- Email: [email protected]
Human cooperation is more than by-product mutualism
ANIMAL BEHAVIOUR, 1999, 57, F1–F3 Article No. anbe.1998.0987, available online at http://www.idealibrary.com on
FORUM Human cooperation is more than by-product mutualism BORIS PALAMETA* & WILLIAM MICHAEL BROWN†
*Department of Psychology, St Thomas University †Department of Psychology, Dalhousie University (Received 24 August 1998; initial acceptance 14 September 1998; final acceptance 7 October 1998; MS. number: AF-2)
known to conform to a Prisoner’s Dilemma’ (page 533). Although we agree wholeheartedly that evidence of reciprocity requires experimental rigour, the experiments have been done and the evidence is there, at least for humans. There is an extensive literature in social psychology and experimental economics on human behaviour in mixedmotive games like PD. Many studies have found that stable cooperation is a common outcome of two-party and multiparty games (Pruitt & Kimmel 1977), especially when most players use TFT (Manstead & Hewstone 1995). Rapaport & Chammah (1965) are cited by Clements & Stephens (1995) and Stephens et al. (1997) as providing evidence that most PD games stabilize into mutual defection. In fact Rapaport & Chammah (1965) show that although cooperation initially declines in the first 30 trials, it ascends over the final 30 trials so that even players who appeared to be locked into mutual defection often switch to mutual cooperation. Results like this probably led Anatol Rapaport to conceptualize the strategy TFT, the eventual winner of Axelrod & Hamilton’s (1981) celebrated tournament. In this commentary, we shall briefly review some recent studies that consider mixed-motive games from an explicitly evolutionary perspective. The standard PD game has two possible strategies (cooperate or defect) and four possible outcomes, two symmetric (i.e. players play the same strategy and receive equal payoffs) and two asymmetric (i.e. players play different strategies and receive unequal payoffs). The PD is structured so that defection always yields higher payoffs than cooperation regardless of what the other player does. Therefore, if both players use the principle of dominance, that is, choose the strategy that yields the higher potential payoff, mutual defection is the endpoint. If players use reciprocity though, only symmetric outcomes are possible, and mutual cooperation is the endpoint because it is more profitable than mutual defection. Mixed motives arise from the fact that dominance
he study of cooperation between unrelated individuals received a major boost from Axelrod & Hamilton’s (1981) classic formulation of an evolutionarily stable solution (Tit for Tat, TFT) to the Prisoner’s Dilemma (PD). More recent models have shown that TFT is not always evolutionarily stable (Nowak & Sigmund 1993): indeed one can always find a mixed strategy capable of invading any pure strategy, including TFT (Dugatkin 1997). However, TFT has remained the most frequently investigated strategy in experimental studies of cooperation. Observations consistent with TFT have been reported for various animal species (Lombardo 1985; Milinski 1987; Dugatkin 1988; Milinski et al. 1990; Huntingford et al. 1994). Recently, Clements & Stephens (1995) disputed such claims on the grounds that (1) the payoff structures of the behaviours in question did not necessarily conform to PD, and (2) alternative explanations were not considered. Clements & Stephens (1995) recorded the behaviour of blue jays playing carefully controlled PD and mutualism games, and found stable cooperation only in the latter situation. Because the payoff structure of mutualism games ensures that cooperation is the best choice regardless of what your partner/ opponent does, the occurrence of cooperation in such games is simply a by-product of individual animals maximizing their own immediate rewards without reacting to each other’s behaviour at all (Roberts 1997). Thus there is no need to invoke reciprocity or TFT as a necessary route to cooperation. Clements & Stephens (1995) and Stephens et al. (1997) suggest that by-product mutualism is the most parsimonious explanation for the occurrence of cooperation in a number of species, including humans. Clements & Stephens (1995) go so far as to state that ‘there is no empirical evidence of nonkin cooperation in a situation, natural or contrived, where the payoffs are
T
Correspondence: B. Palameta, Department of Psychology, St Thomas University, Fredericton, New Brunswick, E3B 5G3, Canada (email: [email protected]). 0003–3472/99/020F01+03 $30.00/0
F1
1999 The Association for the Study of Animal Behaviour
F2
ANIMAL BEHAVIOUR, 57, 2
