Reciprocity in Ranked Relationships - Springer Link

Journal of Bioeconomics 3: 149±170, 2001 # 2002 Kluwer Academic Publishers. Printed in the Netherlands.

Reciprocity in Ranked Relationships: Does Social Structure In¯uence Social Reasoning? LAURENCE FIDDICK ESRC Centre for Economic Learning and Social Evolution, University College London, UK (l.®[email protected]) DENISE DELLAROSA CUMMINS Department of Philosophy, University of California, Davis, USA ([email protected]) Synopsis: Many economic and evolutionary theories have modeled cooperation as the evolutionary outcome of decisions made by autonomous, self-interested agents operating in a social vacuum. In this paper we consider the implications for cooperative interactions when prior social structures and corresponding social norms exist. In particular we investigate the in¯uence of social rank/status on perceptions of fairness and tolerance of cheating. We review evidence from a series of experiments employing the Wason selection task (a test of conditional reasoning) and the ledger task (a decision making task) suggesting that people cued to adopt a perspective of high social rank are more tolerant of cheating and simultaneously believe that they have been more fairly treated (even when cheated) than people cued to adopt a perspective of low social rank. However, the evidence also suggests interesting cross-cultural differences in perceptions of fairness and tolerance of cheating in ranked relationships. Key words: cooperation, cheater detection, cross-cultural differences, dominance theory, evolutionary psychology, hierarchy, norms, relative deprivation, social contract theory, status, Wason selection task JEL classi®cation:

A13, D63, D64, Z13, C91

Introduction Individual self-interest is a fundamental assumption of both economics and evolutionary studies and, yet, there is ample evidence of cooperative behavior that would seem to challenge this deeply held assumption (Camerer & Thaler 1995, Dawes & Thaler 1988, Thaler 1988). A common explanation for such ®ndings in both economics and evolutionary studies is reciprocity, rewarding kindness with kindness and nastiness with nastiness, whether reciprocity be the product of evolution (e.g., Axelrod 1984, Axelrod & Hamilton 1981, Ben±Ner & Putterman 2000, Berg et al. 1995, Cosmides 1989, Cosmides & Tooby 1989, 1992, 1994, Gintis 2000, GuÈth 1995, GuÈth & Yaari 1992, Hoffman et al. 1997, 1998, McCabe et al. 1996, Trivers 1971), nonstandard utility functions incorporating preferences for fairness (Bolton & Ockenfels 2000, Fehr & GaÈchter 1998, 2000, Fehr et al. 1997, Rabin 1993) or simply the social learning of cultural norms (Camerer & Thaler 1995). In general, however, these various accounts of reciprocity have tended to analyze reciprocity as though it existed in a social vacuum

150

FIDDICK AND CUMMINS

in which two anonymous agents with equal exogenous status and no prior social history reach a fair outcome. In this paper, however, we review some of the studies that we have conducted investigating people's reasoning about reciprocal exchanges when the people involved differ in their social status and the implications these have for some prominent research programs in evolutionary psychology and economics. Experimental economics games Among the most striking demonstrations of costly self-sacri®ce are results from experimental economics games where all that would appear to be at stake is ®nancial gain, pure and simple, and the interactions are between anonymous strangers. In the dictator game, for example, two people are assigned a provisional $10. One person, the dictator, then decides how the money is to be split between the two. While some dictators will give themselves the whole $10, as standard self-interested economic analyses would suggest, a signi®cant number of dictators and in many cases the majority will give the other person a nontrivial amount of the money (e.g., Forsythe et al. 1994, Hoffman et al. 1994, 1996, Johannesson & Persson 2000). Moreover, in the ultimatum game, where the second player, the responder, has the opportunity to either accept the proposed split or turn it down causing both players to walk away with nothing, signi®cantly more proposers (equivalent to dictators in the dictator game) offer the responder a nontrivial amount of the money with the modal offer usually being a 50:50 split. Yet, according to standard game-theoretic analyses, the addition of this second phase of play should make little difference to the proposed divisions. By backwards induction, responders should favor a single penny over nothing (the payoff when the responder declines) and the proposers knowing this should offer the responders a penny and keep $9.99 for themselves. Psychological adaptations for reciprocity as an explanation for cooperation Although game-theoretic analyses of repeated games have long since shown that cooperation in the form of reciprocity may be the equilibrium outcome of rationally self-interested play, the perceived problem in applying these analyses to the experimental results is that (1) cooperative outcomes in these experiments typically do not depend upon repeated play, and (2) according to the so-called folk theorem, reciprocity is simply one among many equilibrium outcomes. More recently, however, several economists have attempted to circumvent these dif®culties by adopting an evolutionary approach (e.g., Ben±Ner & Putterman 2000, Berg et al. 1995, Gintis 2000, GuÈth 1995, GuÈth & Yaari 1992, Hoffman et al. 1997, 1998, McCabe et al. 1996). One way around the multiple equilibria problem is to cast payoffs in terms of biological ®tness and analyze the game in terms of its evolutionary dynamics (Maynard Smith 1982). While there may be many equilibrium points under standard analyses, some are more ®tness enhancing and evolutionarily stable than others

RECIPROCITY IN RANKED RELATIONSHIPS

151

(Axelrod 1984, Axelrod & Hamilton 1981, Gintis 2000, GuÈth 1995, GuÈth & Yaari 1992, Trivers 1971). In his seminal essay on reciprocal altruism, Trivers (1971) hypothesized that due to ancestral humans' long lifespans, low dispersal rates and mutual dependence, there would have been many repeated opportunities for mutually cooperative interactions, and hence the evolution of reciprocity. Based upon Trivers' and subsequent evolutionary analyses of reciprocity, Cosmides & Tooby (1989, 1992, Cosmides 1989) proposed an adaptationist account of social exchangeÐthe Computational Theory of Social Exchange (also known as Social Contract Theory)Ðin which they hypothesize that humans have evolved psychological adaptations for engaging in mutually bene®cial, reciprocal exchanges. If humans possess evolved predispositions for engaging in social exchange, these tendencies are likely to be invoked in modern, experimental economics games where there is the potential for cooperative interaction. Although experimental settings often are contrived to be one-shot games, the psychological mechanisms that are invoked are hypothesized to have evolved in repeated game situations.1 The theory, therefore, has the potential to explain widespread cooperative behavior in experimental economics games and, hence, has found favor with some experimental economists (e.g. Ben-Ner & Putterman 2000, Berg et al. 1995, Hoffman et al. 1997, 1998, McCabe et al. 1996). Cheater detection and the Wason selection task One of the key predictions of Social Contract Theory (SCT) theory is that humans possess a `look for cheaters' algorithm that is activated whenever one encounters a cooperative interaction in which there is the potential for cheating, the rationale being that cooperation couldn't evolve unless one were able to detect and punish cheaters (Axelrod 1984, Axelrod & Hamilton 1981, Triver 1971). Indeed, studies by Fehr and his colleagues suggest that the ability to detect and punish cheaters has a large in¯uence in producing cooperative outcomes that deviate from standard gametheoretic predictions (Fehr & GaÈchter 2000, Fehr et al. 1997, see also GuÈth & van Damme 1998). Psychological evidence cited in support of SCT comes from a series of experiments employing the Wason selection task (Cosmides 1989, Cosmides & Tooby 1992, Fiddick et al. 2000, Gigerenzer & Hug 1992, Platt & Griggs 1993). In the Wason selection task (Wason 1968), people are given a rule of the form: If P then Q. Accompanying the rule is four cards to which the rule applies. The cards have information on both sides, but only one side of each is showing. On one side of the card it says whether or not P is true (P vs. not-P) and on the other side of the card it says whether or not Q is true (Q vs. not-Q). The visible sides of the cards represent the four possible contingencies: P, not-P, Q, and not-Q. The person's task is to indicate which of the four cards need to be turned over in order to determine whether the rule has been violated. The solution is straightforward. A rule such as If it is a dog then it is an animal can be falsi®ed only when an instance is discovered that is a dog but is NOT an animal. More generally,

