, 77 were readily understood as mothers feeding offspring, and many other exchanges involved genetic relatives. Some cases, however, involved unrelated bats, which could not be explained simply by kinship.

Those unrelated partners tended to be frequent roostmates, giving them opportunities for repeated encounters. Wilkinson also compared the costs and benefits of donations in terms of hours of life. The same quantity of blood adds more time to the life of a severely starved bat than to that of a less hungry recipient. Donating carries a cost to the donor, but the recipient’s gain in survival can be larger than that cost. For a donor, the blood is less precious on a lucky night than it would be to the recipient on an unlucky one; the exchange can therefore fit the payoff structure of a Prisoner’s Dilemma.

Repeated interaction alone is not enough if the participants cannot distinguish partners. Wilkinson tested recognition by taking one captive bat away for a night and starving it while the others were fed, then returning it to the roost. The experiment used bats from two separate groups taken from caves many miles apart and was repeated with different bats taking turns as the starved individual. Of 13 donations, 12 came from an old friend from the same cave, and one from a new friend from the other cave. The odds of that pattern occurring by chance were calculated as less than one in 500, supporting the conclusion that the bats preferentially helped familiar roostmates.

These cases sit within a broader argument about individual incentives rather than an appeal to the good of a group or species. A conspiracy of Doves might give every participant a better average payoff than a population at the stable Hawk–Dove ratio, but a lone Hawk could exploit the arrangement, making it vulnerable to treachery. Likewise, a welfare state can sustain children who would otherwise exceed their parents’ resources, but the account argues that such support requires birth control if it is not to be abused. These comparisons underline a recurring distinction: collective benefit does not by itself ensure stability. Reciprocal altruism is an account of how cooperation can persist despite individual incentives to cheat, not a claim that every apparently generous act is secretly repayment in disguise. The discussion explicitly resists reducing unpaid blood donation to simple reciprocity, while treating the bat exchanges as a stronger fit with the repeated-game model.