competes with reciprocity; dominance is the only way to achieve the highest possible payoff (defecting while the other cooperates), while reciprocity yields the most profitable symmetric payoff (mutual cooperation). The PD allows each player only one choice (cooperate or defect) per interaction. Recent research has focused on games that offer players several choices per interaction, while maintaining a payoff structure that is similar to the original PD. For instance, McCabe et al. (1996) investigated two-person, extensive-form bargaining games, in which players alternate moves along one of two branches of a game tree. At the pivotal point of the game, player 1 must choose between the branch that offers the highest individual payoff (dominance) and the branch that offers the highest symmetric payoff (reciprocity). Choosing dominance will most often lead player 2 to do likewise, thus terminating the game at a symmetric payoff lower than the one that could have been attained through reciprocity. Choosing reciprocity is risky though because player 2 then has the choice of cooperating by terminating the game at the point of highest symmetric payoff, or defecting by moving towards a higher asymmetric payoff. In some versions of the game, player 1 is given the option of responding to player 2’s defection by either accepting it (i.e. accepting the low portion of an asymmetric payoff) or punishing player 2 at some cost to itself (i.e. reducing both players’ payoffs to the lowest possible point). In this context, punishment is a form of escalated defection that costs both parties, but costs the initial defector more. Like PD, bargaining games can be played on a one-shot basis or with repeated interactions; repeated interactions can be with the same player or with different players each time. Unlike PD, bargaining games may add the option of punishing defection. Regardless of which version is played, game theory predicts a noncooperative outcome. Yet, as we describe below, cooperation occurs far more frequently than expected. Furthermore, the frequency of cooperation varies in accordance with reciprocity theory as the parameters of the game are changed. First, consider the effect of punishment. It is not difficult to see how the option to punish can enhance the effectiveness of TFT. The option of punishing a player who fails to reciprocate cooperation acts as a buffer against exploitation. Reciprocity theory predicts that cooperation will be initiated more frequently and develop more readily when defection can be punished. McCabe et al.’s (1996) results support the prediction. For instance, when subjects played repeatedly, each time with a different counterpart, the percentage of defections that were punished rose from 42% in the first five trials to 81% in the last five trials. This was accompanied by a concurrent increase in mutually cooperative outcomes, from 40 to 57%. When defections could not be punished, mutual cooperation was less frequent (less than 25% of outcomes). Another prediction from reciprocity theory is that mutual cooperation is more likely to arise when the same players encounter each other repeatedly (Trivers 1971). Again, data from bargaining games support the prediction. When players were paired with one counterpart for repeated play, the percentage of mutually cooperative
outcomes rose to 77% in the last five trials (McCabe et al. 1996). In the situation most similar to Clements & Stephens’ (1995) paradigm (i.e. repeated play with the same partner, no opportunity to punish defection), mutual cooperation was initially difficult to achieve, occurring in only 29% of the first five trials. However, mutual cooperation rose to 68% in the last five trials. This implies that cooperation is not merely contingent upon the threat of strategic retaliation, but that it can also be fostered by the sense of reciprocity that develops when two players encounter each other repeatedly. There is evidence that some individuals try to cooperate even when told they will be playing a randomly selected counterpart only once. The game theoretic prediction is that 100% of such one-shot games should end in mutual defection. Contrary to this prediction, in McCabe et al.’s (1996) study, almost 50% of the players attempted cooperation. Furthermore, attempts at cooperation were defected on relatively few trials (50% of attempts when no punishment was possible, and only 23% of attempts when punishment was possible). To achieve mutual cooperation in games with PD-like payoff structures, both players in turn must refrain from seeking to maximize individual payoffs. Cooperative plays are more frequently reciprocated when backed by the threat of retaliation, but reciprocity also occurs in one-shot games where cooperation is backed by nothing more than trust. The potential benefits of trust are