152

FIDDICK AND CUMMINS

a rule of the form: If P then Q, is violated by instances of P & not-Q. So in order to solve the task, one must look for instances of P & not-Q and since only the P and not-Q cards can conceivably be instances of P & not-Q, these are the cards that need to be selected. Surprisingly, despite the apparent simplicity of the task, typically fewer than 10% of people solve standard versions of the task correctly (Evans et al. 1993). However, a different pattern of results emerges when the task requires testing compliance with a social rule as opposed to testing the truth of the rule (Cheng & Holyoak 1985, Cheng et al. 1986, Griggs & Cox 1982, Johnson-Laird et al. 1972). When the task requires reasoners to ®nd instances of cheating (non-compliance), they can do so on average 70% of the time (see Cummins 1996a for a review of this literature). Cosmides and Tooby explained the striking difference in performance patterns on the two versions of the card selection task through appeal to evolved algorithms for reciprocity (Cosmides 1989, Cosmides & Tooby 1989, 1992). They argued that an offer to engage in social exchange can be expressed by a rule of the form: If Bene®t [to Party X, from Party Y] then Bene®t [to Party Y, from Party X]. People encountering such a rule should perceive that individuals who have accepted a bene®t without returning a bene®t are cheating. From Party X's perspective, the rule has the cost/bene®t structure of a social contract: If Bene®t Accepted then Cost Paid, where the four cards would now represent the contingencies: Bene®t Accepted (P), Bene®t not-Accepted (not-P), Cost Paid (Q), and Cost not-Paid (not-Q). According to Cosmides (1989), all of the social rules employed in selection tasks that had previously been found to elicit high levels correct performance were social contracts of this form. Social Contract Theory predicts that people looking to see if Party X cheated will select the Bene®t Accepted (P) and Cost not-Paid (not-Q) cards. This is exactly what Cosmides (1989) observed in her test of SCT, and has subsequently been replicated in several studies (Cosmides & Tooby 1992, Fiddick et al. 2000, Gigerenzer & Hug 1992, Platt & Griggs 1993). Alternative accounts of cheater detection Although SCT has been favorably received by some economists, the reaction within the psychological, reasoning literature has been less enthusiastic. The most frequent objection is that the scope of the theoryÐreciprocal exchangeÐis unnecessarily restrictive (e.g., Cheng & Holyoak 1989, Girotto et al. 1989, Manktelow & Over 1990, Politzer & Nguyen-Xuan 1992) owing to the fact that people are also good at detecting violations of nonsocial contract versions of the Wason selection task (e.g., Cheng & Holyoak 1989, Cummins 1996c, 1999, Fiddick 1998, Fiddick et al. 2000, Girotto et al. 1989, Manktelow & Over 1990). A common feature of the rules eliciting correct performance on the selection task, including social contracts, is that they tend to be deontic rules, rules regulating what one is permitted, obligated, or forbidden to do. This has led many reasoning researchers to explain performance on deontic versions of the selection task, including those employing social contracts, either in terms of a more general competence for deontic reasoning (Cheng & Holyoak 1985, 1989, Cummins 1996b, c, 1999,


153

Holyoak & Cheng 1995, Girotto et al. 1989, Politzer & Nguyen-Xuan 1992) or in terms of utilitarian decision-making (Kirby 1994, Manktelow & Over 1990, 1991, 1995, Oaksford & Chater 1994). This re¯ects a similar divide within economics to explain reciprocity more broadly, either in terms of social norms (e.g. Camerer & Thaler 1995) or in terms of nonstandard utility functions (e.g. Bolton & Ockenfels 2000, Fehr & Schmidt 1999, Rabin 1993), respectively. The deontic point of view has been criticized for invoking without explaining the deontic concepts of obligation and entitlement that appear to be pivotal in people's reasoning (Manktelow & Over 1991). Decision-theoretic accounts attempt to cash out these intuitions in terms of subjective expected utility (Kirby 1994, Manktelow & Over 1990, 1991, 1995, Oaksford & Chater 1994), but fail to explain why people value the social arrangements regulated by these rules. Cummins (1996a, b, 1998, 1999), however, has argued that our deontic intuitions can be accounted for by our evolutionary heritage as a social species. She has proposed a descriptive evolutionary theory of deontic reasoning called Dominance Theory. The key concepts in the theory are social regulation (or social norm), and social status. We will brie¯y summarize the theory here. Dominance Theory: social norms as an adaptive problem Despite the clear bene®ts that derive from sociality, living in social groups also imposes costs in terms constraints on the behavior of individuals. These constraints are termed social norms. Social norms appear not just in the societies of humans but in the societies of non-human animals as well, where they constrain virtually every activity, including who is allowed to sit next to, play with, share food with, groom, and mate with whom (Aruguete 1994, Hall 1964). In order to avoid agonistic encounters and ostracism, members of a social group must learn which behaviors are permitted, prohibited, and obligated under which conditions. Flouting these norms carries great risk; in fact, perceived violations of the `social code' has been designated as the single most common cause of aggression in primate groups (Hall 1964). From Cummins' perspective, the `deontic effect' in human cognition re¯ects the operation of primitive cognitive functions involved in monitoring compliance with social norms. The deontic concepts of permission, obligation, and prohibition that underlie social norms are hypothesized to be primitives in the cognitive architecture of humans (and other intelligent highly-social animals). Because they are primitives, the theory predicts that (a) young children should show a precocity for cognizing the deontic concepts of permission, obligation, and prohibition, as well as a precocity for detecting violations of the social rules that embody them, and (b) that an advantage for reasoning about these concepts should be evident in the reasoning of adults when compared to other types of reasoning of equivalent complexity. The ®rst prediction is supported by research showing that deontic terms emerge early in language development, having been observed in the speech of toddlers as young as 2 years of age (Dunn 1988). Moreover, unlike other reasoning skills, which show