considerable, provided that a trusting player is paired with a trustworthy counterpart. Frank (1987, 1988) developed the theory that cooperation based on trust could evolve as long as cooperators could accurately predict the trustworthiness of their fellow players before deciding whether or not to play. Frank et al. (1993) tested the theory by having people who had never met interact for 30 min before playing a traditional (i.e. no option to punish) one-shot PD. Despite the fact that players knew that neither they nor the experimenters would be permitted to discover any individual subject’s choice, a cooperation rate of over 70% was obtained. Furthermore, players were able to predict their partner’s choice with significantly higher than chance accuracy. Other researchers have tested the limits of the human capacity for cooperation by creating games in which cooperation is more difficult to achieve than it is in PD. These games are structured to allow one player more decision-making power than the other. For example, ultimatum games allow one player to make a unilateral decision on how a resource (e.g. $10) is to be divided. The second player may either accept the offer or reject it, in which case neither player gets any money. The game theoretic prediction is that the first player should offer the smallest amount possible ($1), and the second player should accept it because $1 is better than nothing. In fact, results of ultimatum games usually deviate from this pattern. Offers to split the resources evenly are often the modal result, and offers of uneven splits are sometimes rejected (Guth et al. 1982; Forsythe et al. 1994). Rejection in this context is similar to the strategic punishment discussed earlier; both players lose, but the player who made the initial offer loses more.
FORUM
Dictator games are like ultimatum games with the threat of punishment removed, that is, offers cannot be refused. Game theory predicts that the player making the offer (the dictator) should maximize payoffs by keeping all the money and offering the player nothing. As expected, offers are generally lower in dictator games than in ultimatum games (Forsythe et al. 1994). Nevertheless, the proportion of dictators that offer a 50–50 split (about 20%) is equal to the proportion that offer nothing (Forsythe et al. 1994). Hoffman et al. (1994) attribute these results to the fact that dictator experiments rarely offer subjects complete anonymity. Even when betweensubjects anonymity is assured, experimenters are still aware of subjects’ behaviour. Hoffman et al. (1994) allowed subjects to make decisions in complete social isolation, and found that 50–50 offers declined to about 10%, while offers of nothing increased to around 60%. This kind of behaviour can be interpreted in terms of concerns about reputation. Subjects who choose to give up half the money in dictator games may perceive that they are registering reputational gains. When social scrutiny is completely removed, most (but not all) subjects choose to maximize financial gains. Gains in reputation may be beneficial if people with reputations for honesty are chosen as preferred partners for ventures that require mutual cooperation (Frank 1987, 1988). Human economic decision-making is rarely based on simple payoff maximization. Humans bargain. They anticipate and try to influence each other’s behaviour. Strategies based on reputation building, trust, scorekeeping and punishment can flourish under conditions where the short-term costs of cooperation are outweighed by its long-term benefits. The extent to which such strategies represent uniquely human adaptive specializations remains to be determined. If PD-like payoff matrices rarely occur in nature, strategies based on reciprocity may also be rare. For example, unconditional strategies such as ‘cooperate all the time’ may arise if exploiting others imposes a net cost on the actor (Pusey & Packer 1997) or, conversely, if resisting exploitation imposes a net cost (Winterhalder 1996). Even in cases involving meticulously documented reciprocal exchange in nonhuman animals (Wilkinson 1984; de Waal 1989), it is not entirely clear whether the underlying payoff structures conform to PD. Indeed it is particularly difficult to establish experimental conditions under which animals are given the opportunity to play PD. It is clear, however, that humans can be induced to cooperate at unexpectedly high frequencies under a wide variety of experimental conditions. We are grateful to Craig Packer for constructive criticism of an earlier version of this manuscript. W.M.B. was supported by a postgraduate scholarship from NSERC (Canada).
References Axelrod, R. & Hamilton, W. 1981. The evolution of cooperation. Science, 211, 1390–1396.
Clements, K. C. & Stephens, D. W. 1995. Testing models of non-kin cooperation: mutualism and the Prisoner’s Dilemma. Animal Behaviour, 50, 527–535. Dugatkin, L. A. 1988. Do guppies play Tit for Tat during predator inspection visits? Behavioral Ecology and Sociobiology, 23, 395– 399. Dugatkin, L. A. 1997. Cooperation Among Animals: An Evolutionary Perspective. Oxford: Oxford University Press. Forsythe, R., Horowitz, J., Savin, N. & Sefton, M. 1994. Replicability, fairness and pay in experiments with simple bargaining games. Games and Economic Behavior, 6, 347–369. Frank, R. H. 1987. If Homo Economicus could choose his own utility function, would he want one with a conscience? American Economic Review, 77, 593–604. Frank, R. H. 1988. Passions Within Reason: The Strategic Role of the Emotions. New York: W.W. Norton. Frank, R. H., Gilovich, T. & Regan, D. T. 1993. The evolution of one-shot cooperation: an experiment. Ethology and Sociobiology, 14, 247–256. Guth, W., Schmittberger, R. & Schwarze, B. 1982. An experimental analysis of ultimatum bargaining. Journal of Economic Behavior and Organization, 3, 367–388. Hoffman, E., McCabe, K., Shachat, K. & Smith, V. 1994. Preferences, property rights, and anonymity in bargaining games. Games and Economic Behavior, 7, 346–380. Huntingford, F. A., Lazarus, J., Barrie, B. D. & Webb, S. A. 1994. A dynamic analysis of cooperative predator inspection in sticklebacks. Animal Behaviour, 47, 413–423. Lombardo, M. 1985. Mutual restraint in tree swallows: a test of the Tit for Tat model of reciprocity. Science, 227, 1363–1365. McCabe, K. A., Rassenti, S. J. & Smith, V. L. 1996. Game theory and reciprocity in some extensive form experimental games. Proceedings of the National Academy of Sciences U.S.A., 93, 13 421– 13 428. Manstead, A. S. R. & Hewstone, M. (Eds) 1995. The Blackwell Encyclopedia of Social Psychology. Oxford: Blackwell. Milinski, M. 1987. Tit for Tat and the evolution of cooperation in sticklebacks. Nature, 325, 433–435. Milinski, M., Ku ¨ lling, D. & Kettler, R. 1990. Tit for Tat: sticklebacks (Gasterosteus aculeatus) ‘trusting’ a cooperating partner. Behavioral Ecology, 1, 7–11. Nowak, M. & Sigmund, K. 1993. A strategy of win-stay, lose-shift that outperforms tit-for-tat in the Prisoner’s Dilemma game. Nature, 364, 56–58. Pruitt, D. G. & Kimmel, M. J. 1977. Twenty years of gaming: critique, synthesis, and suggestions for the future. Annual Review of Psychology, 28, 363–392. Pusey, A. E. & Packer, C. 1997. The ecology of relationships. In: Behavioural Ecology: An Evolutionary Approach (Ed. by J. R. Krebs & N. B. Davies), pp. 254–283. Oxford: Blackwell. Rapoport, A. & Chammah, A. M. 1965. Prisoner’s Dilemma: A Study in Conflict and Cooperation. Ann Arbor: University of Michigan Press. Roberts, G. 1997. Testing mutualism: a commentary on Clements & Stephens. Animal Behaviour, 53, 1361–1362. Stephens, D. W., Anderson, J. P. & Benson, K. E. 1997. On the spurious occurrence of Tit for Tat in pairs of predator-approaching fish. Animal Behaviour, 53, 113–131. Trivers, R. L. 1971. The evolution of reciprocal altruism. Quarterly Review of Biology, 46, 35–57. de Waal, F. 1989. Food sharing and reciprocal obligations among chimpanzees. Journal of Human Evolution, 18, 433–459. Wilkinson, G. S. 1984. Reciprocal food sharing in the vampire bat. Nature, 308, 181–184. Winterhalder, B. 1996. A marginal model of tolerated theft. Ethology and Sociobiology, 17, 37–53.