154

FIDDICK AND CUMMINS

pronounced performance improvements during development, children as young as three years of age show the same `deontic effect' that is apparent in adult reasoners (Cummins 1996d, Harris & Nunez 1996) suggesting a biological preparedness to develop these types of cognitive skills. The second prediction is supported by the observation mentioned earlier that adults show superior performance on deontic versions of the selection task when compared to other versions of the task. Dominance Theory makes one further prediction, namely, that the status of group members is taken into consideration when monitoring compliance with social norms. The algorithms responsible for monitoring compliance with social norms are hypothesized to contain parameters pertaining to the relative status of the individual doing the monitoring and the individuals' monitored. An individual's status in¯uences expectations concerning what constitutes an acceptable level of compliance with social norms as well as expectations concerning enforcement of social norms. For example, in many animal societies, competition and cooperation among conspeci®cs produces a complex social structure called a dominance hierarchy. High-status individuals typically take on the role of protecting the implicit social code of the group, aggressing against those who violate it and breaking up disputes between lower-status individuals (see, e.g., Boehm 1992). Further, what counts as suf®cient reciprocation often appears to depend on one's status; high-status individuals don't need to reciprocate as frequently as low-status individuals in order to maintain alliances, presumably because of the greater bene®ts that derive from their interventions (Chapais 1992, Cheney 1983). There is ample evidence that status is a salient characteristic among humans as well. Status hierarchies emerge early in human development having been observed in the playgroups of children as young as two years of age (Frankel & Arbel 1980). By four years of age, children can reliably report the structure of these hierarchies indicating conscious awareness of status differences (Smith 1988). Status hierarchies are also apparent in our social and political institutions where they are variously referred to as monarchies, caste systems, and socio-economic class. There is also evidence supporting Dominance Theory's prediction that this salient characteristic modi®es the way we monitor compliance with social norms. Using a picture recognition task, Mealey et al. (1996) found that people are more likely to remember low-status cheaters than high-status cheaters or non-cheaters of either status. Cummins (1999) manipulated status perspective on the truth-testing and deontic versions of the Wason card selection task. The deontic version embedded nonreciprocal social rules with no explicit costs or bene®ts. The truth-testing version embedded the same rules as utterances overheard by a witness. The deontic version required testing compliance with the rule while the truth-testing version required discovering the truth status of the utterance. Status perspective was found to impact performance only on the deontic version of the task, and the nature of that effect (across two experiments) was that people evidenced higher levels of violation detection when they believed themselves to be of higher status than the individuals whose behavior they were monitoring (Cummins 1999).


155

From social norms to social exchange Cummins' (1999) demonstration that high-ranking individuals are more vigilant with respect to monitoring compliance with social norms than are low-ranking individuals tells us nothing, however, about how status impacts reasoning about reciprocal personal exchanges. Several other studies, however, have inadvertently manipulated social rank on social exchange versions of the Wason selection task, but the results of these studies are confusing and dif®cult to interpret. Invariably, these have been investigations of the in¯uence of social perspective on selection task performanceÐit has been found that participants cued to adopt the perspective of one party to a social exchange select one set of cards, P & not-Q, for example, whereas participants cued to adopt the perspective of the other party to the exchange select the complementary set of cards, not-P & Q. This `perspective effect' has generated considerable interest in the selection task literature and could conceivably throw some light on the issues raised here because, in the process of manipulating social perspective, studies investigating the effect have also tended to manipulate social rank. The ®rst such study was reported by Manktelow & Over (1991). In their investigation of social perspective (Experiments 1 & 2), Manktelow & Over also manipulated the social rank participants were cued to adopt: mother vs. child. In Experiment 1, they found a clear rank effect with participants cued into the perspective of the mother doing better, but they attributed this to the in¯uence of a third party observer present in the mother perspective but absent in the child perspective. When they corrected this in Experiment 2 by having a (low rank) third party observe that actions of the mother, namely a brother of the other child, there was no appreciable difference in levels. In a third experiment, Manktelow & Over (1991) employed a scenario in which a ®rm makes customers the offer: If you spend more than £100, then you may take a free gift. Participants cued to look for cheating on the part of the customers (i.e., spent less than £100 & took a free gift) performed substantially better, 85% correct, than participants cued to look for cheating on the part of the shop, 62% correct.2 Assuming that sellers have more power than buyers (cf. Hoffman et al. 1994), participants performed better when looking for cheating on the part of the lower ranked customers. One complication is that, although participants were cued to look for violations by different parties, in both cases this was from the perspective of a company manager called in to oversee the enforcement of the rule. Hence, performance may have been lower when looking to see if the shop is cheating customers because cheater-detection con¯icted with self-interest in this condition. This confound was removed in a subsequent study conducted by Politzer & NguyanXuan (1992) who independently employed practically the same scenario except that participants in their experiment were also cued to adopt different roles, a consumer's advocate vs. the store manager, when looking for violations on the part of the store and customers, respectively. In contrast to Manktelow & Over's ®ndings, Politzer & Nguyen-Xuan (p. 415) observed that Ìn both experiments [conducted by Politzer & Nguyen-Xuan], and in both scenarios, the facilitation effect was weaker in the manager role than in the consumer role . . . in both cases, the manager role yielded fewer correct

156

FIDDICK AND CUMMINS

responses than the consumer role.' So while Manktelow & Over found a pattern of rank effects that was in general accord with that found by Cummins (1999), Politzer & Nguyen-Xuan (1992) found exactly the opposite pattern using roughly the same materials. Other studies that have unwittingly manipulated status perspective have likewise yielded mixed bag of results. Gigerenzer & Hug (1992), for example, unintentionally employed a similar status manipulation in their perspective switching study. The rules they used had clear cost-bene®t structures, and participants were deliberately cued into the possibility of that cheating was taking place. Their results show no effect of status perspective for their day-off problem (75% correct for low status vs. 61% for high status) or their pension problem (64% correct for low status and 70% for high status), but, like Cummins (1999), signi®cantly greater cheater detection was observed in the high status perspective for their subsidy problem (59% correct for low status vs. 81% for high status). Holyoak & Cheng (1995) employed modi®ed versions of Gigerenzer & Hug's day-off and subsidy problems. They found no consistent difference in performance across the two perspectives. On the day-off problem, they reported 23% for low ranking vs. 56% for high ranking (signi®cant) and on the subsidy problem, 46% correct for low ranking vs. 41% for high ranking (not signi®cant). Pooling across all studies of perspective effects on the selection task, there is no consistent pattern of rank effects on social exchange versions of the Wason selection task one way or the other. Moreover, each study contains a potentially crucial confound: Some of the differences may be due to the fact that the perspective switches in these studies also required a switch in selection pattern. For example, in Gigerenzer & Hug's (1992) study the correct response alternated between P & not-Q and not-P & Q depending upon the way the rule was formulated and the perspective participants were cued to adopt. In every instance, participants performed better on the perspective for which P& not-Q was the correct response. However, Holyoak & Cheng (1995) found no similar or consistent pattern employing nearly identical scenarios. In order to properly assess the in¯uence of social rank on cheater-detection in reciprocal exchanges, we devised our own scenarios that 1) kept the predicted pattern of card selections constant (not-P & Q), and 2) cued participants into the perspective of the party that would have a self-interest in detecting the violation in question. Our scenarios described a carpooling arrangement between factory owner and his employee in which the offer made is: I'll drive you to work if you pay for the gas. In the high-ranking version, participants were cued into the perspective of a factory owner who drives and makes his employee the offer. In the low-ranking version, participants were cued into the perspective of the employee who drives and makes his boss the offer. Hence, in both conditions, cheating corresponded to the cards: drove to work (Q) and did not pay for gas (not-P) and the party that participants were cued to identify with had a self-interested motive to detect violations by the other party. The problems were run twice: in Bonn, Germany, and Davis, California, and, again, the results were mixed. In Bonn we observed no signi®cant difference in performance on the two perspectives (63% correct, low rank vs. 59% correct, high rank), whereas in Davis we observed a signi®cant rank effect with participants performing better in the low-rank condition (70% correct) than