F3
FORUM Human cooperation is more than by-product mutualism BORIS PALAMETA* & WILLIAM MICHAEL BROWN†
*Department of Psychology, St Thomas University †Department of Psychology, Dalhousie University (Received 24 August 1998; initial acceptance 14 September 1998; final acceptance 7 October 1998; MS. number: AF-2)
known to conform to a Prisoner’s Dilemma’ (page 533). Although we agree wholeheartedly that evidence of reciprocity requires experimental rigour, the experiments have been done and the evidence is there, at least for humans. There is an extensive literature in social psychology and experimental economics on human behaviour in mixedmotive games like PD. Many studies have found that stable cooperation is a common outcome of two-party and multiparty games (Pruitt & Kimmel 1977), especially when most players use TFT (Manstead & Hewstone 1995). Rapaport & Chammah (1965) are cited by Clements & Stephens (1995) and Stephens et al. (1997) as providing evidence that most PD games stabilize into mutual defection. In fact Rapaport & Chammah (1965) show that although cooperation initially declines in the first 30 trials, it ascends over the final 30 trials so that even players who appeared to be locked into mutual defection often switch to mutual cooperation. Results like this probably led Anatol Rapaport to conceptualize the strategy TFT, the eventual winner of Axelrod & Hamilton’s (1981) celebrated tournament. In this commentary, we shall briefly review some recent studies that consider mixed-motive games from an explicitly evolutionary perspective. The standard PD game has two possible strategies (cooperate or defect) and four possible outcomes, two symmetric (i.e. players play the same strategy and receive equal payoffs) and two asymmetric (i.e. players play different strategies and receive unequal payoffs). The PD is structured so that defection always yields higher payoffs than cooperation regardless of what the other player does. Therefore, if both players use the principle of dominance, that is, choose the strategy that yields the higher potential payoff, mutual defection is the endpoint. If players use reciprocity though, only symmetric outcomes are possible, and mutual cooperation is the endpoint because it is more profitable than mutual defection. Mixed motives arise from the fact that dominance
he study of cooperation between unrelated individuals received a major boost from Axelrod & Hamilton’s (1981) classic formulation of an evolutionarily stable solution (Tit for Tat, TFT) to the Prisoner’s Dilemma (PD). More recent models have shown that TFT is not always evolutionarily stable (Nowak & Sigmund 1993): indeed one can always find a mixed strategy capable of invading any pure strategy, including TFT (Dugatkin 1997). However, TFT has remained the most frequently investigated strategy in experimental studies of cooperation. Observations consistent with TFT have been reported for various animal species (Lombardo 1985; Milinski 1987; Dugatkin 1988; Milinski et al. 1990; Huntingford et al. 1994). Recently, Clements & Stephens (1995) disputed such claims on the grounds that (1) the payoff structures of the behaviours in question did not necessarily conform to PD, and (2) alternative explanations were not considered. Clements & Stephens (1995) recorded the behaviour of blue jays playing carefully controlled PD and mutualism games, and found stable cooperation only in the latter situation. Because the payoff structure of mutualism games ensures that cooperation is the best choice regardless of what your partner/ opponent does, the occurrence of cooperation in such games is simply a by-product of individual animals maximizing their own immediate rewards without reacting to each other’s behaviour at all (Roberts 1997). Thus there is no need to invoke reciprocity or TFT as a necessary route to cooperation. Clements & Stephens (1995) and Stephens et al. (1997) suggest that by-product mutualism is the most parsimonious explanation for the occurrence of cooperation in a number of species, including humans. Clements & Stephens (1995) go so far as to state that ‘there is no empirical evidence of nonkin cooperation in a situation, natural or contrived, where the payoffs are
T
Correspondence: B. Palameta, Department of Psychology, St Thomas University, Fredericton, New Brunswick, E3B 5G3, Canada (email: [email protected]). 0003–3472/99/020F01+03 $30.00/0
F1
1999 The Association for the Study of Animal Behaviour
F2
ANIMAL BEHAVIOUR, 57, 2
competes with reciprocity; dominance is the only way to achieve the highest possible payoff (defecting while the other cooperates), while reciprocity yields the most profitable symmetric payoff (mutual cooperation). The PD allows each player only one choice (cooperate or defect) per interaction. Recent research has focused on games that offer players several choices per interaction, while maintaining a payoff structure that is similar to the original PD. For instance, McCabe et al. (1996) investigated two-person, extensive-form bargaining games, in which players alternate moves along one of two branches of a game tree. At the pivotal point of the game, player 1 must choose between the branch that offers the highest individual payoff (dominance) and the branch that offers the highest symmetric payoff (reciprocity). Choosing dominance will most often lead player 2 to do likewise, thus terminating the game at a symmetric payoff lower than the one that could have been attained through reciprocity. Choosing reciprocity is risky though because player 2 then has the choice of cooperating by terminating the game at the point of highest symmetric payoff, or defecting by moving towards a higher asymmetric payoff. In some versions of the game, player 1 is given the option of responding to player 2’s defection by either accepting it (i.e. accepting the low portion of an asymmetric payoff) or punishing player 2 at some cost to itself (i.e. reducing both players’ payoffs to the lowest possible point). In this context, punishment is a form of escalated defection that costs both parties, but costs the initial defector more. Like PD, bargaining games can be played on a one-shot basis or with repeated interactions; repeated interactions can be with the same player or with different players each time. Unlike PD, bargaining games may add the option of punishing defection. Regardless of which version is played, game theory predicts a noncooperative outcome. Yet, as we describe below, cooperation occurs far more frequently than expected. Furthermore, the frequency of cooperation varies in accordance with reciprocity theory as the parameters of the game are changed. First, consider the effect of punishment. It is not difficult to see how the option to punish can enhance the effectiveness of TFT. The option of punishing a player who fails to reciprocate cooperation acts as a buffer against exploitation. Reciprocity theory predicts that cooperation will be initiated more frequently and develop more readily when defection can be punished. McCabe et al.’s (1996) results support the prediction. For instance, when subjects played repeatedly, each time with a different counterpart, the percentage of defections that were punished rose from 42% in the first five trials to 81% in the last five trials. This was accompanied by a concurrent increase in mutually cooperative outcomes, from 40 to 57%. When defections could not be punished, mutual cooperation was less frequent (less than 25% of outcomes). Another prediction from reciprocity theory is that mutual cooperation is more likely to arise when the same players encounter each other repeatedly (Trivers 1971). Again, data from bargaining games support the prediction. When players were paired with one counterpart for repeated play, the percentage of mutually cooperative
outcomes rose to 77% in the last five trials (McCabe et al. 1996). In the situation most similar to Clements & Stephens’ (1995) paradigm (i.e. repeated play with the same partner, no opportunity to punish defection), mutual cooperation was initially difficult to achieve, occurring in only 29% of the first five trials. However, mutual cooperation rose to 68% in the last five trials. This implies that cooperation is not merely contingent upon the threat of strategic retaliation, but that it can also be fostered by the sense of reciprocity that develops when two players encounter each other repeatedly. There is evidence that some individuals try to cooperate even when told they will be playing a randomly selected counterpart only once. The game theoretic prediction is that 100% of such one-shot games should end in mutual defection. Contrary to this prediction, in McCabe et al.’s (1996) study, almost 50% of the players attempted cooperation. Furthermore, attempts at cooperation were defected on relatively few trials (50% of attempts when no punishment was possible, and only 23% of attempts when punishment was possible). To achieve mutual cooperation in games with PD-like payoff structures, both players in turn must refrain from seeking to maximize individual payoffs. Cooperative plays are more frequently reciprocated when backed by the threat of retaliation, but reciprocity also occurs in one-shot games where cooperation is backed by nothing more than trust. The potential benefits of trust are considerable, provided that a trusting player is paired with a