157

in the high-rank condition (45% correct). Cross-cultural differences, as we shall see, may be a factor in explaining this disparate pattern of results, but at present all that can be said is that further investigation is required to sort out this confusing pattern of results. From cheater-detection to tolerance of cheating The pattern of selection task results is mixed and inconclusive, but the very method itself, the selection task may be the source of much of the confusion. Beside the fact that the task is notoriously dif®cult to solve, the abstract structure of the task itself can be construed in a myriad of different ways producing individual differences in performance (see Stenning & van Lambalgen 2001) that are orthogonal to the variables of interest here. Given our suspicions about the task, we devised a much simpler task, the ledger task, to assess people's tolerance of cheating in social exchanges when social rank varies (Cummins & Fiddick, under review). In the ledger task (see Figures 1 & 2), people are presented with a social exchange scenario with an embedded conditional rule describing the exchange, much as is the case with the Wason selection task. However, rather than presenting participants with an array of four cards and having them look for violations, they are presented with a ledger detailing a history of past interactions governed by the agreement. Indeed, the speci®c scenario that we employed was identical to carpool selection-task scenario that we described above except that now the offer was: I'll drive you to work if you pay me DM 20 weekly to cover the cost of gasoline (the tasks were originally devised to be administered in Germany). We de®ned cheating on this task the same way it was de®ned in Trivers (1971) and Axelrod & Hamilton (1981), namely, as failure to reciprocate. A ledger followed detailing the payments made for each week of a three-month period. Four different levels of payment were employed, 100% compliance with the agreement, 75%, 58% and 33%. The 100% compliance scenario constituted a complete absence of cheating, or, conversely, perfect compliance. The remaining ledgers depicted different levels of cheating (failure to reciprocate), or, conversely, different degrees of compliance. The participants' task was straightforward: For each level of compliance they had to indicate how likely it is that they would continue the carpool arrangement using a 5-point Likert scale ranging from `Very unlikely' (scored as 1) through Ùnlikely' (scored as 2), Ùncertain' (scored as 3), `Likely' (scored as 4), to `Very Likely' (scored as 5). Their Likert ratings are readily interpreted in terms of tolerance toward cheating: To the degree that participants are willing to continue the arrangement, they evidence a willingness to overlook their partner's failure to reciprocate. To the degree that they are unwilling to continue the arrangement, they evidence intolerance toward cheating. While some participants raised objections about aspects of the story (e.g., some questioned the fairness of the proposed deal), none expressed confusion about the task itself. This is in stark contrast to the selection task where participants often have questions about the nature of the task itself (e.g. what they are supposed to do). Hence, extensive personal experience conducting experiments using both the selection

158

FIDDICK AND CUMMINS

Boss Scenario Imagine that you own a factory in the developing world. You and one of your employees at the factory are both from the same rural village. Ordinarily both you and your employee would take the train to work you traveling first class and your employee traveling second class. Unfortunately, service to your village has recently been stopped while repairs are made to a rail bridge, so you are forced to drive your car to work. Gasoline is expensive; you estimate that it costs about DM 20 weekly to drive to and from work every week. You decide to see whether your employee would like to carpool, so you make him the following offer: I'll drive you to work if you pay me DM 20 weekly to cover the cost of gasoline. Your employee agreed. Below is ledger showing the payments made for three months.

Date: 7-14-99 7-21-99 7-28-99 8-04-99 8-11-99 8-18-99 8-25-99 9-02-99 9-09-99 9-16-99 9-23-99 9-30-99

Payment Received DM 20 DM 20 DM 20 DM 20 DM 20 DM 20 DM 20 DM 20 DM 20 DM 20 DM 20 DM 20

Review the ledger carefully, then mark the scale below to indicate the likelihood that you will continue this carpooling arrangement:

Very Unlikely

Unlikely

Uncertain

Likely

Very Likely

Figure 1. High-ranking (Boss) Version of the Ledger Task.

task and the ledger task suggests to us that the ledger task is far easier for people to understand. Indeed, we would be surprised if it was not easier to understand. Our goal was not to demonstrate a facilatory effect for this or that scenario, but to investigate varying levels of tolerance across different scenarios. But what levels of tolerance might one predict? In their current formulations, SCT and DT make no explicit predictions concerning the impact of rank on reasoning about reciprocal exchanges. Boyd's (1992) model, however, does address this issue, and

159


Employee Scenario Imagine that you are a worker in a factory in the developing world. You and your boss at the factory are both from the same rural village. Ordinarily both you and your boss would take the train to work your boss traveling first class and you traveling second class. Unfortunately, service to your village has recently been stopped while repairs are made to a rail bridge, so you are forced to drive your car to work. Gasoline is expensive; you estimate that it costs about DM 20 weekly to drive to and from work every week. You decide to see whether your boss would like to carpool, so you make him the following offer: I'll drive you to work if you pay me DM 20 weekly to cover the cost of gasoline. Your boss agreed. Below is ledger showing the payments made for three months.

Date: 7-14-99 7-21-99 7-28-99 8-04-99 8-11-99 8-18-99 8-25-99 9-02-99 9-09-99 9-16-99 9-23-99 9-30-99

Payment Received DM 20 DM 20 DM 20 DM 20 DM 20 DM 20 DM 20 DM 20 DM 20 DM 20 DM 20 DM 20

Review the ledger carefully, then mark the scale below to indicate the likelihood that you will continue this carpooling arrangement:

Very Unlikely

Unlikely

Uncertain

Likely

Very Likely

Figure 2. Low-ranking (Employee) Version of the Ledger Task.

analyzes status in terms of asymmetrical costs and bene®ts. Consider, for example, agonistic encounters. In these situations, for a given cost, greater bene®ts are conferred when a dominant aids a subordinate than vice versa. Why would the dominant aid the subordinate in such cases? Boyd shows that even if the bene®t received by the dominant does not compensate for the cost of giving aid, it may be an evolutionarily stable strategy for the dominant to cooperate if he or she cooperates only infrequently while the subordinate cooperates frequently. In situations where the dominant's costs exceed the bene®ts received, therefore, we would expect more cheating (failure to reciprocate)

160

FIDDICK AND CUMMINS

from the dominant than the subordinate. The situation reverses, however, when dominants provide a bene®t at lower cost than do subordinates. In these cases, Boyd's model predicts that it will be the subordinate that will reciprocate infrequently compared to the dominant. Applying this analysis to our ledger task, we can predict that tolerance for failures to reciprocate should vary as a function of the direction of perceived status-based imbalances between costs and bene®ts. The key question is whether perceived costs and bene®ts can be de®ned directly in terms of monetary commitments or whether status itself impacts these perceptions. Status and social exchange: studies conducted with the ledger task From the very ®rst studies conducted with the ledger task (Cummins & Fiddick 2001, under review), it was clear that status in¯uenced tolerance toward cheating. Our initial ®ndings revealed that participants cued into the perspective of the high-ranking individual (the boss/factory owner) were more tolerant of cheating (see Figure 3), an effect we have called the noblesse oblige effect. The noblesse oblige effect was indicated by a signi®cant rank x compliance level interaction. That the point of maximal difference should coincide with 33% compliance was unanticipated since this would depend upon the shape of the tolerance function, which we had not predicted in advance. For example, it would have been entirely consistent with our analysis if tolerance of cheating declined as a step function with most participants in both conditions indicating that they were `Very Likely' to continue the carpool at the 100% compliance level

Noblesse Oblige Effect

Likelihood of Continuing Carpool

5 4.5 4 3.5 Boss Employee

3 2.5 2 1.5 1 100% Figure 3.