trustworthy counterpart. Frank (1987, 1988) developed the theory that cooperation based on trust could evolve as long as cooperators could accurately predict the trustworthiness of their fellow players before deciding whether or not to play. Frank et al. (1993) tested the theory by having people who had never met interact for 30 min before playing a traditional (i.e. no option to punish) one-shot PD. Despite the fact that players knew that neither they nor the experimenters would be permitted to discover any individual subject’s choice, a cooperation rate of over 70% was obtained. Furthermore, players were able to predict their partner’s choice with significantly higher than chance accuracy. Other researchers have tested the limits of the human capacity for cooperation by creating games in which cooperation is more difficult to achieve than it is in PD. These games are structured to allow one player more decision-making power than the other. For example, ultimatum games allow one player to make a unilateral decision on how a resource (e.g. $10) is to be divided. The second player may either accept the offer or reject it, in which case neither player gets any money. The game theoretic prediction is that the first player should offer the smallest amount possible ($1), and the second player should accept it because $1 is better than nothing. In fact, results of ultimatum games usually deviate from this pattern. Offers to split the resources evenly are often the modal result, and offers of uneven splits are sometimes rejected (Guth et al. 1982; Forsythe et al. 1994). Rejection in this context is similar to the strategic punishment discussed earlier; both players lose, but the player who made the initial offer loses more.
FORUM
Dictator games are like ultimatum games with the threat of punishment removed, that is, offers cannot be refused. Game theory predicts that the player making the offer (the dictator) should maximize payoffs by keeping all the money and offering the player nothing. As expected, offers are generally lower in dictator games than in ultimatum games (Forsythe et al. 1994). Nevertheless, the proportion of dictators that offer a 50–50 split (about 20%) is equal to the proportion that offer nothing (Forsythe et al. 1994). Hoffman et al. (1994) attribute these results to the fact that dictator experiments rarely offer subjects complete anonymity. Even when betweensubjects anonymity is assured, experimenters are still aware of subjects’ behaviour. Hoffman et al. (1994) allowed subjects to make decisions in complete social isolation, and found that 50–50 offers declined to about 10%, while offers of nothing increased to around 60%. This kind of behaviour can be interpreted in terms of concerns about reputation. Subjects who choose to give up half the money in dictator games may perceive that they are registering reputational gains. When social scrutiny is completely removed, most (but not all) subjects choose to maximize financial gains. Gains in reputation may be beneficial if people with reputations for honesty are chosen as preferred partners for ventures that require mutual cooperation (Frank 1987, 1988). Human economic decision-making is rarely based on simple payoff maximization. Humans bargain. They anticipate and try to influence each other’s behaviour. Strategies based on reputation building, trust, scorekeeping and punishment can flourish under conditions where the short-term costs of cooperation are outweighed by its long-term benefits. The extent to which such strategies represent uniquely human adaptive specializations remains to be determined. If PD-like payoff matrices rarely occur in nature, strategies based on reciprocity may also be rare. For example, unconditional strategies such as ‘cooperate all the time’ may arise if exploiting others imposes a net cost on the actor (Pusey & Packer 1997) or, conversely, if resisting exploitation imposes a net cost (Winterhalder 1996). Even in cases involving meticulously documented reciprocal exchange in nonhuman animals (Wilkinson 1984; de Waal 1989), it is not entirely clear whether the underlying payoff structures conform to PD. Indeed it is particularly difficult to establish experimental conditions under which animals are given the opportunity to play PD. It is clear, however, that humans can be induced to cooperate at unexpectedly high frequencies under a wide variety of experimental conditions. We are grateful to Craig Packer for constructive criticism of an earlier version of this manuscript. W.M.B. was supported by a postgraduate scholarship from NSERC (Canada).
References Axelrod, R. & Hamilton, W. 1981. The evolution of cooperation. Science, 211, 1390–1396.