75% 58% Level of Compliance

33%

Rank Effect on Tolerance of Cheating. (Error bars represent standard errors).


161

and `Very Unlikely' to continue the carpool at the 33% compliance level, with a sudden transition from tolerance to intolerance occurring at 75% compliance in one condition and 50% compliance in another. Although we had no ®rm theoretical reason to predict the speci®c slope of the tolerance functions, we were, nonetheless, surprised by the high levels of tolerance reported by participants. This was especially evident in the high-ranking condition, where even at 33% compliance, the median tolerance level indicated was Ùncertain' suggesting participants were indifferent to continuing the carpool even though they were being cheated more than half of the time. One explanation for these high levels of tolerance is that given one had to drive to work anyway, why not tolerate some cheating if it means that one will recoup some of the expenses as opposed to none. To rule out this possible explanation for the high tolerance levels, we had included transportation options as a factor in the design. Participants saw either a version which made no mention of alternative sources of transportation or a version in which it was stated that bus service was available. This factor made no difference. It had been suggested to us that in Japan, a different pattern of results might obtain, with employees feeling unfairly treated but suppressing their feelings and driving their bosses to work anyway.3 Concerned that our choice of question focusing on actions might be obscuring how our German participants felt about the situation, we decided to replicate the ®rst study on a third group of participants asking them both how likely it is that they would continue the carpool and how fairly they felt that they had been treated (Fiddick et al., in prep.). If our action-based question was concealing how participants really felt about the situationÐif they were suppressing their feelings and driving anyway (in either the boss or the employee condition), then one would expect some interaction between the answers to the two questions: likelihood of continuing the carpool and fairness of treatment. However, we found no interaction between these two questions. Yet again, however, there was a signi®cant noblesse oblige effect. Is the noblesse oblige effect the result of social rank differences? After producing a signi®cant noblesse oblige effect on two separate occasions employing two different participant populations, we had reason to believe that the effect was robust and reliable. However, we were concerned that the effect may not have been directly caused by our social rank manipulation, but may instead have been caused by some correlated factor, like implicit income differences. In order to assess whether social rank or implicit income differences are driving the noblesse oblige effect, we conducted a further study that independently manipulated these two variables. Again, the scenario described a situation in which an employee and his boss carpool to work, with either the employee or the boss driving. Additionally, it was explained that the employee was the top salesman at his boss's shop such that in the employee-makes-more conditions, he actually made more money on commissions than his boss's yearly income. If implicit income differences are the cause of the noblesse oblige effect, then the effect should disappear with participants cued into the party of

162

FIDDICK AND CUMMINS

No Income Effect (Collapsed accross rank)


5 4.5 4 3.5 Earn More

3

Earn Less

2.5 2 1.5 1 100%

75%

50%

25%

Level of Compliance Figure 4. There is No Income Effect on Tolerance of Cheating. The values plotted are the averages for the person making more vs. less, regardless of whether that person is the boss or employee (Error bars represent standard errors).

Rank Effect (Collapsed across income) 5


4.5 4 3.5 Boss

3

Employee

2.5 2 1.5 1 100%

75%

50%

25%

Level of Compliance Figure 5. The Noblesse Oblige Effect Persists When Earned Income is Controlled For. The values plotted are the averages for the boss vs. employee, regardless of whether the boss makes more or less than the employee. (Error bars represent standard errors).


163

the person making more money, be it the employee or his boss, displaying higher levels of tolerance. No effects of income were found (see Figure 4). The only signi®cant effects were a main effect of compliance level and a main effect of social rank (see Figure 5). These results strongly imply that social rank and not implicit income differences are responsible for the noblesse oblige effect. Indeed, the importance of the social relation as opposed to income differences is further highlighted by another study manipulating the social bond between the carpool partners. In this study, an employee and a boss carpool to work, but they are not each other's employee and boss. Instead, they work at unrelated factories and met through a classi®ed and in which the carpool arrangement was offered. In contrast to the other boss/employee scenarios, there was only a signi®cant effect of compliance level and no interaction or effect involving social rank. To summarize the ledger task results, then: (1) tolerance of cheating appears to decline gradually and approximately linearly with decreasing compliance with an agreed social exchange; (2) participants cued into the perspective of a high-ranking individual indicate more tolerance of cheating than participants cued into the perspective of a low-ranking individual, i.e. there is a noblesse oblige effect; (3) this noblesse oblige effect is found both for judgments of the likelihood of continuing the exchange relation and in perceptions of fairness; and (4) the noblesse oblige effect appears to be driven by the type of social relationÐa ranked relationÐand not income differences. Furthermore, the results of the ledger task studies appear to be more robust and reliable than studies conducted with the Wason selection task employing nearly identical scenarios. The implications for Social Contract Theory and Dominance Theory So where do these ledger task results leave SCT and DT? Currently, SCT de®nes reciprocity in terms of reciprocal exchange of bene®ts, without reference to the social status of the participants. This means that reasoners should be intolerant of cheating, regardless of parties' relative social standing. Our results are problematic for SCT in two respects. First, reasoners showed remarkable tolerance for cheating. Our subjects were willing to overlook signi®cant departures from contract compliance. We interpret this to mean that, while people are very good at detecting cheating (as evidenced by the Wason task literature), they also exhibit considerable tolerance towards it. Just how much cheating a person is willing to tolerate surely has important ®tness consequences, but SCT, as currently formulated, makes no speci®c predictions about the degree of cheating that will be tolerated before an arrangement is terminated. Its primary focus has been predicting which reasoning scenarios will evoke a cheater detection strategy and which will not. The second respect in which our results are problematic is the strong effect we observed of relative status on tolerance. Our data suggest that this factor is noted and processed when evaluating compliance with an informal contract, an effect that is neither predicted nor well-explained by SCT as it is currently formulated. To accommodate these results, some principled explanation