Clements, K. C. & Stephens, D. W. 1995. Testing models of non-kin cooperation: mutualism and the Prisoner’s Dilemma. Animal Behaviour, 50, 527–535. Dugatkin, L. A. 1988. Do guppies play Tit for Tat during predator inspection visits? Behavioral Ecology and Sociobiology, 23, 395– 399. Dugatkin, L. A. 1997. Cooperation Among Animals: An Evolutionary Perspective. Oxford: Oxford University Press. Forsythe, R., Horowitz, J., Savin, N. & Sefton, M. 1994. Replicability, fairness and pay in experiments with simple bargaining games. Games and Economic Behavior, 6, 347–369. Frank, R. H. 1987. If Homo Economicus could choose his own utility function, would he want one with a conscience? American Economic Review, 77, 593–604. Frank, R. H. 1988. Passions Within Reason: The Strategic Role of the Emotions. New York: W.W. Norton. Frank, R. H., Gilovich, T. & Regan, D. T. 1993. The evolution of one-shot cooperation: an experiment. Ethology and Sociobiology, 14, 247–256. Guth, W., Schmittberger, R. & Schwarze, B. 1982. An experimental analysis of ultimatum bargaining. Journal of Economic Behavior and Organization, 3, 367–388. Hoffman, E., McCabe, K., Shachat, K. & Smith, V. 1994. Preferences, property rights, and anonymity in bargaining games. Games and Economic Behavior, 7, 346–380. Huntingford, F. A., Lazarus, J., Barrie, B. D. & Webb, S. A. 1994. A dynamic analysis of cooperative predator inspection in sticklebacks. Animal Behaviour, 47, 413–423. Lombardo, M. 1985. Mutual restraint in tree swallows: a test of the Tit for Tat model of reciprocity. Science, 227, 1363–1365. McCabe, K. A., Rassenti, S. J. & Smith, V. L. 1996. Game theory and reciprocity in some extensive form experimental games. Proceedings of the National Academy of Sciences U.S.A., 93, 13 421– 13 428. Manstead, A. S. R. & Hewstone, M. (Eds) 1995. The Blackwell Encyclopedia of Social Psychology. Oxford: Blackwell. Milinski, M. 1987. Tit for Tat and the evolution of cooperation in sticklebacks. Nature, 325, 433–435. Milinski, M., Ku ¨ lling, D. & Kettler, R. 1990. Tit for Tat: sticklebacks (Gasterosteus aculeatus) ‘trusting’ a cooperating partner. Behavioral Ecology, 1, 7–11. Nowak, M. & Sigmund, K. 1993. A strategy of win-stay, lose-shift that outperforms tit-for-tat in the Prisoner’s Dilemma game. Nature, 364, 56–58. Pruitt, D. G. & Kimmel, M. J. 1977. Twenty years of gaming: critique, synthesis, and suggestions for the future. Annual Review of Psychology, 28, 363–392. Pusey, A. E. & Packer, C. 1997. The ecology of relationships. In: Behavioural Ecology: An Evolutionary Approach (Ed. by J. R. Krebs & N. B. Davies), pp. 254–283. Oxford: Blackwell. Rapoport, A. & Chammah, A. M. 1965. Prisoner’s Dilemma: A Study in Conflict and Cooperation. Ann Arbor: University of Michigan Press. Roberts, G. 1997. Testing mutualism: a commentary on Clements & Stephens. Animal Behaviour, 53, 1361–1362. Stephens, D. W., Anderson, J. P. & Benson, K. E. 1997. On the spurious occurrence of Tit for Tat in pairs of predator-approaching fish. Animal Behaviour, 53, 113–131. Trivers, R. L. 1971. The evolution of reciprocal altruism. Quarterly Review of Biology, 46, 35–57. de Waal, F. 1989. Food sharing and reciprocal obligations among chimpanzees. Journal of Human Evolution, 18, 433–459. Wilkinson, G. S. 1984. Reciprocal food sharing in the vampire bat. Nature, 308, 181–184. Winterhalder, B. 1996. A marginal model of tolerated theft. Ethology and Sociobiology, 17, 37–53.
F3