164

FIDDICK AND CUMMINS

must be offered whereby differences in relative standing in¯uence perceived bene®ts of the exchange. In contrast, DT predicts that relative status is a crucial characteristic that agents participating in an exchange readily note and process when evaluating compliance. With respect to cheater-detection, DT predicts greater cheater detection when reasoners adopt a high-status perspective because high-status individuals are typically assigned responsibility for preserving social norms/laws and monitoring compliance. This prediction is supported by the results of Cummins (1999) and Mealey et al. (1996). Tolerance toward cheating, however, is less straightforward. In its current formulation, DT simply predicts that perceived relative status will in¯uence tolerance toward cheating, but the direction of this in¯uence is not addressed, nor is the process by which status modi®es decision-making explicated. This is because DT has not been formalized in terms of a computational model, so precise predictions concerning the degree and direction of status-based effects cannot be made with any degree of rigor. As such, DT as currently formulated is more a framework than a precisely-speci®ed theory. Further, our results suggest that status is not readily translatable in terms of asymmetrical costs and bene®ts. Participants judged the employee to derive more of a bene®t from the carpooling arrangement than his boss, and there is no difference in the perceived costs paid by the employee and his boss (Fiddick et al., in prep.). Under these conditions, Boyd's model predicts more cheating on the part of the boss, and so we would expect greater tolerance for cheating on the part of the boss. But we observed the opposite: Employees were less tolerant of cheating on the part of bosses, and, conversely, bosses were more tolerant of cheating on the part of employees. Boyd's model predicts greater cheating on the part of the subordinate (and, by extension greater tolerance for subordinates' cheating) when the dominant provides the same degree of bene®t at lower cost than the subordinate does. This implies that we should have observed noblesse oblige only in our experiment in which subordinates made more than their bosses. But this was not the case; in fact, differences in income (and hence costs) had little impact on tolerance toward cheating. Second, when the implicit costs and bene®ts are held constant but the prior social relation is removed (as in the classi®ed ad study), the noblesse oblige effect disappears suggesting that it is asymmetries in the social relation and not asymmetries in costs and bene®ts that underlies the effect. These predictions and interpretation of the data, however, carry an implicit assumption that may not be warranted±that model-based predictions concerning cheating translate freely into predictions of tolerance for cheating. This may not be the case. More cheating does not necessarily mean higher tolerance of cheating by others.4 Instead of asymmetrical costs and bene®ts, our results suggest that status impacts expectations concerning appropriate behavior. Cheating a person of lower-status appears to be more unacceptable than cheating a person of higher-status. Together with the results of Cummins (1999) and Mealey et al. (1996), these results suggest that high-status carries with it an expectation of pastoral responsibility (Bugental 2000, Fiske 1991)Ðmonitoring compliance with laws and contracts, yet showing tolerance during enforcement, particularly if the miscreant is of lower status than the cheated


165

individual. It is too soon, however, to draw any de®nitive conclusions about the role played by status with this data base. More research on this topic is needed. Our cross-cultural results are problematic for both DT and SCT. All of the ledger task studies that we have presented above were conducted in Germany (in Berlin and Bonn). However, when similar studies are conducted in North America, the noblesse oblige effect is attenuated or absent. In Vancouver, Canada, for example, the noblesse oblige effect is found for perceptions of fairness, but not for judgments of the likelihood of continuing the carpool; whereas, in Davis, California, neither perceptions of fairness nor likelihood of continuing judgments showed the noblesse oblige effect (Fiddick et al., in prep.). Such cultural variability could be interpreted to mean that cultural evolution as opposed to the biological evolution as the source of social norms (cf. Henrich 2000). Cultural variability, however, is not inherently problematic for evolutionary theorizing. A simple explanation for intra-species variation is that the trait in question is facultative, i.e. adaptively activated in some environments but not others. For example, many species alternate between territoriality and dominance hierarchies. Territoriality is an alternative means of regulating access to resources in which an individual (or group) defends a territory, within the boundaries of which the individual enjoys priority of access to resources. Territoriality can break down, for example, when the population density increases to the point where it no longer pays individuals to defend territories, at which point populations often shift to a hierarchical form of social organization. Given that humans also defend territories, one might postulate that both forms of social structure are within the repertoire of evolved human social cognition, with reasoning in terms of dominance rank being activated in some populations/contexts, but not others. Though highly speculative, the higher population densities of Europe might promote a tendency to reason in terms of dominance hierarchies while the sparser population densities of western North America might promote a tendency to reason in terms of personal territory. We offer these speculations merely to point out that the cultural variation that we observed is not prima facie evidence against an evolutionary interpretation of the results. Obviously, further investigation will be needed to elucidate the precise cause of these cross-cultural differences. More problematic is the direction of the observed effect, the noblesse oblige effect, which goes contrary to predictions made by Boyd's model. Clearly, models of reciprocation must be developed which explicitly de®ne the role played by status in the evaluation of reciprocal arrangements. The implications for experimental economics Given the apparently disparate results produced by the selection task and our ledger task, with the former eliciting highly variable performance and the latter eliciting a more consistent pattern of judgments, it is worth considering just how generalizable the noblesse oblige effect is methodologically. Therefore, in this ®nal section we will review the results of some recent studies of the ultimatum and dictator games that parallel our

166

FIDDICK AND CUMMINS

own ®ndings with the ledger task and consider the implications this has for some recent moves to provide a uni®ed account for these two games. The in¯uence of social dominance on people's economic decision-making can potentially be seen in the contrast between the dictator and the ultimatum game. Of the two, the dictator game most closely models a dominance hierarchy, with the dictator possessing priority of access to resources. The ultimatum game, on the other hand, provides the responder with the opportunity to punish an unfair proposer by declining the offer and, hence, there is a more equal sharing of power in the ultimatum game. Two experimental results, therefore, are worth noting. First, as reviewed previously, social dominance does not entail the merciless exploitation of others. A clear majority of dictators routinely allocate a signi®cant share of the available resources to their powerless partners. Second, social dominance appears to suppress the exploitation of others vis-aÁ-vis the ultimatum game when the threat of punishment is reduced. This is suggested by a recent set of experiments conducted by van Dijk & Vermunt (2000). Van Dijk & Vermunt (2000) had people play the dictator and ultimatum game under conditions of symmetric and asymmetric information. Under the condition of symmetric information, both players knew that the proposers/dictators would receive twice as much money for each of the tokens being divided. Under the condition of asymmetric information, only the proposers/dictators knew that they would receive twice as much money for each token. In the ultimatum game, proposers made modal offers of two thirds of the tokens (an equal monetary value distribution) for their opponents in the symmetric information condition, but they exploited their opponents' ignorance in the asymmetric information condition by making a seemingly fair offer to split the tokens in half (with more of the actual monetary surplus going to the proposers). Conversely, in the dictator game, dictators made a modal offer of two thirds of the tokens for their opponents under both information conditions. As van Dijk & Vermunt interpreted these results, differences in the distribution of power between these two games were responsible for participants acting more generously in the dictator game than in the ultimatum game. These results would appear to converge with those we observed with the ledger task, where participants cued into the perspective of the more powerful boss are more generous (in tolerating their partner's cheating) than participants cued into the perspective of the less powerful employee.5 The role of power asymmetries in eliciting a fair division of resources is complicated, as it would also appear to depend upon the particular institution in which they occur. For example, GuÈth & van Damme (1998) have observed that powerless third parties in a three-player version of the ultimatum game are routinely excluded from a share of the resources by the two players who do have the power to in¯uence the distribution. Why is it, therefore, that in one institutional context, the dictator game, the powerful act fairly, while in another institutional context, the three-player ultimatum game, the powerful act unfairly? It would appear that experimental economists are faced with a similar dif®culty to the one that we encountered comparing our German and North American results, only here cultural differences cannot be responsible since both studies were conducted in the Netherlands.


167

There may be a common explanation for all these results in that offered by van Dijk & Vermunt (2000) for their own results. Namely, they suggest that in situations involving strong power asymmetries, normative considerations will predominate, whereas in situations involving more equal power relations, strategic considerations will predominate. In the ultimatum game, the responder's ability to reject an unfair offer forces the proposer to reason strategically about his offer, lest it be rejected, regardless of whether he is playing the standard two player version of the game or GuÈth & van Damme's three-player variant. The dictator game imposes no such strategic concerns, so normative considerations of fairness dominate decision-making instead. To the extent that Germans and North Americans attribute different levels of power asymmetry to the boss/employee relationÐi.e. to the extent that Germans view the relationship more hierarchicallyÐnormative concerns could have played a larger role in our German participants' performance on the ledger task, with strategic concerns playing a larger role in our North American participants' performance. Of course, this explanation would still require some account of why normative considerations would produce the noblesse oblige effect, i.e. greater tolerance of cheating on the part of high-ranking individuals, but at least it provides means of resolving the paradox by suggesting that strategic considerations may predominate in interactions among social equals while normative considerations may predominate in interactions between individuals of unequal social status.

Acknowledgements The writing of this paper was initiated while Laurence Fiddick was a visiting research fellow at the Max Planck Project Group: Law, Politics and Economics in Bonn, Germany. We would like to thank them for their generous support. We would also like to thank the ABC group at the Max Planck Institute for Human Development, Charles Crawford and the SFU evolutionary psychology lab, Christoph Engel, Maria Janicki, Dennis Krebs, Oded Stark, XT Wang and two anonymous reviewers for helpful discussions and comments.

Notes 1. Adaptations are designed solutions for past problems, not present problems (Tooby & Cosmides 1990). Therefore, the psychological mechanisms invoked by present problems will tend to anticipate the ancestral structure of a problem, not its current structure. 2. Given the small sample sizes, N 13 in both conditions, the difference is only marginally signi®cant. 3. We would like to thank Masanori Takezawa for bringing this to our attention. 4. We are indebted to X.T. Wang for pointing out this possibility. 5. It should also be noted that van Dijk & Vermunt conducted their studies with European participants. Whether their observations would hold for a North American participant population is an open question.

168

FIDDICK AND CUMMINS

References cited Aruguete, Mara. 1994. Cognition, tradition, and the explanation of social behavior in non-human primates. [Review of Social Processes and Mental Abilities in Non Human Primates] American Journal of Primatology 33:71±74 Axelrod, Robert & William D. Hamilton. 1981. The evolution of cooperation. Science 211:1390±1396. Axelrod, Robert. 1984. The Evolution of Cooperation. Basic Books, New York. Ben-Ner, Avner & Louis Putterman. 2000. On some implications of evolutionary psychology for the study of preferences and institutions. Journal of Economic Behavior & Organization 43:91±99. Berg, Joyce, John Dickhaut & Kevin McCabe. 1995. Trust, reciprocity, and social history. Games and Economic Behavior 10:122±142. Boehm, Christopher. 1992. Segmentary `warfare' and the management of con¯ict: comparison of East African chimpanzees and patrilineal-patrilocal humans. Pp. 137±175 in A. Harcourt & F.B.M. de Waal (ed.) Coalitions and Alliances in Humans and Other Animals, Oxford University Press, Oxford. Bolton, Gary E. & Axel Ockenfels. 2000. ERC: A theory of equity, reciprocity, and competition. American Economic Review 90:166±193. Boyd, Robert. 1992. The evolution of reciprocity when conditions vary. Pp. 473±489 in A. Harcourt & F.B.M. de Waal (ed.), Coalitions and Alliances in Humans and Other Animals, Oxford University Press, Oxford. Bugental, Daphne B. 2000. Acquisition of the algorithms of social life: a domain-based approach. Psychological Bulletin 126:187±219. Camerer, Colin & Richard H. Thaler. 1995. Anomalies: ultimatums, dictators and manners. Journal of Economic Perspectives 9:209±219. Chapais, Bernard. 1992. Role of alliances in the social inheritance of rank among female primates. Pp. 29±60 in A. Harcourt & F.B.M de Waal, (ed.) Coalitions and Alliances in Humans and Other Animals, Oxford University Press, Oxford. Cheney, Dorothy L. 1983. Extra-familial alliances among vervet monkeys. Pp. 278±286 in R.A. Hinde (ed.) Primate Social Relationships, Blackwell, Oxford. Cheng, Patricia W. & Keith J. Holyoak. 1985. Pragmatic reasoning schemas. Cognitive Psychology 17:391±416. Cheng, Patricia W. & Keith J. Holyoak. 1989. On the natural selection of reasoning theories. Cognition 33:285±313. Cheng, Patricia W., Keith J. Holyoak, Richard E. Nisbett & Lindsay M. Oliver. 1986. Pragmatic versus syntactic approaches to training deductive reasoning. Cognitive Psychology 18:293±328. Clutton-Brock, Tim H. & Paul H Harvey. 1976. Evolutionary rules and primate societies. Pp. 195±238 in P.P. G. Bateson & R.A. Hinde, (ed.) Growing Points in Ethology, Cambridge University Press, Cambridge. Cosmides, Leda. 1989. The logic of social exchange: has natural selection shaped how humans reason? Studies with the Wason selection task. Cognition 31:187±276. Cosmides, Leda & John Tooby. 1989. Evolutionary psychology and the generation of culture, part II. Case study: a computational theory of social exchange. Ethology and Sociobiology 10:51±97. Cosmides, Leda & John Tooby. 1992. Cognitive adaptations for social exchange. Pp. 163±228 in J. Barkow, L. Cosmides, & J. Tooby (ed.) The Adapted Mind. Oxford University Press, New York. Cosmides, Leda & John Tooby. 1994. Better than rational: evolutionary psychology and the invisible hand. American Economic Review 84:327±32. Cummins, Denise Dellarosa. 1996a. Dominance hierarchies and the evolution of human reasoning. Minds & Machines 6:463±480. Cummins, Denise Dellarosa. 1996b. Evidence for the innateness of deontic reasoning. Mind & Language 11:160±190. Cummins, Denise Dellarosa. 1996c. Evidence of deontic reasoning in 3- and 4-year-olds. Memory & Cognition 24:823±829. Cummins, Denise Dellarosa. 1998. Social norms and other minds: the evolutionary roots of higher cognition. Pp. 30±50 in D. D. Cummins, & C. Allen (ed.) The Evolution of Mind, Oxford University Press, New York.


169

Cummins, Denise Dellarosa. 1999. Cheater detection is modi®ed by social rank: the impact of dominance on the evolution of cognitive functions. Evolution & Human Behavior 20:229±248. Cummins, Denise Dellarosa & Laurence Fiddick. Under review. Noblesse oblige: greater tolerance for cheating among subordinates in reciprocal social contracts. Manuscript submitted for publication. Dawes, Robyn M. & Richard H. Thaler. 1988. Anomalies: cooperation. Journal of Economic Perspectives 2:187±197. Dunn, Judy. 1988. The beginnings of social understanding. Basil Blackwell, Oxford. Evans, Jonathan St. B. T., Stephen E. Newstead & Ruth M. J. Byrne. 1993. Human reasoning: The psychology of deduction. Lawrence Erlbaum Associates, Hove, UK. Fehr, Ernst & Simon GaÈchter. 1998. Reciprocity and economics: the economic implications of Homo Reciprocans. European Economic Review 42:845±859. Fehr, Ernst & Simon GaÈchter. 2000. Cooperation and punishment in public goods experiments. American Economic Review 90:980±994. Fehr, Ernst, Simon GaÈchter & Georg Kirchsteiger. 1997. Reciprocity as a contract enforcement device: experimental evidence. Econometrica 65:833±860. Fehr, Ernst & Klaus M. Schmidt. 1999. A theory of fairness, competition, and cooperation. Quarterly Journal of Economics 114:817±868. Fiddick, Laurence. 1998. The deal and the danger: an evolutionary analysis of deontic reasoning. Unpublished doctoral dissertation. University of California Santa Barbara. Fiddick, Laurence, Leda Cosmides & John Tooby. 2000. No interpretation without representation: the role of domain-speci®c representations and inferences in the Wason selection task. Cognition 77:1±79. Fiddick, Laurence, Denise Dellarosa Cummins & Maria Janicki. In prep. Cultural differences in tolerance of cheating among partners with unequal social rank. Fiske, Alan P. 1991. Structures of social life: the four elementary forms of human relations. Free Press, New York. Forsythe, Robert, Joel L. Howowitz, N.E. Savin & Martin Sefton. 1994. Fairness in simple bargaining experiments. Games & Economic Behavior 6:347±369. Frankel, Daniel G. & Tali Arbel. 1980. Group formation by two-year-olds. International Journal of Behavioral Development 3:287±298. Gigerenzer, Gerd & Klaus Hug. 1992. Domain speci®c reasoning: social contracts, cheating, and perspective change. Cognition 43:127±171. Gintis, Herbert. 2000. Strong reciprocity and human sociality. Journal of Theoretical Biology 206:169±179. Girotto, Vittorio, AgneÁs Blaye & Fernand Farioli. 1989. A reason to reason: pragmatic basis of children's search for counterexamples. Cahiers de Psychologie Cognitive 9:297±321. Griggs, Richard A. & James R. Cox. 1982. The elusive thematic-materials effect in Wason's selection task. British Journal of Psychology 73:407±420. GuÈth, Werner. 1995. An evolutionary approach to explaining cooperative behavior by reciprocal incentives. International Journal of Game Theory 24:323±344. GuÈth, Werner & Eric van Damme. 1998. Information, strategic behavior, and fairness in ultimatum bargaining: An experimental study. Journal of Mathematical Psychology 42:227±247. GuÈth, Werner & Menahem E. Yaari. 1992. Explaining reciprocal behavior in simple strategic games: an evolutionary approach. Pp. 23±34 in U. Witt (ed.) Explaining Process and Change: Approaches to Evolutionary Economics, University of Michigan Press, Ann Arbor, MI. Hall, K.R.L. 1964. Aggression in monkey and ape societies. Pp. 51±64 in J. Carthy & F. Ebling (ed.) The Natural History of Aggression, Academic Press, London. Harris, Paul L. & Maria NunÄez. 1996. Understanding of permission rules by preschool children. Child Development 67:1572±1591. Henrich, Joseph. 2000. Does culture matter in economic behavior? Ultimatum game bargaining among the Machiguenga of the Peruvian Amazon. American Economic Review 90:973±979. Hoffman, Elizabeth, Kevin McCabe, Jason Shachat & Vernon Smith. 1994. Preferences, property rights, and anonymity in bargaining games. Games & Economic Behavior 7:346±380.

170

FIDDICK AND CUMMINS

Hoffman, Elizabeth, Kevin McCabe & Vernon Smith. 1996. Social distance and other-regarding behavior in dictator games. American Economic Review 86:653±660. Hoffman, Elizabeth, Kevin McCabe & Vernon Smith. 1997. Reciprocity: The behavioral foundations of socio-economic games. Pp. 328±344 in W. Albers, W. GuÈth, P. Hammerstein, B. Moldovanu, & E. van Damme (ed.) Understanding Strategic Interaction: Essays in Honor of Reinhard Selten, SpringerVerlag, London. Hoffman, Elizabeth, Kevin McCabe & Vernon Smith. 1998. Behavioral foundations of reciprocity: experimental economics and evolutionary psychology. Economic Inquiry 36:335±352. Holyoak, Keith J. & Patricia Cheng. 1995. Pragmatic reasoning with a point of view. Thinking and Reasoning 1:289±313. Johannesson, Magnus & BjoÈrn Persson. 2000. Non-reciprocal altruism in dictator games. Economics Letters 69:137±142. Johnson-Laird, Philip N., Legrenzi, Paolo & Legrenzi, Maria S. 1972. Reasoning and a sense of reality. British Journal of Psychology 63:395±400. Kirby, Kris N. 1994. Probabilities and utilities of ®ctional outcomes in Wason's four-card selection task. Cognition 51:1±28. Manktelow, Ken I. & David E. Over. 1990. Deontic thought and the selection task. Pp. 153±164 in K. Gilhooly, M. Keane, R. Logie, & G. Erdos (ed.) Lines of Thought: Re¯ections of the Psychology of Thinking, Wiley, London. Manktelow, Ken I. & David E. Over. 1991. Social roles and utilities in reasoning with deontic conditionals. Cognition 39:85±105. Manktelow, Ken I. & David E. Over. 1995. Deontic reasoning. Pp. 91±114 in S. Newstead & J. St. B.T. Evans (ed.) Perspectives on Thinking and Reasoning: Essays in Honour of Peter Wason, Erlbaum, Hove, England. Maynard Smith, John. 1982. Evolution and the theory of games. Cambridge University Press, Cambridge. McCabe, Kevin, Stephen Rassenti & Vernon Smith. 1998. Reciprocity, trust, and payoff privacy in extensive form bargaining. Games & Economic Behavior 24:10±24. Mealey, Linda, Christopher Daood & Michael Krage. 1996. Enhanced memory for faces of cheaters. Ethology and Sociobiology 17:119±128. Oaksford, Mike & Nick Chater. 1994. A rational analysis of the selection task as optimal data selection. Psychological Review 101:608±631. Platt, Richard D. & Richard A. Griggs. 1993. Darwinian algorithms and the Wason selection task: a factorial analysis of social contract selection task problems. Cognition 48:163±192. Politzer, Guy & Anh Nguyen-Xuan. 1992. Reasoning about conditional promises and warnings: Darwinian algorithms, mental models, relevance judgments or pragmatic schemas? Quarterly Journal of Experimental Psychology 44A:401±421. Rabin, Matthew. 1993. Incorporating fairness into game theory and economics. American Economic Review 83:1281±1302. Smith, Peter K. 1988. The cognitive demands of children's social interactions with peers. Pp. 94±110 in R.W. Byrne & A. Whiten (ed.) Machiavellian Intelligence: Social Expertise and the Evolution of Intellect in Monkeys, Apes, and Humans, Oxford University Press, Oxford. Sperber, Dan, Francesco Cara & Vittorio Girotto. 1995. Relevance theory explains the selection task. Cognition 57:31±95. Stenning, Keith & Michiel van Lambalgen. 2001. Semantics as a foundation for psychology: a case study of Wason's selection task. Journal of Logic, Language and Information 10:273±317. Thaler, Richard H. 1988. Anomalies: the ultimatum game. Journal of Economic Perspectives 2:195±206. Tooby, John & Leda Cosmides. 1990. The past explains the present: emotional adaptations and the structure of ancestral environments. Ethology and Sociobiology, 11:375±424. Trivers, Robert. 1971. The evolution of reciprocal altruism. Quarterly Review of Biology 46:35±57. van Dijk, Eric & RieÈl Vermunt. 2000. Strategy and fairness in social decision making: sometimes it pays to be powerless. Journal of Experimental Social Psychology 36:1±25. Wason, Peter. 1968. Reasoning about a rule. Quarterly Journal of Experimental Psychology 20:273±281.